{
    "segments": [
        {
            "id": 1,
            "seek": 2642,
            "start": 2.53,
            "end": 8.71,
            "text": " Welcome to the last talk of the AI village from Fahid.",
            "tokens": [
                50404,
                4027,
                281,
                264,
                1036,
                751,
                295,
                264,
                7318,
                7288,
                490,
                479,
                545,
                327,
                13,
                50680
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2901785845284934,
            "compression_ratio": 1.5458515283842795,
            "no_speech_prob": 0.1287841796875,
            "words": [
                {
                    "start": 2.53,
                    "end": 3.05,
                    "word": " Welcome",
                    "probability": 0.42822265625
                },
                {
                    "start": 3.05,
                    "end": 3.57,
                    "word": " to",
                    "probability": 0.90185546875
                },
                {
                    "start": 3.57,
                    "end": 4.29,
                    "word": " the",
                    "probability": 0.8212890625
                },
                {
                    "start": 4.29,
                    "end": 5.05,
                    "word": " last",
                    "probability": 0.80126953125
                },
                {
                    "start": 5.05,
                    "end": 5.49,
                    "word": " talk",
                    "probability": 0.822265625
                },
                {
                    "start": 5.49,
                    "end": 5.83,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 5.83,
                    "end": 5.99,
                    "word": " the",
                    "probability": 0.744140625
                },
                {
                    "start": 5.99,
                    "end": 6.17,
                    "word": " AI",
                    "probability": 0.8955078125
                },
                {
                    "start": 6.17,
                    "end": 6.59,
                    "word": " village",
                    "probability": 0.3642578125
                },
                {
                    "start": 6.59,
                    "end": 7.45,
                    "word": " from",
                    "probability": 0.5986328125
                },
                {
                    "start": 7.45,
                    "end": 8.19,
                    "word": " Fahid",
                    "probability": 0.7713216145833334
                },
                {
                    "start": 8.19,
                    "end": 8.71,
                    "word": ".",
                    "probability": 0.68798828125
                }
            ]
        },
        {
            "id": 2,
            "seek": 2642,
            "start": 11.71,
            "end": 14.05,
            "text": " I'll just let him take it away.",
            "tokens": [
                50868,
                286,
                603,
                445,
                718,
                796,
                747,
                309,
                1314,
                13,
                50948
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2901785845284934,
            "compression_ratio": 1.5458515283842795,
            "no_speech_prob": 0.1287841796875,
            "words": [
                {
                    "start": 11.71,
                    "end": 12.23,
                    "word": " I'll",
                    "probability": 0.9501953125
                },
                {
                    "start": 12.23,
                    "end": 12.43,
                    "word": " just",
                    "probability": 0.8896484375
                },
                {
                    "start": 12.43,
                    "end": 12.67,
                    "word": " let",
                    "probability": 0.9013671875
                },
                {
                    "start": 12.67,
                    "end": 12.89,
                    "word": " him",
                    "probability": 0.90478515625
                },
                {
                    "start": 12.89,
                    "end": 13.15,
                    "word": " take",
                    "probability": 0.8037109375
                },
                {
                    "start": 13.15,
                    "end": 13.27,
                    "word": " it",
                    "probability": 0.900390625
                },
                {
                    "start": 13.27,
                    "end": 13.53,
                    "word": " away",
                    "probability": 0.8076171875
                },
                {
                    "start": 13.53,
                    "end": 14.05,
                    "word": ".",
                    "probability": 0.98779296875
                }
            ]
        },
        {
            "id": 3,
            "seek": 2642,
            "start": 19.41,
            "end": 23.75,
            "text": " Hello and thank you for joining this talk on the security issues and challenges in deep",
            "tokens": [
                51036,
                2425,
                293,
                1309,
                291,
                337,
                5549,
                341,
                751,
                322,
                264,
                3825,
                2663,
                293,
                4759,
                294,
                2452,
                51240
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2901785845284934,
            "compression_ratio": 1.5458515283842795,
            "no_speech_prob": 0.1287841796875,
            "words": [
                {
                    "start": 19.41,
                    "end": 19.93,
                    "word": " Hello",
                    "probability": 0.307861328125
                },
                {
                    "start": 19.93,
                    "end": 20.13,
                    "word": " and",
                    "probability": 0.498046875
                },
                {
                    "start": 20.13,
                    "end": 20.35,
                    "word": " thank",
                    "probability": 0.8212890625
                },
                {
                    "start": 20.35,
                    "end": 20.43,
                    "word": " you",
                    "probability": 0.8935546875
                },
                {
                    "start": 20.43,
                    "end": 20.71,
                    "word": " for",
                    "probability": 0.896484375
                },
                {
                    "start": 20.71,
                    "end": 21.17,
                    "word": " joining",
                    "probability": 0.88720703125
                },
                {
                    "start": 21.17,
                    "end": 21.37,
                    "word": " this",
                    "probability": 0.85888671875
                },
                {
                    "start": 21.37,
                    "end": 21.69,
                    "word": " talk",
                    "probability": 0.80322265625
                },
                {
                    "start": 21.69,
                    "end": 22.29,
                    "word": " on",
                    "probability": 0.888671875
                },
                {
                    "start": 22.29,
                    "end": 22.45,
                    "word": " the",
                    "probability": 0.74658203125
                },
                {
                    "start": 22.45,
                    "end": 22.71,
                    "word": " security",
                    "probability": 0.775390625
                },
                {
                    "start": 22.71,
                    "end": 22.99,
                    "word": " issues",
                    "probability": 0.7822265625
                },
                {
                    "start": 22.99,
                    "end": 23.15,
                    "word": " and",
                    "probability": 0.89208984375
                },
                {
                    "start": 23.15,
                    "end": 23.45,
                    "word": " challenges",
                    "probability": 0.73779296875
                },
                {
                    "start": 23.45,
                    "end": 23.61,
                    "word": " in",
                    "probability": 0.7822265625
                },
                {
                    "start": 23.61,
                    "end": 23.75,
                    "word": " deep",
                    "probability": 0.85302734375
                }
            ]
        },
        {
            "id": 4,
            "seek": 2642,
            "start": 23.75,
            "end": 29.19,
            "text": " reinforcement learning. I'm Fahid Behzadan, I'm an assistant professor of computer science and",
            "tokens": [
                51240,
                29280,
                2539,
                13,
                286,
                478,
                479,
                545,
                327,
                13068,
                89,
                345,
                282,
                11,
                286,
                478,
                364,
                10994,
                8304,
                295,
                3820,
                3497,
                293,
                51508
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2901785845284934,
            "compression_ratio": 1.5458515283842795,
            "no_speech_prob": 0.1287841796875,
            "words": [
                {
                    "start": 23.75,
                    "end": 24.15,
                    "word": " reinforcement",
                    "probability": 0.86181640625
                },
                {
                    "start": 24.15,
                    "end": 24.61,
                    "word": " learning",
                    "probability": 0.8271484375
                },
                {
                    "start": 24.61,
                    "end": 25.13,
                    "word": ".",
                    "probability": 0.94384765625
                },
                {
                    "start": 25.93,
                    "end": 26.45,
                    "word": " I'm",
                    "probability": 0.983642578125
                },
                {
                    "start": 26.45,
                    "end": 26.71,
                    "word": " Fahid",
                    "probability": 0.9737955729166666
                },
                {
                    "start": 26.71,
                    "end": 27.13,
                    "word": " Behzadan",
                    "probability": 0.56475830078125
                },
                {
                    "start": 27.13,
                    "end": 27.19,
                    "word": ",",
                    "probability": 0.430419921875
                },
                {
                    "start": 27.19,
                    "end": 27.29,
                    "word": " I'm",
                    "probability": 0.881103515625
                },
                {
                    "start": 27.29,
                    "end": 27.39,
                    "word": " an",
                    "probability": 0.8837890625
                },
                {
                    "start": 27.39,
                    "end": 27.59,
                    "word": " assistant",
                    "probability": 0.62744140625
                },
                {
                    "start": 27.59,
                    "end": 27.93,
                    "word": " professor",
                    "probability": 0.765625
                },
                {
                    "start": 27.93,
                    "end": 28.17,
                    "word": " of",
                    "probability": 0.91064453125
                },
                {
                    "start": 28.17,
                    "end": 28.47,
                    "word": " computer",
                    "probability": 0.55078125
                },
                {
                    "start": 28.47,
                    "end": 28.93,
                    "word": " science",
                    "probability": 0.853515625
                },
                {
                    "start": 28.93,
                    "end": 29.19,
                    "word": " and",
                    "probability": 0.890625
                }
            ]
        },
        {
            "id": 5,
            "seek": 2642,
            "start": 29.19,
            "end": 32.55,
            "text": " data science at the University of New Haven and I also direct the Secure and Assured",
            "tokens": [
                51508,
                1412,
                3497,
                412,
                264,
                3535,
                295,
                1873,
                23770,
                293,
                286,
                611,
                2047,
                264,
                3306,
                540,
                293,
                6281,
                3831,
                51700
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2901785845284934,
            "compression_ratio": 1.5458515283842795,
            "no_speech_prob": 0.1287841796875,
            "words": [
                {
                    "start": 29.19,
                    "end": 29.35,
                    "word": " data",
                    "probability": 0.763671875
                },
                {
                    "start": 29.35,
                    "end": 29.57,
                    "word": " science",
                    "probability": 0.83837890625
                },
                {
                    "start": 29.57,
                    "end": 29.69,
                    "word": " at",
                    "probability": 0.90478515625
                },
                {
                    "start": 29.69,
                    "end": 29.77,
                    "word": " the",
                    "probability": 0.787109375
                },
                {
                    "start": 29.77,
                    "end": 30.01,
                    "word": " University",
                    "probability": 0.6591796875
                },
                {
                    "start": 30.01,
                    "end": 30.19,
                    "word": " of",
                    "probability": 0.9052734375
                },
                {
                    "start": 30.19,
                    "end": 30.31,
                    "word": " New",
                    "probability": 0.884765625
                },
                {
                    "start": 30.31,
                    "end": 30.55,
                    "word": " Haven",
                    "probability": 0.6962890625
                },
                {
                    "start": 30.55,
                    "end": 30.83,
                    "word": " and",
                    "probability": 0.37744140625
                },
                {
                    "start": 30.83,
                    "end": 30.97,
                    "word": " I",
                    "probability": 0.9951171875
                },
                {
                    "start": 30.97,
                    "end": 31.23,
                    "word": " also",
                    "probability": 0.81396484375
                },
                {
                    "start": 31.23,
                    "end": 31.57,
                    "word": " direct",
                    "probability": 0.7451171875
                },
                {
                    "start": 31.57,
                    "end": 31.81,
                    "word": " the",
                    "probability": 0.77880859375
                },
                {
                    "start": 31.81,
                    "end": 32.11,
                    "word": " Secure",
                    "probability": 0.697021484375
                },
                {
                    "start": 32.11,
                    "end": 32.21,
                    "word": " and",
                    "probability": 0.5810546875
                },
                {
                    "start": 32.21,
                    "end": 32.55,
                    "word": " Assured",
                    "probability": 0.775146484375
                }
            ]
        },
        {
            "id": 6,
            "seek": 5518,
            "start": 32.55,
            "end": 38.33,
            "text": " Intelligent Learning Lab, or SAIL for short, working on AI safety, security and applications",
            "tokens": [
                50376,
                18762,
                25002,
                15205,
                10137,
                11,
                420,
                16482,
                4620,
                337,
                2099,
                11,
                1364,
                322,
                7318,
                4514,
                11,
                3825,
                293,
                5821,
                50656
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22024600699226907,
            "compression_ratio": 1.7581967213114753,
            "no_speech_prob": 0.0236663818359375,
            "words": [
                {
                    "start": 32.55,
                    "end": 33.39,
                    "word": " Intelligent",
                    "probability": 0.51153564453125
                },
                {
                    "start": 33.39,
                    "end": 33.63,
                    "word": " Learning",
                    "probability": 0.83447265625
                },
                {
                    "start": 33.63,
                    "end": 33.95,
                    "word": " Lab",
                    "probability": 0.88037109375
                },
                {
                    "start": 33.95,
                    "end": 34.25,
                    "word": ",",
                    "probability": 0.5595703125
                },
                {
                    "start": 34.25,
                    "end": 34.33,
                    "word": " or",
                    "probability": 0.8837890625
                },
                {
                    "start": 34.33,
                    "end": 34.71,
                    "word": " SAIL",
                    "probability": 0.93017578125
                },
                {
                    "start": 34.71,
                    "end": 34.89,
                    "word": " for",
                    "probability": 0.87060546875
                },
                {
                    "start": 34.89,
                    "end": 35.19,
                    "word": " short",
                    "probability": 0.90478515625
                },
                {
                    "start": 35.19,
                    "end": 35.85,
                    "word": ",",
                    "probability": 0.80126953125
                },
                {
                    "start": 35.85,
                    "end": 36.53,
                    "word": " working",
                    "probability": 0.70068359375
                },
                {
                    "start": 36.53,
                    "end": 36.73,
                    "word": " on",
                    "probability": 0.9033203125
                },
                {
                    "start": 36.73,
                    "end": 36.95,
                    "word": " AI",
                    "probability": 0.8388671875
                },
                {
                    "start": 36.95,
                    "end": 37.31,
                    "word": " safety",
                    "probability": 0.8154296875
                },
                {
                    "start": 37.31,
                    "end": 37.63,
                    "word": ",",
                    "probability": 0.8408203125
                },
                {
                    "start": 37.63,
                    "end": 37.81,
                    "word": " security",
                    "probability": 0.927734375
                },
                {
                    "start": 37.81,
                    "end": 37.95,
                    "word": " and",
                    "probability": 0.43310546875
                },
                {
                    "start": 37.95,
                    "end": 38.33,
                    "word": " applications",
                    "probability": 0.73974609375
                }
            ]
        },
        {
            "id": 7,
            "seek": 5518,
            "start": 38.33,
            "end": 43.41,
            "text": " of machine learning to cybersecurity and safety of complex systems.",
            "tokens": [
                50656,
                295,
                3479,
                2539,
                281,
                38765,
                293,
                4514,
                295,
                3997,
                3652,
                13,
                50872
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22024600699226907,
            "compression_ratio": 1.7581967213114753,
            "no_speech_prob": 0.0236663818359375,
            "words": [
                {
                    "start": 38.33,
                    "end": 38.77,
                    "word": " of",
                    "probability": 0.876953125
                },
                {
                    "start": 38.77,
                    "end": 39.01,
                    "word": " machine",
                    "probability": 0.6943359375
                },
                {
                    "start": 39.01,
                    "end": 39.31,
                    "word": " learning",
                    "probability": 0.85791015625
                },
                {
                    "start": 39.31,
                    "end": 39.49,
                    "word": " to",
                    "probability": 0.8720703125
                },
                {
                    "start": 39.49,
                    "end": 40.07,
                    "word": " cybersecurity",
                    "probability": 0.59375
                },
                {
                    "start": 40.07,
                    "end": 41.09,
                    "word": " and",
                    "probability": 0.7392578125
                },
                {
                    "start": 41.09,
                    "end": 41.59,
                    "word": " safety",
                    "probability": 0.8974609375
                },
                {
                    "start": 41.59,
                    "end": 41.85,
                    "word": " of",
                    "probability": 0.90869140625
                },
                {
                    "start": 41.85,
                    "end": 42.27,
                    "word": " complex",
                    "probability": 0.71875
                },
                {
                    "start": 42.27,
                    "end": 42.81,
                    "word": " systems",
                    "probability": 0.81298828125
                },
                {
                    "start": 42.81,
                    "end": 43.41,
                    "word": ".",
                    "probability": 0.98974609375
                }
            ]
        },
        {
            "id": 8,
            "seek": 5518,
            "start": 45.71,
            "end": 51.31,
            "text": " So here's the outline of my talk. I'm going to quickly go over the basics of deep reinforcement",
            "tokens": [
                50988,
                407,
                510,
                311,
                264,
                16387,
                295,
                452,
                751,
                13,
                286,
                478,
                516,
                281,
                2661,
                352,
                670,
                264,
                14688,
                295,
                2452,
                29280,
                51300
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22024600699226907,
            "compression_ratio": 1.7581967213114753,
            "no_speech_prob": 0.0236663818359375,
            "words": [
                {
                    "start": 45.71,
                    "end": 46.31,
                    "word": " So",
                    "probability": 0.818359375
                },
                {
                    "start": 46.31,
                    "end": 46.91,
                    "word": " here's",
                    "probability": 0.7470703125
                },
                {
                    "start": 46.91,
                    "end": 47.23,
                    "word": " the",
                    "probability": 0.82080078125
                },
                {
                    "start": 47.23,
                    "end": 47.71,
                    "word": " outline",
                    "probability": 0.765625
                },
                {
                    "start": 47.71,
                    "end": 48.17,
                    "word": " of",
                    "probability": 0.91162109375
                },
                {
                    "start": 48.17,
                    "end": 48.53,
                    "word": " my",
                    "probability": 0.90673828125
                },
                {
                    "start": 48.53,
                    "end": 48.85,
                    "word": " talk",
                    "probability": 0.8134765625
                },
                {
                    "start": 48.85,
                    "end": 48.89,
                    "word": ".",
                    "probability": 0.759765625
                },
                {
                    "start": 48.89,
                    "end": 49.01,
                    "word": " I'm",
                    "probability": 0.997802734375
                },
                {
                    "start": 49.01,
                    "end": 49.13,
                    "word": " going",
                    "probability": 0.85400390625
                },
                {
                    "start": 49.13,
                    "end": 49.33,
                    "word": " to",
                    "probability": 0.90869140625
                },
                {
                    "start": 49.33,
                    "end": 49.57,
                    "word": " quickly",
                    "probability": 0.875
                },
                {
                    "start": 49.57,
                    "end": 49.85,
                    "word": " go",
                    "probability": 0.89501953125
                },
                {
                    "start": 49.85,
                    "end": 50.15,
                    "word": " over",
                    "probability": 0.880859375
                },
                {
                    "start": 50.15,
                    "end": 50.33,
                    "word": " the",
                    "probability": 0.8154296875
                },
                {
                    "start": 50.33,
                    "end": 50.61,
                    "word": " basics",
                    "probability": 0.89501953125
                },
                {
                    "start": 50.61,
                    "end": 50.81,
                    "word": " of",
                    "probability": 0.90087890625
                },
                {
                    "start": 50.81,
                    "end": 50.95,
                    "word": " deep",
                    "probability": 0.677734375
                },
                {
                    "start": 50.95,
                    "end": 51.31,
                    "word": " reinforcement",
                    "probability": 0.8779296875
                }
            ]
        },
        {
            "id": 9,
            "seek": 5518,
            "start": 51.31,
            "end": 56.85,
            "text": " learning and reinforcement learning. There will be some math but I'll keep it to a minimum.",
            "tokens": [
                51300,
                2539,
                293,
                29280,
                2539,
                13,
                821,
                486,
                312,
                512,
                5221,
                457,
                286,
                603,
                1066,
                309,
                281,
                257,
                7285,
                13,
                51576
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22024600699226907,
            "compression_ratio": 1.7581967213114753,
            "no_speech_prob": 0.0236663818359375,
            "words": [
                {
                    "start": 51.31,
                    "end": 51.61,
                    "word": " learning",
                    "probability": 0.8203125
                },
                {
                    "start": 51.61,
                    "end": 51.75,
                    "word": " and",
                    "probability": 0.8720703125
                },
                {
                    "start": 51.75,
                    "end": 52.03,
                    "word": " reinforcement",
                    "probability": 0.80078125
                },
                {
                    "start": 52.03,
                    "end": 52.45,
                    "word": " learning",
                    "probability": 0.833984375
                },
                {
                    "start": 52.45,
                    "end": 52.87,
                    "word": ".",
                    "probability": 0.9326171875
                },
                {
                    "start": 52.87,
                    "end": 53.03,
                    "word": " There",
                    "probability": 0.80322265625
                },
                {
                    "start": 53.03,
                    "end": 53.15,
                    "word": " will",
                    "probability": 0.8046875
                },
                {
                    "start": 53.15,
                    "end": 53.29,
                    "word": " be",
                    "probability": 0.90966796875
                },
                {
                    "start": 53.29,
                    "end": 53.53,
                    "word": " some",
                    "probability": 0.8134765625
                },
                {
                    "start": 53.53,
                    "end": 53.81,
                    "word": " math",
                    "probability": 0.89453125
                },
                {
                    "start": 53.81,
                    "end": 54.01,
                    "word": " but",
                    "probability": 0.24365234375
                },
                {
                    "start": 54.01,
                    "end": 54.35,
                    "word": " I'll",
                    "probability": 0.96484375
                },
                {
                    "start": 54.35,
                    "end": 55.03,
                    "word": " keep",
                    "probability": 0.7939453125
                },
                {
                    "start": 55.03,
                    "end": 55.17,
                    "word": " it",
                    "probability": 0.8994140625
                },
                {
                    "start": 55.17,
                    "end": 55.29,
                    "word": " to",
                    "probability": 0.8896484375
                },
                {
                    "start": 55.97,
                    "end": 56.27,
                    "word": " a minimum",
                    "probability": 0.9150390625
                },
                {
                    "start": 56.27,
                    "end": 56.85,
                    "word": ".",
                    "probability": 0.88623046875
                }
            ]
        },
        {
            "id": 10,
            "seek": 5518,
            "start": 56.85,
            "end": 61.31,
            "text": " Then we'll talk about vulnerabilities of deep reinforcement learning and whether",
            "tokens": [
                51576,
                1396,
                321,
                603,
                751,
                466,
                37633,
                295,
                2452,
                29280,
                2539,
                293,
                1968,
                51816
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22024600699226907,
            "compression_ratio": 1.7581967213114753,
            "no_speech_prob": 0.0236663818359375,
            "words": [
                {
                    "start": 56.85,
                    "end": 57.15,
                    "word": " Then",
                    "probability": 0.8095703125
                },
                {
                    "start": 57.15,
                    "end": 57.31,
                    "word": " we'll",
                    "probability": 0.91162109375
                },
                {
                    "start": 57.31,
                    "end": 57.55,
                    "word": " talk",
                    "probability": 0.8125
                },
                {
                    "start": 57.55,
                    "end": 58.01,
                    "word": " about",
                    "probability": 0.80615234375
                },
                {
                    "start": 58.01,
                    "end": 58.77,
                    "word": " vulnerabilities",
                    "probability": 0.77099609375
                },
                {
                    "start": 58.77,
                    "end": 59.19,
                    "word": " of",
                    "probability": 0.9033203125
                },
                {
                    "start": 59.19,
                    "end": 59.39,
                    "word": " deep",
                    "probability": 0.82275390625
                },
                {
                    "start": 59.39,
                    "end": 59.73,
                    "word": " reinforcement",
                    "probability": 0.84423828125
                },
                {
                    "start": 59.73,
                    "end": 60.13,
                    "word": " learning",
                    "probability": 0.845703125
                },
                {
                    "start": 60.13,
                    "end": 60.63,
                    "word": " and",
                    "probability": 0.80126953125
                },
                {
                    "start": 60.63,
                    "end": 61.31,
                    "word": " whether",
                    "probability": 0.8349609375
                }
            ]
        },
        {
            "id": 11,
            "seek": 7646,
            "start": 61.31,
            "end": 68.07,
            "text": " deep RL is susceptible to classical adversarial machine learning attacks like adversarial examples",
            "tokens": [
                50380,
                2452,
                497,
                43,
                307,
                31249,
                281,
                13735,
                17641,
                44745,
                3479,
                2539,
                8122,
                411,
                17641,
                44745,
                5110,
                50700
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20106907685597739,
            "compression_ratio": 1.6331360946745561,
            "no_speech_prob": 8.219480514526367e-05,
            "words": [
                {
                    "start": 61.31,
                    "end": 61.95,
                    "word": " deep",
                    "probability": 0.07830810546875
                },
                {
                    "start": 61.95,
                    "end": 62.29,
                    "word": " RL",
                    "probability": 0.921630859375
                },
                {
                    "start": 62.29,
                    "end": 62.85,
                    "word": " is",
                    "probability": 0.8984375
                },
                {
                    "start": 62.85,
                    "end": 64.13,
                    "word": " susceptible",
                    "probability": 0.671875
                },
                {
                    "start": 64.13,
                    "end": 64.63,
                    "word": " to",
                    "probability": 0.91064453125
                },
                {
                    "start": 64.63,
                    "end": 65.55,
                    "word": " classical",
                    "probability": 0.90771484375
                },
                {
                    "start": 65.55,
                    "end": 66.01,
                    "word": " adversarial",
                    "probability": 0.927001953125
                },
                {
                    "start": 66.01,
                    "end": 66.29,
                    "word": " machine",
                    "probability": 0.64501953125
                },
                {
                    "start": 66.29,
                    "end": 66.49,
                    "word": " learning",
                    "probability": 0.8046875
                },
                {
                    "start": 66.49,
                    "end": 66.81,
                    "word": " attacks",
                    "probability": 0.74462890625
                },
                {
                    "start": 66.81,
                    "end": 67.11,
                    "word": " like",
                    "probability": 0.61767578125
                },
                {
                    "start": 67.11,
                    "end": 67.57,
                    "word": " adversarial",
                    "probability": 0.940185546875
                },
                {
                    "start": 67.57,
                    "end": 68.07,
                    "word": " examples",
                    "probability": 0.7998046875
                }
            ]
        },
        {
            "id": 12,
            "seek": 7646,
            "start": 68.07,
            "end": 73.65,
            "text": " and such. We'll talk about and develop a threat model for deep reinforcement learning. We'll",
            "tokens": [
                50700,
                293,
                1270,
                13,
                492,
                603,
                751,
                466,
                293,
                1499,
                257,
                4734,
                2316,
                337,
                2452,
                29280,
                2539,
                13,
                492,
                603,
                50976
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20106907685597739,
            "compression_ratio": 1.6331360946745561,
            "no_speech_prob": 8.219480514526367e-05,
            "words": [
                {
                    "start": 68.07,
                    "end": 68.25,
                    "word": " and",
                    "probability": 0.873046875
                },
                {
                    "start": 68.25,
                    "end": 68.51,
                    "word": " such",
                    "probability": 0.87744140625
                },
                {
                    "start": 68.51,
                    "end": 69.01,
                    "word": ".",
                    "probability": 0.9443359375
                },
                {
                    "start": 69.01,
                    "end": 69.35,
                    "word": " We'll",
                    "probability": 0.732421875
                },
                {
                    "start": 69.35,
                    "end": 69.61,
                    "word": " talk",
                    "probability": 0.77880859375
                },
                {
                    "start": 69.61,
                    "end": 70.01,
                    "word": " about",
                    "probability": 0.8193359375
                },
                {
                    "start": 70.01,
                    "end": 70.71,
                    "word": " and",
                    "probability": 0.8251953125
                },
                {
                    "start": 70.71,
                    "end": 71.17,
                    "word": " develop",
                    "probability": 0.67919921875
                },
                {
                    "start": 71.51,
                    "end": 71.75,
                    "word": " a threat",
                    "probability": 0.5283203125
                },
                {
                    "start": 71.75,
                    "end": 72.09,
                    "word": " model",
                    "probability": 0.7900390625
                },
                {
                    "start": 72.09,
                    "end": 72.53,
                    "word": " for",
                    "probability": 0.88671875
                },
                {
                    "start": 72.53,
                    "end": 72.69,
                    "word": " deep",
                    "probability": 0.8671875
                },
                {
                    "start": 72.69,
                    "end": 73.07,
                    "word": " reinforcement",
                    "probability": 0.830078125
                },
                {
                    "start": 73.07,
                    "end": 73.43,
                    "word": " learning",
                    "probability": 0.8046875
                },
                {
                    "start": 73.43,
                    "end": 73.55,
                    "word": ".",
                    "probability": 0.7265625
                },
                {
                    "start": 73.55,
                    "end": 73.65,
                    "word": " We'll",
                    "probability": 0.8330078125
                }
            ]
        },
        {
            "id": 13,
            "seek": 7646,
            "start": 73.65,
            "end": 82.59,
            "text": " identify different attack models, attack surfaces, we'll identify different types of",
            "tokens": [
                50976,
                5876,
                819,
                2690,
                5245,
                11,
                2690,
                16130,
                11,
                321,
                603,
                5876,
                819,
                3467,
                295,
                51428
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20106907685597739,
            "compression_ratio": 1.6331360946745561,
            "no_speech_prob": 8.219480514526367e-05,
            "words": [
                {
                    "start": 73.65,
                    "end": 74.19,
                    "word": " identify",
                    "probability": 0.88330078125
                },
                {
                    "start": 74.19,
                    "end": 75.83,
                    "word": " different",
                    "probability": 0.6435546875
                },
                {
                    "start": 75.83,
                    "end": 78.39,
                    "word": " attack",
                    "probability": 0.66650390625
                },
                {
                    "start": 78.39,
                    "end": 78.93,
                    "word": " models",
                    "probability": 0.8408203125
                },
                {
                    "start": 78.93,
                    "end": 79.15,
                    "word": ",",
                    "probability": 0.9326171875
                },
                {
                    "start": 79.15,
                    "end": 79.35,
                    "word": " attack",
                    "probability": 0.802734375
                },
                {
                    "start": 79.35,
                    "end": 79.79,
                    "word": " surfaces",
                    "probability": 0.8232421875
                },
                {
                    "start": 79.79,
                    "end": 80.19,
                    "word": ",",
                    "probability": 0.383056640625
                },
                {
                    "start": 80.19,
                    "end": 80.55,
                    "word": " we'll",
                    "probability": 0.90625
                },
                {
                    "start": 80.55,
                    "end": 81.11,
                    "word": " identify",
                    "probability": 0.9052734375
                },
                {
                    "start": 81.11,
                    "end": 81.59,
                    "word": " different",
                    "probability": 0.69287109375
                },
                {
                    "start": 81.59,
                    "end": 81.93,
                    "word": " types",
                    "probability": 0.73681640625
                },
                {
                    "start": 81.93,
                    "end": 82.59,
                    "word": " of",
                    "probability": 0.9091796875
                }
            ]
        },
        {
            "id": 14,
            "seek": 10410,
            "start": 82.59,
            "end": 92.59,
            "text": " vulnerabilities, and we'll introduce a number of attack mechanisms and corresponding defenses",
            "tokens": [
                50364,
                37633,
                11,
                293,
                321,
                603,
                5366,
                257,
                1230,
                295,
                2690,
                15902,
                293,
                11760,
                35989,
                50864
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21428571523181975,
            "compression_ratio": 1.4742268041237114,
            "no_speech_prob": 0.0010318756103515625,
            "words": [
                {
                    "start": 82.59,
                    "end": 83.13,
                    "word": " vulnerabilities",
                    "probability": 0.43310546875
                },
                {
                    "start": 83.13,
                    "end": 84.13,
                    "word": ",",
                    "probability": 0.2130126953125
                },
                {
                    "start": 84.13,
                    "end": 85.13,
                    "word": " and",
                    "probability": 0.8671875
                },
                {
                    "start": 85.13,
                    "end": 85.97,
                    "word": " we'll",
                    "probability": 0.70947265625
                },
                {
                    "start": 85.97,
                    "end": 86.31,
                    "word": " introduce",
                    "probability": 0.68408203125
                },
                {
                    "start": 86.55,
                    "end": 86.85,
                    "word": " a number",
                    "probability": 0.875
                },
                {
                    "start": 86.85,
                    "end": 87.25,
                    "word": " of",
                    "probability": 0.90625
                },
                {
                    "start": 87.25,
                    "end": 87.67,
                    "word": " attack",
                    "probability": 0.8095703125
                },
                {
                    "start": 87.67,
                    "end": 88.17,
                    "word": " mechanisms",
                    "probability": 0.6416015625
                },
                {
                    "start": 88.17,
                    "end": 91.07,
                    "word": " and",
                    "probability": 0.7783203125
                },
                {
                    "start": 91.07,
                    "end": 92.07,
                    "word": " corresponding",
                    "probability": 0.6689453125
                },
                {
                    "start": 92.07,
                    "end": 92.59,
                    "word": " defenses",
                    "probability": 0.67138671875
                }
            ]
        },
        {
            "id": 15,
            "seek": 10410,
            "start": 92.59,
            "end": 101.75,
            "text": " that have been developed in recent years. And we'll talk about the frontiers and areas of future",
            "tokens": [
                50864,
                300,
                362,
                668,
                4743,
                294,
                5162,
                924,
                13,
                400,
                321,
                603,
                751,
                466,
                264,
                1868,
                4890,
                293,
                3179,
                295,
                2027,
                51320
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21428571523181975,
            "compression_ratio": 1.4742268041237114,
            "no_speech_prob": 0.0010318756103515625,
            "words": [
                {
                    "start": 92.59,
                    "end": 92.85,
                    "word": " that",
                    "probability": 0.8759765625
                },
                {
                    "start": 92.85,
                    "end": 92.95,
                    "word": " have",
                    "probability": 0.88232421875
                },
                {
                    "start": 92.95,
                    "end": 93.07,
                    "word": " been",
                    "probability": 0.8203125
                },
                {
                    "start": 93.07,
                    "end": 93.45,
                    "word": " developed",
                    "probability": 0.65234375
                },
                {
                    "start": 93.45,
                    "end": 94.53,
                    "word": " in",
                    "probability": 0.8984375
                },
                {
                    "start": 94.53,
                    "end": 96.33,
                    "word": " recent",
                    "probability": 0.5556640625
                },
                {
                    "start": 96.33,
                    "end": 96.85,
                    "word": " years",
                    "probability": 0.87939453125
                },
                {
                    "start": 96.85,
                    "end": 97.27,
                    "word": ".",
                    "probability": 0.410888671875
                },
                {
                    "start": 97.27,
                    "end": 97.99,
                    "word": " And",
                    "probability": 0.73486328125
                },
                {
                    "start": 97.99,
                    "end": 98.17,
                    "word": " we'll",
                    "probability": 0.8955078125
                },
                {
                    "start": 98.17,
                    "end": 98.39,
                    "word": " talk",
                    "probability": 0.81787109375
                },
                {
                    "start": 98.39,
                    "end": 98.71,
                    "word": " about",
                    "probability": 0.81982421875
                },
                {
                    "start": 98.71,
                    "end": 98.97,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 98.97,
                    "end": 99.67,
                    "word": " frontiers",
                    "probability": 0.875732421875
                },
                {
                    "start": 99.67,
                    "end": 100.43,
                    "word": " and",
                    "probability": 0.85888671875
                },
                {
                    "start": 100.43,
                    "end": 101.11,
                    "word": " areas",
                    "probability": 0.8857421875
                },
                {
                    "start": 101.11,
                    "end": 101.41,
                    "word": " of",
                    "probability": 0.90625
                },
                {
                    "start": 101.41,
                    "end": 101.75,
                    "word": " future",
                    "probability": 0.9267578125
                }
            ]
        },
        {
            "id": 16,
            "seek": 10410,
            "start": 101.75,
            "end": 110.23,
            "text": " research and work in this area. So just a quick overview. I assume most of the audience here is",
            "tokens": [
                51320,
                2132,
                293,
                589,
                294,
                341,
                1859,
                13,
                407,
                445,
                257,
                1702,
                12492,
                13,
                286,
                6552,
                881,
                295,
                264,
                4034,
                510,
                307,
                51740
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21428571523181975,
            "compression_ratio": 1.4742268041237114,
            "no_speech_prob": 0.0010318756103515625,
            "words": [
                {
                    "start": 101.75,
                    "end": 102.19,
                    "word": " research",
                    "probability": 0.7880859375
                },
                {
                    "start": 102.19,
                    "end": 102.33,
                    "word": " and",
                    "probability": 0.8779296875
                },
                {
                    "start": 102.33,
                    "end": 102.55,
                    "word": " work",
                    "probability": 0.79833984375
                },
                {
                    "start": 102.55,
                    "end": 102.79,
                    "word": " in",
                    "probability": 0.86767578125
                },
                {
                    "start": 102.79,
                    "end": 104.27,
                    "word": " this",
                    "probability": 0.873046875
                },
                {
                    "start": 104.27,
                    "end": 104.63,
                    "word": " area",
                    "probability": 0.7998046875
                },
                {
                    "start": 104.63,
                    "end": 105.35,
                    "word": ".",
                    "probability": 0.978515625
                },
                {
                    "start": 105.35,
                    "end": 105.51,
                    "word": " So",
                    "probability": 0.81005859375
                },
                {
                    "start": 105.51,
                    "end": 106.63,
                    "word": " just",
                    "probability": 0.41650390625
                },
                {
                    "start": 106.75,
                    "end": 106.95,
                    "word": " a quick",
                    "probability": 0.9189453125
                },
                {
                    "start": 106.95,
                    "end": 107.39,
                    "word": " overview",
                    "probability": 0.822265625
                },
                {
                    "start": 107.39,
                    "end": 108.13,
                    "word": ".",
                    "probability": 0.6025390625
                },
                {
                    "start": 108.19,
                    "end": 108.79,
                    "word": " I",
                    "probability": 0.99365234375
                },
                {
                    "start": 108.79,
                    "end": 109.05,
                    "word": " assume",
                    "probability": 0.79638671875
                },
                {
                    "start": 109.05,
                    "end": 109.33,
                    "word": " most",
                    "probability": 0.78466796875
                },
                {
                    "start": 109.33,
                    "end": 109.43,
                    "word": " of",
                    "probability": 0.9013671875
                },
                {
                    "start": 109.43,
                    "end": 109.53,
                    "word": " the",
                    "probability": 0.80859375
                },
                {
                    "start": 109.53,
                    "end": 109.77,
                    "word": " audience",
                    "probability": 0.8203125
                },
                {
                    "start": 109.77,
                    "end": 110.11,
                    "word": " here",
                    "probability": 0.80615234375
                },
                {
                    "start": 110.11,
                    "end": 110.23,
                    "word": " is",
                    "probability": 0.89990234375
                }
            ]
        },
        {
            "id": 17,
            "seek": 12926,
            "start": 110.23,
            "end": 117.03,
            "text": " already familiar with the terminology used here. We can classify machine learning algorithms as",
            "tokens": [
                50364,
                1217,
                4963,
                365,
                264,
                27575,
                1143,
                510,
                13,
                492,
                393,
                33872,
                3479,
                2539,
                14642,
                382,
                50704
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24185739142794005,
            "compression_ratio": 1.8426395939086295,
            "no_speech_prob": 0.0008296966552734375,
            "words": [
                {
                    "start": 110.23,
                    "end": 110.39,
                    "word": " already",
                    "probability": 0.189208984375
                },
                {
                    "start": 110.39,
                    "end": 110.79,
                    "word": " familiar",
                    "probability": 0.76611328125
                },
                {
                    "start": 110.79,
                    "end": 111.35,
                    "word": " with",
                    "probability": 0.8115234375
                },
                {
                    "start": 111.35,
                    "end": 112.43,
                    "word": " the",
                    "probability": 0.7783203125
                },
                {
                    "start": 112.43,
                    "end": 112.83,
                    "word": " terminology",
                    "probability": 0.90673828125
                },
                {
                    "start": 112.83,
                    "end": 113.21,
                    "word": " used",
                    "probability": 0.78662109375
                },
                {
                    "start": 113.21,
                    "end": 113.43,
                    "word": " here",
                    "probability": 0.80810546875
                },
                {
                    "start": 113.43,
                    "end": 113.79,
                    "word": ".",
                    "probability": 0.43798828125
                },
                {
                    "start": 113.79,
                    "end": 114.47,
                    "word": " We",
                    "probability": 0.8623046875
                },
                {
                    "start": 114.47,
                    "end": 114.63,
                    "word": " can",
                    "probability": 0.89111328125
                },
                {
                    "start": 114.63,
                    "end": 115.01,
                    "word": " classify",
                    "probability": 0.9521484375
                },
                {
                    "start": 115.01,
                    "end": 115.37,
                    "word": " machine",
                    "probability": 0.63720703125
                },
                {
                    "start": 115.37,
                    "end": 115.57,
                    "word": " learning",
                    "probability": 0.826171875
                },
                {
                    "start": 115.57,
                    "end": 116.01,
                    "word": " algorithms",
                    "probability": 0.8505859375
                },
                {
                    "start": 116.01,
                    "end": 117.03,
                    "word": " as",
                    "probability": 0.8779296875
                }
            ]
        },
        {
            "id": 18,
            "seek": 12926,
            "start": 117.03,
            "end": 121.53,
            "text": " supervised, unsupervised, and reinforcement learning algorithms. Supervised learning",
            "tokens": [
                50704,
                46533,
                11,
                2693,
                12879,
                24420,
                11,
                293,
                29280,
                2539,
                14642,
                13,
                4548,
                24420,
                2539,
                50924
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24185739142794005,
            "compression_ratio": 1.8426395939086295,
            "no_speech_prob": 0.0008296966552734375,
            "words": [
                {
                    "start": 117.03,
                    "end": 117.57,
                    "word": " supervised",
                    "probability": 0.78515625
                },
                {
                    "start": 117.57,
                    "end": 117.73,
                    "word": ",",
                    "probability": 0.8818359375
                },
                {
                    "start": 117.73,
                    "end": 118.33,
                    "word": " unsupervised",
                    "probability": 0.8914388020833334
                },
                {
                    "start": 118.33,
                    "end": 118.45,
                    "word": ",",
                    "probability": 0.5537109375
                },
                {
                    "start": 118.45,
                    "end": 118.51,
                    "word": " and",
                    "probability": 0.884765625
                },
                {
                    "start": 118.51,
                    "end": 118.99,
                    "word": " reinforcement",
                    "probability": 0.7919921875
                },
                {
                    "start": 118.99,
                    "end": 119.53,
                    "word": " learning",
                    "probability": 0.83203125
                },
                {
                    "start": 119.53,
                    "end": 120.33,
                    "word": " algorithms",
                    "probability": 0.90087890625
                },
                {
                    "start": 120.33,
                    "end": 120.69,
                    "word": ".",
                    "probability": 0.90234375
                },
                {
                    "start": 120.69,
                    "end": 121.23,
                    "word": " Supervised",
                    "probability": 0.884765625
                },
                {
                    "start": 121.23,
                    "end": 121.53,
                    "word": " learning",
                    "probability": 0.84130859375
                }
            ]
        },
        {
            "id": 19,
            "seek": 12926,
            "start": 121.53,
            "end": 127.45,
            "text": " algorithms are those where the training data set includes labeled data, meaning that",
            "tokens": [
                50924,
                14642,
                366,
                729,
                689,
                264,
                3097,
                1412,
                992,
                5974,
                21335,
                1412,
                11,
                3620,
                300,
                51228
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24185739142794005,
            "compression_ratio": 1.8426395939086295,
            "no_speech_prob": 0.0008296966552734375,
            "words": [
                {
                    "start": 121.53,
                    "end": 121.91,
                    "word": " algorithms",
                    "probability": 0.84765625
                },
                {
                    "start": 121.91,
                    "end": 122.07,
                    "word": " are",
                    "probability": 0.86962890625
                },
                {
                    "start": 122.07,
                    "end": 122.31,
                    "word": " those",
                    "probability": 0.830078125
                },
                {
                    "start": 122.31,
                    "end": 122.53,
                    "word": " where",
                    "probability": 0.89208984375
                },
                {
                    "start": 122.53,
                    "end": 122.75,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 122.75,
                    "end": 123.09,
                    "word": " training",
                    "probability": 0.8427734375
                },
                {
                    "start": 123.09,
                    "end": 123.99,
                    "word": " data",
                    "probability": 0.378662109375
                },
                {
                    "start": 123.99,
                    "end": 124.27,
                    "word": " set",
                    "probability": 0.845703125
                },
                {
                    "start": 124.27,
                    "end": 124.87,
                    "word": " includes",
                    "probability": 0.8271484375
                },
                {
                    "start": 124.87,
                    "end": 125.93,
                    "word": " labeled",
                    "probability": 0.56396484375
                },
                {
                    "start": 125.93,
                    "end": 126.43,
                    "word": " data",
                    "probability": 0.80712890625
                },
                {
                    "start": 126.43,
                    "end": 126.93,
                    "word": ",",
                    "probability": 0.74365234375
                },
                {
                    "start": 126.93,
                    "end": 127.15,
                    "word": " meaning",
                    "probability": 0.8076171875
                },
                {
                    "start": 127.15,
                    "end": 127.45,
                    "word": " that",
                    "probability": 0.87255859375
                }
            ]
        },
        {
            "id": 20,
            "seek": 12926,
            "start": 128.59,
            "end": 135.39,
            "text": " each data point in that data set comes with the correct label or the correct output expected from",
            "tokens": [
                51304,
                1184,
                1412,
                935,
                294,
                300,
                1412,
                992,
                1487,
                365,
                264,
                3006,
                7645,
                420,
                264,
                3006,
                5598,
                5176,
                490,
                51624
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24185739142794005,
            "compression_ratio": 1.8426395939086295,
            "no_speech_prob": 0.0008296966552734375,
            "words": [
                {
                    "start": 128.59,
                    "end": 129.33,
                    "word": " each",
                    "probability": 0.66259765625
                },
                {
                    "start": 129.33,
                    "end": 129.63,
                    "word": " data",
                    "probability": 0.80029296875
                },
                {
                    "start": 129.63,
                    "end": 129.85,
                    "word": " point",
                    "probability": 0.86767578125
                },
                {
                    "start": 129.85,
                    "end": 129.95,
                    "word": " in",
                    "probability": 0.892578125
                },
                {
                    "start": 129.95,
                    "end": 130.09,
                    "word": " that",
                    "probability": 0.85888671875
                },
                {
                    "start": 130.09,
                    "end": 130.31,
                    "word": " data",
                    "probability": 0.72607421875
                },
                {
                    "start": 130.31,
                    "end": 130.51,
                    "word": " set",
                    "probability": 0.875
                },
                {
                    "start": 130.51,
                    "end": 130.97,
                    "word": " comes",
                    "probability": 0.814453125
                },
                {
                    "start": 130.97,
                    "end": 131.55,
                    "word": " with",
                    "probability": 0.81103515625
                },
                {
                    "start": 131.55,
                    "end": 132.45,
                    "word": " the",
                    "probability": 0.81494140625
                },
                {
                    "start": 132.45,
                    "end": 132.83,
                    "word": " correct",
                    "probability": 0.87939453125
                },
                {
                    "start": 132.83,
                    "end": 133.13,
                    "word": " label",
                    "probability": 0.78955078125
                },
                {
                    "start": 133.13,
                    "end": 133.43,
                    "word": " or",
                    "probability": 0.76904296875
                },
                {
                    "start": 133.43,
                    "end": 133.69,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 133.69,
                    "end": 134.13,
                    "word": " correct",
                    "probability": 0.8681640625
                },
                {
                    "start": 134.13,
                    "end": 134.65,
                    "word": " output",
                    "probability": 0.8349609375
                },
                {
                    "start": 134.65,
                    "end": 135.03,
                    "word": " expected",
                    "probability": 0.75390625
                },
                {
                    "start": 135.03,
                    "end": 135.39,
                    "word": " from",
                    "probability": 0.8173828125
                }
            ]
        },
        {
            "id": 21,
            "seek": 15674,
            "start": 135.65,
            "end": 142.69,
            "text": " a model trained on that data set. Then there is unsupervised learning where clustering",
            "tokens": [
                50364,
                257,
                2316,
                8895,
                322,
                300,
                1412,
                992,
                13,
                1396,
                456,
                307,
                2693,
                12879,
                24420,
                2539,
                689,
                596,
                48673,
                50748
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2173996854711462,
            "compression_ratio": 1.7045454545454546,
            "no_speech_prob": 0.0006771087646484375,
            "words": [
                {
                    "start": 135.65,
                    "end": 136.15,
                    "word": " a model",
                    "probability": 0.8251953125
                },
                {
                    "start": 136.15,
                    "end": 136.51,
                    "word": " trained",
                    "probability": 0.79541015625
                },
                {
                    "start": 136.51,
                    "end": 136.73,
                    "word": " on",
                    "probability": 0.900390625
                },
                {
                    "start": 136.73,
                    "end": 137.13,
                    "word": " that",
                    "probability": 0.88232421875
                },
                {
                    "start": 137.13,
                    "end": 137.35,
                    "word": " data",
                    "probability": 0.4482421875
                },
                {
                    "start": 137.35,
                    "end": 137.63,
                    "word": " set",
                    "probability": 0.873046875
                },
                {
                    "start": 137.63,
                    "end": 138.23,
                    "word": ".",
                    "probability": 0.94873046875
                },
                {
                    "start": 138.55,
                    "end": 138.81,
                    "word": " Then",
                    "probability": 0.80078125
                },
                {
                    "start": 138.81,
                    "end": 138.99,
                    "word": " there",
                    "probability": 0.787109375
                },
                {
                    "start": 138.99,
                    "end": 139.29,
                    "word": " is",
                    "probability": 0.89794921875
                },
                {
                    "start": 139.29,
                    "end": 139.89,
                    "word": " unsupervised",
                    "probability": 0.9138997395833334
                },
                {
                    "start": 139.89,
                    "end": 140.29,
                    "word": " learning",
                    "probability": 0.83349609375
                },
                {
                    "start": 140.29,
                    "end": 140.73,
                    "word": " where",
                    "probability": 0.48583984375
                },
                {
                    "start": 140.73,
                    "end": 142.69,
                    "word": " clustering",
                    "probability": 0.533935546875
                }
            ]
        },
        {
            "id": 22,
            "seek": 15674,
            "start": 142.69,
            "end": 147.91,
            "text": " and anomaly detection and some other algorithms fall under where there are no labels available",
            "tokens": [
                50748,
                293,
                42737,
                17784,
                293,
                512,
                661,
                14642,
                2100,
                833,
                689,
                456,
                366,
                572,
                16949,
                2435,
                50988
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2173996854711462,
            "compression_ratio": 1.7045454545454546,
            "no_speech_prob": 0.0006771087646484375,
            "words": [
                {
                    "start": 142.69,
                    "end": 143.65,
                    "word": " and",
                    "probability": 0.7705078125
                },
                {
                    "start": 143.65,
                    "end": 144.69,
                    "word": " anomaly",
                    "probability": 0.80419921875
                },
                {
                    "start": 144.69,
                    "end": 145.07,
                    "word": " detection",
                    "probability": 0.73876953125
                },
                {
                    "start": 145.07,
                    "end": 145.39,
                    "word": " and",
                    "probability": 0.849609375
                },
                {
                    "start": 145.39,
                    "end": 145.55,
                    "word": " some",
                    "probability": 0.81982421875
                },
                {
                    "start": 145.55,
                    "end": 145.71,
                    "word": " other",
                    "probability": 0.7724609375
                },
                {
                    "start": 145.71,
                    "end": 146.09,
                    "word": " algorithms",
                    "probability": 0.84814453125
                },
                {
                    "start": 146.09,
                    "end": 146.33,
                    "word": " fall",
                    "probability": 0.7294921875
                },
                {
                    "start": 146.33,
                    "end": 146.53,
                    "word": " under",
                    "probability": 0.79931640625
                },
                {
                    "start": 146.53,
                    "end": 146.77,
                    "word": " where",
                    "probability": 0.6923828125
                },
                {
                    "start": 146.77,
                    "end": 146.97,
                    "word": " there",
                    "probability": 0.85009765625
                },
                {
                    "start": 146.97,
                    "end": 147.11,
                    "word": " are",
                    "probability": 0.892578125
                },
                {
                    "start": 147.11,
                    "end": 147.27,
                    "word": " no",
                    "probability": 0.90576171875
                },
                {
                    "start": 147.27,
                    "end": 147.57,
                    "word": " labels",
                    "probability": 0.8359375
                },
                {
                    "start": 147.57,
                    "end": 147.91,
                    "word": " available",
                    "probability": 0.75341796875
                }
            ]
        },
        {
            "id": 23,
            "seek": 15674,
            "start": 147.91,
            "end": 154.67,
            "text": " for data points. There's no feedback available on the data. However, the goal is to find some",
            "tokens": [
                50988,
                337,
                1412,
                2793,
                13,
                821,
                311,
                572,
                5824,
                2435,
                322,
                264,
                1412,
                13,
                2908,
                11,
                264,
                3387,
                307,
                281,
                915,
                512,
                51324
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2173996854711462,
            "compression_ratio": 1.7045454545454546,
            "no_speech_prob": 0.0006771087646484375,
            "words": [
                {
                    "start": 147.91,
                    "end": 148.13,
                    "word": " for",
                    "probability": 0.89892578125
                },
                {
                    "start": 148.13,
                    "end": 148.31,
                    "word": " data",
                    "probability": 0.8017578125
                },
                {
                    "start": 148.31,
                    "end": 148.71,
                    "word": " points",
                    "probability": 0.7763671875
                },
                {
                    "start": 148.71,
                    "end": 149.01,
                    "word": ".",
                    "probability": 0.58251953125
                },
                {
                    "start": 149.01,
                    "end": 149.33,
                    "word": " There's",
                    "probability": 0.849853515625
                },
                {
                    "start": 149.33,
                    "end": 149.47,
                    "word": " no",
                    "probability": 0.9189453125
                },
                {
                    "start": 149.47,
                    "end": 149.75,
                    "word": " feedback",
                    "probability": 0.85107421875
                },
                {
                    "start": 149.75,
                    "end": 150.21,
                    "word": " available",
                    "probability": 0.7412109375
                },
                {
                    "start": 150.21,
                    "end": 150.93,
                    "word": " on",
                    "probability": 0.9033203125
                },
                {
                    "start": 150.93,
                    "end": 151.39,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 151.39,
                    "end": 151.69,
                    "word": " data",
                    "probability": 0.81884765625
                },
                {
                    "start": 151.69,
                    "end": 152.29,
                    "word": ".",
                    "probability": 0.85693359375
                },
                {
                    "start": 152.53,
                    "end": 153.13,
                    "word": " However",
                    "probability": 0.8251953125
                },
                {
                    "start": 153.13,
                    "end": 153.21,
                    "word": ",",
                    "probability": 0.88525390625
                },
                {
                    "start": 153.21,
                    "end": 153.39,
                    "word": " the",
                    "probability": 0.80908203125
                },
                {
                    "start": 153.39,
                    "end": 153.67,
                    "word": " goal",
                    "probability": 0.90869140625
                },
                {
                    "start": 153.67,
                    "end": 154.03,
                    "word": " is",
                    "probability": 0.90673828125
                },
                {
                    "start": 154.03,
                    "end": 154.17,
                    "word": " to",
                    "probability": 0.9091796875
                },
                {
                    "start": 154.17,
                    "end": 154.43,
                    "word": " find",
                    "probability": 0.81787109375
                },
                {
                    "start": 154.43,
                    "end": 154.67,
                    "word": " some",
                    "probability": 0.814453125
                }
            ]
        },
        {
            "id": 24,
            "seek": 15674,
            "start": 154.67,
            "end": 162.87,
            "text": " underlying structure in the data. And then there is reinforcement learning, which is concerned with",
            "tokens": [
                51324,
                14217,
                3877,
                294,
                264,
                1412,
                13,
                400,
                550,
                456,
                307,
                29280,
                2539,
                11,
                597,
                307,
                5922,
                365,
                51724
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2173996854711462,
            "compression_ratio": 1.7045454545454546,
            "no_speech_prob": 0.0006771087646484375,
            "words": [
                {
                    "start": 154.67,
                    "end": 155.07,
                    "word": " underlying",
                    "probability": 0.912109375
                },
                {
                    "start": 155.07,
                    "end": 155.61,
                    "word": " structure",
                    "probability": 0.71826171875
                },
                {
                    "start": 155.61,
                    "end": 156.11,
                    "word": " in",
                    "probability": 0.9052734375
                },
                {
                    "start": 156.11,
                    "end": 156.21,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 156.21,
                    "end": 156.47,
                    "word": " data",
                    "probability": 0.806640625
                },
                {
                    "start": 156.47,
                    "end": 157.07,
                    "word": ".",
                    "probability": 0.95361328125
                },
                {
                    "start": 157.79,
                    "end": 157.95,
                    "word": " And",
                    "probability": 0.82470703125
                },
                {
                    "start": 157.95,
                    "end": 158.13,
                    "word": " then",
                    "probability": 0.728515625
                },
                {
                    "start": 158.13,
                    "end": 158.27,
                    "word": " there",
                    "probability": 0.85205078125
                },
                {
                    "start": 158.27,
                    "end": 158.49,
                    "word": " is",
                    "probability": 0.89794921875
                },
                {
                    "start": 158.49,
                    "end": 159.07,
                    "word": " reinforcement",
                    "probability": 0.7802734375
                },
                {
                    "start": 159.07,
                    "end": 159.47,
                    "word": " learning",
                    "probability": 0.830078125
                },
                {
                    "start": 159.47,
                    "end": 159.83,
                    "word": ",",
                    "probability": 0.708984375
                },
                {
                    "start": 159.83,
                    "end": 160.91,
                    "word": " which",
                    "probability": 0.9091796875
                },
                {
                    "start": 160.91,
                    "end": 161.25,
                    "word": " is",
                    "probability": 0.90380859375
                },
                {
                    "start": 161.25,
                    "end": 162.09,
                    "word": " concerned",
                    "probability": 0.83154296875
                },
                {
                    "start": 162.09,
                    "end": 162.87,
                    "word": " with",
                    "probability": 0.81787109375
                }
            ]
        },
        {
            "id": 25,
            "seek": 18262,
            "start": 163.35,
            "end": 170.13,
            "text": " the problem of sequential decision making. There's a reward system included in reinforcement learning,",
            "tokens": [
                50392,
                264,
                1154,
                295,
                42881,
                3537,
                1455,
                13,
                821,
                311,
                257,
                7782,
                1185,
                5556,
                294,
                29280,
                2539,
                11,
                50724
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19645833293596904,
            "compression_ratio": 1.7616822429906542,
            "no_speech_prob": 0.0004172325134277344,
            "words": [
                {
                    "start": 163.35,
                    "end": 163.69,
                    "word": " the",
                    "probability": 0.38720703125
                },
                {
                    "start": 163.69,
                    "end": 164.15,
                    "word": " problem",
                    "probability": 0.73486328125
                },
                {
                    "start": 164.15,
                    "end": 164.51,
                    "word": " of",
                    "probability": 0.9091796875
                },
                {
                    "start": 164.51,
                    "end": 164.99,
                    "word": " sequential",
                    "probability": 0.9189453125
                },
                {
                    "start": 164.99,
                    "end": 165.71,
                    "word": " decision",
                    "probability": 0.85791015625
                },
                {
                    "start": 165.71,
                    "end": 166.07,
                    "word": " making",
                    "probability": 0.475341796875
                },
                {
                    "start": 166.07,
                    "end": 166.57,
                    "word": ".",
                    "probability": 0.93603515625
                },
                {
                    "start": 166.57,
                    "end": 166.89,
                    "word": " There's",
                    "probability": 0.74072265625
                },
                {
                    "start": 167.01,
                    "end": 167.31,
                    "word": " a reward",
                    "probability": 0.87646484375
                },
                {
                    "start": 167.31,
                    "end": 167.77,
                    "word": " system",
                    "probability": 0.87158203125
                },
                {
                    "start": 167.77,
                    "end": 169.01,
                    "word": " included",
                    "probability": 0.7607421875
                },
                {
                    "start": 169.01,
                    "end": 169.21,
                    "word": " in",
                    "probability": 0.8955078125
                },
                {
                    "start": 169.21,
                    "end": 169.59,
                    "word": " reinforcement",
                    "probability": 0.77099609375
                },
                {
                    "start": 169.59,
                    "end": 169.95,
                    "word": " learning",
                    "probability": 0.82373046875
                },
                {
                    "start": 169.95,
                    "end": 170.13,
                    "word": ",",
                    "probability": 0.63427734375
                }
            ]
        },
        {
            "id": 26,
            "seek": 18262,
            "start": 170.13,
            "end": 175.55,
            "text": " but there are fundamental differences between the settings of reinforcement learning and supervised",
            "tokens": [
                50724,
                457,
                456,
                366,
                8088,
                7300,
                1296,
                264,
                6257,
                295,
                29280,
                2539,
                293,
                46533,
                51024
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19645833293596904,
            "compression_ratio": 1.7616822429906542,
            "no_speech_prob": 0.0004172325134277344,
            "words": [
                {
                    "start": 170.13,
                    "end": 170.41,
                    "word": " but",
                    "probability": 0.896484375
                },
                {
                    "start": 170.41,
                    "end": 170.57,
                    "word": " there",
                    "probability": 0.84765625
                },
                {
                    "start": 170.57,
                    "end": 170.91,
                    "word": " are",
                    "probability": 0.884765625
                },
                {
                    "start": 170.91,
                    "end": 172.31,
                    "word": " fundamental",
                    "probability": 0.7001953125
                },
                {
                    "start": 172.31,
                    "end": 172.93,
                    "word": " differences",
                    "probability": 0.705078125
                },
                {
                    "start": 172.93,
                    "end": 173.51,
                    "word": " between",
                    "probability": 0.77880859375
                },
                {
                    "start": 173.51,
                    "end": 174.09,
                    "word": " the",
                    "probability": 0.80029296875
                },
                {
                    "start": 174.09,
                    "end": 174.33,
                    "word": " settings",
                    "probability": 0.81201171875
                },
                {
                    "start": 174.33,
                    "end": 174.45,
                    "word": " of",
                    "probability": 0.80029296875
                },
                {
                    "start": 174.45,
                    "end": 174.83,
                    "word": " reinforcement",
                    "probability": 0.78271484375
                },
                {
                    "start": 174.83,
                    "end": 175.07,
                    "word": " learning",
                    "probability": 0.8232421875
                },
                {
                    "start": 175.07,
                    "end": 175.25,
                    "word": " and",
                    "probability": 0.876953125
                },
                {
                    "start": 175.25,
                    "end": 175.55,
                    "word": " supervised",
                    "probability": 0.73291015625
                }
            ]
        },
        {
            "id": 27,
            "seek": 18262,
            "start": 176.07,
            "end": 181.41,
            "text": " learning, and that's why it merits its own category. We'll talk about those differences",
            "tokens": [
                51024,
                2539,
                11,
                293,
                300,
                311,
                983,
                309,
                40923,
                1080,
                1065,
                7719,
                13,
                492,
                603,
                751,
                466,
                729,
                7300,
                51296
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19645833293596904,
            "compression_ratio": 1.7616822429906542,
            "no_speech_prob": 0.0004172325134277344,
            "words": [
                {
                    "start": 176.07,
                    "end": 176.45,
                    "word": " learning",
                    "probability": 0.84912109375
                },
                {
                    "start": 176.45,
                    "end": 176.57,
                    "word": ",",
                    "probability": 0.42919921875
                },
                {
                    "start": 176.57,
                    "end": 176.65,
                    "word": " and",
                    "probability": 0.89453125
                },
                {
                    "start": 176.65,
                    "end": 176.95,
                    "word": " that's",
                    "probability": 0.939208984375
                },
                {
                    "start": 176.95,
                    "end": 177.07,
                    "word": " why",
                    "probability": 0.814453125
                },
                {
                    "start": 177.07,
                    "end": 178.25,
                    "word": " it",
                    "probability": 0.859375
                },
                {
                    "start": 178.25,
                    "end": 178.77,
                    "word": " merits",
                    "probability": 0.85107421875
                },
                {
                    "start": 178.77,
                    "end": 179.27,
                    "word": " its",
                    "probability": 0.798828125
                },
                {
                    "start": 179.27,
                    "end": 179.65,
                    "word": " own",
                    "probability": 0.818359375
                },
                {
                    "start": 179.65,
                    "end": 180.31,
                    "word": " category",
                    "probability": 0.82421875
                },
                {
                    "start": 180.31,
                    "end": 180.45,
                    "word": ".",
                    "probability": 0.970703125
                },
                {
                    "start": 180.45,
                    "end": 180.57,
                    "word": " We'll",
                    "probability": 0.882080078125
                },
                {
                    "start": 180.57,
                    "end": 180.77,
                    "word": " talk",
                    "probability": 0.8095703125
                },
                {
                    "start": 180.77,
                    "end": 180.95,
                    "word": " about",
                    "probability": 0.82470703125
                },
                {
                    "start": 180.95,
                    "end": 181.11,
                    "word": " those",
                    "probability": 0.849609375
                },
                {
                    "start": 181.11,
                    "end": 181.41,
                    "word": " differences",
                    "probability": 0.69921875
                }
            ]
        },
        {
            "id": 28,
            "seek": 18262,
            "start": 181.41,
            "end": 188.75,
            "text": " in the next few slides. So you can see the general settings of reinforcement learning,",
            "tokens": [
                51296,
                294,
                264,
                958,
                1326,
                9788,
                13,
                407,
                291,
                393,
                536,
                264,
                2674,
                6257,
                295,
                29280,
                2539,
                11,
                51652
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19645833293596904,
            "compression_ratio": 1.7616822429906542,
            "no_speech_prob": 0.0004172325134277344,
            "words": [
                {
                    "start": 181.41,
                    "end": 181.75,
                    "word": " in",
                    "probability": 0.90234375
                },
                {
                    "start": 181.75,
                    "end": 181.85,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 181.85,
                    "end": 181.99,
                    "word": " next",
                    "probability": 0.861328125
                },
                {
                    "start": 181.99,
                    "end": 182.13,
                    "word": " few",
                    "probability": 0.8125
                },
                {
                    "start": 182.13,
                    "end": 182.51,
                    "word": " slides",
                    "probability": 0.89990234375
                },
                {
                    "start": 182.51,
                    "end": 183.19,
                    "word": ".",
                    "probability": 0.99609375
                },
                {
                    "start": 183.47,
                    "end": 183.81,
                    "word": " So",
                    "probability": 0.85009765625
                },
                {
                    "start": 183.81,
                    "end": 184.23,
                    "word": " you",
                    "probability": 0.66162109375
                },
                {
                    "start": 184.23,
                    "end": 184.37,
                    "word": " can",
                    "probability": 0.89599609375
                },
                {
                    "start": 184.37,
                    "end": 184.71,
                    "word": " see",
                    "probability": 0.81591796875
                },
                {
                    "start": 184.71,
                    "end": 185.21,
                    "word": " the",
                    "probability": 0.80908203125
                },
                {
                    "start": 185.21,
                    "end": 185.79,
                    "word": " general",
                    "probability": 0.7880859375
                },
                {
                    "start": 185.79,
                    "end": 186.85,
                    "word": " settings",
                    "probability": 0.8095703125
                },
                {
                    "start": 186.85,
                    "end": 187.83,
                    "word": " of",
                    "probability": 0.90625
                },
                {
                    "start": 187.83,
                    "end": 188.45,
                    "word": " reinforcement",
                    "probability": 0.80029296875
                },
                {
                    "start": 188.45,
                    "end": 188.71,
                    "word": " learning",
                    "probability": 0.8291015625
                },
                {
                    "start": 188.71,
                    "end": 188.75,
                    "word": ",",
                    "probability": 0.51806640625
                }
            ]
        },
        {
            "id": 29,
            "seek": 21198,
            "start": 188.75,
            "end": 198.91,
            "text": " or in general, sensory motor agent problems and settings. In RL, we have an agent which is to be",
            "tokens": [
                50364,
                420,
                294,
                2674,
                11,
                27233,
                5932,
                9461,
                2740,
                293,
                6257,
                13,
                682,
                497,
                43,
                11,
                321,
                362,
                364,
                9461,
                597,
                307,
                281,
                312,
                50868
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23270089764680182,
            "compression_ratio": 1.7027027027027026,
            "no_speech_prob": 4.988908767700195e-05,
            "words": [
                {
                    "start": 188.75,
                    "end": 188.83,
                    "word": " or",
                    "probability": 0.73974609375
                },
                {
                    "start": 188.83,
                    "end": 188.97,
                    "word": " in",
                    "probability": 0.84814453125
                },
                {
                    "start": 188.97,
                    "end": 189.43,
                    "word": " general",
                    "probability": 0.806640625
                },
                {
                    "start": 189.43,
                    "end": 189.93,
                    "word": ",",
                    "probability": 0.498291015625
                },
                {
                    "start": 189.93,
                    "end": 190.19,
                    "word": " sensory",
                    "probability": 0.52099609375
                },
                {
                    "start": 190.19,
                    "end": 190.43,
                    "word": " motor",
                    "probability": 0.86181640625
                },
                {
                    "start": 190.43,
                    "end": 191.47,
                    "word": " agent",
                    "probability": 0.6162109375
                },
                {
                    "start": 191.47,
                    "end": 193.07,
                    "word": " problems",
                    "probability": 0.55126953125
                },
                {
                    "start": 193.07,
                    "end": 193.41,
                    "word": " and",
                    "probability": 0.88232421875
                },
                {
                    "start": 193.41,
                    "end": 193.77,
                    "word": " settings",
                    "probability": 0.826171875
                },
                {
                    "start": 193.77,
                    "end": 194.49,
                    "word": ".",
                    "probability": 0.845703125
                },
                {
                    "start": 194.61,
                    "end": 194.89,
                    "word": " In",
                    "probability": 0.9111328125
                },
                {
                    "start": 194.89,
                    "end": 195.35,
                    "word": " RL",
                    "probability": 0.957763671875
                },
                {
                    "start": 195.35,
                    "end": 196.07,
                    "word": ",",
                    "probability": 0.771484375
                },
                {
                    "start": 196.07,
                    "end": 196.31,
                    "word": " we",
                    "probability": 0.8994140625
                },
                {
                    "start": 196.31,
                    "end": 196.53,
                    "word": " have",
                    "probability": 0.88037109375
                },
                {
                    "start": 196.53,
                    "end": 196.71,
                    "word": " an",
                    "probability": 0.89990234375
                },
                {
                    "start": 196.71,
                    "end": 197.09,
                    "word": " agent",
                    "probability": 0.80859375
                },
                {
                    "start": 197.09,
                    "end": 198.13,
                    "word": " which",
                    "probability": 0.4228515625
                },
                {
                    "start": 198.13,
                    "end": 198.61,
                    "word": " is",
                    "probability": 0.87939453125
                },
                {
                    "start": 198.61,
                    "end": 198.73,
                    "word": " to",
                    "probability": 0.90771484375
                },
                {
                    "start": 198.73,
                    "end": 198.91,
                    "word": " be",
                    "probability": 0.9091796875
                }
            ]
        },
        {
            "id": 30,
            "seek": 21198,
            "start": 198.91,
            "end": 203.87,
            "text": " trained by a reinforcement learning algorithm. This agent interacts with the environment by",
            "tokens": [
                50868,
                8895,
                538,
                257,
                29280,
                2539,
                9284,
                13,
                639,
                9461,
                43582,
                365,
                264,
                2823,
                538,
                51116
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23270089764680182,
            "compression_ratio": 1.7027027027027026,
            "no_speech_prob": 4.988908767700195e-05,
            "words": [
                {
                    "start": 198.91,
                    "end": 199.25,
                    "word": " trained",
                    "probability": 0.80810546875
                },
                {
                    "start": 199.25,
                    "end": 199.51,
                    "word": " by",
                    "probability": 0.90380859375
                },
                {
                    "start": 199.69,
                    "end": 200.03,
                    "word": " a reinforcement",
                    "probability": 0.7861328125
                },
                {
                    "start": 200.03,
                    "end": 200.35,
                    "word": " learning",
                    "probability": 0.8056640625
                },
                {
                    "start": 200.35,
                    "end": 200.81,
                    "word": " algorithm",
                    "probability": 0.80810546875
                },
                {
                    "start": 200.81,
                    "end": 201.23,
                    "word": ".",
                    "probability": 0.93359375
                },
                {
                    "start": 201.23,
                    "end": 201.53,
                    "word": " This",
                    "probability": 0.8095703125
                },
                {
                    "start": 201.53,
                    "end": 201.89,
                    "word": " agent",
                    "probability": 0.8173828125
                },
                {
                    "start": 201.89,
                    "end": 202.53,
                    "word": " interacts",
                    "probability": 0.73779296875
                },
                {
                    "start": 202.53,
                    "end": 202.71,
                    "word": " with",
                    "probability": 0.81396484375
                },
                {
                    "start": 202.71,
                    "end": 202.87,
                    "word": " the",
                    "probability": 0.806640625
                },
                {
                    "start": 202.87,
                    "end": 203.37,
                    "word": " environment",
                    "probability": 0.7861328125
                },
                {
                    "start": 203.37,
                    "end": 203.87,
                    "word": " by",
                    "probability": 0.89892578125
                }
            ]
        },
        {
            "id": 31,
            "seek": 21198,
            "start": 203.87,
            "end": 210.77,
            "text": " performing some action. This action causes the state of the environment to change. Then the new",
            "tokens": [
                51116,
                10205,
                512,
                3069,
                13,
                639,
                3069,
                7700,
                264,
                1785,
                295,
                264,
                2823,
                281,
                1319,
                13,
                1396,
                264,
                777,
                51456
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23270089764680182,
            "compression_ratio": 1.7027027027027026,
            "no_speech_prob": 4.988908767700195e-05,
            "words": [
                {
                    "start": 203.87,
                    "end": 204.25,
                    "word": " performing",
                    "probability": 0.69921875
                },
                {
                    "start": 204.25,
                    "end": 204.53,
                    "word": " some",
                    "probability": 0.8193359375
                },
                {
                    "start": 204.53,
                    "end": 204.89,
                    "word": " action",
                    "probability": 0.80517578125
                },
                {
                    "start": 204.89,
                    "end": 205.35,
                    "word": ".",
                    "probability": 0.9765625
                },
                {
                    "start": 205.35,
                    "end": 205.73,
                    "word": " This",
                    "probability": 0.81298828125
                },
                {
                    "start": 205.73,
                    "end": 206.23,
                    "word": " action",
                    "probability": 0.80078125
                },
                {
                    "start": 206.23,
                    "end": 206.87,
                    "word": " causes",
                    "probability": 0.8037109375
                },
                {
                    "start": 206.87,
                    "end": 207.11,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 207.11,
                    "end": 207.41,
                    "word": " state",
                    "probability": 0.8818359375
                },
                {
                    "start": 207.41,
                    "end": 207.53,
                    "word": " of",
                    "probability": 0.9140625
                },
                {
                    "start": 207.53,
                    "end": 207.67,
                    "word": " the",
                    "probability": 0.79541015625
                },
                {
                    "start": 207.67,
                    "end": 208.03,
                    "word": " environment",
                    "probability": 0.76904296875
                },
                {
                    "start": 208.03,
                    "end": 208.29,
                    "word": " to",
                    "probability": 0.90625
                },
                {
                    "start": 208.29,
                    "end": 208.87,
                    "word": " change",
                    "probability": 0.8349609375
                },
                {
                    "start": 208.87,
                    "end": 209.59,
                    "word": ".",
                    "probability": 0.7939453125
                },
                {
                    "start": 209.59,
                    "end": 209.87,
                    "word": " Then",
                    "probability": 0.80908203125
                },
                {
                    "start": 209.87,
                    "end": 210.27,
                    "word": " the",
                    "probability": 0.67041015625
                },
                {
                    "start": 210.27,
                    "end": 210.77,
                    "word": " new",
                    "probability": 0.8232421875
                }
            ]
        },
        {
            "id": 32,
            "seek": 21198,
            "start": 210.77,
            "end": 218.11,
            "text": " state is observed by the agent. There's also some sort of reward associated with this change,",
            "tokens": [
                51484,
                1785,
                307,
                13095,
                538,
                264,
                9461,
                13,
                821,
                311,
                611,
                512,
                1333,
                295,
                7782,
                6615,
                365,
                341,
                1319,
                11,
                51816
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23270089764680182,
            "compression_ratio": 1.7027027027027026,
            "no_speech_prob": 4.988908767700195e-05,
            "words": [
                {
                    "start": 210.77,
                    "end": 211.69,
                    "word": " state",
                    "probability": 0.828125
                },
                {
                    "start": 211.69,
                    "end": 212.47,
                    "word": " is",
                    "probability": 0.90283203125
                },
                {
                    "start": 212.47,
                    "end": 212.89,
                    "word": " observed",
                    "probability": 0.771484375
                },
                {
                    "start": 212.89,
                    "end": 213.15,
                    "word": " by",
                    "probability": 0.904296875
                },
                {
                    "start": 213.15,
                    "end": 213.33,
                    "word": " the",
                    "probability": 0.82080078125
                },
                {
                    "start": 213.33,
                    "end": 213.65,
                    "word": " agent",
                    "probability": 0.796875
                },
                {
                    "start": 213.65,
                    "end": 213.95,
                    "word": ".",
                    "probability": 0.90771484375
                },
                {
                    "start": 214.05,
                    "end": 214.77,
                    "word": " There's",
                    "probability": 0.835205078125
                },
                {
                    "start": 214.77,
                    "end": 215.15,
                    "word": " also",
                    "probability": 0.81884765625
                },
                {
                    "start": 215.15,
                    "end": 215.47,
                    "word": " some",
                    "probability": 0.81884765625
                },
                {
                    "start": 215.47,
                    "end": 215.67,
                    "word": " sort",
                    "probability": 0.80859375
                },
                {
                    "start": 215.67,
                    "end": 215.81,
                    "word": " of",
                    "probability": 0.904296875
                },
                {
                    "start": 215.81,
                    "end": 216.33,
                    "word": " reward",
                    "probability": 0.84130859375
                },
                {
                    "start": 216.33,
                    "end": 216.93,
                    "word": " associated",
                    "probability": 0.845703125
                },
                {
                    "start": 216.93,
                    "end": 217.17,
                    "word": " with",
                    "probability": 0.81298828125
                },
                {
                    "start": 217.17,
                    "end": 217.35,
                    "word": " this",
                    "probability": 0.88134765625
                },
                {
                    "start": 217.35,
                    "end": 217.75,
                    "word": " change",
                    "probability": 0.8388671875
                },
                {
                    "start": 217.75,
                    "end": 218.11,
                    "word": ",",
                    "probability": 0.279296875
                }
            ]
        },
        {
            "id": 33,
            "seek": 24024,
            "start": 218.81,
            "end": 223.73,
            "text": " provided or inferred by the agent based on this change in state.",
            "tokens": [
                50400,
                5649,
                420,
                13596,
                986,
                538,
                264,
                9461,
                2361,
                322,
                341,
                1319,
                294,
                1785,
                13,
                50612
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20105421650840574,
            "compression_ratio": 1.83419689119171,
            "no_speech_prob": 0.00327301025390625,
            "words": [
                {
                    "start": 218.81,
                    "end": 219.39,
                    "word": " provided",
                    "probability": 0.095458984375
                },
                {
                    "start": 219.39,
                    "end": 219.83,
                    "word": " or",
                    "probability": 0.77685546875
                },
                {
                    "start": 219.83,
                    "end": 220.37,
                    "word": " inferred",
                    "probability": 0.906005859375
                },
                {
                    "start": 220.37,
                    "end": 220.55,
                    "word": " by",
                    "probability": 0.89453125
                },
                {
                    "start": 220.55,
                    "end": 220.69,
                    "word": " the",
                    "probability": 0.81103515625
                },
                {
                    "start": 220.69,
                    "end": 220.95,
                    "word": " agent",
                    "probability": 0.8037109375
                },
                {
                    "start": 220.95,
                    "end": 221.25,
                    "word": " based",
                    "probability": 0.73291015625
                },
                {
                    "start": 221.25,
                    "end": 221.41,
                    "word": " on",
                    "probability": 0.91015625
                },
                {
                    "start": 221.41,
                    "end": 221.73,
                    "word": " this",
                    "probability": 0.85302734375
                },
                {
                    "start": 221.73,
                    "end": 222.61,
                    "word": " change",
                    "probability": 0.79052734375
                },
                {
                    "start": 222.61,
                    "end": 222.77,
                    "word": " in",
                    "probability": 0.89501953125
                },
                {
                    "start": 222.77,
                    "end": 223.15,
                    "word": " state",
                    "probability": 0.85302734375
                },
                {
                    "start": 223.15,
                    "end": 223.73,
                    "word": ".",
                    "probability": 0.6259765625
                }
            ]
        },
        {
            "id": 34,
            "seek": 24024,
            "start": 225.79,
            "end": 232.57,
            "text": " Well, you can think of this in terms of playing a game. The environment can be the game environment",
            "tokens": [
                50760,
                1042,
                11,
                291,
                393,
                519,
                295,
                341,
                294,
                2115,
                295,
                2433,
                257,
                1216,
                13,
                440,
                2823,
                393,
                312,
                264,
                1216,
                2823,
                51088
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20105421650840574,
            "compression_ratio": 1.83419689119171,
            "no_speech_prob": 0.00327301025390625,
            "words": [
                {
                    "start": 225.79,
                    "end": 226.37,
                    "word": " Well",
                    "probability": 0.55126953125
                },
                {
                    "start": 226.37,
                    "end": 226.61,
                    "word": ",",
                    "probability": 0.85498046875
                },
                {
                    "start": 226.61,
                    "end": 227.33,
                    "word": " you",
                    "probability": 0.8896484375
                },
                {
                    "start": 227.33,
                    "end": 227.61,
                    "word": " can",
                    "probability": 0.90087890625
                },
                {
                    "start": 227.61,
                    "end": 228.57,
                    "word": " think",
                    "probability": 0.7958984375
                },
                {
                    "start": 228.57,
                    "end": 228.67,
                    "word": " of",
                    "probability": 0.9033203125
                },
                {
                    "start": 228.67,
                    "end": 228.83,
                    "word": " this",
                    "probability": 0.87841796875
                },
                {
                    "start": 228.83,
                    "end": 228.95,
                    "word": " in",
                    "probability": 0.892578125
                },
                {
                    "start": 228.95,
                    "end": 229.13,
                    "word": " terms",
                    "probability": 0.7783203125
                },
                {
                    "start": 229.13,
                    "end": 229.29,
                    "word": " of",
                    "probability": 0.908203125
                },
                {
                    "start": 229.29,
                    "end": 229.51,
                    "word": " playing",
                    "probability": 0.79931640625
                },
                {
                    "start": 229.69,
                    "end": 230.03,
                    "word": " a game",
                    "probability": 0.8173828125
                },
                {
                    "start": 230.03,
                    "end": 230.57,
                    "word": ".",
                    "probability": 0.74365234375
                },
                {
                    "start": 230.57,
                    "end": 231.01,
                    "word": " The",
                    "probability": 0.787109375
                },
                {
                    "start": 231.01,
                    "end": 231.35,
                    "word": " environment",
                    "probability": 0.734375
                },
                {
                    "start": 231.35,
                    "end": 231.57,
                    "word": " can",
                    "probability": 0.9013671875
                },
                {
                    "start": 231.57,
                    "end": 231.73,
                    "word": " be",
                    "probability": 0.90673828125
                },
                {
                    "start": 231.73,
                    "end": 231.93,
                    "word": " the",
                    "probability": 0.7978515625
                },
                {
                    "start": 231.93,
                    "end": 232.19,
                    "word": " game",
                    "probability": 0.79443359375
                },
                {
                    "start": 232.19,
                    "end": 232.57,
                    "word": " environment",
                    "probability": 0.751953125
                }
            ]
        },
        {
            "id": 35,
            "seek": 24024,
            "start": 232.57,
            "end": 241.75,
            "text": " where different actions may result in scoring or loss of score, and the actions are, well,",
            "tokens": [
                51088,
                689,
                819,
                5909,
                815,
                1874,
                294,
                22358,
                420,
                4470,
                295,
                6175,
                11,
                293,
                264,
                5909,
                366,
                11,
                731,
                11,
                51536
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20105421650840574,
            "compression_ratio": 1.83419689119171,
            "no_speech_prob": 0.00327301025390625,
            "words": [
                {
                    "start": 232.57,
                    "end": 232.93,
                    "word": " where",
                    "probability": 0.74462890625
                },
                {
                    "start": 232.93,
                    "end": 233.67,
                    "word": " different",
                    "probability": 0.66552734375
                },
                {
                    "start": 233.67,
                    "end": 234.07,
                    "word": " actions",
                    "probability": 0.8671875
                },
                {
                    "start": 234.07,
                    "end": 234.23,
                    "word": " may",
                    "probability": 0.89306640625
                },
                {
                    "start": 234.23,
                    "end": 234.63,
                    "word": " result",
                    "probability": 0.8583984375
                },
                {
                    "start": 234.63,
                    "end": 235.11,
                    "word": " in",
                    "probability": 0.90576171875
                },
                {
                    "start": 235.11,
                    "end": 235.71,
                    "word": " scoring",
                    "probability": 0.865234375
                },
                {
                    "start": 235.71,
                    "end": 236.35,
                    "word": " or",
                    "probability": 0.86181640625
                },
                {
                    "start": 236.35,
                    "end": 237.37,
                    "word": " loss",
                    "probability": 0.826171875
                },
                {
                    "start": 237.37,
                    "end": 237.93,
                    "word": " of",
                    "probability": 0.908203125
                },
                {
                    "start": 237.93,
                    "end": 239.45,
                    "word": " score",
                    "probability": 0.73583984375
                },
                {
                    "start": 239.45,
                    "end": 239.83,
                    "word": ",",
                    "probability": 0.29150390625
                },
                {
                    "start": 239.83,
                    "end": 240.31,
                    "word": " and",
                    "probability": 0.89111328125
                },
                {
                    "start": 240.31,
                    "end": 240.45,
                    "word": " the",
                    "probability": 0.80908203125
                },
                {
                    "start": 240.45,
                    "end": 240.81,
                    "word": " actions",
                    "probability": 0.86962890625
                },
                {
                    "start": 240.81,
                    "end": 241.23,
                    "word": " are",
                    "probability": 0.88037109375
                },
                {
                    "start": 241.23,
                    "end": 241.43,
                    "word": ",",
                    "probability": 0.5908203125
                },
                {
                    "start": 241.43,
                    "end": 241.69,
                    "word": " well",
                    "probability": 0.859375
                },
                {
                    "start": 241.69,
                    "end": 241.75,
                    "word": ",",
                    "probability": 0.849609375
                }
            ]
        },
        {
            "id": 36,
            "seek": 24024,
            "start": 241.75,
            "end": 246.37,
            "text": " essentially the actions that the player can take. The state of the environment can be the state of",
            "tokens": [
                51536,
                4476,
                264,
                5909,
                300,
                264,
                4256,
                393,
                747,
                13,
                440,
                1785,
                295,
                264,
                2823,
                393,
                312,
                264,
                1785,
                295,
                51772
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20105421650840574,
            "compression_ratio": 1.83419689119171,
            "no_speech_prob": 0.00327301025390625,
            "words": [
                {
                    "start": 241.75,
                    "end": 241.99,
                    "word": " essentially",
                    "probability": 0.66064453125
                },
                {
                    "start": 241.99,
                    "end": 242.15,
                    "word": " the",
                    "probability": 0.69091796875
                },
                {
                    "start": 242.15,
                    "end": 242.39,
                    "word": " actions",
                    "probability": 0.865234375
                },
                {
                    "start": 242.39,
                    "end": 242.69,
                    "word": " that",
                    "probability": 0.86865234375
                },
                {
                    "start": 242.69,
                    "end": 243.01,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 243.01,
                    "end": 243.27,
                    "word": " player",
                    "probability": 0.779296875
                },
                {
                    "start": 243.27,
                    "end": 243.49,
                    "word": " can",
                    "probability": 0.89599609375
                },
                {
                    "start": 243.49,
                    "end": 243.85,
                    "word": " take",
                    "probability": 0.8173828125
                },
                {
                    "start": 243.85,
                    "end": 244.19,
                    "word": ".",
                    "probability": 0.93505859375
                },
                {
                    "start": 244.19,
                    "end": 244.41,
                    "word": " The",
                    "probability": 0.814453125
                },
                {
                    "start": 244.41,
                    "end": 244.67,
                    "word": " state",
                    "probability": 0.880859375
                },
                {
                    "start": 244.67,
                    "end": 244.79,
                    "word": " of",
                    "probability": 0.9091796875
                },
                {
                    "start": 244.79,
                    "end": 244.93,
                    "word": " the",
                    "probability": 0.80615234375
                },
                {
                    "start": 244.93,
                    "end": 245.35,
                    "word": " environment",
                    "probability": 0.74853515625
                },
                {
                    "start": 245.35,
                    "end": 245.81,
                    "word": " can",
                    "probability": 0.8935546875
                },
                {
                    "start": 245.81,
                    "end": 245.91,
                    "word": " be",
                    "probability": 0.90673828125
                },
                {
                    "start": 245.91,
                    "end": 246.07,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 246.07,
                    "end": 246.25,
                    "word": " state",
                    "probability": 0.884765625
                },
                {
                    "start": 246.25,
                    "end": 246.37,
                    "word": " of",
                    "probability": 0.9091796875
                }
            ]
        },
        {
            "id": 37,
            "seek": 26758,
            "start": 246.37,
            "end": 254.53,
            "text": " the game. For example, in Breakout, the configuration of these breaks can be one of",
            "tokens": [
                50364,
                264,
                1216,
                13,
                1171,
                1365,
                11,
                294,
                16925,
                346,
                11,
                264,
                11694,
                295,
                613,
                9857,
                393,
                312,
                472,
                295,
                50768
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21101973841064856,
            "compression_ratio": 1.6473214285714286,
            "no_speech_prob": 0.0037078857421875,
            "words": [
                {
                    "start": 246.37,
                    "end": 246.47,
                    "word": " the",
                    "probability": 0.2071533203125
                },
                {
                    "start": 246.47,
                    "end": 246.83,
                    "word": " game",
                    "probability": 0.56396484375
                },
                {
                    "start": 246.83,
                    "end": 247.05,
                    "word": ".",
                    "probability": 0.63330078125
                },
                {
                    "start": 247.05,
                    "end": 247.17,
                    "word": " For",
                    "probability": 0.79736328125
                },
                {
                    "start": 247.17,
                    "end": 247.53,
                    "word": " example",
                    "probability": 0.921875
                },
                {
                    "start": 247.53,
                    "end": 247.77,
                    "word": ",",
                    "probability": 0.513671875
                },
                {
                    "start": 247.77,
                    "end": 247.87,
                    "word": " in",
                    "probability": 0.85205078125
                },
                {
                    "start": 247.87,
                    "end": 248.37,
                    "word": " Breakout",
                    "probability": 0.833251953125
                },
                {
                    "start": 248.37,
                    "end": 249.05,
                    "word": ",",
                    "probability": 0.6416015625
                },
                {
                    "start": 249.05,
                    "end": 251.01,
                    "word": " the",
                    "probability": 0.79248046875
                },
                {
                    "start": 251.01,
                    "end": 252.01,
                    "word": " configuration",
                    "probability": 0.837890625
                },
                {
                    "start": 252.01,
                    "end": 252.41,
                    "word": " of",
                    "probability": 0.890625
                },
                {
                    "start": 252.41,
                    "end": 252.61,
                    "word": " these",
                    "probability": 0.646484375
                },
                {
                    "start": 252.61,
                    "end": 253.05,
                    "word": " breaks",
                    "probability": 0.70751953125
                },
                {
                    "start": 253.05,
                    "end": 254.11,
                    "word": " can",
                    "probability": 0.876953125
                },
                {
                    "start": 254.11,
                    "end": 254.27,
                    "word": " be",
                    "probability": 0.8916015625
                },
                {
                    "start": 254.27,
                    "end": 254.43,
                    "word": " one",
                    "probability": 0.72314453125
                },
                {
                    "start": 254.43,
                    "end": 254.53,
                    "word": " of",
                    "probability": 0.88818359375
                }
            ]
        },
        {
            "id": 38,
            "seek": 26758,
            "start": 254.53,
            "end": 259.01,
            "text": " those where the agent is, where the ball is. If you recall the game of Breakout, of course, if you",
            "tokens": [
                50768,
                729,
                689,
                264,
                9461,
                307,
                11,
                689,
                264,
                2594,
                307,
                13,
                759,
                291,
                9901,
                264,
                1216,
                295,
                16925,
                346,
                11,
                295,
                1164,
                11,
                498,
                291,
                50992
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21101973841064856,
            "compression_ratio": 1.6473214285714286,
            "no_speech_prob": 0.0037078857421875,
            "words": [
                {
                    "start": 254.53,
                    "end": 254.71,
                    "word": " those",
                    "probability": 0.80322265625
                },
                {
                    "start": 254.71,
                    "end": 254.89,
                    "word": " where",
                    "probability": 0.59619140625
                },
                {
                    "start": 254.89,
                    "end": 255.05,
                    "word": " the",
                    "probability": 0.7998046875
                },
                {
                    "start": 255.05,
                    "end": 255.29,
                    "word": " agent",
                    "probability": 0.76025390625
                },
                {
                    "start": 255.29,
                    "end": 255.67,
                    "word": " is",
                    "probability": 0.89794921875
                },
                {
                    "start": 255.67,
                    "end": 255.81,
                    "word": ",",
                    "probability": 0.9296875
                },
                {
                    "start": 255.81,
                    "end": 255.97,
                    "word": " where",
                    "probability": 0.8984375
                },
                {
                    "start": 255.97,
                    "end": 256.13,
                    "word": " the",
                    "probability": 0.81494140625
                },
                {
                    "start": 256.13,
                    "end": 256.33,
                    "word": " ball",
                    "probability": 0.83447265625
                },
                {
                    "start": 256.33,
                    "end": 256.65,
                    "word": " is",
                    "probability": 0.9013671875
                },
                {
                    "start": 256.65,
                    "end": 256.99,
                    "word": ".",
                    "probability": 0.47021484375
                },
                {
                    "start": 256.99,
                    "end": 257.11,
                    "word": " If",
                    "probability": 0.9013671875
                },
                {
                    "start": 257.11,
                    "end": 257.21,
                    "word": " you",
                    "probability": 0.89453125
                },
                {
                    "start": 257.21,
                    "end": 257.47,
                    "word": " recall",
                    "probability": 0.75634765625
                },
                {
                    "start": 257.47,
                    "end": 257.61,
                    "word": " the",
                    "probability": 0.80322265625
                },
                {
                    "start": 257.61,
                    "end": 257.77,
                    "word": " game",
                    "probability": 0.787109375
                },
                {
                    "start": 257.77,
                    "end": 257.95,
                    "word": " of",
                    "probability": 0.77001953125
                },
                {
                    "start": 257.95,
                    "end": 258.29,
                    "word": " Breakout",
                    "probability": 0.90380859375
                },
                {
                    "start": 258.29,
                    "end": 258.39,
                    "word": ",",
                    "probability": 0.73193359375
                },
                {
                    "start": 258.39,
                    "end": 258.45,
                    "word": " of",
                    "probability": 0.8447265625
                },
                {
                    "start": 258.45,
                    "end": 258.67,
                    "word": " course",
                    "probability": 0.8662109375
                },
                {
                    "start": 258.67,
                    "end": 258.79,
                    "word": ",",
                    "probability": 0.84326171875
                },
                {
                    "start": 258.79,
                    "end": 258.91,
                    "word": " if",
                    "probability": 0.87109375
                },
                {
                    "start": 258.91,
                    "end": 259.01,
                    "word": " you",
                    "probability": 0.89208984375
                }
            ]
        },
        {
            "id": 39,
            "seek": 26758,
            "start": 259.01,
            "end": 266.07,
            "text": " are old enough to remember Breakout, you're probably familiar with the dynamics. But you can",
            "tokens": [
                50992,
                366,
                1331,
                1547,
                281,
                1604,
                16925,
                346,
                11,
                291,
                434,
                1391,
                4963,
                365,
                264,
                15679,
                13,
                583,
                291,
                393,
                51348
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21101973841064856,
            "compression_ratio": 1.6473214285714286,
            "no_speech_prob": 0.0037078857421875,
            "words": [
                {
                    "start": 259.01,
                    "end": 259.11,
                    "word": " are",
                    "probability": 0.576171875
                },
                {
                    "start": 259.11,
                    "end": 259.31,
                    "word": " old",
                    "probability": 0.86962890625
                },
                {
                    "start": 259.31,
                    "end": 259.53,
                    "word": " enough",
                    "probability": 0.78515625
                },
                {
                    "start": 259.53,
                    "end": 259.75,
                    "word": " to",
                    "probability": 0.8994140625
                },
                {
                    "start": 259.75,
                    "end": 260.41,
                    "word": " remember",
                    "probability": 0.79638671875
                },
                {
                    "start": 260.41,
                    "end": 262.13,
                    "word": " Breakout",
                    "probability": 0.89013671875
                },
                {
                    "start": 262.13,
                    "end": 262.93,
                    "word": ",",
                    "probability": 0.87353515625
                },
                {
                    "start": 262.93,
                    "end": 263.67,
                    "word": " you're",
                    "probability": 0.694091796875
                },
                {
                    "start": 263.67,
                    "end": 263.99,
                    "word": " probably",
                    "probability": 0.91943359375
                },
                {
                    "start": 263.99,
                    "end": 264.37,
                    "word": " familiar",
                    "probability": 0.798828125
                },
                {
                    "start": 264.37,
                    "end": 264.85,
                    "word": " with",
                    "probability": 0.8193359375
                },
                {
                    "start": 264.85,
                    "end": 265.15,
                    "word": " the",
                    "probability": 0.8173828125
                },
                {
                    "start": 265.15,
                    "end": 265.55,
                    "word": " dynamics",
                    "probability": 0.7119140625
                },
                {
                    "start": 265.55,
                    "end": 265.75,
                    "word": ".",
                    "probability": 0.49658203125
                },
                {
                    "start": 265.75,
                    "end": 265.87,
                    "word": " But",
                    "probability": 0.876953125
                },
                {
                    "start": 265.87,
                    "end": 265.97,
                    "word": " you",
                    "probability": 0.87158203125
                },
                {
                    "start": 265.97,
                    "end": 266.07,
                    "word": " can",
                    "probability": 0.89794921875
                }
            ]
        },
        {
            "id": 40,
            "seek": 26758,
            "start": 266.07,
            "end": 273.71,
            "text": " think of any other game, and this setting still applies. What's the goal here? The goal is to",
            "tokens": [
                51348,
                519,
                295,
                604,
                661,
                1216,
                11,
                293,
                341,
                3287,
                920,
                13165,
                13,
                708,
                311,
                264,
                3387,
                510,
                30,
                440,
                3387,
                307,
                281,
                51744
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21101973841064856,
            "compression_ratio": 1.6473214285714286,
            "no_speech_prob": 0.0037078857421875,
            "words": [
                {
                    "start": 266.07,
                    "end": 266.29,
                    "word": " think",
                    "probability": 0.79833984375
                },
                {
                    "start": 266.29,
                    "end": 266.51,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 266.51,
                    "end": 268.13,
                    "word": " any",
                    "probability": 0.78955078125
                },
                {
                    "start": 268.13,
                    "end": 268.37,
                    "word": " other",
                    "probability": 0.76708984375
                },
                {
                    "start": 268.37,
                    "end": 268.73,
                    "word": " game",
                    "probability": 0.8232421875
                },
                {
                    "start": 268.73,
                    "end": 269.29,
                    "word": ",",
                    "probability": 0.35888671875
                },
                {
                    "start": 269.29,
                    "end": 269.61,
                    "word": " and",
                    "probability": 0.89990234375
                },
                {
                    "start": 269.61,
                    "end": 269.79,
                    "word": " this",
                    "probability": 0.46435546875
                },
                {
                    "start": 269.79,
                    "end": 270.01,
                    "word": " setting",
                    "probability": 0.83251953125
                },
                {
                    "start": 270.01,
                    "end": 270.27,
                    "word": " still",
                    "probability": 0.892578125
                },
                {
                    "start": 270.27,
                    "end": 270.79,
                    "word": " applies",
                    "probability": 0.80322265625
                },
                {
                    "start": 270.79,
                    "end": 271.23,
                    "word": ".",
                    "probability": 0.99560546875
                },
                {
                    "start": 271.27,
                    "end": 271.61,
                    "word": " What's",
                    "probability": 0.893798828125
                },
                {
                    "start": 271.61,
                    "end": 271.75,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 271.75,
                    "end": 272.01,
                    "word": " goal",
                    "probability": 0.912109375
                },
                {
                    "start": 272.01,
                    "end": 272.29,
                    "word": " here",
                    "probability": 0.80615234375
                },
                {
                    "start": 272.29,
                    "end": 272.41,
                    "word": "?",
                    "probability": 0.9912109375
                },
                {
                    "start": 272.41,
                    "end": 272.63,
                    "word": " The",
                    "probability": 0.814453125
                },
                {
                    "start": 272.63,
                    "end": 272.89,
                    "word": " goal",
                    "probability": 0.90869140625
                },
                {
                    "start": 272.89,
                    "end": 273.45,
                    "word": " is",
                    "probability": 0.9052734375
                },
                {
                    "start": 273.45,
                    "end": 273.71,
                    "word": " to",
                    "probability": 0.9033203125
                }
            ]
        },
        {
            "id": 41,
            "seek": 29194,
            "start": 273.71,
            "end": 280.57,
            "text": " learn how to take actions in order to maximize cumulative rewards, not instantaneous rewards,",
            "tokens": [
                50376,
                1466,
                577,
                281,
                747,
                5909,
                294,
                1668,
                281,
                19874,
                38379,
                17203,
                11,
                406,
                45596,
                17203,
                11,
                50700
            ],
            "temperature": 0.0,
            "avg_logprob": -0.18840580055679101,
            "compression_ratio": 1.660919540229885,
            "no_speech_prob": 0.0006561279296875,
            "words": [
                {
                    "start": 273.71,
                    "end": 274.29,
                    "word": " learn",
                    "probability": 0.297119140625
                },
                {
                    "start": 274.29,
                    "end": 274.59,
                    "word": " how",
                    "probability": 0.8994140625
                },
                {
                    "start": 274.59,
                    "end": 274.75,
                    "word": " to",
                    "probability": 0.90625
                },
                {
                    "start": 274.75,
                    "end": 274.99,
                    "word": " take",
                    "probability": 0.7939453125
                },
                {
                    "start": 274.99,
                    "end": 275.61,
                    "word": " actions",
                    "probability": 0.78369140625
                },
                {
                    "start": 275.61,
                    "end": 276.53,
                    "word": " in",
                    "probability": 0.8486328125
                },
                {
                    "start": 276.53,
                    "end": 276.81,
                    "word": " order",
                    "probability": 0.84423828125
                },
                {
                    "start": 276.81,
                    "end": 277.07,
                    "word": " to",
                    "probability": 0.9091796875
                },
                {
                    "start": 277.07,
                    "end": 277.71,
                    "word": " maximize",
                    "probability": 0.6943359375
                },
                {
                    "start": 277.71,
                    "end": 278.69,
                    "word": " cumulative",
                    "probability": 0.64501953125
                },
                {
                    "start": 278.69,
                    "end": 279.07,
                    "word": " rewards",
                    "probability": 0.65869140625
                },
                {
                    "start": 279.07,
                    "end": 279.19,
                    "word": ",",
                    "probability": 0.39599609375
                },
                {
                    "start": 279.19,
                    "end": 279.27,
                    "word": " not",
                    "probability": 0.890625
                },
                {
                    "start": 279.27,
                    "end": 279.75,
                    "word": " instantaneous",
                    "probability": 0.89306640625
                },
                {
                    "start": 279.75,
                    "end": 280.25,
                    "word": " rewards",
                    "probability": 0.6728515625
                },
                {
                    "start": 280.25,
                    "end": 280.57,
                    "word": ",",
                    "probability": 0.6943359375
                }
            ]
        },
        {
            "id": 42,
            "seek": 29194,
            "start": 280.57,
            "end": 288.51,
            "text": " but cumulative rewards. So an RL agent is not just concerned with maximizing the current score.",
            "tokens": [
                50700,
                457,
                38379,
                17203,
                13,
                407,
                364,
                497,
                43,
                9461,
                307,
                406,
                445,
                5922,
                365,
                5138,
                3319,
                264,
                2190,
                6175,
                13,
                51108
            ],
            "temperature": 0.0,
            "avg_logprob": -0.18840580055679101,
            "compression_ratio": 1.660919540229885,
            "no_speech_prob": 0.0006561279296875,
            "words": [
                {
                    "start": 280.57,
                    "end": 280.75,
                    "word": " but",
                    "probability": 0.90283203125
                },
                {
                    "start": 280.75,
                    "end": 281.37,
                    "word": " cumulative",
                    "probability": 0.7138671875
                },
                {
                    "start": 281.37,
                    "end": 281.89,
                    "word": " rewards",
                    "probability": 0.67822265625
                },
                {
                    "start": 281.89,
                    "end": 282.49,
                    "word": ".",
                    "probability": 0.892578125
                },
                {
                    "start": 282.49,
                    "end": 282.63,
                    "word": " So",
                    "probability": 0.81494140625
                },
                {
                    "start": 282.63,
                    "end": 282.93,
                    "word": " an",
                    "probability": 0.779296875
                },
                {
                    "start": 282.93,
                    "end": 283.17,
                    "word": " RL",
                    "probability": 0.882568359375
                },
                {
                    "start": 283.17,
                    "end": 283.45,
                    "word": " agent",
                    "probability": 0.80126953125
                },
                {
                    "start": 283.45,
                    "end": 283.79,
                    "word": " is",
                    "probability": 0.90234375
                },
                {
                    "start": 283.79,
                    "end": 284.09,
                    "word": " not",
                    "probability": 0.8984375
                },
                {
                    "start": 284.09,
                    "end": 284.35,
                    "word": " just",
                    "probability": 0.88720703125
                },
                {
                    "start": 284.35,
                    "end": 284.83,
                    "word": " concerned",
                    "probability": 0.84326171875
                },
                {
                    "start": 284.83,
                    "end": 285.53,
                    "word": " with",
                    "probability": 0.8251953125
                },
                {
                    "start": 285.53,
                    "end": 286.41,
                    "word": " maximizing",
                    "probability": 0.935302734375
                },
                {
                    "start": 286.41,
                    "end": 287.55,
                    "word": " the",
                    "probability": 0.802734375
                },
                {
                    "start": 287.55,
                    "end": 287.83,
                    "word": " current",
                    "probability": 0.79443359375
                },
                {
                    "start": 287.83,
                    "end": 288.25,
                    "word": " score",
                    "probability": 0.8134765625
                },
                {
                    "start": 288.25,
                    "end": 288.51,
                    "word": ".",
                    "probability": 0.56494140625
                }
            ]
        },
        {
            "id": 43,
            "seek": 29194,
            "start": 288.51,
            "end": 298.07,
            "text": " It wants to learn how to act so that at the end of the game, its total sum of rewards is maximized.",
            "tokens": [
                51108,
                467,
                2738,
                281,
                1466,
                577,
                281,
                605,
                370,
                300,
                412,
                264,
                917,
                295,
                264,
                1216,
                11,
                1080,
                3217,
                2408,
                295,
                17203,
                307,
                5138,
                1602,
                13,
                51544
            ],
            "temperature": 0.0,
            "avg_logprob": -0.18840580055679101,
            "compression_ratio": 1.660919540229885,
            "no_speech_prob": 0.0006561279296875,
            "words": [
                {
                    "start": 288.51,
                    "end": 288.85,
                    "word": " It",
                    "probability": 0.90283203125
                },
                {
                    "start": 288.85,
                    "end": 289.13,
                    "word": " wants",
                    "probability": 0.73193359375
                },
                {
                    "start": 289.13,
                    "end": 289.35,
                    "word": " to",
                    "probability": 0.904296875
                },
                {
                    "start": 289.35,
                    "end": 289.69,
                    "word": " learn",
                    "probability": 0.91748046875
                },
                {
                    "start": 289.69,
                    "end": 290.27,
                    "word": " how",
                    "probability": 0.89990234375
                },
                {
                    "start": 290.27,
                    "end": 290.45,
                    "word": " to",
                    "probability": 0.904296875
                },
                {
                    "start": 290.45,
                    "end": 290.81,
                    "word": " act",
                    "probability": 0.8271484375
                },
                {
                    "start": 290.81,
                    "end": 291.27,
                    "word": " so",
                    "probability": 0.76708984375
                },
                {
                    "start": 291.27,
                    "end": 291.43,
                    "word": " that",
                    "probability": 0.8828125
                },
                {
                    "start": 291.43,
                    "end": 291.53,
                    "word": " at",
                    "probability": 0.8740234375
                },
                {
                    "start": 291.53,
                    "end": 291.71,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 291.71,
                    "end": 291.93,
                    "word": " end",
                    "probability": 0.806640625
                },
                {
                    "start": 291.93,
                    "end": 292.13,
                    "word": " of",
                    "probability": 0.9091796875
                },
                {
                    "start": 292.13,
                    "end": 292.27,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 292.27,
                    "end": 292.61,
                    "word": " game",
                    "probability": 0.8115234375
                },
                {
                    "start": 292.61,
                    "end": 293.21,
                    "word": ",",
                    "probability": 0.4931640625
                },
                {
                    "start": 293.21,
                    "end": 293.49,
                    "word": " its",
                    "probability": 0.72509765625
                },
                {
                    "start": 293.49,
                    "end": 293.99,
                    "word": " total",
                    "probability": 0.81201171875
                },
                {
                    "start": 293.99,
                    "end": 294.39,
                    "word": " sum",
                    "probability": 0.912109375
                },
                {
                    "start": 294.39,
                    "end": 294.99,
                    "word": " of",
                    "probability": 0.90625
                },
                {
                    "start": 294.99,
                    "end": 295.99,
                    "word": " rewards",
                    "probability": 0.5986328125
                },
                {
                    "start": 295.99,
                    "end": 296.41,
                    "word": " is",
                    "probability": 0.90380859375
                },
                {
                    "start": 296.41,
                    "end": 297.39,
                    "word": " maximized",
                    "probability": 0.92236328125
                },
                {
                    "start": 297.39,
                    "end": 298.07,
                    "word": ".",
                    "probability": 0.98095703125
                }
            ]
        },
        {
            "id": 44,
            "seek": 31716,
            "start": 298.07,
            "end": 305.49,
            "text": " What are the different applications of RL? Of course, game playing is well publicized because",
            "tokens": [
                50380,
                708,
                366,
                264,
                819,
                5821,
                295,
                497,
                43,
                30,
                2720,
                1164,
                11,
                1216,
                2433,
                307,
                731,
                1908,
                1602,
                570,
                50728
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21602471207463464,
            "compression_ratio": 1.6,
            "no_speech_prob": 0.0142822265625,
            "words": [
                {
                    "start": 298.07,
                    "end": 298.75,
                    "word": " What",
                    "probability": 0.6435546875
                },
                {
                    "start": 298.75,
                    "end": 298.85,
                    "word": " are",
                    "probability": 0.89794921875
                },
                {
                    "start": 298.85,
                    "end": 298.95,
                    "word": " the",
                    "probability": 0.80859375
                },
                {
                    "start": 298.95,
                    "end": 299.13,
                    "word": " different",
                    "probability": 0.68505859375
                },
                {
                    "start": 299.13,
                    "end": 299.65,
                    "word": " applications",
                    "probability": 0.76123046875
                },
                {
                    "start": 299.65,
                    "end": 299.91,
                    "word": " of",
                    "probability": 0.90234375
                },
                {
                    "start": 299.91,
                    "end": 300.03,
                    "word": " RL",
                    "probability": 0.964111328125
                },
                {
                    "start": 300.03,
                    "end": 300.11,
                    "word": "?",
                    "probability": 0.94970703125
                },
                {
                    "start": 300.11,
                    "end": 300.21,
                    "word": " Of",
                    "probability": 0.9267578125
                },
                {
                    "start": 300.21,
                    "end": 300.41,
                    "word": " course",
                    "probability": 0.86474609375
                },
                {
                    "start": 300.41,
                    "end": 300.65,
                    "word": ",",
                    "probability": 0.65966796875
                },
                {
                    "start": 300.65,
                    "end": 300.81,
                    "word": " game",
                    "probability": 0.6279296875
                },
                {
                    "start": 300.81,
                    "end": 301.11,
                    "word": " playing",
                    "probability": 0.587890625
                },
                {
                    "start": 301.11,
                    "end": 302.71,
                    "word": " is",
                    "probability": 0.89013671875
                },
                {
                    "start": 302.71,
                    "end": 303.81,
                    "word": " well",
                    "probability": 0.8583984375
                },
                {
                    "start": 303.81,
                    "end": 305.27,
                    "word": " publicized",
                    "probability": 0.8037109375
                },
                {
                    "start": 305.27,
                    "end": 305.49,
                    "word": " because",
                    "probability": 0.71240234375
                }
            ]
        },
        {
            "id": 45,
            "seek": 31716,
            "start": 305.49,
            "end": 311.37,
            "text": " it's a very good testbed, and it has become one of the common testbeds and experimental",
            "tokens": [
                50728,
                309,
                311,
                257,
                588,
                665,
                1500,
                2883,
                11,
                293,
                309,
                575,
                1813,
                472,
                295,
                264,
                2689,
                1500,
                2883,
                82,
                293,
                17069,
                51032
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21602471207463464,
            "compression_ratio": 1.6,
            "no_speech_prob": 0.0142822265625,
            "words": [
                {
                    "start": 305.49,
                    "end": 305.65,
                    "word": " it's",
                    "probability": 0.93017578125
                },
                {
                    "start": 305.75,
                    "end": 305.85,
                    "word": " a very",
                    "probability": 0.72216796875
                },
                {
                    "start": 305.85,
                    "end": 306.03,
                    "word": " good",
                    "probability": 0.8896484375
                },
                {
                    "start": 306.03,
                    "end": 306.61,
                    "word": " testbed",
                    "probability": 0.5914306640625
                },
                {
                    "start": 306.61,
                    "end": 306.79,
                    "word": ",",
                    "probability": 0.2108154296875
                },
                {
                    "start": 306.79,
                    "end": 306.89,
                    "word": " and",
                    "probability": 0.892578125
                },
                {
                    "start": 306.89,
                    "end": 306.99,
                    "word": " it",
                    "probability": 0.896484375
                },
                {
                    "start": 306.99,
                    "end": 307.13,
                    "word": " has",
                    "probability": 0.888671875
                },
                {
                    "start": 307.13,
                    "end": 307.53,
                    "word": " become",
                    "probability": 0.83740234375
                },
                {
                    "start": 307.53,
                    "end": 308.05,
                    "word": " one",
                    "probability": 0.8115234375
                },
                {
                    "start": 308.05,
                    "end": 308.15,
                    "word": " of",
                    "probability": 0.90380859375
                },
                {
                    "start": 308.15,
                    "end": 308.71,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 308.71,
                    "end": 309.43,
                    "word": " common",
                    "probability": 0.7294921875
                },
                {
                    "start": 309.43,
                    "end": 310.79,
                    "word": " testbeds",
                    "probability": 0.8704427083333334
                },
                {
                    "start": 310.79,
                    "end": 310.89,
                    "word": " and",
                    "probability": 0.78369140625
                },
                {
                    "start": 310.89,
                    "end": 311.37,
                    "word": " experimental",
                    "probability": 0.76904296875
                }
            ]
        },
        {
            "id": 46,
            "seek": 31716,
            "start": 312.25,
            "end": 318.23,
            "text": " settings for RL research. However, there are some major real-world applications for reinforcement",
            "tokens": [
                51076,
                6257,
                337,
                497,
                43,
                2132,
                13,
                2908,
                11,
                456,
                366,
                512,
                2563,
                957,
                12,
                13217,
                5821,
                337,
                29280,
                51372
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21602471207463464,
            "compression_ratio": 1.6,
            "no_speech_prob": 0.0142822265625,
            "words": [
                {
                    "start": 312.25,
                    "end": 312.73,
                    "word": " settings",
                    "probability": 0.82958984375
                },
                {
                    "start": 312.73,
                    "end": 313.13,
                    "word": " for",
                    "probability": 0.89697265625
                },
                {
                    "start": 313.13,
                    "end": 313.47,
                    "word": " RL",
                    "probability": 0.962890625
                },
                {
                    "start": 313.47,
                    "end": 313.85,
                    "word": " research",
                    "probability": 0.78369140625
                },
                {
                    "start": 313.85,
                    "end": 314.03,
                    "word": ".",
                    "probability": 0.92333984375
                },
                {
                    "start": 314.03,
                    "end": 314.19,
                    "word": " However",
                    "probability": 0.82861328125
                },
                {
                    "start": 314.19,
                    "end": 314.23,
                    "word": ",",
                    "probability": 0.9755859375
                },
                {
                    "start": 314.23,
                    "end": 314.39,
                    "word": " there",
                    "probability": 0.853515625
                },
                {
                    "start": 314.39,
                    "end": 314.51,
                    "word": " are",
                    "probability": 0.88818359375
                },
                {
                    "start": 314.51,
                    "end": 314.89,
                    "word": " some",
                    "probability": 0.81640625
                },
                {
                    "start": 314.89,
                    "end": 316.17,
                    "word": " major",
                    "probability": 0.779296875
                },
                {
                    "start": 316.17,
                    "end": 316.93,
                    "word": " real",
                    "probability": 0.84912109375
                },
                {
                    "start": 316.93,
                    "end": 317.11,
                    "word": "-world",
                    "probability": 0.701171875
                },
                {
                    "start": 317.11,
                    "end": 317.59,
                    "word": " applications",
                    "probability": 0.71728515625
                },
                {
                    "start": 317.59,
                    "end": 317.83,
                    "word": " for",
                    "probability": 0.89697265625
                },
                {
                    "start": 317.83,
                    "end": 318.23,
                    "word": " reinforcement",
                    "probability": 0.78759765625
                }
            ]
        },
        {
            "id": 47,
            "seek": 31716,
            "start": 318.23,
            "end": 323.29,
            "text": " learning. In essence, RL, or reinforcement learning, is the machine learning response to",
            "tokens": [
                51372,
                2539,
                13,
                682,
                12801,
                11,
                497,
                43,
                11,
                420,
                29280,
                2539,
                11,
                307,
                264,
                3479,
                2539,
                4134,
                281,
                51612
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21602471207463464,
            "compression_ratio": 1.6,
            "no_speech_prob": 0.0142822265625,
            "words": [
                {
                    "start": 318.23,
                    "end": 318.55,
                    "word": " learning",
                    "probability": 0.81005859375
                },
                {
                    "start": 318.55,
                    "end": 318.83,
                    "word": ".",
                    "probability": 0.9755859375
                },
                {
                    "start": 318.83,
                    "end": 318.97,
                    "word": " In",
                    "probability": 0.91455078125
                },
                {
                    "start": 318.97,
                    "end": 319.39,
                    "word": " essence",
                    "probability": 0.83349609375
                },
                {
                    "start": 319.39,
                    "end": 319.57,
                    "word": ",",
                    "probability": 0.94775390625
                },
                {
                    "start": 319.57,
                    "end": 319.87,
                    "word": " RL",
                    "probability": 0.95751953125
                },
                {
                    "start": 319.87,
                    "end": 319.99,
                    "word": ",",
                    "probability": 0.505859375
                },
                {
                    "start": 319.99,
                    "end": 320.05,
                    "word": " or",
                    "probability": 0.876953125
                },
                {
                    "start": 320.05,
                    "end": 320.41,
                    "word": " reinforcement",
                    "probability": 0.7626953125
                },
                {
                    "start": 320.41,
                    "end": 320.77,
                    "word": " learning",
                    "probability": 0.83349609375
                },
                {
                    "start": 320.77,
                    "end": 320.99,
                    "word": ",",
                    "probability": 0.95556640625
                },
                {
                    "start": 320.99,
                    "end": 321.53,
                    "word": " is",
                    "probability": 0.900390625
                },
                {
                    "start": 321.53,
                    "end": 321.71,
                    "word": " the",
                    "probability": 0.80615234375
                },
                {
                    "start": 321.71,
                    "end": 321.95,
                    "word": " machine",
                    "probability": 0.72265625
                },
                {
                    "start": 321.95,
                    "end": 322.17,
                    "word": " learning",
                    "probability": 0.71044921875
                },
                {
                    "start": 322.17,
                    "end": 322.79,
                    "word": " response",
                    "probability": 0.87451171875
                },
                {
                    "start": 322.79,
                    "end": 323.29,
                    "word": " to",
                    "probability": 0.90380859375
                }
            ]
        },
        {
            "id": 48,
            "seek": 34450,
            "start": 324.01,
            "end": 330.73,
            "text": " the need for data-driven control problems. Where do we encounter such problems? Well,",
            "tokens": [
                50412,
                264,
                643,
                337,
                1412,
                12,
                25456,
                1969,
                2740,
                13,
                2305,
                360,
                321,
                8593,
                1270,
                2740,
                30,
                1042,
                11,
                50728
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2054036408662796,
            "compression_ratio": 1.667857142857143,
            "no_speech_prob": 0.01690673828125,
            "words": [
                {
                    "start": 324.01,
                    "end": 324.57,
                    "word": " the",
                    "probability": 0.3515625
                },
                {
                    "start": 324.57,
                    "end": 324.81,
                    "word": " need",
                    "probability": 0.83740234375
                },
                {
                    "start": 324.81,
                    "end": 325.01,
                    "word": " for",
                    "probability": 0.8974609375
                },
                {
                    "start": 325.01,
                    "end": 325.49,
                    "word": " data",
                    "probability": 0.8193359375
                },
                {
                    "start": 325.49,
                    "end": 325.81,
                    "word": "-driven",
                    "probability": 0.73876953125
                },
                {
                    "start": 325.81,
                    "end": 326.89,
                    "word": " control",
                    "probability": 0.650390625
                },
                {
                    "start": 326.89,
                    "end": 328.17,
                    "word": " problems",
                    "probability": 0.6630859375
                },
                {
                    "start": 328.17,
                    "end": 328.51,
                    "word": ".",
                    "probability": 0.93798828125
                },
                {
                    "start": 328.59,
                    "end": 329.15,
                    "word": " Where",
                    "probability": 0.72802734375
                },
                {
                    "start": 329.15,
                    "end": 329.27,
                    "word": " do",
                    "probability": 0.908203125
                },
                {
                    "start": 329.27,
                    "end": 329.43,
                    "word": " we",
                    "probability": 0.9013671875
                },
                {
                    "start": 329.43,
                    "end": 329.79,
                    "word": " encounter",
                    "probability": 0.69189453125
                },
                {
                    "start": 329.79,
                    "end": 330.03,
                    "word": " such",
                    "probability": 0.8818359375
                },
                {
                    "start": 330.03,
                    "end": 330.37,
                    "word": " problems",
                    "probability": 0.705078125
                },
                {
                    "start": 330.37,
                    "end": 330.53,
                    "word": "?",
                    "probability": 0.955078125
                },
                {
                    "start": 330.53,
                    "end": 330.63,
                    "word": " Well",
                    "probability": 0.671875
                },
                {
                    "start": 330.63,
                    "end": 330.73,
                    "word": ",",
                    "probability": 0.83544921875
                }
            ]
        },
        {
            "id": 49,
            "seek": 34450,
            "start": 330.73,
            "end": 334.85,
            "text": " in robotics, like autonomous navigation, object manipulation, and such. Algorithmic trading,",
            "tokens": [
                50728,
                294,
                34145,
                11,
                411,
                23797,
                17346,
                11,
                2657,
                26475,
                11,
                293,
                1270,
                13,
                35014,
                6819,
                13195,
                9529,
                11,
                50932
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2054036408662796,
            "compression_ratio": 1.667857142857143,
            "no_speech_prob": 0.01690673828125,
            "words": [
                {
                    "start": 330.73,
                    "end": 330.83,
                    "word": " in",
                    "probability": 0.83251953125
                },
                {
                    "start": 330.83,
                    "end": 331.17,
                    "word": " robotics",
                    "probability": 0.72216796875
                },
                {
                    "start": 331.17,
                    "end": 331.31,
                    "word": ",",
                    "probability": 0.51416015625
                },
                {
                    "start": 331.31,
                    "end": 331.45,
                    "word": " like",
                    "probability": 0.85498046875
                },
                {
                    "start": 331.45,
                    "end": 331.75,
                    "word": " autonomous",
                    "probability": 0.8447265625
                },
                {
                    "start": 331.75,
                    "end": 332.27,
                    "word": " navigation",
                    "probability": 0.802734375
                },
                {
                    "start": 332.27,
                    "end": 332.49,
                    "word": ",",
                    "probability": 0.9638671875
                },
                {
                    "start": 332.49,
                    "end": 332.69,
                    "word": " object",
                    "probability": 0.9208984375
                },
                {
                    "start": 332.69,
                    "end": 333.15,
                    "word": " manipulation",
                    "probability": 0.94482421875
                },
                {
                    "start": 333.15,
                    "end": 333.43,
                    "word": ",",
                    "probability": 0.396240234375
                },
                {
                    "start": 333.43,
                    "end": 333.51,
                    "word": " and",
                    "probability": 0.890625
                },
                {
                    "start": 333.51,
                    "end": 333.73,
                    "word": " such",
                    "probability": 0.89501953125
                },
                {
                    "start": 333.73,
                    "end": 333.99,
                    "word": ".",
                    "probability": 0.67041015625
                },
                {
                    "start": 333.99,
                    "end": 334.45,
                    "word": " Algorithmic",
                    "probability": 0.95361328125
                },
                {
                    "start": 334.45,
                    "end": 334.75,
                    "word": " trading",
                    "probability": 0.57568359375
                },
                {
                    "start": 334.75,
                    "end": 334.85,
                    "word": ",",
                    "probability": 0.61328125
                }
            ]
        },
        {
            "id": 50,
            "seek": 34450,
            "start": 334.85,
            "end": 339.97,
            "text": " this is one of the areas that my research group has recently become active in. Critical infrastructure",
            "tokens": [
                50932,
                341,
                307,
                472,
                295,
                264,
                3179,
                300,
                452,
                2132,
                1594,
                575,
                3938,
                1813,
                4967,
                294,
                13,
                39482,
                6896,
                51200
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2054036408662796,
            "compression_ratio": 1.667857142857143,
            "no_speech_prob": 0.01690673828125,
            "words": [
                {
                    "start": 334.85,
                    "end": 334.97,
                    "word": " this",
                    "probability": 0.876953125
                },
                {
                    "start": 334.97,
                    "end": 335.05,
                    "word": " is",
                    "probability": 0.9072265625
                },
                {
                    "start": 335.05,
                    "end": 335.17,
                    "word": " one",
                    "probability": 0.81884765625
                },
                {
                    "start": 335.17,
                    "end": 335.25,
                    "word": " of",
                    "probability": 0.8984375
                },
                {
                    "start": 335.25,
                    "end": 335.35,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 335.35,
                    "end": 335.67,
                    "word": " areas",
                    "probability": 0.8798828125
                },
                {
                    "start": 335.67,
                    "end": 335.99,
                    "word": " that",
                    "probability": 0.87841796875
                },
                {
                    "start": 335.99,
                    "end": 336.23,
                    "word": " my",
                    "probability": 0.89990234375
                },
                {
                    "start": 336.23,
                    "end": 336.49,
                    "word": " research",
                    "probability": 0.77490234375
                },
                {
                    "start": 336.49,
                    "end": 336.65,
                    "word": " group",
                    "probability": 0.8662109375
                },
                {
                    "start": 336.65,
                    "end": 336.77,
                    "word": " has",
                    "probability": 0.89404296875
                },
                {
                    "start": 336.77,
                    "end": 337.05,
                    "word": " recently",
                    "probability": 0.80224609375
                },
                {
                    "start": 337.05,
                    "end": 337.39,
                    "word": " become",
                    "probability": 0.8310546875
                },
                {
                    "start": 337.39,
                    "end": 338.23,
                    "word": " active",
                    "probability": 0.84521484375
                },
                {
                    "start": 338.23,
                    "end": 338.49,
                    "word": " in",
                    "probability": 0.890625
                },
                {
                    "start": 338.49,
                    "end": 339.05,
                    "word": ".",
                    "probability": 0.96484375
                },
                {
                    "start": 339.09,
                    "end": 339.37,
                    "word": " Critical",
                    "probability": 0.82568359375
                },
                {
                    "start": 339.37,
                    "end": 339.97,
                    "word": " infrastructure",
                    "probability": 0.88623046875
                }
            ]
        },
        {
            "id": 51,
            "seek": 34450,
            "start": 339.97,
            "end": 344.53,
            "text": " controlling smart cities, resource allocation in smart cities, traffic management, intelligent",
            "tokens": [
                51200,
                14905,
                4069,
                6486,
                11,
                7684,
                27599,
                294,
                4069,
                6486,
                11,
                6419,
                4592,
                11,
                13232,
                51424
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2054036408662796,
            "compression_ratio": 1.667857142857143,
            "no_speech_prob": 0.01690673828125,
            "words": [
                {
                    "start": 339.97,
                    "end": 340.51,
                    "word": " controlling",
                    "probability": 0.327392578125
                },
                {
                    "start": 340.51,
                    "end": 340.95,
                    "word": " smart",
                    "probability": 0.73388671875
                },
                {
                    "start": 340.95,
                    "end": 341.27,
                    "word": " cities",
                    "probability": 0.7490234375
                },
                {
                    "start": 341.27,
                    "end": 341.45,
                    "word": ",",
                    "probability": 0.8388671875
                },
                {
                    "start": 341.45,
                    "end": 341.73,
                    "word": " resource",
                    "probability": 0.82861328125
                },
                {
                    "start": 341.73,
                    "end": 342.19,
                    "word": " allocation",
                    "probability": 0.8203125
                },
                {
                    "start": 342.19,
                    "end": 342.55,
                    "word": " in",
                    "probability": 0.85888671875
                },
                {
                    "start": 342.55,
                    "end": 342.77,
                    "word": " smart",
                    "probability": 0.87255859375
                },
                {
                    "start": 342.77,
                    "end": 343.07,
                    "word": " cities",
                    "probability": 0.76953125
                },
                {
                    "start": 343.07,
                    "end": 343.31,
                    "word": ",",
                    "probability": 0.93115234375
                },
                {
                    "start": 343.31,
                    "end": 343.59,
                    "word": " traffic",
                    "probability": 0.7880859375
                },
                {
                    "start": 343.59,
                    "end": 344.09,
                    "word": " management",
                    "probability": 0.732421875
                },
                {
                    "start": 344.09,
                    "end": 344.29,
                    "word": ",",
                    "probability": 0.982421875
                },
                {
                    "start": 344.29,
                    "end": 344.53,
                    "word": " intelligent",
                    "probability": 0.7607421875
                }
            ]
        },
        {
            "id": 52,
            "seek": 34450,
            "start": 344.53,
            "end": 350.63,
            "text": " traffic systems, smart grid management, and control. Healthcare, such as clinical decision",
            "tokens": [
                51424,
                6419,
                3652,
                11,
                4069,
                10748,
                4592,
                11,
                293,
                1969,
                13,
                45548,
                11,
                1270,
                382,
                9115,
                3537,
                51728
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2054036408662796,
            "compression_ratio": 1.667857142857143,
            "no_speech_prob": 0.01690673828125,
            "words": [
                {
                    "start": 344.53,
                    "end": 344.81,
                    "word": " traffic",
                    "probability": 0.7744140625
                },
                {
                    "start": 344.81,
                    "end": 345.29,
                    "word": " systems",
                    "probability": 0.8173828125
                },
                {
                    "start": 345.29,
                    "end": 345.91,
                    "word": ",",
                    "probability": 0.9541015625
                },
                {
                    "start": 345.91,
                    "end": 346.15,
                    "word": " smart",
                    "probability": 0.88232421875
                },
                {
                    "start": 346.15,
                    "end": 346.53,
                    "word": " grid",
                    "probability": 0.73046875
                },
                {
                    "start": 346.53,
                    "end": 347.27,
                    "word": " management",
                    "probability": 0.56005859375
                },
                {
                    "start": 347.27,
                    "end": 347.59,
                    "word": ",",
                    "probability": 0.40966796875
                },
                {
                    "start": 347.59,
                    "end": 347.77,
                    "word": " and",
                    "probability": 0.8935546875
                },
                {
                    "start": 347.77,
                    "end": 348.19,
                    "word": " control",
                    "probability": 0.658203125
                },
                {
                    "start": 348.19,
                    "end": 348.75,
                    "word": ".",
                    "probability": 0.90966796875
                },
                {
                    "start": 349.09,
                    "end": 349.45,
                    "word": " Healthcare",
                    "probability": 0.77001953125
                },
                {
                    "start": 349.45,
                    "end": 349.73,
                    "word": ",",
                    "probability": 0.8857421875
                },
                {
                    "start": 349.73,
                    "end": 349.87,
                    "word": " such",
                    "probability": 0.8701171875
                },
                {
                    "start": 349.87,
                    "end": 350.01,
                    "word": " as",
                    "probability": 0.9130859375
                },
                {
                    "start": 350.01,
                    "end": 350.27,
                    "word": " clinical",
                    "probability": 0.900390625
                },
                {
                    "start": 350.27,
                    "end": 350.63,
                    "word": " decision",
                    "probability": 0.853515625
                }
            ]
        },
        {
            "id": 53,
            "seek": 37412,
            "start": 350.63,
            "end": 358.51,
            "text": " making, this is actually one of the better known applications of RL in the real world.",
            "tokens": [
                50364,
                1455,
                11,
                341,
                307,
                767,
                472,
                295,
                264,
                1101,
                2570,
                5821,
                295,
                497,
                43,
                294,
                264,
                957,
                1002,
                13,
                50720
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24013158522154154,
            "compression_ratio": 1.5636363636363637,
            "no_speech_prob": 0.00020992755889892578,
            "words": [
                {
                    "start": 350.63,
                    "end": 350.89,
                    "word": " making",
                    "probability": 0.177001953125
                },
                {
                    "start": 350.89,
                    "end": 350.99,
                    "word": ",",
                    "probability": 0.01371002197265625
                },
                {
                    "start": 350.99,
                    "end": 351.09,
                    "word": " this",
                    "probability": 0.81591796875
                },
                {
                    "start": 351.09,
                    "end": 351.21,
                    "word": " is",
                    "probability": 0.90380859375
                },
                {
                    "start": 351.21,
                    "end": 351.45,
                    "word": " actually",
                    "probability": 0.77978515625
                },
                {
                    "start": 351.45,
                    "end": 351.69,
                    "word": " one",
                    "probability": 0.806640625
                },
                {
                    "start": 351.69,
                    "end": 351.85,
                    "word": " of",
                    "probability": 0.8974609375
                },
                {
                    "start": 351.85,
                    "end": 352.15,
                    "word": " the",
                    "probability": 0.80517578125
                },
                {
                    "start": 352.15,
                    "end": 355.71,
                    "word": " better",
                    "probability": 0.489990234375
                },
                {
                    "start": 355.71,
                    "end": 356.11,
                    "word": " known",
                    "probability": 0.5556640625
                },
                {
                    "start": 356.11,
                    "end": 356.61,
                    "word": " applications",
                    "probability": 0.7939453125
                },
                {
                    "start": 356.61,
                    "end": 356.87,
                    "word": " of",
                    "probability": 0.9072265625
                },
                {
                    "start": 356.87,
                    "end": 357.09,
                    "word": " RL",
                    "probability": 0.954833984375
                },
                {
                    "start": 357.09,
                    "end": 357.21,
                    "word": " in",
                    "probability": 0.89501953125
                },
                {
                    "start": 357.21,
                    "end": 357.31,
                    "word": " the",
                    "probability": 0.81103515625
                },
                {
                    "start": 357.31,
                    "end": 357.49,
                    "word": " real",
                    "probability": 0.88134765625
                },
                {
                    "start": 357.49,
                    "end": 357.85,
                    "word": " world",
                    "probability": 0.869140625
                },
                {
                    "start": 357.85,
                    "end": 358.51,
                    "word": ".",
                    "probability": 0.1207275390625
                }
            ]
        },
        {
            "id": 54,
            "seek": 37412,
            "start": 358.67,
            "end": 364.14,
            "text": " And other types of resource management and applications in operations research and such.",
            "tokens": [
                50760,
                400,
                661,
                3467,
                295,
                7684,
                4592,
                293,
                5821,
                294,
                7705,
                2132,
                293,
                1270,
                13,
                51024
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24013158522154154,
            "compression_ratio": 1.5636363636363637,
            "no_speech_prob": 0.00020992755889892578,
            "words": [
                {
                    "start": 358.67,
                    "end": 358.81,
                    "word": " And",
                    "probability": 0.787109375
                },
                {
                    "start": 358.81,
                    "end": 359.09,
                    "word": " other",
                    "probability": 0.7626953125
                },
                {
                    "start": 359.09,
                    "end": 359.61,
                    "word": " types",
                    "probability": 0.65869140625
                },
                {
                    "start": 359.61,
                    "end": 359.71,
                    "word": " of",
                    "probability": 0.89697265625
                },
                {
                    "start": 359.71,
                    "end": 359.97,
                    "word": " resource",
                    "probability": 0.8037109375
                },
                {
                    "start": 359.97,
                    "end": 360.43,
                    "word": " management",
                    "probability": 0.7490234375
                },
                {
                    "start": 360.43,
                    "end": 361.43,
                    "word": " and",
                    "probability": 0.791015625
                },
                {
                    "start": 361.43,
                    "end": 362.29,
                    "word": " applications",
                    "probability": 0.7548828125
                },
                {
                    "start": 362.29,
                    "end": 362.65,
                    "word": " in",
                    "probability": 0.896484375
                },
                {
                    "start": 362.65,
                    "end": 363.01,
                    "word": " operations",
                    "probability": 0.8173828125
                },
                {
                    "start": 363.01,
                    "end": 363.43,
                    "word": " research",
                    "probability": 0.71875
                },
                {
                    "start": 363.43,
                    "end": 363.57,
                    "word": " and",
                    "probability": 0.83251953125
                },
                {
                    "start": 363.57,
                    "end": 363.81,
                    "word": " such",
                    "probability": 0.84814453125
                },
                {
                    "start": 363.81,
                    "end": 364.14,
                    "word": ".",
                    "probability": 0.9755859375
                }
            ]
        },
        {
            "id": 55,
            "seek": 37412,
            "start": 364.77,
            "end": 373.43,
            "text": " So, RL is either envisioned or is already heavily adopted in various industries,",
            "tokens": [
                51064,
                407,
                11,
                497,
                43,
                307,
                2139,
                47733,
                420,
                307,
                1217,
                10950,
                12175,
                294,
                3683,
                13284,
                11,
                51500
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24013158522154154,
            "compression_ratio": 1.5636363636363637,
            "no_speech_prob": 0.00020992755889892578,
            "words": [
                {
                    "start": 364.77,
                    "end": 364.99,
                    "word": " So",
                    "probability": 0.80615234375
                },
                {
                    "start": 364.99,
                    "end": 365.39,
                    "word": ",",
                    "probability": 0.2410888671875
                },
                {
                    "start": 365.39,
                    "end": 365.53,
                    "word": " RL",
                    "probability": 0.967529296875
                },
                {
                    "start": 365.53,
                    "end": 365.75,
                    "word": " is",
                    "probability": 0.90380859375
                },
                {
                    "start": 365.75,
                    "end": 365.97,
                    "word": " either",
                    "probability": 0.876953125
                },
                {
                    "start": 365.97,
                    "end": 366.51,
                    "word": " envisioned",
                    "probability": 0.81982421875
                },
                {
                    "start": 366.51,
                    "end": 366.87,
                    "word": " or",
                    "probability": 0.8212890625
                },
                {
                    "start": 366.87,
                    "end": 367.07,
                    "word": " is",
                    "probability": 0.8427734375
                },
                {
                    "start": 367.07,
                    "end": 367.99,
                    "word": " already",
                    "probability": 0.8369140625
                },
                {
                    "start": 367.99,
                    "end": 370.37,
                    "word": " heavily",
                    "probability": 0.7080078125
                },
                {
                    "start": 370.37,
                    "end": 371.03,
                    "word": " adopted",
                    "probability": 0.84814453125
                },
                {
                    "start": 371.03,
                    "end": 371.99,
                    "word": " in",
                    "probability": 0.90478515625
                },
                {
                    "start": 371.99,
                    "end": 372.61,
                    "word": " various",
                    "probability": 0.880859375
                },
                {
                    "start": 372.61,
                    "end": 373.19,
                    "word": " industries",
                    "probability": 0.72607421875
                },
                {
                    "start": 373.19,
                    "end": 373.43,
                    "word": ",",
                    "probability": 0.7958984375
                }
            ]
        },
        {
            "id": 56,
            "seek": 37412,
            "start": 373.43,
            "end": 380.25,
            "text": " many of which are critical and may become targets of malicious actions. Let's formalize",
            "tokens": [
                51500,
                867,
                295,
                597,
                366,
                4924,
                293,
                815,
                1813,
                12911,
                295,
                33496,
                5909,
                13,
                961,
                311,
                9860,
                1125,
                51856
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24013158522154154,
            "compression_ratio": 1.5636363636363637,
            "no_speech_prob": 0.00020992755889892578,
            "words": [
                {
                    "start": 373.43,
                    "end": 373.63,
                    "word": " many",
                    "probability": 0.8154296875
                },
                {
                    "start": 373.63,
                    "end": 373.81,
                    "word": " of",
                    "probability": 0.89794921875
                },
                {
                    "start": 373.81,
                    "end": 374.03,
                    "word": " which",
                    "probability": 0.91845703125
                },
                {
                    "start": 374.03,
                    "end": 374.33,
                    "word": " are",
                    "probability": 0.8955078125
                },
                {
                    "start": 374.33,
                    "end": 374.85,
                    "word": " critical",
                    "probability": 0.8583984375
                },
                {
                    "start": 374.85,
                    "end": 375.25,
                    "word": " and",
                    "probability": 0.85107421875
                },
                {
                    "start": 375.25,
                    "end": 375.47,
                    "word": " may",
                    "probability": 0.88720703125
                },
                {
                    "start": 375.47,
                    "end": 375.71,
                    "word": " become",
                    "probability": 0.84619140625
                },
                {
                    "start": 375.71,
                    "end": 376.11,
                    "word": " targets",
                    "probability": 0.59765625
                },
                {
                    "start": 376.11,
                    "end": 376.57,
                    "word": " of",
                    "probability": 0.908203125
                },
                {
                    "start": 376.57,
                    "end": 377.27,
                    "word": " malicious",
                    "probability": 0.845703125
                },
                {
                    "start": 377.27,
                    "end": 377.77,
                    "word": " actions",
                    "probability": 0.61181640625
                },
                {
                    "start": 377.77,
                    "end": 378.43,
                    "word": ".",
                    "probability": 0.99755859375
                },
                {
                    "start": 378.77,
                    "end": 379.43,
                    "word": " Let's",
                    "probability": 0.941162109375
                },
                {
                    "start": 379.43,
                    "end": 380.25,
                    "word": " formalize",
                    "probability": 0.7568359375
                }
            ]
        },
        {
            "id": 57,
            "seek": 39934,
            "start": 380.25,
            "end": 385.33,
            "text": " the RL problem a little more. I'm just going to quickly go over this. There will be some math,",
            "tokens": [
                50388,
                264,
                497,
                43,
                1154,
                257,
                707,
                544,
                13,
                286,
                478,
                445,
                516,
                281,
                2661,
                352,
                670,
                341,
                13,
                821,
                486,
                312,
                512,
                5221,
                11,
                50616
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1992187567853502,
            "compression_ratio": 1.7326007326007327,
            "no_speech_prob": 0.014495849609375,
            "words": [
                {
                    "start": 380.25,
                    "end": 380.67,
                    "word": " the",
                    "probability": 0.1361083984375
                },
                {
                    "start": 380.67,
                    "end": 381.03,
                    "word": " RL",
                    "probability": 0.826416015625
                },
                {
                    "start": 381.03,
                    "end": 381.35,
                    "word": " problem",
                    "probability": 0.74365234375
                },
                {
                    "start": 381.47,
                    "end": 381.59,
                    "word": " a little",
                    "probability": 0.80615234375
                },
                {
                    "start": 381.59,
                    "end": 381.91,
                    "word": " more",
                    "probability": 0.86962890625
                },
                {
                    "start": 381.91,
                    "end": 382.27,
                    "word": ".",
                    "probability": 0.7216796875
                },
                {
                    "start": 382.27,
                    "end": 382.57,
                    "word": " I'm",
                    "probability": 0.951904296875
                },
                {
                    "start": 382.57,
                    "end": 382.73,
                    "word": " just",
                    "probability": 0.86865234375
                },
                {
                    "start": 382.73,
                    "end": 382.91,
                    "word": " going",
                    "probability": 0.525390625
                },
                {
                    "start": 382.91,
                    "end": 383.05,
                    "word": " to",
                    "probability": 0.89990234375
                },
                {
                    "start": 383.05,
                    "end": 383.33,
                    "word": " quickly",
                    "probability": 0.875
                },
                {
                    "start": 383.33,
                    "end": 383.55,
                    "word": " go",
                    "probability": 0.89501953125
                },
                {
                    "start": 383.55,
                    "end": 383.71,
                    "word": " over",
                    "probability": 0.89697265625
                },
                {
                    "start": 383.71,
                    "end": 383.97,
                    "word": " this",
                    "probability": 0.875
                },
                {
                    "start": 383.97,
                    "end": 384.21,
                    "word": ".",
                    "probability": 0.802734375
                },
                {
                    "start": 384.21,
                    "end": 384.27,
                    "word": " There",
                    "probability": 0.79931640625
                },
                {
                    "start": 384.27,
                    "end": 384.39,
                    "word": " will",
                    "probability": 0.806640625
                },
                {
                    "start": 384.39,
                    "end": 384.53,
                    "word": " be",
                    "probability": 0.9111328125
                },
                {
                    "start": 384.53,
                    "end": 384.77,
                    "word": " some",
                    "probability": 0.81005859375
                },
                {
                    "start": 384.77,
                    "end": 385.07,
                    "word": " math",
                    "probability": 0.87451171875
                },
                {
                    "start": 385.07,
                    "end": 385.33,
                    "word": ",",
                    "probability": 0.62158203125
                }
            ]
        },
        {
            "id": 58,
            "seek": 39934,
            "start": 385.33,
            "end": 392.01,
            "text": " but this is just to introduce the basic settings. The underlying framework to formulate the RL",
            "tokens": [
                50616,
                457,
                341,
                307,
                445,
                281,
                5366,
                264,
                3875,
                6257,
                13,
                440,
                14217,
                8388,
                281,
                47881,
                264,
                497,
                43,
                50964
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1992187567853502,
            "compression_ratio": 1.7326007326007327,
            "no_speech_prob": 0.014495849609375,
            "words": [
                {
                    "start": 385.33,
                    "end": 385.43,
                    "word": " but",
                    "probability": 0.89306640625
                },
                {
                    "start": 385.43,
                    "end": 385.97,
                    "word": " this",
                    "probability": 0.84619140625
                },
                {
                    "start": 385.97,
                    "end": 386.05,
                    "word": " is",
                    "probability": 0.81982421875
                },
                {
                    "start": 386.05,
                    "end": 386.17,
                    "word": " just",
                    "probability": 0.87744140625
                },
                {
                    "start": 386.17,
                    "end": 386.31,
                    "word": " to",
                    "probability": 0.8505859375
                },
                {
                    "start": 386.31,
                    "end": 386.71,
                    "word": " introduce",
                    "probability": 0.68408203125
                },
                {
                    "start": 386.71,
                    "end": 386.97,
                    "word": " the",
                    "probability": 0.80126953125
                },
                {
                    "start": 386.97,
                    "end": 387.25,
                    "word": " basic",
                    "probability": 0.91015625
                },
                {
                    "start": 387.25,
                    "end": 387.67,
                    "word": " settings",
                    "probability": 0.8134765625
                },
                {
                    "start": 387.67,
                    "end": 388.19,
                    "word": ".",
                    "probability": 0.96728515625
                },
                {
                    "start": 388.97,
                    "end": 389.49,
                    "word": " The",
                    "probability": 0.796875
                },
                {
                    "start": 389.49,
                    "end": 390.07,
                    "word": " underlying",
                    "probability": 0.85546875
                },
                {
                    "start": 390.07,
                    "end": 390.87,
                    "word": " framework",
                    "probability": 0.76904296875
                },
                {
                    "start": 390.87,
                    "end": 391.17,
                    "word": " to",
                    "probability": 0.892578125
                },
                {
                    "start": 391.17,
                    "end": 391.57,
                    "word": " formulate",
                    "probability": 0.8818359375
                },
                {
                    "start": 391.57,
                    "end": 391.81,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 391.81,
                    "end": 392.01,
                    "word": " RL",
                    "probability": 0.96630859375
                }
            ]
        },
        {
            "id": 59,
            "seek": 39934,
            "start": 392.01,
            "end": 397.03,
            "text": " problem and provide a framework to think about and reason about the RL problem is the Markov",
            "tokens": [
                50964,
                1154,
                293,
                2893,
                257,
                8388,
                281,
                519,
                466,
                293,
                1778,
                466,
                264,
                497,
                43,
                1154,
                307,
                264,
                3934,
                5179,
                51216
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1992187567853502,
            "compression_ratio": 1.7326007326007327,
            "no_speech_prob": 0.014495849609375,
            "words": [
                {
                    "start": 392.01,
                    "end": 392.37,
                    "word": " problem",
                    "probability": 0.7763671875
                },
                {
                    "start": 392.37,
                    "end": 392.79,
                    "word": " and",
                    "probability": 0.81884765625
                },
                {
                    "start": 392.79,
                    "end": 393.23,
                    "word": " provide",
                    "probability": 0.88232421875
                },
                {
                    "start": 393.45,
                    "end": 393.71,
                    "word": " a framework",
                    "probability": 0.72021484375
                },
                {
                    "start": 393.71,
                    "end": 393.89,
                    "word": " to",
                    "probability": 0.89453125
                },
                {
                    "start": 393.89,
                    "end": 394.11,
                    "word": " think",
                    "probability": 0.80712890625
                },
                {
                    "start": 394.11,
                    "end": 394.51,
                    "word": " about",
                    "probability": 0.8212890625
                },
                {
                    "start": 394.51,
                    "end": 394.87,
                    "word": " and",
                    "probability": 0.31982421875
                },
                {
                    "start": 394.87,
                    "end": 395.09,
                    "word": " reason",
                    "probability": 0.92236328125
                },
                {
                    "start": 395.09,
                    "end": 395.35,
                    "word": " about",
                    "probability": 0.80908203125
                },
                {
                    "start": 395.35,
                    "end": 395.49,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 395.49,
                    "end": 395.67,
                    "word": " RL",
                    "probability": 0.96728515625
                },
                {
                    "start": 395.67,
                    "end": 396.03,
                    "word": " problem",
                    "probability": 0.77001953125
                },
                {
                    "start": 396.03,
                    "end": 396.51,
                    "word": " is",
                    "probability": 0.83984375
                },
                {
                    "start": 396.51,
                    "end": 396.77,
                    "word": " the",
                    "probability": 0.8095703125
                },
                {
                    "start": 396.77,
                    "end": 397.03,
                    "word": " Markov",
                    "probability": 0.92822265625
                }
            ]
        },
        {
            "id": 60,
            "seek": 39934,
            "start": 397.03,
            "end": 400.85,
            "text": " decision process. Why do we call it the Markov decision process? Because it's based on the",
            "tokens": [
                51216,
                3537,
                1399,
                13,
                1545,
                360,
                321,
                818,
                309,
                264,
                3934,
                5179,
                3537,
                1399,
                30,
                1436,
                309,
                311,
                2361,
                322,
                264,
                51404
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1992187567853502,
            "compression_ratio": 1.7326007326007327,
            "no_speech_prob": 0.014495849609375,
            "words": [
                {
                    "start": 397.03,
                    "end": 397.31,
                    "word": " decision",
                    "probability": 0.6845703125
                },
                {
                    "start": 397.31,
                    "end": 397.69,
                    "word": " process",
                    "probability": 0.85400390625
                },
                {
                    "start": 397.69,
                    "end": 397.77,
                    "word": ".",
                    "probability": 0.9853515625
                },
                {
                    "start": 397.77,
                    "end": 397.85,
                    "word": " Why",
                    "probability": 0.78759765625
                },
                {
                    "start": 397.85,
                    "end": 397.91,
                    "word": " do",
                    "probability": 0.888671875
                },
                {
                    "start": 397.91,
                    "end": 398.01,
                    "word": " we",
                    "probability": 0.89892578125
                },
                {
                    "start": 398.01,
                    "end": 398.13,
                    "word": " call",
                    "probability": 0.80615234375
                },
                {
                    "start": 398.13,
                    "end": 398.19,
                    "word": " it",
                    "probability": 0.8408203125
                },
                {
                    "start": 398.19,
                    "end": 398.29,
                    "word": " the",
                    "probability": 0.75927734375
                },
                {
                    "start": 398.29,
                    "end": 398.53,
                    "word": " Markov",
                    "probability": 0.905517578125
                },
                {
                    "start": 398.53,
                    "end": 398.83,
                    "word": " decision",
                    "probability": 0.85107421875
                },
                {
                    "start": 398.83,
                    "end": 399.17,
                    "word": " process",
                    "probability": 0.86474609375
                },
                {
                    "start": 399.17,
                    "end": 399.19,
                    "word": "?",
                    "probability": 0.91943359375
                },
                {
                    "start": 399.19,
                    "end": 399.43,
                    "word": " Because",
                    "probability": 0.892578125
                },
                {
                    "start": 399.43,
                    "end": 399.63,
                    "word": " it's",
                    "probability": 0.946044921875
                },
                {
                    "start": 399.63,
                    "end": 399.97,
                    "word": " based",
                    "probability": 0.818359375
                },
                {
                    "start": 399.97,
                    "end": 400.33,
                    "word": " on",
                    "probability": 0.9033203125
                },
                {
                    "start": 400.33,
                    "end": 400.85,
                    "word": " the",
                    "probability": 0.8125
                }
            ]
        },
        {
            "id": 61,
            "seek": 39934,
            "start": 400.85,
            "end": 405.47,
            "text": " Markovian assumption, or the Markov property, which says the current state completely characterizes",
            "tokens": [
                51404,
                3934,
                5179,
                952,
                15302,
                11,
                420,
                264,
                3934,
                5179,
                4707,
                11,
                597,
                1619,
                264,
                2190,
                1785,
                2584,
                2517,
                5660,
                51640
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1992187567853502,
            "compression_ratio": 1.7326007326007327,
            "no_speech_prob": 0.014495849609375,
            "words": [
                {
                    "start": 400.85,
                    "end": 401.35,
                    "word": " Markovian",
                    "probability": 0.9265950520833334
                },
                {
                    "start": 401.35,
                    "end": 401.73,
                    "word": " assumption",
                    "probability": 0.81884765625
                },
                {
                    "start": 401.73,
                    "end": 401.93,
                    "word": ",",
                    "probability": 0.1949462890625
                },
                {
                    "start": 401.93,
                    "end": 402.03,
                    "word": " or",
                    "probability": 0.9072265625
                },
                {
                    "start": 402.03,
                    "end": 402.15,
                    "word": " the",
                    "probability": 0.81103515625
                },
                {
                    "start": 402.15,
                    "end": 402.45,
                    "word": " Markov",
                    "probability": 0.902587890625
                },
                {
                    "start": 402.45,
                    "end": 402.83,
                    "word": " property",
                    "probability": 0.8212890625
                },
                {
                    "start": 402.83,
                    "end": 402.99,
                    "word": ",",
                    "probability": 0.89794921875
                },
                {
                    "start": 402.99,
                    "end": 403.11,
                    "word": " which",
                    "probability": 0.90576171875
                },
                {
                    "start": 403.11,
                    "end": 403.31,
                    "word": " says",
                    "probability": 0.80224609375
                },
                {
                    "start": 403.31,
                    "end": 403.43,
                    "word": " the",
                    "probability": 0.7431640625
                },
                {
                    "start": 403.43,
                    "end": 403.63,
                    "word": " current",
                    "probability": 0.7177734375
                },
                {
                    "start": 403.63,
                    "end": 404.01,
                    "word": " state",
                    "probability": 0.83349609375
                },
                {
                    "start": 404.01,
                    "end": 404.87,
                    "word": " completely",
                    "probability": 0.59326171875
                },
                {
                    "start": 404.87,
                    "end": 405.47,
                    "word": " characterizes",
                    "probability": 0.841552734375
                }
            ]
        },
        {
            "id": 62,
            "seek": 42500,
            "start": 405.47,
            "end": 412.05,
            "text": " the state of the world. So, if you have the current state and you perform an action,",
            "tokens": [
                50364,
                264,
                1785,
                295,
                264,
                1002,
                13,
                407,
                11,
                498,
                291,
                362,
                264,
                2190,
                1785,
                293,
                291,
                2042,
                364,
                3069,
                11,
                50656
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1771399433846059,
            "compression_ratio": 1.7440758293838863,
            "no_speech_prob": 0.0010166168212890625,
            "words": [
                {
                    "start": 405.47,
                    "end": 405.65,
                    "word": " the",
                    "probability": 0.30322265625
                },
                {
                    "start": 405.65,
                    "end": 405.93,
                    "word": " state",
                    "probability": 0.8544921875
                },
                {
                    "start": 405.93,
                    "end": 406.05,
                    "word": " of",
                    "probability": 0.9072265625
                },
                {
                    "start": 406.05,
                    "end": 406.17,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 406.17,
                    "end": 406.39,
                    "word": " world",
                    "probability": 0.86865234375
                },
                {
                    "start": 406.39,
                    "end": 406.91,
                    "word": ".",
                    "probability": 0.845703125
                },
                {
                    "start": 407.11,
                    "end": 407.45,
                    "word": " So",
                    "probability": 0.85986328125
                },
                {
                    "start": 407.45,
                    "end": 407.61,
                    "word": ",",
                    "probability": 0.317138671875
                },
                {
                    "start": 407.61,
                    "end": 408.15,
                    "word": " if",
                    "probability": 0.896484375
                },
                {
                    "start": 408.15,
                    "end": 408.27,
                    "word": " you",
                    "probability": 0.896484375
                },
                {
                    "start": 408.27,
                    "end": 408.53,
                    "word": " have",
                    "probability": 0.88330078125
                },
                {
                    "start": 408.53,
                    "end": 408.73,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 408.73,
                    "end": 408.93,
                    "word": " current",
                    "probability": 0.736328125
                },
                {
                    "start": 408.93,
                    "end": 409.41,
                    "word": " state",
                    "probability": 0.87548828125
                },
                {
                    "start": 409.41,
                    "end": 410.03,
                    "word": " and",
                    "probability": 0.72412109375
                },
                {
                    "start": 410.03,
                    "end": 410.29,
                    "word": " you",
                    "probability": 0.89501953125
                },
                {
                    "start": 410.29,
                    "end": 410.67,
                    "word": " perform",
                    "probability": 0.79150390625
                },
                {
                    "start": 410.67,
                    "end": 410.91,
                    "word": " an",
                    "probability": 0.9013671875
                },
                {
                    "start": 410.91,
                    "end": 411.35,
                    "word": " action",
                    "probability": 0.7978515625
                },
                {
                    "start": 411.35,
                    "end": 412.05,
                    "word": ",",
                    "probability": 0.79931640625
                }
            ]
        },
        {
            "id": 63,
            "seek": 42500,
            "start": 412.05,
            "end": 419.41,
            "text": " the next state is only going to depend or be a function of the previous state and nothing",
            "tokens": [
                50692,
                264,
                958,
                1785,
                307,
                787,
                516,
                281,
                5672,
                420,
                312,
                257,
                2445,
                295,
                264,
                3894,
                1785,
                293,
                1825,
                51056
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1771399433846059,
            "compression_ratio": 1.7440758293838863,
            "no_speech_prob": 0.0010166168212890625,
            "words": [
                {
                    "start": 412.05,
                    "end": 412.27,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 412.27,
                    "end": 412.55,
                    "word": " next",
                    "probability": 0.8662109375
                },
                {
                    "start": 412.55,
                    "end": 412.93,
                    "word": " state",
                    "probability": 0.88427734375
                },
                {
                    "start": 412.93,
                    "end": 413.27,
                    "word": " is",
                    "probability": 0.89892578125
                },
                {
                    "start": 413.27,
                    "end": 413.59,
                    "word": " only",
                    "probability": 0.85498046875
                },
                {
                    "start": 413.59,
                    "end": 413.91,
                    "word": " going",
                    "probability": 0.8564453125
                },
                {
                    "start": 413.91,
                    "end": 414.27,
                    "word": " to",
                    "probability": 0.904296875
                },
                {
                    "start": 414.27,
                    "end": 415.07,
                    "word": " depend",
                    "probability": 0.80078125
                },
                {
                    "start": 415.07,
                    "end": 415.31,
                    "word": " or",
                    "probability": 0.7392578125
                },
                {
                    "start": 415.31,
                    "end": 415.49,
                    "word": " be",
                    "probability": 0.90087890625
                },
                {
                    "start": 415.81,
                    "end": 416.23,
                    "word": " a function",
                    "probability": 0.8779296875
                },
                {
                    "start": 416.23,
                    "end": 417.15,
                    "word": " of",
                    "probability": 0.90380859375
                },
                {
                    "start": 417.15,
                    "end": 417.95,
                    "word": " the",
                    "probability": 0.81884765625
                },
                {
                    "start": 417.95,
                    "end": 418.33,
                    "word": " previous",
                    "probability": 0.76171875
                },
                {
                    "start": 418.33,
                    "end": 418.75,
                    "word": " state",
                    "probability": 0.81494140625
                },
                {
                    "start": 418.75,
                    "end": 419.15,
                    "word": " and",
                    "probability": 0.77685546875
                },
                {
                    "start": 419.15,
                    "end": 419.41,
                    "word": " nothing",
                    "probability": 0.84375
                }
            ]
        },
        {
            "id": 64,
            "seek": 42500,
            "start": 419.41,
            "end": 423.35,
            "text": " before it. You don't need to know the history of the environment. All you need to know is what the",
            "tokens": [
                51056,
                949,
                309,
                13,
                509,
                500,
                380,
                643,
                281,
                458,
                264,
                2503,
                295,
                264,
                2823,
                13,
                1057,
                291,
                643,
                281,
                458,
                307,
                437,
                264,
                51252
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1771399433846059,
            "compression_ratio": 1.7440758293838863,
            "no_speech_prob": 0.0010166168212890625,
            "words": [
                {
                    "start": 419.41,
                    "end": 419.67,
                    "word": " before",
                    "probability": 0.79052734375
                },
                {
                    "start": 419.67,
                    "end": 419.77,
                    "word": " it",
                    "probability": 0.444580078125
                },
                {
                    "start": 419.77,
                    "end": 419.79,
                    "word": ".",
                    "probability": 0.814453125
                },
                {
                    "start": 419.79,
                    "end": 419.83,
                    "word": " You",
                    "probability": 0.884765625
                },
                {
                    "start": 419.83,
                    "end": 419.95,
                    "word": " don't",
                    "probability": 0.94287109375
                },
                {
                    "start": 419.95,
                    "end": 420.07,
                    "word": " need",
                    "probability": 0.8115234375
                },
                {
                    "start": 420.07,
                    "end": 420.17,
                    "word": " to",
                    "probability": 0.90625
                },
                {
                    "start": 420.17,
                    "end": 420.27,
                    "word": " know",
                    "probability": 0.82373046875
                },
                {
                    "start": 420.27,
                    "end": 420.43,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 420.43,
                    "end": 420.83,
                    "word": " history",
                    "probability": 0.85546875
                },
                {
                    "start": 420.83,
                    "end": 421.71,
                    "word": " of",
                    "probability": 0.90576171875
                },
                {
                    "start": 421.71,
                    "end": 421.87,
                    "word": " the",
                    "probability": 0.806640625
                },
                {
                    "start": 421.87,
                    "end": 422.17,
                    "word": " environment",
                    "probability": 0.7763671875
                },
                {
                    "start": 422.17,
                    "end": 422.33,
                    "word": ".",
                    "probability": 0.65087890625
                },
                {
                    "start": 422.33,
                    "end": 422.43,
                    "word": " All",
                    "probability": 0.89794921875
                },
                {
                    "start": 422.43,
                    "end": 422.53,
                    "word": " you",
                    "probability": 0.89599609375
                },
                {
                    "start": 422.53,
                    "end": 422.67,
                    "word": " need",
                    "probability": 0.8154296875
                },
                {
                    "start": 422.67,
                    "end": 422.77,
                    "word": " to",
                    "probability": 0.904296875
                },
                {
                    "start": 422.77,
                    "end": 422.93,
                    "word": " know",
                    "probability": 0.81884765625
                },
                {
                    "start": 422.93,
                    "end": 423.09,
                    "word": " is",
                    "probability": 0.9052734375
                },
                {
                    "start": 423.09,
                    "end": 423.23,
                    "word": " what",
                    "probability": 0.87353515625
                },
                {
                    "start": 423.23,
                    "end": 423.35,
                    "word": " the",
                    "probability": 0.8193359375
                }
            ]
        },
        {
            "id": 65,
            "seek": 42500,
            "start": 423.35,
            "end": 431.13,
            "text": " previous state has been to infer what the coming state is going to be. Now, in Markov decision",
            "tokens": [
                51252,
                3894,
                1785,
                575,
                668,
                281,
                13596,
                437,
                264,
                1348,
                1785,
                307,
                516,
                281,
                312,
                13,
                823,
                11,
                294,
                3934,
                5179,
                3537,
                51648
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1771399433846059,
            "compression_ratio": 1.7440758293838863,
            "no_speech_prob": 0.0010166168212890625,
            "words": [
                {
                    "start": 423.35,
                    "end": 423.67,
                    "word": " previous",
                    "probability": 0.75830078125
                },
                {
                    "start": 423.67,
                    "end": 423.95,
                    "word": " state",
                    "probability": 0.76025390625
                },
                {
                    "start": 423.95,
                    "end": 424.09,
                    "word": " has",
                    "probability": 0.85595703125
                },
                {
                    "start": 424.09,
                    "end": 424.25,
                    "word": " been",
                    "probability": 0.826171875
                },
                {
                    "start": 424.25,
                    "end": 425.43,
                    "word": " to",
                    "probability": 0.62744140625
                },
                {
                    "start": 425.43,
                    "end": 425.75,
                    "word": " infer",
                    "probability": 0.88720703125
                },
                {
                    "start": 425.75,
                    "end": 425.93,
                    "word": " what",
                    "probability": 0.87744140625
                },
                {
                    "start": 425.93,
                    "end": 426.21,
                    "word": " the",
                    "probability": 0.73291015625
                },
                {
                    "start": 426.21,
                    "end": 427.23,
                    "word": " coming",
                    "probability": 0.7587890625
                },
                {
                    "start": 427.23,
                    "end": 427.63,
                    "word": " state",
                    "probability": 0.8427734375
                },
                {
                    "start": 427.63,
                    "end": 428.19,
                    "word": " is",
                    "probability": 0.90234375
                },
                {
                    "start": 428.19,
                    "end": 428.41,
                    "word": " going",
                    "probability": 0.857421875
                },
                {
                    "start": 428.41,
                    "end": 428.59,
                    "word": " to",
                    "probability": 0.904296875
                },
                {
                    "start": 428.59,
                    "end": 428.75,
                    "word": " be",
                    "probability": 0.9013671875
                },
                {
                    "start": 428.75,
                    "end": 429.27,
                    "word": ".",
                    "probability": 0.9853515625
                },
                {
                    "start": 429.53,
                    "end": 429.93,
                    "word": " Now",
                    "probability": 0.88232421875
                },
                {
                    "start": 429.93,
                    "end": 430.09,
                    "word": ",",
                    "probability": 0.8154296875
                },
                {
                    "start": 430.09,
                    "end": 430.51,
                    "word": " in",
                    "probability": 0.908203125
                },
                {
                    "start": 430.51,
                    "end": 430.83,
                    "word": " Markov",
                    "probability": 0.897705078125
                },
                {
                    "start": 430.83,
                    "end": 431.13,
                    "word": " decision",
                    "probability": 0.79248046875
                }
            ]
        },
        {
            "id": 66,
            "seek": 45014,
            "start": 431.13,
            "end": 439.49,
            "text": " processes, a problem or setting is formulated by a tuple of five main parameters. One is the",
            "tokens": [
                50364,
                7555,
                11,
                257,
                1154,
                420,
                3287,
                307,
                48936,
                538,
                257,
                2604,
                781,
                295,
                1732,
                2135,
                9834,
                13,
                1485,
                307,
                264,
                50772
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19196428440429353,
            "compression_ratio": 1.6233183856502242,
            "no_speech_prob": 0.02227783203125,
            "words": [
                {
                    "start": 431.13,
                    "end": 431.63,
                    "word": " processes",
                    "probability": 0.33447265625
                },
                {
                    "start": 431.63,
                    "end": 432.13,
                    "word": ",",
                    "probability": 0.380126953125
                },
                {
                    "start": 432.39,
                    "end": 433.31,
                    "word": " a problem",
                    "probability": 0.70947265625
                },
                {
                    "start": 433.31,
                    "end": 433.49,
                    "word": " or",
                    "probability": 0.86328125
                },
                {
                    "start": 433.49,
                    "end": 433.77,
                    "word": " setting",
                    "probability": 0.7919921875
                },
                {
                    "start": 433.77,
                    "end": 433.93,
                    "word": " is",
                    "probability": 0.82470703125
                },
                {
                    "start": 433.93,
                    "end": 434.33,
                    "word": " formulated",
                    "probability": 0.86279296875
                },
                {
                    "start": 434.33,
                    "end": 434.67,
                    "word": " by",
                    "probability": 0.9033203125
                },
                {
                    "start": 434.83,
                    "end": 435.09,
                    "word": " a tuple",
                    "probability": 0.89892578125
                },
                {
                    "start": 435.09,
                    "end": 435.65,
                    "word": " of",
                    "probability": 0.9052734375
                },
                {
                    "start": 435.65,
                    "end": 436.61,
                    "word": " five",
                    "probability": 0.728515625
                },
                {
                    "start": 436.61,
                    "end": 437.97,
                    "word": " main",
                    "probability": 0.845703125
                },
                {
                    "start": 437.97,
                    "end": 438.43,
                    "word": " parameters",
                    "probability": 0.83154296875
                },
                {
                    "start": 438.43,
                    "end": 438.75,
                    "word": ".",
                    "probability": 0.90478515625
                },
                {
                    "start": 438.75,
                    "end": 439.15,
                    "word": " One",
                    "probability": 0.8798828125
                },
                {
                    "start": 439.15,
                    "end": 439.37,
                    "word": " is",
                    "probability": 0.8994140625
                },
                {
                    "start": 439.37,
                    "end": 439.49,
                    "word": " the",
                    "probability": 0.798828125
                }
            ]
        },
        {
            "id": 67,
            "seek": 45014,
            "start": 439.49,
            "end": 446.83,
            "text": " set of possible states, S. Set of possible actions, A. Distribution of reward given state",
            "tokens": [
                50772,
                992,
                295,
                1944,
                4368,
                11,
                318,
                13,
                8928,
                295,
                1944,
                5909,
                11,
                316,
                13,
                9840,
                30783,
                295,
                7782,
                2212,
                1785,
                51148
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19196428440429353,
            "compression_ratio": 1.6233183856502242,
            "no_speech_prob": 0.02227783203125,
            "words": [
                {
                    "start": 439.49,
                    "end": 439.63,
                    "word": " set",
                    "probability": 0.88818359375
                },
                {
                    "start": 439.63,
                    "end": 439.75,
                    "word": " of",
                    "probability": 0.8955078125
                },
                {
                    "start": 439.75,
                    "end": 440.07,
                    "word": " possible",
                    "probability": 0.86279296875
                },
                {
                    "start": 440.07,
                    "end": 440.49,
                    "word": " states",
                    "probability": 0.87939453125
                },
                {
                    "start": 440.49,
                    "end": 440.79,
                    "word": ",",
                    "probability": 0.37060546875
                },
                {
                    "start": 440.79,
                    "end": 440.91,
                    "word": " S",
                    "probability": 0.77099609375
                },
                {
                    "start": 440.91,
                    "end": 441.31,
                    "word": ".",
                    "probability": 0.343505859375
                },
                {
                    "start": 441.31,
                    "end": 441.69,
                    "word": " Set",
                    "probability": 0.8408203125
                },
                {
                    "start": 441.69,
                    "end": 441.79,
                    "word": " of",
                    "probability": 0.884765625
                },
                {
                    "start": 441.79,
                    "end": 442.07,
                    "word": " possible",
                    "probability": 0.86376953125
                },
                {
                    "start": 442.07,
                    "end": 442.55,
                    "word": " actions",
                    "probability": 0.841796875
                },
                {
                    "start": 442.55,
                    "end": 442.81,
                    "word": ",",
                    "probability": 0.88623046875
                },
                {
                    "start": 442.81,
                    "end": 443.03,
                    "word": " A",
                    "probability": 0.93798828125
                },
                {
                    "start": 443.03,
                    "end": 443.55,
                    "word": ".",
                    "probability": 0.9326171875
                },
                {
                    "start": 443.91,
                    "end": 444.43,
                    "word": " Distribution",
                    "probability": 0.92529296875
                },
                {
                    "start": 444.43,
                    "end": 445.31,
                    "word": " of",
                    "probability": 0.9052734375
                },
                {
                    "start": 445.31,
                    "end": 445.95,
                    "word": " reward",
                    "probability": 0.86376953125
                },
                {
                    "start": 445.95,
                    "end": 446.41,
                    "word": " given",
                    "probability": 0.72998046875
                },
                {
                    "start": 446.41,
                    "end": 446.83,
                    "word": " state",
                    "probability": 0.70556640625
                }
            ]
        },
        {
            "id": 68,
            "seek": 45014,
            "start": 446.83,
            "end": 451.09,
            "text": " action pair, or R. This distribution can also be a function. It doesn't necessarily have to",
            "tokens": [
                51148,
                3069,
                6119,
                11,
                420,
                497,
                13,
                639,
                7316,
                393,
                611,
                312,
                257,
                2445,
                13,
                467,
                1177,
                380,
                4725,
                362,
                281,
                51356
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19196428440429353,
            "compression_ratio": 1.6233183856502242,
            "no_speech_prob": 0.02227783203125,
            "words": [
                {
                    "start": 446.83,
                    "end": 447.23,
                    "word": " action",
                    "probability": 0.42138671875
                },
                {
                    "start": 447.23,
                    "end": 447.53,
                    "word": " pair",
                    "probability": 0.818359375
                },
                {
                    "start": 447.53,
                    "end": 447.71,
                    "word": ",",
                    "probability": 0.58984375
                },
                {
                    "start": 447.71,
                    "end": 447.85,
                    "word": " or",
                    "probability": 0.83447265625
                },
                {
                    "start": 447.85,
                    "end": 448.15,
                    "word": " R",
                    "probability": 0.9306640625
                },
                {
                    "start": 448.15,
                    "end": 448.43,
                    "word": ".",
                    "probability": 0.94775390625
                },
                {
                    "start": 448.43,
                    "end": 448.65,
                    "word": " This",
                    "probability": 0.8046875
                },
                {
                    "start": 448.65,
                    "end": 449.11,
                    "word": " distribution",
                    "probability": 0.69775390625
                },
                {
                    "start": 449.11,
                    "end": 449.39,
                    "word": " can",
                    "probability": 0.8994140625
                },
                {
                    "start": 449.39,
                    "end": 449.63,
                    "word": " also",
                    "probability": 0.82470703125
                },
                {
                    "start": 449.63,
                    "end": 449.85,
                    "word": " be",
                    "probability": 0.9013671875
                },
                {
                    "start": 450.01,
                    "end": 450.25,
                    "word": " a function",
                    "probability": 0.873046875
                },
                {
                    "start": 450.25,
                    "end": 450.37,
                    "word": ".",
                    "probability": 0.393798828125
                },
                {
                    "start": 450.37,
                    "end": 450.39,
                    "word": " It",
                    "probability": 0.89892578125
                },
                {
                    "start": 450.39,
                    "end": 450.53,
                    "word": " doesn't",
                    "probability": 0.93701171875
                },
                {
                    "start": 450.53,
                    "end": 450.87,
                    "word": " necessarily",
                    "probability": 0.779296875
                },
                {
                    "start": 450.87,
                    "end": 451.03,
                    "word": " have",
                    "probability": 0.88525390625
                },
                {
                    "start": 451.03,
                    "end": 451.09,
                    "word": " to",
                    "probability": 0.90673828125
                }
            ]
        },
        {
            "id": 69,
            "seek": 45014,
            "start": 451.09,
            "end": 456.27,
            "text": " be probabilistic. It can be just a function that tells you if at a particular state, S,",
            "tokens": [
                51356,
                312,
                31959,
                3142,
                13,
                467,
                393,
                312,
                445,
                257,
                2445,
                300,
                5112,
                291,
                498,
                412,
                257,
                1729,
                1785,
                11,
                318,
                11,
                51592
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19196428440429353,
            "compression_ratio": 1.6233183856502242,
            "no_speech_prob": 0.02227783203125,
            "words": [
                {
                    "start": 451.09,
                    "end": 451.21,
                    "word": " be",
                    "probability": 0.9033203125
                },
                {
                    "start": 451.21,
                    "end": 451.75,
                    "word": " probabilistic",
                    "probability": 0.94482421875
                },
                {
                    "start": 451.75,
                    "end": 451.85,
                    "word": ".",
                    "probability": 0.81005859375
                },
                {
                    "start": 451.85,
                    "end": 451.89,
                    "word": " It",
                    "probability": 0.9013671875
                },
                {
                    "start": 451.89,
                    "end": 452.03,
                    "word": " can",
                    "probability": 0.8984375
                },
                {
                    "start": 452.03,
                    "end": 452.15,
                    "word": " be",
                    "probability": 0.87744140625
                },
                {
                    "start": 452.15,
                    "end": 452.31,
                    "word": " just",
                    "probability": 0.88525390625
                },
                {
                    "start": 452.43,
                    "end": 452.61,
                    "word": " a function",
                    "probability": 0.87109375
                },
                {
                    "start": 452.61,
                    "end": 452.79,
                    "word": " that",
                    "probability": 0.7197265625
                },
                {
                    "start": 452.79,
                    "end": 452.91,
                    "word": " tells",
                    "probability": 0.8046875
                },
                {
                    "start": 452.91,
                    "end": 453.03,
                    "word": " you",
                    "probability": 0.8583984375
                },
                {
                    "start": 453.03,
                    "end": 453.19,
                    "word": " if",
                    "probability": 0.7421875
                },
                {
                    "start": 453.19,
                    "end": 454.39,
                    "word": " at",
                    "probability": 0.303466796875
                },
                {
                    "start": 454.73,
                    "end": 455.13,
                    "word": " a particular",
                    "probability": 0.7978515625
                },
                {
                    "start": 455.13,
                    "end": 455.51,
                    "word": " state",
                    "probability": 0.86474609375
                },
                {
                    "start": 455.51,
                    "end": 455.67,
                    "word": ",",
                    "probability": 0.41650390625
                },
                {
                    "start": 455.67,
                    "end": 455.89,
                    "word": " S",
                    "probability": 0.62744140625
                },
                {
                    "start": 455.89,
                    "end": 456.27,
                    "word": ",",
                    "probability": 0.7841796875
                }
            ]
        },
        {
            "id": 70,
            "seek": 47368,
            "start": 456.27,
            "end": 463.41,
            "text": " I, an action AI is taken, then the reward is going to be, let's say, RI. Then there's transition",
            "tokens": [
                50364,
                286,
                11,
                364,
                3069,
                7318,
                307,
                2726,
                11,
                550,
                264,
                7782,
                307,
                516,
                281,
                312,
                11,
                718,
                311,
                584,
                11,
                30474,
                13,
                1396,
                456,
                311,
                6034,
                50720
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2365756979290868,
            "compression_ratio": 1.6136363636363635,
            "no_speech_prob": 0.0094146728515625,
            "words": [
                {
                    "start": 456.27,
                    "end": 456.67,
                    "word": " I",
                    "probability": 0.04071044921875
                },
                {
                    "start": 456.67,
                    "end": 457.13,
                    "word": ",",
                    "probability": 0.477783203125
                },
                {
                    "start": 457.13,
                    "end": 457.31,
                    "word": " an",
                    "probability": 0.681640625
                },
                {
                    "start": 457.31,
                    "end": 457.75,
                    "word": " action",
                    "probability": 0.76318359375
                },
                {
                    "start": 457.75,
                    "end": 458.33,
                    "word": " AI",
                    "probability": 0.46142578125
                },
                {
                    "start": 458.33,
                    "end": 458.67,
                    "word": " is",
                    "probability": 0.779296875
                },
                {
                    "start": 458.67,
                    "end": 458.95,
                    "word": " taken",
                    "probability": 0.72216796875
                },
                {
                    "start": 458.95,
                    "end": 459.15,
                    "word": ",",
                    "probability": 0.69287109375
                },
                {
                    "start": 459.15,
                    "end": 459.23,
                    "word": " then",
                    "probability": 0.68212890625
                },
                {
                    "start": 459.23,
                    "end": 459.35,
                    "word": " the",
                    "probability": 0.80029296875
                },
                {
                    "start": 459.35,
                    "end": 459.65,
                    "word": " reward",
                    "probability": 0.8486328125
                },
                {
                    "start": 459.65,
                    "end": 459.79,
                    "word": " is",
                    "probability": 0.89404296875
                },
                {
                    "start": 459.79,
                    "end": 459.93,
                    "word": " going",
                    "probability": 0.84375
                },
                {
                    "start": 459.93,
                    "end": 460.07,
                    "word": " to",
                    "probability": 0.9033203125
                },
                {
                    "start": 460.07,
                    "end": 460.37,
                    "word": " be",
                    "probability": 0.9052734375
                },
                {
                    "start": 460.37,
                    "end": 460.71,
                    "word": ",",
                    "probability": 0.54736328125
                },
                {
                    "start": 460.71,
                    "end": 461.03,
                    "word": " let's",
                    "probability": 0.915771484375
                },
                {
                    "start": 461.03,
                    "end": 461.17,
                    "word": " say",
                    "probability": 0.8935546875
                },
                {
                    "start": 461.17,
                    "end": 461.25,
                    "word": ",",
                    "probability": 0.69482421875
                },
                {
                    "start": 461.25,
                    "end": 461.55,
                    "word": " RI",
                    "probability": 0.3828125
                },
                {
                    "start": 461.55,
                    "end": 462.15,
                    "word": ".",
                    "probability": 0.912109375
                },
                {
                    "start": 462.33,
                    "end": 462.67,
                    "word": " Then",
                    "probability": 0.7900390625
                },
                {
                    "start": 462.67,
                    "end": 462.95,
                    "word": " there's",
                    "probability": 0.65283203125
                },
                {
                    "start": 462.95,
                    "end": 463.41,
                    "word": " transition",
                    "probability": 0.83544921875
                }
            ]
        },
        {
            "id": 71,
            "seek": 47368,
            "start": 463.41,
            "end": 474.91,
            "text": " probability, or transition dynamics, P, which represents the dynamics of the environment.",
            "tokens": [
                50720,
                8482,
                11,
                420,
                6034,
                15679,
                11,
                430,
                11,
                597,
                8855,
                264,
                15679,
                295,
                264,
                2823,
                13,
                51268
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2365756979290868,
            "compression_ratio": 1.6136363636363635,
            "no_speech_prob": 0.0094146728515625,
            "words": [
                {
                    "start": 463.41,
                    "end": 463.99,
                    "word": " probability",
                    "probability": 0.8984375
                },
                {
                    "start": 463.99,
                    "end": 464.63,
                    "word": ",",
                    "probability": 0.47509765625
                },
                {
                    "start": 464.63,
                    "end": 464.77,
                    "word": " or",
                    "probability": 0.8798828125
                },
                {
                    "start": 464.77,
                    "end": 465.47,
                    "word": " transition",
                    "probability": 0.90771484375
                },
                {
                    "start": 465.47,
                    "end": 466.17,
                    "word": " dynamics",
                    "probability": 0.75341796875
                },
                {
                    "start": 466.17,
                    "end": 466.49,
                    "word": ",",
                    "probability": 0.7939453125
                },
                {
                    "start": 466.49,
                    "end": 466.67,
                    "word": " P",
                    "probability": 0.8828125
                },
                {
                    "start": 466.67,
                    "end": 467.27,
                    "word": ",",
                    "probability": 0.96533203125
                },
                {
                    "start": 467.27,
                    "end": 467.71,
                    "word": " which",
                    "probability": 0.92138671875
                },
                {
                    "start": 467.71,
                    "end": 469.35,
                    "word": " represents",
                    "probability": 0.7880859375
                },
                {
                    "start": 469.35,
                    "end": 470.85,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 470.85,
                    "end": 473.59,
                    "word": " dynamics",
                    "probability": 0.673828125
                },
                {
                    "start": 473.59,
                    "end": 473.95,
                    "word": " of",
                    "probability": 0.90869140625
                },
                {
                    "start": 473.95,
                    "end": 474.05,
                    "word": " the",
                    "probability": 0.7958984375
                },
                {
                    "start": 474.05,
                    "end": 474.35,
                    "word": " environment",
                    "probability": 0.75244140625
                },
                {
                    "start": 474.35,
                    "end": 474.91,
                    "word": ".",
                    "probability": 0.50439453125
                }
            ]
        },
        {
            "id": 72,
            "seek": 47368,
            "start": 474.91,
            "end": 479.81,
            "text": " Which state are we going to end up in if we are in a particular state and we perform a particular",
            "tokens": [
                51304,
                3013,
                1785,
                366,
                321,
                516,
                281,
                917,
                493,
                294,
                498,
                321,
                366,
                294,
                257,
                1729,
                1785,
                293,
                321,
                2042,
                257,
                1729,
                51544
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2365756979290868,
            "compression_ratio": 1.6136363636363635,
            "no_speech_prob": 0.0094146728515625,
            "words": [
                {
                    "start": 474.91,
                    "end": 475.45,
                    "word": " Which",
                    "probability": 0.810546875
                },
                {
                    "start": 475.45,
                    "end": 475.83,
                    "word": " state",
                    "probability": 0.83447265625
                },
                {
                    "start": 475.83,
                    "end": 475.95,
                    "word": " are",
                    "probability": 0.90283203125
                },
                {
                    "start": 475.95,
                    "end": 476.03,
                    "word": " we",
                    "probability": 0.9013671875
                },
                {
                    "start": 476.03,
                    "end": 476.15,
                    "word": " going",
                    "probability": 0.85009765625
                },
                {
                    "start": 476.15,
                    "end": 476.29,
                    "word": " to",
                    "probability": 0.90869140625
                },
                {
                    "start": 476.29,
                    "end": 476.49,
                    "word": " end",
                    "probability": 0.81494140625
                },
                {
                    "start": 476.49,
                    "end": 476.67,
                    "word": " up",
                    "probability": 0.90625
                },
                {
                    "start": 476.67,
                    "end": 476.77,
                    "word": " in",
                    "probability": 0.87451171875
                },
                {
                    "start": 476.77,
                    "end": 476.91,
                    "word": " if",
                    "probability": 0.744140625
                },
                {
                    "start": 476.91,
                    "end": 477.01,
                    "word": " we",
                    "probability": 0.83154296875
                },
                {
                    "start": 477.01,
                    "end": 477.11,
                    "word": " are",
                    "probability": 0.86767578125
                },
                {
                    "start": 477.11,
                    "end": 477.19,
                    "word": " in",
                    "probability": 0.90283203125
                },
                {
                    "start": 477.29,
                    "end": 477.57,
                    "word": " a particular",
                    "probability": 0.80517578125
                },
                {
                    "start": 477.57,
                    "end": 478.05,
                    "word": " state",
                    "probability": 0.86279296875
                },
                {
                    "start": 478.05,
                    "end": 478.39,
                    "word": " and",
                    "probability": 0.791015625
                },
                {
                    "start": 478.39,
                    "end": 478.65,
                    "word": " we",
                    "probability": 0.873046875
                },
                {
                    "start": 478.65,
                    "end": 479.41,
                    "word": " perform",
                    "probability": 0.76953125
                },
                {
                    "start": 479.53,
                    "end": 479.81,
                    "word": " a particular",
                    "probability": 0.80810546875
                }
            ]
        },
        {
            "id": 73,
            "seek": 49708,
            "start": 479.81,
            "end": 486.05,
            "text": " action. And there's finally a discount factor which defines how myopic the agent is, how much",
            "tokens": [
                50368,
                3069,
                13,
                400,
                456,
                311,
                2721,
                257,
                11635,
                5952,
                597,
                23122,
                577,
                452,
                40216,
                264,
                9461,
                307,
                11,
                577,
                709,
                50676
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26980634306518125,
            "compression_ratio": 1.4816753926701571,
            "no_speech_prob": 0.03021240234375,
            "words": [
                {
                    "start": 479.81,
                    "end": 480.21,
                    "word": " action",
                    "probability": 0.2034912109375
                },
                {
                    "start": 480.21,
                    "end": 480.47,
                    "word": ".",
                    "probability": 0.482666015625
                },
                {
                    "start": 480.47,
                    "end": 480.53,
                    "word": " And",
                    "probability": 0.79052734375
                },
                {
                    "start": 480.53,
                    "end": 480.73,
                    "word": " there's",
                    "probability": 0.86865234375
                },
                {
                    "start": 480.73,
                    "end": 480.97,
                    "word": " finally",
                    "probability": 0.794921875
                },
                {
                    "start": 481.15,
                    "end": 481.41,
                    "word": " a discount",
                    "probability": 0.8798828125
                },
                {
                    "start": 481.41,
                    "end": 481.81,
                    "word": " factor",
                    "probability": 0.72216796875
                },
                {
                    "start": 481.81,
                    "end": 482.67,
                    "word": " which",
                    "probability": 0.6748046875
                },
                {
                    "start": 482.67,
                    "end": 483.55,
                    "word": " defines",
                    "probability": 0.72998046875
                },
                {
                    "start": 483.55,
                    "end": 483.85,
                    "word": " how",
                    "probability": 0.8984375
                },
                {
                    "start": 483.85,
                    "end": 484.33,
                    "word": " myopic",
                    "probability": 0.959716796875
                },
                {
                    "start": 484.33,
                    "end": 484.59,
                    "word": " the",
                    "probability": 0.80810546875
                },
                {
                    "start": 484.59,
                    "end": 484.87,
                    "word": " agent",
                    "probability": 0.75341796875
                },
                {
                    "start": 484.87,
                    "end": 485.17,
                    "word": " is",
                    "probability": 0.90185546875
                },
                {
                    "start": 485.17,
                    "end": 485.45,
                    "word": ",",
                    "probability": 0.60791015625
                },
                {
                    "start": 485.45,
                    "end": 485.57,
                    "word": " how",
                    "probability": 0.8916015625
                },
                {
                    "start": 485.57,
                    "end": 486.05,
                    "word": " much",
                    "probability": 0.67578125
                }
            ]
        },
        {
            "id": 74,
            "seek": 49708,
            "start": 486.05,
            "end": 496.09,
            "text": " values, rewards that may occur further in the future, further down the line. Now, in general,",
            "tokens": [
                50676,
                4190,
                11,
                17203,
                300,
                815,
                5160,
                3052,
                294,
                264,
                2027,
                11,
                3052,
                760,
                264,
                1622,
                13,
                823,
                11,
                294,
                2674,
                11,
                51172
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26980634306518125,
            "compression_ratio": 1.4816753926701571,
            "no_speech_prob": 0.03021240234375,
            "words": [
                {
                    "start": 486.05,
                    "end": 486.77,
                    "word": " values",
                    "probability": 0.082275390625
                },
                {
                    "start": 486.77,
                    "end": 487.47,
                    "word": ",",
                    "probability": 0.392333984375
                },
                {
                    "start": 487.47,
                    "end": 488.89,
                    "word": " rewards",
                    "probability": 0.68017578125
                },
                {
                    "start": 488.89,
                    "end": 489.09,
                    "word": " that",
                    "probability": 0.71630859375
                },
                {
                    "start": 489.09,
                    "end": 489.21,
                    "word": " may",
                    "probability": 0.8701171875
                },
                {
                    "start": 489.21,
                    "end": 489.45,
                    "word": " occur",
                    "probability": 0.8408203125
                },
                {
                    "start": 489.45,
                    "end": 489.95,
                    "word": " further",
                    "probability": 0.8212890625
                },
                {
                    "start": 489.95,
                    "end": 490.39,
                    "word": " in",
                    "probability": 0.87353515625
                },
                {
                    "start": 490.39,
                    "end": 490.53,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 490.53,
                    "end": 490.87,
                    "word": " future",
                    "probability": 0.9384765625
                },
                {
                    "start": 490.87,
                    "end": 491.03,
                    "word": ",",
                    "probability": 0.8466796875
                },
                {
                    "start": 491.03,
                    "end": 491.21,
                    "word": " further",
                    "probability": 0.853515625
                },
                {
                    "start": 491.21,
                    "end": 491.43,
                    "word": " down",
                    "probability": 0.82568359375
                },
                {
                    "start": 491.43,
                    "end": 491.55,
                    "word": " the",
                    "probability": 0.806640625
                },
                {
                    "start": 491.55,
                    "end": 491.79,
                    "word": " line",
                    "probability": 0.8740234375
                },
                {
                    "start": 491.79,
                    "end": 492.31,
                    "word": ".",
                    "probability": 0.9912109375
                },
                {
                    "start": 493.19,
                    "end": 493.71,
                    "word": " Now",
                    "probability": 0.8720703125
                },
                {
                    "start": 493.71,
                    "end": 493.89,
                    "word": ",",
                    "probability": 0.66162109375
                },
                {
                    "start": 493.89,
                    "end": 495.07,
                    "word": " in",
                    "probability": 0.9033203125
                },
                {
                    "start": 495.07,
                    "end": 495.69,
                    "word": " general",
                    "probability": 0.806640625
                },
                {
                    "start": 495.69,
                    "end": 496.09,
                    "word": ",",
                    "probability": 0.85302734375
                }
            ]
        },
        {
            "id": 75,
            "seek": 49708,
            "start": 496.09,
            "end": 503.21,
            "text": " the RL setting is based on the following process. At time step T equals zero, in the beginning,",
            "tokens": [
                51172,
                264,
                497,
                43,
                3287,
                307,
                2361,
                322,
                264,
                3480,
                1399,
                13,
                1711,
                565,
                1823,
                314,
                6915,
                4018,
                11,
                294,
                264,
                2863,
                11,
                51496
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26980634306518125,
            "compression_ratio": 1.4816753926701571,
            "no_speech_prob": 0.03021240234375,
            "words": [
                {
                    "start": 496.09,
                    "end": 496.29,
                    "word": " the",
                    "probability": 0.8173828125
                },
                {
                    "start": 496.29,
                    "end": 496.67,
                    "word": " RL",
                    "probability": 0.91015625
                },
                {
                    "start": 496.67,
                    "end": 497.05,
                    "word": " setting",
                    "probability": 0.80859375
                },
                {
                    "start": 497.05,
                    "end": 498.45,
                    "word": " is",
                    "probability": 0.8916015625
                },
                {
                    "start": 498.45,
                    "end": 498.99,
                    "word": " based",
                    "probability": 0.80810546875
                },
                {
                    "start": 498.99,
                    "end": 499.25,
                    "word": " on",
                    "probability": 0.90673828125
                },
                {
                    "start": 499.25,
                    "end": 499.43,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 499.43,
                    "end": 499.67,
                    "word": " following",
                    "probability": 0.84619140625
                },
                {
                    "start": 499.67,
                    "end": 500.17,
                    "word": " process",
                    "probability": 0.8603515625
                },
                {
                    "start": 500.17,
                    "end": 500.51,
                    "word": ".",
                    "probability": 0.76953125
                },
                {
                    "start": 500.51,
                    "end": 500.65,
                    "word": " At",
                    "probability": 0.9033203125
                },
                {
                    "start": 500.65,
                    "end": 500.91,
                    "word": " time",
                    "probability": 0.72021484375
                },
                {
                    "start": 500.91,
                    "end": 501.15,
                    "word": " step",
                    "probability": 0.6923828125
                },
                {
                    "start": 501.15,
                    "end": 501.37,
                    "word": " T",
                    "probability": 0.252197265625
                },
                {
                    "start": 501.37,
                    "end": 501.57,
                    "word": " equals",
                    "probability": 0.382080078125
                },
                {
                    "start": 501.57,
                    "end": 501.81,
                    "word": " zero",
                    "probability": 0.61328125
                },
                {
                    "start": 501.81,
                    "end": 501.85,
                    "word": ",",
                    "probability": 0.419921875
                },
                {
                    "start": 501.85,
                    "end": 501.99,
                    "word": " in",
                    "probability": 0.8935546875
                },
                {
                    "start": 501.99,
                    "end": 502.15,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 502.15,
                    "end": 502.45,
                    "word": " beginning",
                    "probability": 0.80322265625
                },
                {
                    "start": 502.45,
                    "end": 503.21,
                    "word": ",",
                    "probability": 0.79541015625
                }
            ]
        },
        {
            "id": 76,
            "seek": 52330,
            "start": 503.21,
            "end": 510.69,
            "text": " the environment samples initial state S0. Then, for T equals zero, this is a loop, until done,",
            "tokens": [
                50364,
                264,
                2823,
                10938,
                5883,
                1785,
                318,
                15,
                13,
                1396,
                11,
                337,
                314,
                6915,
                4018,
                11,
                341,
                307,
                257,
                6367,
                11,
                1826,
                1096,
                11,
                50728
            ],
            "temperature": 0.0,
            "avg_logprob": -0.271661924706264,
            "compression_ratio": 1.6363636363636365,
            "no_speech_prob": 0.003765106201171875,
            "words": [
                {
                    "start": 503.21,
                    "end": 503.35,
                    "word": " the",
                    "probability": 0.0899658203125
                },
                {
                    "start": 503.35,
                    "end": 503.67,
                    "word": " environment",
                    "probability": 0.66455078125
                },
                {
                    "start": 503.67,
                    "end": 504.21,
                    "word": " samples",
                    "probability": 0.728515625
                },
                {
                    "start": 504.21,
                    "end": 504.71,
                    "word": " initial",
                    "probability": 0.8564453125
                },
                {
                    "start": 504.71,
                    "end": 505.07,
                    "word": " state",
                    "probability": 0.8515625
                },
                {
                    "start": 505.07,
                    "end": 505.81,
                    "word": " S0",
                    "probability": 0.549560546875
                },
                {
                    "start": 505.81,
                    "end": 506.41,
                    "word": ".",
                    "probability": 0.33544921875
                },
                {
                    "start": 506.55,
                    "end": 506.79,
                    "word": " Then",
                    "probability": 0.76611328125
                },
                {
                    "start": 506.79,
                    "end": 507.49,
                    "word": ",",
                    "probability": 0.28466796875
                },
                {
                    "start": 507.49,
                    "end": 507.71,
                    "word": " for",
                    "probability": 0.8974609375
                },
                {
                    "start": 507.71,
                    "end": 508.07,
                    "word": " T",
                    "probability": 0.35009765625
                },
                {
                    "start": 508.07,
                    "end": 508.33,
                    "word": " equals",
                    "probability": 0.3759765625
                },
                {
                    "start": 508.33,
                    "end": 508.59,
                    "word": " zero",
                    "probability": 0.406494140625
                },
                {
                    "start": 508.59,
                    "end": 508.89,
                    "word": ",",
                    "probability": 0.80810546875
                },
                {
                    "start": 508.89,
                    "end": 509.13,
                    "word": " this",
                    "probability": 0.876953125
                },
                {
                    "start": 509.13,
                    "end": 509.21,
                    "word": " is",
                    "probability": 0.904296875
                },
                {
                    "start": 509.31,
                    "end": 509.55,
                    "word": " a loop",
                    "probability": 0.86767578125
                },
                {
                    "start": 509.55,
                    "end": 509.77,
                    "word": ",",
                    "probability": 0.483642578125
                },
                {
                    "start": 509.77,
                    "end": 510.07,
                    "word": " until",
                    "probability": 0.8037109375
                },
                {
                    "start": 510.07,
                    "end": 510.45,
                    "word": " done",
                    "probability": 0.78076171875
                },
                {
                    "start": 510.45,
                    "end": 510.69,
                    "word": ",",
                    "probability": 0.77880859375
                }
            ]
        },
        {
            "id": 77,
            "seek": 52330,
            "start": 510.69,
            "end": 514.91,
            "text": " the agent selects action AT based on some criteria. It can be completely random in the beginning and",
            "tokens": [
                50728,
                264,
                9461,
                3048,
                82,
                3069,
                8872,
                2361,
                322,
                512,
                11101,
                13,
                467,
                393,
                312,
                2584,
                4974,
                294,
                264,
                2863,
                293,
                50944
            ],
            "temperature": 0.0,
            "avg_logprob": -0.271661924706264,
            "compression_ratio": 1.6363636363636365,
            "no_speech_prob": 0.003765106201171875,
            "words": [
                {
                    "start": 510.69,
                    "end": 510.81,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 510.81,
                    "end": 511.05,
                    "word": " agent",
                    "probability": 0.80712890625
                },
                {
                    "start": 511.05,
                    "end": 511.63,
                    "word": " selects",
                    "probability": 0.882568359375
                },
                {
                    "start": 511.63,
                    "end": 512.03,
                    "word": " action",
                    "probability": 0.69384765625
                },
                {
                    "start": 512.03,
                    "end": 512.29,
                    "word": " AT",
                    "probability": 0.5556640625
                },
                {
                    "start": 512.29,
                    "end": 512.59,
                    "word": " based",
                    "probability": 0.640625
                },
                {
                    "start": 512.59,
                    "end": 512.71,
                    "word": " on",
                    "probability": 0.87890625
                },
                {
                    "start": 512.71,
                    "end": 512.89,
                    "word": " some",
                    "probability": 0.81396484375
                },
                {
                    "start": 512.89,
                    "end": 513.27,
                    "word": " criteria",
                    "probability": 0.787109375
                },
                {
                    "start": 513.27,
                    "end": 513.43,
                    "word": ".",
                    "probability": 0.6083984375
                },
                {
                    "start": 513.43,
                    "end": 513.43,
                    "word": " It",
                    "probability": 0.88818359375
                },
                {
                    "start": 513.43,
                    "end": 513.55,
                    "word": " can",
                    "probability": 0.900390625
                },
                {
                    "start": 513.55,
                    "end": 513.67,
                    "word": " be",
                    "probability": 0.9013671875
                },
                {
                    "start": 513.67,
                    "end": 513.93,
                    "word": " completely",
                    "probability": 0.6455078125
                },
                {
                    "start": 513.93,
                    "end": 514.25,
                    "word": " random",
                    "probability": 0.740234375
                },
                {
                    "start": 514.25,
                    "end": 514.41,
                    "word": " in",
                    "probability": 0.880859375
                },
                {
                    "start": 514.41,
                    "end": 514.51,
                    "word": " the",
                    "probability": 0.81640625
                },
                {
                    "start": 514.51,
                    "end": 514.71,
                    "word": " beginning",
                    "probability": 0.79931640625
                },
                {
                    "start": 514.71,
                    "end": 514.91,
                    "word": " and",
                    "probability": 0.6025390625
                }
            ]
        },
        {
            "id": 78,
            "seek": 52330,
            "start": 514.91,
            "end": 521.73,
            "text": " then slowly becomes more targeted and more policy-driven. It performs some action,",
            "tokens": [
                50944,
                550,
                5692,
                3643,
                544,
                15045,
                293,
                544,
                3897,
                12,
                25456,
                13,
                467,
                26213,
                512,
                3069,
                11,
                51284
            ],
            "temperature": 0.0,
            "avg_logprob": -0.271661924706264,
            "compression_ratio": 1.6363636363636365,
            "no_speech_prob": 0.003765106201171875,
            "words": [
                {
                    "start": 514.91,
                    "end": 515.15,
                    "word": " then",
                    "probability": 0.73583984375
                },
                {
                    "start": 515.15,
                    "end": 515.79,
                    "word": " slowly",
                    "probability": 0.80908203125
                },
                {
                    "start": 515.79,
                    "end": 516.13,
                    "word": " becomes",
                    "probability": 0.80810546875
                },
                {
                    "start": 516.13,
                    "end": 516.49,
                    "word": " more",
                    "probability": 0.8818359375
                },
                {
                    "start": 516.49,
                    "end": 517.17,
                    "word": " targeted",
                    "probability": 0.798828125
                },
                {
                    "start": 517.17,
                    "end": 517.51,
                    "word": " and",
                    "probability": 0.876953125
                },
                {
                    "start": 517.51,
                    "end": 517.83,
                    "word": " more",
                    "probability": 0.87060546875
                },
                {
                    "start": 517.83,
                    "end": 519.17,
                    "word": " policy",
                    "probability": 0.81396484375
                },
                {
                    "start": 519.17,
                    "end": 519.45,
                    "word": "-driven",
                    "probability": 0.698974609375
                },
                {
                    "start": 519.45,
                    "end": 519.91,
                    "word": ".",
                    "probability": 0.9716796875
                },
                {
                    "start": 519.91,
                    "end": 520.51,
                    "word": " It",
                    "probability": 0.876953125
                },
                {
                    "start": 520.51,
                    "end": 520.73,
                    "word": " performs",
                    "probability": 0.67138671875
                },
                {
                    "start": 520.73,
                    "end": 520.95,
                    "word": " some",
                    "probability": 0.81982421875
                },
                {
                    "start": 520.95,
                    "end": 521.29,
                    "word": " action",
                    "probability": 0.8017578125
                },
                {
                    "start": 521.29,
                    "end": 521.73,
                    "word": ",",
                    "probability": 0.50927734375
                }
            ]
        },
        {
            "id": 79,
            "seek": 52330,
            "start": 521.73,
            "end": 529.43,
            "text": " the environment produces a reward signal, and the next state, and the agent receives those signals.",
            "tokens": [
                51284,
                264,
                2823,
                14725,
                257,
                7782,
                6358,
                11,
                293,
                264,
                958,
                1785,
                11,
                293,
                264,
                9461,
                20717,
                729,
                12354,
                13,
                51672
            ],
            "temperature": 0.0,
            "avg_logprob": -0.271661924706264,
            "compression_ratio": 1.6363636363636365,
            "no_speech_prob": 0.003765106201171875,
            "words": [
                {
                    "start": 521.73,
                    "end": 521.91,
                    "word": " the",
                    "probability": 0.81494140625
                },
                {
                    "start": 521.91,
                    "end": 522.41,
                    "word": " environment",
                    "probability": 0.75244140625
                },
                {
                    "start": 522.41,
                    "end": 523.27,
                    "word": " produces",
                    "probability": 0.83203125
                },
                {
                    "start": 524.01,
                    "end": 524.35,
                    "word": " a reward",
                    "probability": 0.90087890625
                },
                {
                    "start": 524.35,
                    "end": 524.79,
                    "word": " signal",
                    "probability": 0.78125
                },
                {
                    "start": 524.79,
                    "end": 525.03,
                    "word": ",",
                    "probability": 0.368408203125
                },
                {
                    "start": 525.03,
                    "end": 525.39,
                    "word": " and",
                    "probability": 0.89111328125
                },
                {
                    "start": 525.39,
                    "end": 525.59,
                    "word": " the",
                    "probability": 0.80908203125
                },
                {
                    "start": 525.59,
                    "end": 525.83,
                    "word": " next",
                    "probability": 0.88330078125
                },
                {
                    "start": 525.83,
                    "end": 526.21,
                    "word": " state",
                    "probability": 0.65869140625
                },
                {
                    "start": 526.21,
                    "end": 526.89,
                    "word": ",",
                    "probability": 0.7236328125
                },
                {
                    "start": 526.89,
                    "end": 527.09,
                    "word": " and",
                    "probability": 0.88232421875
                },
                {
                    "start": 527.09,
                    "end": 527.21,
                    "word": " the",
                    "probability": 0.81494140625
                },
                {
                    "start": 527.21,
                    "end": 527.47,
                    "word": " agent",
                    "probability": 0.81298828125
                },
                {
                    "start": 527.47,
                    "end": 528.17,
                    "word": " receives",
                    "probability": 0.740234375
                },
                {
                    "start": 528.17,
                    "end": 528.47,
                    "word": " those",
                    "probability": 0.83447265625
                },
                {
                    "start": 528.47,
                    "end": 528.93,
                    "word": " signals",
                    "probability": 0.91455078125
                },
                {
                    "start": 528.93,
                    "end": 529.43,
                    "word": ".",
                    "probability": 0.73876953125
                }
            ]
        },
        {
            "id": 80,
            "seek": 55164,
            "start": 529.43,
            "end": 535.37,
            "text": " It may be partial, it may be incomplete, or it may be noisy, but the agent receives some signal",
            "tokens": [
                50364,
                467,
                815,
                312,
                14641,
                11,
                309,
                815,
                312,
                31709,
                11,
                420,
                309,
                815,
                312,
                24518,
                11,
                457,
                264,
                9461,
                20717,
                512,
                6358,
                50656
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2221235836094076,
            "compression_ratio": 1.6944444444444444,
            "no_speech_prob": 0.00039196014404296875,
            "words": [
                {
                    "start": 529.43,
                    "end": 529.59,
                    "word": " It",
                    "probability": 0.5029296875
                },
                {
                    "start": 529.59,
                    "end": 529.91,
                    "word": " may",
                    "probability": 0.87890625
                },
                {
                    "start": 529.91,
                    "end": 530.73,
                    "word": " be",
                    "probability": 0.9033203125
                },
                {
                    "start": 530.73,
                    "end": 532.13,
                    "word": " partial",
                    "probability": 0.8818359375
                },
                {
                    "start": 532.13,
                    "end": 532.45,
                    "word": ",",
                    "probability": 0.84814453125
                },
                {
                    "start": 532.45,
                    "end": 532.45,
                    "word": " it",
                    "probability": 0.8251953125
                },
                {
                    "start": 532.45,
                    "end": 532.55,
                    "word": " may",
                    "probability": 0.89501953125
                },
                {
                    "start": 532.55,
                    "end": 532.65,
                    "word": " be",
                    "probability": 0.90478515625
                },
                {
                    "start": 532.65,
                    "end": 533.01,
                    "word": " incomplete",
                    "probability": 0.7783203125
                },
                {
                    "start": 533.01,
                    "end": 533.17,
                    "word": ",",
                    "probability": 0.54052734375
                },
                {
                    "start": 533.17,
                    "end": 533.23,
                    "word": " or",
                    "probability": 0.86474609375
                },
                {
                    "start": 533.23,
                    "end": 533.35,
                    "word": " it",
                    "probability": 0.89599609375
                },
                {
                    "start": 533.35,
                    "end": 533.51,
                    "word": " may",
                    "probability": 0.896484375
                },
                {
                    "start": 533.51,
                    "end": 533.63,
                    "word": " be",
                    "probability": 0.91015625
                },
                {
                    "start": 533.63,
                    "end": 533.89,
                    "word": " noisy",
                    "probability": 0.77734375
                },
                {
                    "start": 533.89,
                    "end": 534.03,
                    "word": ",",
                    "probability": 0.642578125
                },
                {
                    "start": 534.03,
                    "end": 534.07,
                    "word": " but",
                    "probability": 0.89794921875
                },
                {
                    "start": 534.07,
                    "end": 534.17,
                    "word": " the",
                    "probability": 0.7841796875
                },
                {
                    "start": 534.17,
                    "end": 534.31,
                    "word": " agent",
                    "probability": 0.7841796875
                },
                {
                    "start": 534.31,
                    "end": 534.69,
                    "word": " receives",
                    "probability": 0.54296875
                },
                {
                    "start": 534.69,
                    "end": 534.97,
                    "word": " some",
                    "probability": 0.79931640625
                },
                {
                    "start": 534.97,
                    "end": 535.37,
                    "word": " signal",
                    "probability": 0.78271484375
                }
            ]
        },
        {
            "id": 81,
            "seek": 55164,
            "start": 536.39,
            "end": 544.07,
            "text": " resulting from the transition to the new state ST plus one, and a reward that has emerged from",
            "tokens": [
                50720,
                16505,
                490,
                264,
                6034,
                281,
                264,
                777,
                1785,
                4904,
                1804,
                472,
                11,
                293,
                257,
                7782,
                300,
                575,
                20178,
                490,
                51092
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2221235836094076,
            "compression_ratio": 1.6944444444444444,
            "no_speech_prob": 0.00039196014404296875,
            "words": [
                {
                    "start": 536.39,
                    "end": 537.03,
                    "word": " resulting",
                    "probability": 0.73388671875
                },
                {
                    "start": 537.03,
                    "end": 537.61,
                    "word": " from",
                    "probability": 0.82373046875
                },
                {
                    "start": 537.61,
                    "end": 538.95,
                    "word": " the",
                    "probability": 0.8076171875
                },
                {
                    "start": 538.95,
                    "end": 539.37,
                    "word": " transition",
                    "probability": 0.923828125
                },
                {
                    "start": 539.37,
                    "end": 539.59,
                    "word": " to",
                    "probability": 0.89990234375
                },
                {
                    "start": 539.59,
                    "end": 539.71,
                    "word": " the",
                    "probability": 0.814453125
                },
                {
                    "start": 539.71,
                    "end": 539.87,
                    "word": " new",
                    "probability": 0.82861328125
                },
                {
                    "start": 539.87,
                    "end": 540.15,
                    "word": " state",
                    "probability": 0.88818359375
                },
                {
                    "start": 540.15,
                    "end": 540.53,
                    "word": " ST",
                    "probability": 0.1240234375
                },
                {
                    "start": 540.53,
                    "end": 541.15,
                    "word": " plus",
                    "probability": 0.430908203125
                },
                {
                    "start": 541.15,
                    "end": 541.47,
                    "word": " one",
                    "probability": 0.56884765625
                },
                {
                    "start": 541.47,
                    "end": 541.61,
                    "word": ",",
                    "probability": 0.366943359375
                },
                {
                    "start": 541.61,
                    "end": 541.97,
                    "word": " and",
                    "probability": 0.8994140625
                },
                {
                    "start": 542.39,
                    "end": 542.89,
                    "word": " a reward",
                    "probability": 0.865234375
                },
                {
                    "start": 542.89,
                    "end": 543.35,
                    "word": " that",
                    "probability": 0.8701171875
                },
                {
                    "start": 543.35,
                    "end": 543.55,
                    "word": " has",
                    "probability": 0.8935546875
                },
                {
                    "start": 543.55,
                    "end": 543.89,
                    "word": " emerged",
                    "probability": 0.6083984375
                },
                {
                    "start": 543.89,
                    "end": 544.07,
                    "word": " from",
                    "probability": 0.8095703125
                }
            ]
        },
        {
            "id": 82,
            "seek": 55164,
            "start": 544.07,
            "end": 551.27,
            "text": " that transition and has resulted from that transition. Now, we define a new entity here",
            "tokens": [
                51092,
                300,
                6034,
                293,
                575,
                18753,
                490,
                300,
                6034,
                13,
                823,
                11,
                321,
                6964,
                257,
                777,
                13977,
                510,
                51464
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2221235836094076,
            "compression_ratio": 1.6944444444444444,
            "no_speech_prob": 0.00039196014404296875,
            "words": [
                {
                    "start": 544.07,
                    "end": 544.21,
                    "word": " that",
                    "probability": 0.8486328125
                },
                {
                    "start": 544.21,
                    "end": 544.55,
                    "word": " transition",
                    "probability": 0.8994140625
                },
                {
                    "start": 544.55,
                    "end": 544.69,
                    "word": " and",
                    "probability": 0.414794921875
                },
                {
                    "start": 544.69,
                    "end": 544.81,
                    "word": " has",
                    "probability": 0.712890625
                },
                {
                    "start": 544.81,
                    "end": 545.09,
                    "word": " resulted",
                    "probability": 0.79296875
                },
                {
                    "start": 545.09,
                    "end": 545.29,
                    "word": " from",
                    "probability": 0.81884765625
                },
                {
                    "start": 545.29,
                    "end": 545.43,
                    "word": " that",
                    "probability": 0.86767578125
                },
                {
                    "start": 545.43,
                    "end": 545.83,
                    "word": " transition",
                    "probability": 0.91845703125
                },
                {
                    "start": 545.83,
                    "end": 546.47,
                    "word": ".",
                    "probability": 0.9951171875
                },
                {
                    "start": 546.69,
                    "end": 547.13,
                    "word": " Now",
                    "probability": 0.884765625
                },
                {
                    "start": 547.13,
                    "end": 547.19,
                    "word": ",",
                    "probability": 0.3232421875
                },
                {
                    "start": 547.19,
                    "end": 547.27,
                    "word": " we",
                    "probability": 0.8896484375
                },
                {
                    "start": 547.27,
                    "end": 547.61,
                    "word": " define",
                    "probability": 0.88525390625
                },
                {
                    "start": 548.01,
                    "end": 548.77,
                    "word": " a new",
                    "probability": 0.818359375
                },
                {
                    "start": 548.77,
                    "end": 550.91,
                    "word": " entity",
                    "probability": 0.84765625
                },
                {
                    "start": 550.91,
                    "end": 551.27,
                    "word": " here",
                    "probability": 0.81494140625
                }
            ]
        },
        {
            "id": 83,
            "seek": 55164,
            "start": 551.27,
            "end": 557.77,
            "text": " called pi. A policy pi is a function, or a distribution in the probabilistic case, that",
            "tokens": [
                51464,
                1219,
                3895,
                13,
                316,
                3897,
                3895,
                307,
                257,
                2445,
                11,
                420,
                257,
                7316,
                294,
                264,
                31959,
                3142,
                1389,
                11,
                300,
                51784
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2221235836094076,
            "compression_ratio": 1.6944444444444444,
            "no_speech_prob": 0.00039196014404296875,
            "words": [
                {
                    "start": 551.27,
                    "end": 551.81,
                    "word": " called",
                    "probability": 0.74951171875
                },
                {
                    "start": 551.81,
                    "end": 552.43,
                    "word": " pi",
                    "probability": 0.5009765625
                },
                {
                    "start": 552.43,
                    "end": 552.91,
                    "word": ".",
                    "probability": 0.87158203125
                },
                {
                    "start": 552.91,
                    "end": 553.11,
                    "word": " A",
                    "probability": 0.97265625
                },
                {
                    "start": 553.11,
                    "end": 553.55,
                    "word": " policy",
                    "probability": 0.8076171875
                },
                {
                    "start": 553.55,
                    "end": 553.99,
                    "word": " pi",
                    "probability": 0.900390625
                },
                {
                    "start": 553.99,
                    "end": 554.65,
                    "word": " is",
                    "probability": 0.888671875
                },
                {
                    "start": 554.83,
                    "end": 555.31,
                    "word": " a function",
                    "probability": 0.88037109375
                },
                {
                    "start": 555.31,
                    "end": 555.79,
                    "word": ",",
                    "probability": 0.36572265625
                },
                {
                    "start": 555.79,
                    "end": 555.99,
                    "word": " or",
                    "probability": 0.89501953125
                },
                {
                    "start": 556.13,
                    "end": 556.49,
                    "word": " a distribution",
                    "probability": 0.732421875
                },
                {
                    "start": 556.49,
                    "end": 556.67,
                    "word": " in",
                    "probability": 0.82568359375
                },
                {
                    "start": 556.67,
                    "end": 556.75,
                    "word": " the",
                    "probability": 0.669921875
                },
                {
                    "start": 556.75,
                    "end": 557.17,
                    "word": " probabilistic",
                    "probability": 0.95556640625
                },
                {
                    "start": 557.17,
                    "end": 557.55,
                    "word": " case",
                    "probability": 0.8271484375
                },
                {
                    "start": 557.55,
                    "end": 557.61,
                    "word": ",",
                    "probability": 0.869140625
                },
                {
                    "start": 557.61,
                    "end": 557.77,
                    "word": " that",
                    "probability": 0.88232421875
                }
            ]
        },
        {
            "id": 84,
            "seek": 57614,
            "start": 557.77,
            "end": 567.37,
            "text": " maps from state to action. So, it tells the agent what action to perform given any state.",
            "tokens": [
                50364,
                11317,
                490,
                1785,
                281,
                3069,
                13,
                407,
                11,
                309,
                5112,
                264,
                9461,
                437,
                3069,
                281,
                2042,
                2212,
                604,
                1785,
                13,
                50804
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23710316845348903,
            "compression_ratio": 1.5562130177514792,
            "no_speech_prob": 0.0001055598258972168,
            "words": [
                {
                    "start": 557.77,
                    "end": 558.17,
                    "word": " maps",
                    "probability": 0.5478515625
                },
                {
                    "start": 558.17,
                    "end": 558.49,
                    "word": " from",
                    "probability": 0.77685546875
                },
                {
                    "start": 558.49,
                    "end": 559.05,
                    "word": " state",
                    "probability": 0.7978515625
                },
                {
                    "start": 559.05,
                    "end": 559.31,
                    "word": " to",
                    "probability": 0.8974609375
                },
                {
                    "start": 559.31,
                    "end": 559.69,
                    "word": " action",
                    "probability": 0.81298828125
                },
                {
                    "start": 559.69,
                    "end": 560.43,
                    "word": ".",
                    "probability": 0.5751953125
                },
                {
                    "start": 560.43,
                    "end": 560.53,
                    "word": " So",
                    "probability": 0.67724609375
                },
                {
                    "start": 560.53,
                    "end": 560.75,
                    "word": ",",
                    "probability": 0.1854248046875
                },
                {
                    "start": 560.75,
                    "end": 560.81,
                    "word": " it",
                    "probability": 0.89453125
                },
                {
                    "start": 560.81,
                    "end": 561.43,
                    "word": " tells",
                    "probability": 0.73193359375
                },
                {
                    "start": 561.43,
                    "end": 561.65,
                    "word": " the",
                    "probability": 0.80859375
                },
                {
                    "start": 561.65,
                    "end": 562.05,
                    "word": " agent",
                    "probability": 0.7890625
                },
                {
                    "start": 562.05,
                    "end": 563.61,
                    "word": " what",
                    "probability": 0.8447265625
                },
                {
                    "start": 563.61,
                    "end": 564.01,
                    "word": " action",
                    "probability": 0.80810546875
                },
                {
                    "start": 564.01,
                    "end": 564.17,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 564.17,
                    "end": 564.65,
                    "word": " perform",
                    "probability": 0.77294921875
                },
                {
                    "start": 564.65,
                    "end": 565.53,
                    "word": " given",
                    "probability": 0.53173828125
                },
                {
                    "start": 565.53,
                    "end": 566.21,
                    "word": " any",
                    "probability": 0.79443359375
                },
                {
                    "start": 566.21,
                    "end": 566.57,
                    "word": " state",
                    "probability": 0.8818359375
                },
                {
                    "start": 566.57,
                    "end": 567.37,
                    "word": ".",
                    "probability": 0.79248046875
                }
            ]
        },
        {
            "id": 85,
            "seek": 57614,
            "start": 567.87,
            "end": 573.73,
            "text": " And the objective is to find the optimal policy pi star that maximizes the cumulative",
            "tokens": [
                50868,
                400,
                264,
                10024,
                307,
                281,
                915,
                264,
                16252,
                3897,
                3895,
                3543,
                300,
                5138,
                5660,
                264,
                38379,
                51176
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23710316845348903,
            "compression_ratio": 1.5562130177514792,
            "no_speech_prob": 0.0001055598258972168,
            "words": [
                {
                    "start": 567.87,
                    "end": 568.15,
                    "word": " And",
                    "probability": 0.77587890625
                },
                {
                    "start": 568.15,
                    "end": 568.29,
                    "word": " the",
                    "probability": 0.7861328125
                },
                {
                    "start": 568.29,
                    "end": 568.77,
                    "word": " objective",
                    "probability": 0.80859375
                },
                {
                    "start": 568.77,
                    "end": 569.21,
                    "word": " is",
                    "probability": 0.90234375
                },
                {
                    "start": 569.21,
                    "end": 569.41,
                    "word": " to",
                    "probability": 0.89990234375
                },
                {
                    "start": 569.41,
                    "end": 569.89,
                    "word": " find",
                    "probability": 0.80859375
                },
                {
                    "start": 569.89,
                    "end": 570.05,
                    "word": " the",
                    "probability": 0.80859375
                },
                {
                    "start": 570.05,
                    "end": 570.43,
                    "word": " optimal",
                    "probability": 0.75732421875
                },
                {
                    "start": 570.43,
                    "end": 570.99,
                    "word": " policy",
                    "probability": 0.806640625
                },
                {
                    "start": 570.99,
                    "end": 571.29,
                    "word": " pi",
                    "probability": 0.4716796875
                },
                {
                    "start": 571.29,
                    "end": 571.69,
                    "word": " star",
                    "probability": 0.43115234375
                },
                {
                    "start": 571.69,
                    "end": 572.03,
                    "word": " that",
                    "probability": 0.8232421875
                },
                {
                    "start": 572.03,
                    "end": 572.87,
                    "word": " maximizes",
                    "probability": 0.923095703125
                },
                {
                    "start": 572.87,
                    "end": 573.15,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 573.15,
                    "end": 573.73,
                    "word": " cumulative",
                    "probability": 0.81103515625
                }
            ]
        },
        {
            "id": 86,
            "seek": 57614,
            "start": 573.73,
            "end": 582.27,
            "text": " discounted reward. We also call this return. So, cumulative, meaning the sum of rewards",
            "tokens": [
                51176,
                11635,
                292,
                7782,
                13,
                492,
                611,
                818,
                341,
                2736,
                13,
                407,
                11,
                38379,
                11,
                3620,
                264,
                2408,
                295,
                17203,
                51604
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23710316845348903,
            "compression_ratio": 1.5562130177514792,
            "no_speech_prob": 0.0001055598258972168,
            "words": [
                {
                    "start": 573.73,
                    "end": 574.83,
                    "word": " discounted",
                    "probability": 0.92333984375
                },
                {
                    "start": 574.83,
                    "end": 575.45,
                    "word": " reward",
                    "probability": 0.7998046875
                },
                {
                    "start": 575.45,
                    "end": 576.07,
                    "word": ".",
                    "probability": 0.94482421875
                },
                {
                    "start": 576.07,
                    "end": 576.27,
                    "word": " We",
                    "probability": 0.88134765625
                },
                {
                    "start": 576.27,
                    "end": 576.69,
                    "word": " also",
                    "probability": 0.367431640625
                },
                {
                    "start": 576.69,
                    "end": 577.87,
                    "word": " call",
                    "probability": 0.79345703125
                },
                {
                    "start": 577.87,
                    "end": 578.05,
                    "word": " this",
                    "probability": 0.8798828125
                },
                {
                    "start": 578.05,
                    "end": 578.53,
                    "word": " return",
                    "probability": 0.6396484375
                },
                {
                    "start": 578.53,
                    "end": 579.33,
                    "word": ".",
                    "probability": 0.94677734375
                },
                {
                    "start": 579.35,
                    "end": 579.53,
                    "word": " So",
                    "probability": 0.80224609375
                },
                {
                    "start": 579.53,
                    "end": 579.75,
                    "word": ",",
                    "probability": 0.9208984375
                },
                {
                    "start": 579.75,
                    "end": 580.17,
                    "word": " cumulative",
                    "probability": 0.74169921875
                },
                {
                    "start": 580.17,
                    "end": 580.57,
                    "word": ",",
                    "probability": 0.7158203125
                },
                {
                    "start": 580.57,
                    "end": 580.87,
                    "word": " meaning",
                    "probability": 0.84228515625
                },
                {
                    "start": 580.87,
                    "end": 581.39,
                    "word": " the",
                    "probability": 0.76611328125
                },
                {
                    "start": 581.39,
                    "end": 581.61,
                    "word": " sum",
                    "probability": 0.9033203125
                },
                {
                    "start": 581.61,
                    "end": 581.75,
                    "word": " of",
                    "probability": 0.908203125
                },
                {
                    "start": 581.75,
                    "end": 582.27,
                    "word": " rewards",
                    "probability": 0.66650390625
                }
            ]
        },
        {
            "id": 87,
            "seek": 60194,
            "start": 582.27,
            "end": 587.85,
            "text": " in the entire duration of interaction. It can be one episode of a game, it can be throughout the",
            "tokens": [
                50380,
                294,
                264,
                2302,
                16365,
                295,
                9285,
                13,
                467,
                393,
                312,
                472,
                3500,
                295,
                257,
                1216,
                11,
                309,
                393,
                312,
                3710,
                264,
                50632
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21263586794552597,
            "compression_ratio": 1.6478260869565218,
            "no_speech_prob": 0.0321044921875,
            "words": [
                {
                    "start": 582.27,
                    "end": 582.79,
                    "word": " in",
                    "probability": 0.303466796875
                },
                {
                    "start": 582.79,
                    "end": 582.95,
                    "word": " the",
                    "probability": 0.8076171875
                },
                {
                    "start": 582.95,
                    "end": 583.49,
                    "word": " entire",
                    "probability": 0.740234375
                },
                {
                    "start": 583.49,
                    "end": 584.53,
                    "word": " duration",
                    "probability": 0.892578125
                },
                {
                    "start": 584.53,
                    "end": 584.89,
                    "word": " of",
                    "probability": 0.90087890625
                },
                {
                    "start": 584.89,
                    "end": 585.35,
                    "word": " interaction",
                    "probability": 0.73779296875
                },
                {
                    "start": 585.35,
                    "end": 585.49,
                    "word": ".",
                    "probability": 0.7099609375
                },
                {
                    "start": 585.49,
                    "end": 585.53,
                    "word": " It",
                    "probability": 0.71435546875
                },
                {
                    "start": 585.53,
                    "end": 585.63,
                    "word": " can",
                    "probability": 0.904296875
                },
                {
                    "start": 585.63,
                    "end": 585.75,
                    "word": " be",
                    "probability": 0.89697265625
                },
                {
                    "start": 585.75,
                    "end": 585.93,
                    "word": " one",
                    "probability": 0.79443359375
                },
                {
                    "start": 585.93,
                    "end": 586.25,
                    "word": " episode",
                    "probability": 0.72119140625
                },
                {
                    "start": 586.25,
                    "end": 586.41,
                    "word": " of",
                    "probability": 0.89599609375
                },
                {
                    "start": 586.51,
                    "end": 586.71,
                    "word": " a game",
                    "probability": 0.81005859375
                },
                {
                    "start": 586.71,
                    "end": 586.79,
                    "word": ",",
                    "probability": 0.68359375
                },
                {
                    "start": 586.79,
                    "end": 586.87,
                    "word": " it",
                    "probability": 0.86328125
                },
                {
                    "start": 586.87,
                    "end": 586.99,
                    "word": " can",
                    "probability": 0.9013671875
                },
                {
                    "start": 586.99,
                    "end": 587.17,
                    "word": " be",
                    "probability": 0.89306640625
                },
                {
                    "start": 587.17,
                    "end": 587.49,
                    "word": " throughout",
                    "probability": 0.60302734375
                },
                {
                    "start": 587.49,
                    "end": 587.85,
                    "word": " the",
                    "probability": 0.630859375
                }
            ]
        },
        {
            "id": 88,
            "seek": 60194,
            "start": 588.43,
            "end": 595.39,
            "text": " training period or training horizon, and it's discounted. This gamma is the discount factor",
            "tokens": [
                50672,
                3097,
                2896,
                420,
                3097,
                18046,
                11,
                293,
                309,
                311,
                11635,
                292,
                13,
                639,
                15546,
                307,
                264,
                11635,
                5952,
                51016
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21263586794552597,
            "compression_ratio": 1.6478260869565218,
            "no_speech_prob": 0.0321044921875,
            "words": [
                {
                    "start": 588.43,
                    "end": 588.83,
                    "word": " training",
                    "probability": 0.7392578125
                },
                {
                    "start": 588.83,
                    "end": 591.09,
                    "word": " period",
                    "probability": 0.72802734375
                },
                {
                    "start": 591.09,
                    "end": 591.29,
                    "word": " or",
                    "probability": 0.81103515625
                },
                {
                    "start": 591.29,
                    "end": 591.63,
                    "word": " training",
                    "probability": 0.4228515625
                },
                {
                    "start": 591.63,
                    "end": 592.29,
                    "word": " horizon",
                    "probability": 0.80419921875
                },
                {
                    "start": 592.29,
                    "end": 592.67,
                    "word": ",",
                    "probability": 0.3720703125
                },
                {
                    "start": 592.67,
                    "end": 593.17,
                    "word": " and",
                    "probability": 0.88232421875
                },
                {
                    "start": 593.17,
                    "end": 593.39,
                    "word": " it's",
                    "probability": 0.925048828125
                },
                {
                    "start": 593.39,
                    "end": 594.07,
                    "word": " discounted",
                    "probability": 0.89599609375
                },
                {
                    "start": 594.07,
                    "end": 594.13,
                    "word": ".",
                    "probability": 0.904296875
                },
                {
                    "start": 594.13,
                    "end": 594.29,
                    "word": " This",
                    "probability": 0.71728515625
                },
                {
                    "start": 594.29,
                    "end": 594.51,
                    "word": " gamma",
                    "probability": 0.708984375
                },
                {
                    "start": 594.51,
                    "end": 594.69,
                    "word": " is",
                    "probability": 0.89404296875
                },
                {
                    "start": 594.69,
                    "end": 594.83,
                    "word": " the",
                    "probability": 0.496337890625
                },
                {
                    "start": 594.83,
                    "end": 595.11,
                    "word": " discount",
                    "probability": 0.85986328125
                },
                {
                    "start": 595.11,
                    "end": 595.39,
                    "word": " factor",
                    "probability": 0.72412109375
                }
            ]
        },
        {
            "id": 89,
            "seek": 60194,
            "start": 595.39,
            "end": 600.75,
            "text": " that defines how myopic the agent is, or how much it values events that occur further down the line.",
            "tokens": [
                51016,
                300,
                23122,
                577,
                452,
                40216,
                264,
                9461,
                307,
                11,
                420,
                577,
                709,
                309,
                4190,
                3931,
                300,
                5160,
                3052,
                760,
                264,
                1622,
                13,
                51280
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21263586794552597,
            "compression_ratio": 1.6478260869565218,
            "no_speech_prob": 0.0321044921875,
            "words": [
                {
                    "start": 595.39,
                    "end": 595.59,
                    "word": " that",
                    "probability": 0.87939453125
                },
                {
                    "start": 595.59,
                    "end": 595.95,
                    "word": " defines",
                    "probability": 0.74609375
                },
                {
                    "start": 595.95,
                    "end": 596.15,
                    "word": " how",
                    "probability": 0.89404296875
                },
                {
                    "start": 596.15,
                    "end": 596.45,
                    "word": " myopic",
                    "probability": 0.966552734375
                },
                {
                    "start": 596.45,
                    "end": 596.69,
                    "word": " the",
                    "probability": 0.787109375
                },
                {
                    "start": 596.69,
                    "end": 597.03,
                    "word": " agent",
                    "probability": 0.77490234375
                },
                {
                    "start": 597.03,
                    "end": 597.27,
                    "word": " is",
                    "probability": 0.90478515625
                },
                {
                    "start": 597.27,
                    "end": 597.29,
                    "word": ",",
                    "probability": 0.364013671875
                },
                {
                    "start": 597.29,
                    "end": 597.41,
                    "word": " or",
                    "probability": 0.89208984375
                },
                {
                    "start": 597.41,
                    "end": 597.55,
                    "word": " how",
                    "probability": 0.90185546875
                },
                {
                    "start": 597.55,
                    "end": 597.69,
                    "word": " much",
                    "probability": 0.80712890625
                },
                {
                    "start": 597.69,
                    "end": 597.81,
                    "word": " it",
                    "probability": 0.794921875
                },
                {
                    "start": 597.81,
                    "end": 598.25,
                    "word": " values",
                    "probability": 0.8427734375
                },
                {
                    "start": 598.25,
                    "end": 598.55,
                    "word": " events",
                    "probability": 0.75341796875
                },
                {
                    "start": 598.55,
                    "end": 598.91,
                    "word": " that",
                    "probability": 0.88427734375
                },
                {
                    "start": 598.91,
                    "end": 599.67,
                    "word": " occur",
                    "probability": 0.837890625
                },
                {
                    "start": 599.67,
                    "end": 600.07,
                    "word": " further",
                    "probability": 0.83203125
                },
                {
                    "start": 600.07,
                    "end": 600.29,
                    "word": " down",
                    "probability": 0.82275390625
                },
                {
                    "start": 600.29,
                    "end": 600.43,
                    "word": " the",
                    "probability": 0.80615234375
                },
                {
                    "start": 600.43,
                    "end": 600.63,
                    "word": " line",
                    "probability": 0.87646484375
                },
                {
                    "start": 600.63,
                    "end": 600.75,
                    "word": ".",
                    "probability": 0.9794921875
                }
            ]
        },
        {
            "id": 90,
            "seek": 60194,
            "start": 600.75,
            "end": 608.07,
            "text": " It's typically a constant value between zero and one, and as you can see, as t increases,",
            "tokens": [
                51280,
                467,
                311,
                5850,
                257,
                5754,
                2158,
                1296,
                4018,
                293,
                472,
                11,
                293,
                382,
                291,
                393,
                536,
                11,
                382,
                256,
                8637,
                11,
                51648
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21263586794552597,
            "compression_ratio": 1.6478260869565218,
            "no_speech_prob": 0.0321044921875,
            "words": [
                {
                    "start": 600.75,
                    "end": 600.95,
                    "word": " It's",
                    "probability": 0.94287109375
                },
                {
                    "start": 600.95,
                    "end": 601.63,
                    "word": " typically",
                    "probability": 0.81103515625
                },
                {
                    "start": 602.09,
                    "end": 602.51,
                    "word": " a constant",
                    "probability": 0.8486328125
                },
                {
                    "start": 602.51,
                    "end": 602.97,
                    "word": " value",
                    "probability": 0.919921875
                },
                {
                    "start": 602.97,
                    "end": 603.23,
                    "word": " between",
                    "probability": 0.7646484375
                },
                {
                    "start": 603.23,
                    "end": 603.59,
                    "word": " zero",
                    "probability": 0.474609375
                },
                {
                    "start": 603.59,
                    "end": 603.81,
                    "word": " and",
                    "probability": 0.89794921875
                },
                {
                    "start": 603.81,
                    "end": 604.17,
                    "word": " one",
                    "probability": 0.802734375
                },
                {
                    "start": 604.17,
                    "end": 604.75,
                    "word": ",",
                    "probability": 0.56298828125
                },
                {
                    "start": 604.75,
                    "end": 605.39,
                    "word": " and",
                    "probability": 0.89013671875
                },
                {
                    "start": 605.39,
                    "end": 605.73,
                    "word": " as",
                    "probability": 0.78955078125
                },
                {
                    "start": 605.73,
                    "end": 605.81,
                    "word": " you",
                    "probability": 0.8896484375
                },
                {
                    "start": 605.81,
                    "end": 605.93,
                    "word": " can",
                    "probability": 0.89794921875
                },
                {
                    "start": 605.93,
                    "end": 606.25,
                    "word": " see",
                    "probability": 0.81982421875
                },
                {
                    "start": 606.25,
                    "end": 606.61,
                    "word": ",",
                    "probability": 0.9267578125
                },
                {
                    "start": 606.61,
                    "end": 606.85,
                    "word": " as",
                    "probability": 0.9052734375
                },
                {
                    "start": 607.07,
                    "end": 607.59,
                    "word": " t increases",
                    "probability": 0.9501953125
                },
                {
                    "start": 607.59,
                    "end": 608.07,
                    "word": ",",
                    "probability": 0.904296875
                }
            ]
        },
        {
            "id": 91,
            "seek": 62654,
            "start": 608.07,
            "end": 618.77,
            "text": " this gamma to power t decreases. So, the value, the preference, or the observed value of something",
            "tokens": [
                50364,
                341,
                15546,
                281,
                1347,
                256,
                24108,
                13,
                407,
                11,
                264,
                2158,
                11,
                264,
                17502,
                11,
                420,
                264,
                13095,
                2158,
                295,
                746,
                50900
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21489876466737667,
            "compression_ratio": 1.5956284153005464,
            "no_speech_prob": 0.00037407875061035156,
            "words": [
                {
                    "start": 608.07,
                    "end": 608.33,
                    "word": " this",
                    "probability": 0.61865234375
                },
                {
                    "start": 608.33,
                    "end": 608.63,
                    "word": " gamma",
                    "probability": 0.724609375
                },
                {
                    "start": 608.63,
                    "end": 608.79,
                    "word": " to",
                    "probability": 0.83056640625
                },
                {
                    "start": 608.79,
                    "end": 609.03,
                    "word": " power",
                    "probability": 0.7626953125
                },
                {
                    "start": 609.37,
                    "end": 610.45,
                    "word": " t decreases",
                    "probability": 0.91357421875
                },
                {
                    "start": 610.45,
                    "end": 610.79,
                    "word": ".",
                    "probability": 0.75048828125
                },
                {
                    "start": 610.79,
                    "end": 610.81,
                    "word": " So",
                    "probability": 0.8623046875
                },
                {
                    "start": 610.81,
                    "end": 610.93,
                    "word": ",",
                    "probability": 0.0941162109375
                },
                {
                    "start": 610.93,
                    "end": 611.05,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 611.05,
                    "end": 611.43,
                    "word": " value",
                    "probability": 0.87109375
                },
                {
                    "start": 611.43,
                    "end": 611.83,
                    "word": ",",
                    "probability": 0.8388671875
                },
                {
                    "start": 611.83,
                    "end": 612.03,
                    "word": " the",
                    "probability": 0.8173828125
                },
                {
                    "start": 612.03,
                    "end": 614.81,
                    "word": " preference",
                    "probability": 0.60400390625
                },
                {
                    "start": 614.81,
                    "end": 615.41,
                    "word": ",",
                    "probability": 0.4404296875
                },
                {
                    "start": 615.41,
                    "end": 615.73,
                    "word": " or",
                    "probability": 0.9013671875
                },
                {
                    "start": 615.73,
                    "end": 616.43,
                    "word": " the",
                    "probability": 0.79345703125
                },
                {
                    "start": 616.43,
                    "end": 617.59,
                    "word": " observed",
                    "probability": 0.79736328125
                },
                {
                    "start": 617.59,
                    "end": 618.11,
                    "word": " value",
                    "probability": 0.91552734375
                },
                {
                    "start": 618.11,
                    "end": 618.47,
                    "word": " of",
                    "probability": 0.8056640625
                },
                {
                    "start": 618.47,
                    "end": 618.77,
                    "word": " something",
                    "probability": 0.74462890625
                }
            ]
        },
        {
            "id": 92,
            "seek": 62654,
            "start": 618.77,
            "end": 625.89,
            "text": " that happens further down the line, which means with greater t, is going to decrease as t increases.",
            "tokens": [
                50900,
                300,
                2314,
                3052,
                760,
                264,
                1622,
                11,
                597,
                1355,
                365,
                5044,
                256,
                11,
                307,
                516,
                281,
                11514,
                382,
                256,
                8637,
                13,
                51244
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21489876466737667,
            "compression_ratio": 1.5956284153005464,
            "no_speech_prob": 0.00037407875061035156,
            "words": [
                {
                    "start": 618.77,
                    "end": 618.95,
                    "word": " that",
                    "probability": 0.87939453125
                },
                {
                    "start": 618.95,
                    "end": 619.21,
                    "word": " happens",
                    "probability": 0.8984375
                },
                {
                    "start": 619.21,
                    "end": 619.51,
                    "word": " further",
                    "probability": 0.84521484375
                },
                {
                    "start": 619.51,
                    "end": 619.73,
                    "word": " down",
                    "probability": 0.82080078125
                },
                {
                    "start": 619.73,
                    "end": 619.87,
                    "word": " the",
                    "probability": 0.81103515625
                },
                {
                    "start": 619.87,
                    "end": 620.15,
                    "word": " line",
                    "probability": 0.8720703125
                },
                {
                    "start": 620.15,
                    "end": 620.49,
                    "word": ",",
                    "probability": 0.88232421875
                },
                {
                    "start": 620.49,
                    "end": 620.67,
                    "word": " which",
                    "probability": 0.888671875
                },
                {
                    "start": 620.67,
                    "end": 620.89,
                    "word": " means",
                    "probability": 0.88427734375
                },
                {
                    "start": 620.89,
                    "end": 621.95,
                    "word": " with",
                    "probability": 0.67919921875
                },
                {
                    "start": 621.95,
                    "end": 622.47,
                    "word": " greater",
                    "probability": 0.84130859375
                },
                {
                    "start": 622.85,
                    "end": 623.21,
                    "word": " t,",
                    "probability": 0.84326171875
                },
                {
                    "start": 623.21,
                    "end": 623.41,
                    "word": " is",
                    "probability": 0.8857421875
                },
                {
                    "start": 623.41,
                    "end": 623.61,
                    "word": " going",
                    "probability": 0.85888671875
                },
                {
                    "start": 623.61,
                    "end": 623.89,
                    "word": " to",
                    "probability": 0.90625
                },
                {
                    "start": 623.89,
                    "end": 624.51,
                    "word": " decrease",
                    "probability": 0.77783203125
                },
                {
                    "start": 624.51,
                    "end": 625.01,
                    "word": " as",
                    "probability": 0.88330078125
                },
                {
                    "start": 625.17,
                    "end": 625.59,
                    "word": " t increases",
                    "probability": 0.93359375
                },
                {
                    "start": 625.59,
                    "end": 625.89,
                    "word": ".",
                    "probability": 0.99560546875
                }
            ]
        },
        {
            "id": 93,
            "seek": 62654,
            "start": 626.65,
            "end": 632.67,
            "text": " So, again, the objective is to find the optimal policy pi star that maximizes this sum here.",
            "tokens": [
                51292,
                407,
                11,
                797,
                11,
                264,
                10024,
                307,
                281,
                915,
                264,
                16252,
                3897,
                3895,
                3543,
                300,
                5138,
                5660,
                341,
                2408,
                510,
                13,
                51564
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21489876466737667,
            "compression_ratio": 1.5956284153005464,
            "no_speech_prob": 0.00037407875061035156,
            "words": [
                {
                    "start": 626.65,
                    "end": 626.89,
                    "word": " So",
                    "probability": 0.85791015625
                },
                {
                    "start": 626.89,
                    "end": 627.33,
                    "word": ",",
                    "probability": 0.67333984375
                },
                {
                    "start": 627.33,
                    "end": 627.61,
                    "word": " again",
                    "probability": 0.91650390625
                },
                {
                    "start": 627.61,
                    "end": 627.87,
                    "word": ",",
                    "probability": 0.9384765625
                },
                {
                    "start": 627.87,
                    "end": 628.11,
                    "word": " the",
                    "probability": 0.80908203125
                },
                {
                    "start": 628.11,
                    "end": 628.47,
                    "word": " objective",
                    "probability": 0.81201171875
                },
                {
                    "start": 628.47,
                    "end": 628.63,
                    "word": " is",
                    "probability": 0.8935546875
                },
                {
                    "start": 628.63,
                    "end": 628.75,
                    "word": " to",
                    "probability": 0.89208984375
                },
                {
                    "start": 628.75,
                    "end": 628.91,
                    "word": " find",
                    "probability": 0.80615234375
                },
                {
                    "start": 628.91,
                    "end": 629.07,
                    "word": " the",
                    "probability": 0.78564453125
                },
                {
                    "start": 629.07,
                    "end": 629.35,
                    "word": " optimal",
                    "probability": 0.7646484375
                },
                {
                    "start": 629.35,
                    "end": 629.67,
                    "word": " policy",
                    "probability": 0.7802734375
                },
                {
                    "start": 629.67,
                    "end": 629.93,
                    "word": " pi",
                    "probability": 0.75830078125
                },
                {
                    "start": 629.93,
                    "end": 630.19,
                    "word": " star",
                    "probability": 0.65625
                },
                {
                    "start": 630.19,
                    "end": 630.41,
                    "word": " that",
                    "probability": 0.869140625
                },
                {
                    "start": 630.41,
                    "end": 631.09,
                    "word": " maximizes",
                    "probability": 0.919677734375
                },
                {
                    "start": 631.09,
                    "end": 631.39,
                    "word": " this",
                    "probability": 0.71533203125
                },
                {
                    "start": 631.39,
                    "end": 631.71,
                    "word": " sum",
                    "probability": 0.93505859375
                },
                {
                    "start": 631.71,
                    "end": 632.07,
                    "word": " here",
                    "probability": 0.77587890625
                },
                {
                    "start": 632.07,
                    "end": 632.67,
                    "word": ".",
                    "probability": 0.99609375
                }
            ]
        },
        {
            "id": 94,
            "seek": 65530,
            "start": 632.67,
            "end": 639.51,
            "text": " All right, we are going to quickly define two other definitions here. If we want to evaluate",
            "tokens": [
                50372,
                1057,
                558,
                11,
                321,
                366,
                516,
                281,
                2661,
                6964,
                732,
                661,
                21988,
                510,
                13,
                759,
                321,
                528,
                281,
                13059,
                50704
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20689654727091736,
            "compression_ratio": 1.6929824561403508,
            "no_speech_prob": 0.0018968582153320312,
            "words": [
                {
                    "start": 632.67,
                    "end": 633.11,
                    "word": " All",
                    "probability": 0.2198486328125
                },
                {
                    "start": 633.11,
                    "end": 633.27,
                    "word": " right",
                    "probability": 0.8759765625
                },
                {
                    "start": 633.27,
                    "end": 633.29,
                    "word": ",",
                    "probability": 0.68505859375
                },
                {
                    "start": 633.29,
                    "end": 633.43,
                    "word": " we",
                    "probability": 0.86962890625
                },
                {
                    "start": 633.43,
                    "end": 633.53,
                    "word": " are",
                    "probability": 0.432861328125
                },
                {
                    "start": 633.53,
                    "end": 633.65,
                    "word": " going",
                    "probability": 0.84619140625
                },
                {
                    "start": 633.65,
                    "end": 633.79,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 633.79,
                    "end": 634.13,
                    "word": " quickly",
                    "probability": 0.75537109375
                },
                {
                    "start": 634.13,
                    "end": 634.47,
                    "word": " define",
                    "probability": 0.8701171875
                },
                {
                    "start": 634.47,
                    "end": 634.69,
                    "word": " two",
                    "probability": 0.78857421875
                },
                {
                    "start": 634.69,
                    "end": 635.23,
                    "word": " other",
                    "probability": 0.787109375
                },
                {
                    "start": 635.23,
                    "end": 636.99,
                    "word": " definitions",
                    "probability": 0.7216796875
                },
                {
                    "start": 636.99,
                    "end": 637.33,
                    "word": " here",
                    "probability": 0.79248046875
                },
                {
                    "start": 637.33,
                    "end": 638.01,
                    "word": ".",
                    "probability": 0.962890625
                },
                {
                    "start": 638.29,
                    "end": 638.63,
                    "word": " If",
                    "probability": 0.888671875
                },
                {
                    "start": 638.63,
                    "end": 638.73,
                    "word": " we",
                    "probability": 0.3623046875
                },
                {
                    "start": 638.73,
                    "end": 638.95,
                    "word": " want",
                    "probability": 0.818359375
                },
                {
                    "start": 638.95,
                    "end": 639.11,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 639.11,
                    "end": 639.51,
                    "word": " evaluate",
                    "probability": 0.74755859375
                }
            ]
        },
        {
            "id": 95,
            "seek": 65530,
            "start": 639.67,
            "end": 647.01,
            "text": " a particular state in these settings, one approach is to measure the value of that state using the",
            "tokens": [
                50704,
                257,
                1729,
                1785,
                294,
                613,
                6257,
                11,
                472,
                3109,
                307,
                281,
                3481,
                264,
                2158,
                295,
                300,
                1785,
                1228,
                264,
                51076
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20689654727091736,
            "compression_ratio": 1.6929824561403508,
            "no_speech_prob": 0.0018968582153320312,
            "words": [
                {
                    "start": 639.67,
                    "end": 640.05,
                    "word": " a particular",
                    "probability": 0.81103515625
                },
                {
                    "start": 640.05,
                    "end": 640.67,
                    "word": " state",
                    "probability": 0.84814453125
                },
                {
                    "start": 640.67,
                    "end": 641.21,
                    "word": " in",
                    "probability": 0.8740234375
                },
                {
                    "start": 641.21,
                    "end": 641.57,
                    "word": " these",
                    "probability": 0.65380859375
                },
                {
                    "start": 641.57,
                    "end": 642.05,
                    "word": " settings",
                    "probability": 0.81201171875
                },
                {
                    "start": 642.05,
                    "end": 643.03,
                    "word": ",",
                    "probability": 0.90283203125
                },
                {
                    "start": 643.03,
                    "end": 643.39,
                    "word": " one",
                    "probability": 0.8046875
                },
                {
                    "start": 643.39,
                    "end": 643.75,
                    "word": " approach",
                    "probability": 0.78564453125
                },
                {
                    "start": 643.75,
                    "end": 643.95,
                    "word": " is",
                    "probability": 0.90234375
                },
                {
                    "start": 643.95,
                    "end": 644.11,
                    "word": " to",
                    "probability": 0.90087890625
                },
                {
                    "start": 644.11,
                    "end": 644.41,
                    "word": " measure",
                    "probability": 0.8115234375
                },
                {
                    "start": 644.41,
                    "end": 645.11,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 645.11,
                    "end": 645.65,
                    "word": " value",
                    "probability": 0.92919921875
                },
                {
                    "start": 645.65,
                    "end": 645.85,
                    "word": " of",
                    "probability": 0.90625
                },
                {
                    "start": 645.85,
                    "end": 646.03,
                    "word": " that",
                    "probability": 0.8779296875
                },
                {
                    "start": 646.03,
                    "end": 646.37,
                    "word": " state",
                    "probability": 0.87841796875
                },
                {
                    "start": 646.37,
                    "end": 646.73,
                    "word": " using",
                    "probability": 0.7783203125
                },
                {
                    "start": 646.73,
                    "end": 647.01,
                    "word": " the",
                    "probability": 0.81201171875
                }
            ]
        },
        {
            "id": 96,
            "seek": 65530,
            "start": 647.01,
            "end": 654.17,
            "text": " value function. The value function of state s is the expected cumulative reward from following",
            "tokens": [
                51076,
                2158,
                2445,
                13,
                440,
                2158,
                2445,
                295,
                1785,
                262,
                307,
                264,
                5176,
                38379,
                7782,
                490,
                3480,
                51448
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20689654727091736,
            "compression_ratio": 1.6929824561403508,
            "no_speech_prob": 0.0018968582153320312,
            "words": [
                {
                    "start": 647.01,
                    "end": 647.25,
                    "word": " value",
                    "probability": 0.853515625
                },
                {
                    "start": 647.25,
                    "end": 647.67,
                    "word": " function",
                    "probability": 0.87890625
                },
                {
                    "start": 647.67,
                    "end": 647.79,
                    "word": ".",
                    "probability": 0.97119140625
                },
                {
                    "start": 647.79,
                    "end": 647.93,
                    "word": " The",
                    "probability": 0.81591796875
                },
                {
                    "start": 647.93,
                    "end": 648.23,
                    "word": " value",
                    "probability": 0.91748046875
                },
                {
                    "start": 648.23,
                    "end": 648.73,
                    "word": " function",
                    "probability": 0.87939453125
                },
                {
                    "start": 648.73,
                    "end": 649.65,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 649.65,
                    "end": 650.07,
                    "word": " state",
                    "probability": 0.82763671875
                },
                {
                    "start": 650.07,
                    "end": 650.49,
                    "word": " s",
                    "probability": 0.302490234375
                },
                {
                    "start": 650.49,
                    "end": 650.67,
                    "word": " is",
                    "probability": 0.88134765625
                },
                {
                    "start": 650.67,
                    "end": 650.83,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 650.83,
                    "end": 651.27,
                    "word": " expected",
                    "probability": 0.74560546875
                },
                {
                    "start": 651.27,
                    "end": 651.73,
                    "word": " cumulative",
                    "probability": 0.79931640625
                },
                {
                    "start": 651.73,
                    "end": 652.27,
                    "word": " reward",
                    "probability": 0.84228515625
                },
                {
                    "start": 652.27,
                    "end": 653.61,
                    "word": " from",
                    "probability": 0.72900390625
                },
                {
                    "start": 653.61,
                    "end": 654.17,
                    "word": " following",
                    "probability": 0.8330078125
                }
            ]
        },
        {
            "id": 97,
            "seek": 65530,
            "start": 654.17,
            "end": 661.43,
            "text": " the policy pi from this state. So, let's assume we already have a policy. What will be the expected",
            "tokens": [
                51448,
                264,
                3897,
                3895,
                490,
                341,
                1785,
                13,
                407,
                11,
                718,
                311,
                6552,
                321,
                1217,
                362,
                257,
                3897,
                13,
                708,
                486,
                312,
                264,
                5176,
                51796
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20689654727091736,
            "compression_ratio": 1.6929824561403508,
            "no_speech_prob": 0.0018968582153320312,
            "words": [
                {
                    "start": 654.17,
                    "end": 654.65,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 654.65,
                    "end": 655.17,
                    "word": " policy",
                    "probability": 0.8193359375
                },
                {
                    "start": 655.17,
                    "end": 655.81,
                    "word": " pi",
                    "probability": 0.7001953125
                },
                {
                    "start": 655.81,
                    "end": 656.81,
                    "word": " from",
                    "probability": 0.77783203125
                },
                {
                    "start": 656.81,
                    "end": 657.25,
                    "word": " this",
                    "probability": 0.59130859375
                },
                {
                    "start": 657.25,
                    "end": 657.51,
                    "word": " state",
                    "probability": 0.89794921875
                },
                {
                    "start": 657.51,
                    "end": 657.65,
                    "word": ".",
                    "probability": 0.95556640625
                },
                {
                    "start": 657.65,
                    "end": 657.81,
                    "word": " So",
                    "probability": 0.80126953125
                },
                {
                    "start": 657.81,
                    "end": 657.95,
                    "word": ",",
                    "probability": 0.188232421875
                },
                {
                    "start": 657.95,
                    "end": 658.11,
                    "word": " let's",
                    "probability": 0.937744140625
                },
                {
                    "start": 658.11,
                    "end": 658.31,
                    "word": " assume",
                    "probability": 0.818359375
                },
                {
                    "start": 658.31,
                    "end": 658.43,
                    "word": " we",
                    "probability": 0.84912109375
                },
                {
                    "start": 658.43,
                    "end": 658.61,
                    "word": " already",
                    "probability": 0.8427734375
                },
                {
                    "start": 658.61,
                    "end": 658.83,
                    "word": " have",
                    "probability": 0.87939453125
                },
                {
                    "start": 658.95,
                    "end": 659.29,
                    "word": " a policy",
                    "probability": 0.8271484375
                },
                {
                    "start": 659.29,
                    "end": 659.89,
                    "word": ".",
                    "probability": 0.77783203125
                },
                {
                    "start": 659.89,
                    "end": 660.23,
                    "word": " What",
                    "probability": 0.80126953125
                },
                {
                    "start": 660.23,
                    "end": 660.41,
                    "word": " will",
                    "probability": 0.80810546875
                },
                {
                    "start": 660.41,
                    "end": 660.55,
                    "word": " be",
                    "probability": 0.90771484375
                },
                {
                    "start": 660.55,
                    "end": 660.73,
                    "word": " the",
                    "probability": 0.8076171875
                },
                {
                    "start": 660.73,
                    "end": 661.43,
                    "word": " expected",
                    "probability": 0.755859375
                }
            ]
        },
        {
            "id": 98,
            "seek": 68052,
            "start": 661.43,
            "end": 669.75,
            "text": " cumulative reward if you start in state s and keep following policy pi? Now, we also define",
            "tokens": [
                50388,
                38379,
                7782,
                498,
                291,
                722,
                294,
                1785,
                262,
                293,
                1066,
                3480,
                3897,
                3895,
                30,
                823,
                11,
                321,
                611,
                6964,
                50792
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1921672101144667,
            "compression_ratio": 1.5977653631284916,
            "no_speech_prob": 0.0005884170532226562,
            "words": [
                {
                    "start": 661.43,
                    "end": 662.37,
                    "word": " cumulative",
                    "probability": 0.50732421875
                },
                {
                    "start": 662.37,
                    "end": 662.97,
                    "word": " reward",
                    "probability": 0.8583984375
                },
                {
                    "start": 662.97,
                    "end": 663.49,
                    "word": " if",
                    "probability": 0.75927734375
                },
                {
                    "start": 663.49,
                    "end": 663.63,
                    "word": " you",
                    "probability": 0.8486328125
                },
                {
                    "start": 663.63,
                    "end": 663.91,
                    "word": " start",
                    "probability": 0.86572265625
                },
                {
                    "start": 663.91,
                    "end": 664.11,
                    "word": " in",
                    "probability": 0.8173828125
                },
                {
                    "start": 664.11,
                    "end": 664.75,
                    "word": " state",
                    "probability": 0.77734375
                },
                {
                    "start": 664.75,
                    "end": 665.19,
                    "word": " s",
                    "probability": 0.3505859375
                },
                {
                    "start": 665.19,
                    "end": 665.67,
                    "word": " and",
                    "probability": 0.845703125
                },
                {
                    "start": 665.67,
                    "end": 665.95,
                    "word": " keep",
                    "probability": 0.8056640625
                },
                {
                    "start": 665.95,
                    "end": 666.33,
                    "word": " following",
                    "probability": 0.826171875
                },
                {
                    "start": 666.33,
                    "end": 666.77,
                    "word": " policy",
                    "probability": 0.7216796875
                },
                {
                    "start": 666.77,
                    "end": 667.09,
                    "word": " pi",
                    "probability": 0.77392578125
                },
                {
                    "start": 667.09,
                    "end": 667.67,
                    "word": "?",
                    "probability": 0.00846099853515625
                },
                {
                    "start": 667.87,
                    "end": 668.53,
                    "word": " Now",
                    "probability": 0.8408203125
                },
                {
                    "start": 668.53,
                    "end": 668.65,
                    "word": ",",
                    "probability": 0.56689453125
                },
                {
                    "start": 668.65,
                    "end": 668.81,
                    "word": " we",
                    "probability": 0.9052734375
                },
                {
                    "start": 668.81,
                    "end": 669.17,
                    "word": " also",
                    "probability": 0.8154296875
                },
                {
                    "start": 669.17,
                    "end": 669.75,
                    "word": " define",
                    "probability": 0.89111328125
                }
            ]
        },
        {
            "id": 99,
            "seek": 68052,
            "start": 670.31,
            "end": 679.21,
            "text": " a Q value, or a Q function, which tells us how good is a particular action A if it's performed",
            "tokens": [
                50792,
                257,
                1249,
                2158,
                11,
                420,
                257,
                1249,
                2445,
                11,
                597,
                5112,
                505,
                577,
                665,
                307,
                257,
                1729,
                3069,
                316,
                498,
                309,
                311,
                10332,
                51252
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1921672101144667,
            "compression_ratio": 1.5977653631284916,
            "no_speech_prob": 0.0005884170532226562,
            "words": [
                {
                    "start": 670.31,
                    "end": 670.63,
                    "word": " a Q",
                    "probability": 0.662109375
                },
                {
                    "start": 670.63,
                    "end": 670.97,
                    "word": " value",
                    "probability": 0.69677734375
                },
                {
                    "start": 670.97,
                    "end": 671.13,
                    "word": ",",
                    "probability": 0.1142578125
                },
                {
                    "start": 671.13,
                    "end": 671.21,
                    "word": " or",
                    "probability": 0.86181640625
                },
                {
                    "start": 671.33,
                    "end": 671.55,
                    "word": " a Q",
                    "probability": 0.9892578125
                },
                {
                    "start": 671.55,
                    "end": 671.95,
                    "word": " function",
                    "probability": 0.8447265625
                },
                {
                    "start": 671.95,
                    "end": 672.63,
                    "word": ",",
                    "probability": 0.82421875
                },
                {
                    "start": 672.63,
                    "end": 673.09,
                    "word": " which",
                    "probability": 0.9130859375
                },
                {
                    "start": 673.09,
                    "end": 673.89,
                    "word": " tells",
                    "probability": 0.748046875
                },
                {
                    "start": 673.89,
                    "end": 674.29,
                    "word": " us",
                    "probability": 0.90380859375
                },
                {
                    "start": 674.29,
                    "end": 675.13,
                    "word": " how",
                    "probability": 0.83251953125
                },
                {
                    "start": 675.13,
                    "end": 675.59,
                    "word": " good",
                    "probability": 0.8798828125
                },
                {
                    "start": 675.59,
                    "end": 675.97,
                    "word": " is",
                    "probability": 0.8974609375
                },
                {
                    "start": 676.11,
                    "end": 676.47,
                    "word": " a particular",
                    "probability": 0.81103515625
                },
                {
                    "start": 676.47,
                    "end": 676.95,
                    "word": " action",
                    "probability": 0.7861328125
                },
                {
                    "start": 676.95,
                    "end": 677.47,
                    "word": " A",
                    "probability": 0.56005859375
                },
                {
                    "start": 677.47,
                    "end": 678.61,
                    "word": " if",
                    "probability": 0.71630859375
                },
                {
                    "start": 678.61,
                    "end": 678.81,
                    "word": " it's",
                    "probability": 0.93505859375
                },
                {
                    "start": 678.81,
                    "end": 679.21,
                    "word": " performed",
                    "probability": 0.71484375
                }
            ]
        },
        {
            "id": 100,
            "seek": 68052,
            "start": 679.21,
            "end": 686.65,
            "text": " in state s. In other words, if we are in state s, if the agent is in state s and performs action A,",
            "tokens": [
                51252,
                294,
                1785,
                262,
                13,
                682,
                661,
                2283,
                11,
                498,
                321,
                366,
                294,
                1785,
                262,
                11,
                498,
                264,
                9461,
                307,
                294,
                1785,
                262,
                293,
                26213,
                3069,
                316,
                11,
                51592
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1921672101144667,
            "compression_ratio": 1.5977653631284916,
            "no_speech_prob": 0.0005884170532226562,
            "words": [
                {
                    "start": 679.21,
                    "end": 679.47,
                    "word": " in",
                    "probability": 0.90869140625
                },
                {
                    "start": 679.47,
                    "end": 679.69,
                    "word": " state",
                    "probability": 0.87548828125
                },
                {
                    "start": 679.69,
                    "end": 680.19,
                    "word": " s",
                    "probability": 0.9267578125
                },
                {
                    "start": 680.19,
                    "end": 680.49,
                    "word": ".",
                    "probability": 0.69384765625
                },
                {
                    "start": 680.49,
                    "end": 680.67,
                    "word": " In",
                    "probability": 0.896484375
                },
                {
                    "start": 680.67,
                    "end": 680.83,
                    "word": " other",
                    "probability": 0.79345703125
                },
                {
                    "start": 680.83,
                    "end": 681.15,
                    "word": " words",
                    "probability": 0.728515625
                },
                {
                    "start": 681.15,
                    "end": 681.67,
                    "word": ",",
                    "probability": 0.9794921875
                },
                {
                    "start": 681.67,
                    "end": 682.05,
                    "word": " if",
                    "probability": 0.8974609375
                },
                {
                    "start": 682.05,
                    "end": 682.69,
                    "word": " we",
                    "probability": 0.841796875
                },
                {
                    "start": 682.69,
                    "end": 682.85,
                    "word": " are",
                    "probability": 0.8642578125
                },
                {
                    "start": 682.85,
                    "end": 683.03,
                    "word": " in",
                    "probability": 0.9052734375
                },
                {
                    "start": 683.03,
                    "end": 683.21,
                    "word": " state",
                    "probability": 0.87646484375
                },
                {
                    "start": 683.21,
                    "end": 683.47,
                    "word": " s",
                    "probability": 0.9638671875
                },
                {
                    "start": 683.47,
                    "end": 683.51,
                    "word": ",",
                    "probability": 0.9111328125
                },
                {
                    "start": 683.51,
                    "end": 683.59,
                    "word": " if",
                    "probability": 0.8720703125
                },
                {
                    "start": 683.59,
                    "end": 683.69,
                    "word": " the",
                    "probability": 0.8056640625
                },
                {
                    "start": 683.69,
                    "end": 683.93,
                    "word": " agent",
                    "probability": 0.7880859375
                },
                {
                    "start": 683.93,
                    "end": 684.07,
                    "word": " is",
                    "probability": 0.89794921875
                },
                {
                    "start": 684.07,
                    "end": 684.23,
                    "word": " in",
                    "probability": 0.90283203125
                },
                {
                    "start": 684.23,
                    "end": 684.43,
                    "word": " state",
                    "probability": 0.85888671875
                },
                {
                    "start": 684.43,
                    "end": 684.75,
                    "word": " s",
                    "probability": 0.990234375
                },
                {
                    "start": 684.75,
                    "end": 684.89,
                    "word": " and",
                    "probability": 0.826171875
                },
                {
                    "start": 684.89,
                    "end": 685.19,
                    "word": " performs",
                    "probability": 0.76025390625
                },
                {
                    "start": 685.19,
                    "end": 685.65,
                    "word": " action",
                    "probability": 0.791015625
                },
                {
                    "start": 685.65,
                    "end": 686.13,
                    "word": " A",
                    "probability": 0.92626953125
                },
                {
                    "start": 686.13,
                    "end": 686.65,
                    "word": ",",
                    "probability": 0.9501953125
                }
            ]
        },
        {
            "id": 101,
            "seek": 70712,
            "start": 686.65,
            "end": 694.99,
            "text": " the Q value of this setting is the expected cumulative reward from taking action A in state",
            "tokens": [
                50364,
                264,
                1249,
                2158,
                295,
                341,
                3287,
                307,
                264,
                5176,
                38379,
                7782,
                490,
                1940,
                3069,
                316,
                294,
                1785,
                50788
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23281250711944368,
            "compression_ratio": 1.8238095238095238,
            "no_speech_prob": 0.0004513263702392578,
            "words": [
                {
                    "start": 686.65,
                    "end": 686.79,
                    "word": " the",
                    "probability": 0.1455078125
                },
                {
                    "start": 686.79,
                    "end": 687.45,
                    "word": " Q",
                    "probability": 0.732421875
                },
                {
                    "start": 687.45,
                    "end": 688.67,
                    "word": " value",
                    "probability": 0.5869140625
                },
                {
                    "start": 688.67,
                    "end": 689.51,
                    "word": " of",
                    "probability": 0.89697265625
                },
                {
                    "start": 689.51,
                    "end": 690.29,
                    "word": " this",
                    "probability": 0.87646484375
                },
                {
                    "start": 690.29,
                    "end": 690.95,
                    "word": " setting",
                    "probability": 0.8408203125
                },
                {
                    "start": 690.95,
                    "end": 691.23,
                    "word": " is",
                    "probability": 0.82861328125
                },
                {
                    "start": 691.23,
                    "end": 691.33,
                    "word": " the",
                    "probability": 0.80078125
                },
                {
                    "start": 691.33,
                    "end": 691.73,
                    "word": " expected",
                    "probability": 0.73974609375
                },
                {
                    "start": 691.73,
                    "end": 692.25,
                    "word": " cumulative",
                    "probability": 0.72314453125
                },
                {
                    "start": 692.25,
                    "end": 692.71,
                    "word": " reward",
                    "probability": 0.84716796875
                },
                {
                    "start": 692.71,
                    "end": 692.97,
                    "word": " from",
                    "probability": 0.798828125
                },
                {
                    "start": 692.97,
                    "end": 693.33,
                    "word": " taking",
                    "probability": 0.8125
                },
                {
                    "start": 693.33,
                    "end": 693.73,
                    "word": " action",
                    "probability": 0.70654296875
                },
                {
                    "start": 693.73,
                    "end": 694.15,
                    "word": " A",
                    "probability": 0.89697265625
                },
                {
                    "start": 694.15,
                    "end": 694.79,
                    "word": " in",
                    "probability": 0.86767578125
                },
                {
                    "start": 694.79,
                    "end": 694.99,
                    "word": " state",
                    "probability": 0.82568359375
                }
            ]
        },
        {
            "id": 102,
            "seek": 70712,
            "start": 694.99,
            "end": 700.53,
            "text": " s and then following the policy. Remember, the policy is a function that tells the agent what",
            "tokens": [
                50788,
                262,
                293,
                550,
                3480,
                264,
                3897,
                13,
                5459,
                11,
                264,
                3897,
                307,
                257,
                2445,
                300,
                5112,
                264,
                9461,
                437,
                51060
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23281250711944368,
            "compression_ratio": 1.8238095238095238,
            "no_speech_prob": 0.0004513263702392578,
            "words": [
                {
                    "start": 694.99,
                    "end": 695.37,
                    "word": " s",
                    "probability": 0.10882568359375
                },
                {
                    "start": 695.37,
                    "end": 695.63,
                    "word": " and",
                    "probability": 0.60791015625
                },
                {
                    "start": 695.63,
                    "end": 695.83,
                    "word": " then",
                    "probability": 0.72607421875
                },
                {
                    "start": 695.83,
                    "end": 696.23,
                    "word": " following",
                    "probability": 0.81591796875
                },
                {
                    "start": 696.23,
                    "end": 696.43,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 696.43,
                    "end": 696.69,
                    "word": " policy",
                    "probability": 0.8134765625
                },
                {
                    "start": 696.69,
                    "end": 696.81,
                    "word": ".",
                    "probability": 0.8876953125
                },
                {
                    "start": 696.81,
                    "end": 696.99,
                    "word": " Remember",
                    "probability": 0.921875
                },
                {
                    "start": 696.99,
                    "end": 697.11,
                    "word": ",",
                    "probability": 0.489990234375
                },
                {
                    "start": 697.11,
                    "end": 697.19,
                    "word": " the",
                    "probability": 0.80224609375
                },
                {
                    "start": 697.19,
                    "end": 697.47,
                    "word": " policy",
                    "probability": 0.79296875
                },
                {
                    "start": 697.47,
                    "end": 697.61,
                    "word": " is",
                    "probability": 0.869140625
                },
                {
                    "start": 697.69,
                    "end": 697.91,
                    "word": " a function",
                    "probability": 0.8779296875
                },
                {
                    "start": 697.91,
                    "end": 698.09,
                    "word": " that",
                    "probability": 0.88134765625
                },
                {
                    "start": 698.09,
                    "end": 698.29,
                    "word": " tells",
                    "probability": 0.7314453125
                },
                {
                    "start": 698.29,
                    "end": 699.59,
                    "word": " the",
                    "probability": 0.6279296875
                },
                {
                    "start": 699.59,
                    "end": 700.01,
                    "word": " agent",
                    "probability": 0.681640625
                },
                {
                    "start": 700.01,
                    "end": 700.53,
                    "word": " what",
                    "probability": 0.85009765625
                }
            ]
        },
        {
            "id": 103,
            "seek": 70712,
            "start": 700.53,
            "end": 705.91,
            "text": " action to perform given any state. When the agent performs action A in state s, it goes into a state",
            "tokens": [
                51060,
                3069,
                281,
                2042,
                2212,
                604,
                1785,
                13,
                1133,
                264,
                9461,
                26213,
                3069,
                316,
                294,
                1785,
                262,
                11,
                309,
                1709,
                666,
                257,
                1785,
                51336
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23281250711944368,
            "compression_ratio": 1.8238095238095238,
            "no_speech_prob": 0.0004513263702392578,
            "words": [
                {
                    "start": 700.53,
                    "end": 700.87,
                    "word": " action",
                    "probability": 0.79833984375
                },
                {
                    "start": 700.87,
                    "end": 701.01,
                    "word": " to",
                    "probability": 0.90234375
                },
                {
                    "start": 701.01,
                    "end": 701.29,
                    "word": " perform",
                    "probability": 0.77734375
                },
                {
                    "start": 701.29,
                    "end": 701.59,
                    "word": " given",
                    "probability": 0.67822265625
                },
                {
                    "start": 701.59,
                    "end": 701.81,
                    "word": " any",
                    "probability": 0.436767578125
                },
                {
                    "start": 701.81,
                    "end": 702.05,
                    "word": " state",
                    "probability": 0.8896484375
                },
                {
                    "start": 702.05,
                    "end": 702.17,
                    "word": ".",
                    "probability": 0.95458984375
                },
                {
                    "start": 702.17,
                    "end": 702.23,
                    "word": " When",
                    "probability": 0.7978515625
                },
                {
                    "start": 702.23,
                    "end": 702.43,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 702.43,
                    "end": 702.67,
                    "word": " agent",
                    "probability": 0.83056640625
                },
                {
                    "start": 702.67,
                    "end": 703.13,
                    "word": " performs",
                    "probability": 0.7333984375
                },
                {
                    "start": 703.13,
                    "end": 703.45,
                    "word": " action",
                    "probability": 0.77099609375
                },
                {
                    "start": 703.45,
                    "end": 703.69,
                    "word": " A",
                    "probability": 0.953125
                },
                {
                    "start": 703.69,
                    "end": 703.85,
                    "word": " in",
                    "probability": 0.88623046875
                },
                {
                    "start": 703.85,
                    "end": 704.01,
                    "word": " state",
                    "probability": 0.88037109375
                },
                {
                    "start": 704.01,
                    "end": 704.47,
                    "word": " s",
                    "probability": 0.95361328125
                },
                {
                    "start": 704.47,
                    "end": 704.87,
                    "word": ",",
                    "probability": 0.93017578125
                },
                {
                    "start": 704.87,
                    "end": 705.05,
                    "word": " it",
                    "probability": 0.904296875
                },
                {
                    "start": 705.05,
                    "end": 705.25,
                    "word": " goes",
                    "probability": 0.85986328125
                },
                {
                    "start": 705.25,
                    "end": 705.45,
                    "word": " into",
                    "probability": 0.72705078125
                },
                {
                    "start": 705.63,
                    "end": 705.91,
                    "word": " a state",
                    "probability": 0.875
                }
            ]
        },
        {
            "id": 104,
            "seek": 70712,
            "start": 705.91,
            "end": 713.25,
            "text": " s prime, or the next state, and then we can use this policy function here to see what the action",
            "tokens": [
                51336,
                262,
                5835,
                11,
                420,
                264,
                958,
                1785,
                11,
                293,
                550,
                321,
                393,
                764,
                341,
                3897,
                2445,
                510,
                281,
                536,
                437,
                264,
                3069,
                51696
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23281250711944368,
            "compression_ratio": 1.8238095238095238,
            "no_speech_prob": 0.0004513263702392578,
            "words": [
                {
                    "start": 705.91,
                    "end": 706.39,
                    "word": " s",
                    "probability": 0.89208984375
                },
                {
                    "start": 706.39,
                    "end": 706.85,
                    "word": " prime",
                    "probability": 0.54833984375
                },
                {
                    "start": 706.85,
                    "end": 706.95,
                    "word": ",",
                    "probability": 0.17236328125
                },
                {
                    "start": 706.95,
                    "end": 707.07,
                    "word": " or",
                    "probability": 0.89990234375
                },
                {
                    "start": 707.07,
                    "end": 707.23,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 707.23,
                    "end": 707.45,
                    "word": " next",
                    "probability": 0.85546875
                },
                {
                    "start": 707.45,
                    "end": 707.85,
                    "word": " state",
                    "probability": 0.87548828125
                },
                {
                    "start": 707.85,
                    "end": 708.33,
                    "word": ",",
                    "probability": 0.495361328125
                },
                {
                    "start": 708.33,
                    "end": 708.75,
                    "word": " and",
                    "probability": 0.88818359375
                },
                {
                    "start": 708.75,
                    "end": 708.99,
                    "word": " then",
                    "probability": 0.720703125
                },
                {
                    "start": 708.99,
                    "end": 709.49,
                    "word": " we",
                    "probability": 0.87548828125
                },
                {
                    "start": 709.49,
                    "end": 709.65,
                    "word": " can",
                    "probability": 0.89404296875
                },
                {
                    "start": 709.65,
                    "end": 709.93,
                    "word": " use",
                    "probability": 0.8134765625
                },
                {
                    "start": 709.93,
                    "end": 710.15,
                    "word": " this",
                    "probability": 0.884765625
                },
                {
                    "start": 710.15,
                    "end": 710.55,
                    "word": " policy",
                    "probability": 0.80322265625
                },
                {
                    "start": 710.55,
                    "end": 710.93,
                    "word": " function",
                    "probability": 0.89013671875
                },
                {
                    "start": 710.93,
                    "end": 711.31,
                    "word": " here",
                    "probability": 0.8046875
                },
                {
                    "start": 711.31,
                    "end": 711.75,
                    "word": " to",
                    "probability": 0.88623046875
                },
                {
                    "start": 711.75,
                    "end": 712.07,
                    "word": " see",
                    "probability": 0.8095703125
                },
                {
                    "start": 712.07,
                    "end": 712.39,
                    "word": " what",
                    "probability": 0.8798828125
                },
                {
                    "start": 712.39,
                    "end": 712.59,
                    "word": " the",
                    "probability": 0.662109375
                },
                {
                    "start": 712.59,
                    "end": 713.25,
                    "word": " action",
                    "probability": 0.81982421875
                }
            ]
        },
        {
            "id": 105,
            "seek": 73710,
            "start": 713.25,
            "end": 718.09,
            "text": " should be, what action should the agent take at that state, which takes it to another state,",
            "tokens": [
                50412,
                820,
                312,
                11,
                437,
                3069,
                820,
                264,
                9461,
                747,
                412,
                300,
                1785,
                11,
                597,
                2516,
                309,
                281,
                1071,
                1785,
                11,
                50600
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2182348858524155,
            "compression_ratio": 1.7892156862745099,
            "no_speech_prob": 0.0156707763671875,
            "words": [
                {
                    "start": 713.25,
                    "end": 713.47,
                    "word": " should",
                    "probability": 0.29296875
                },
                {
                    "start": 713.47,
                    "end": 713.87,
                    "word": " be",
                    "probability": 0.78271484375
                },
                {
                    "start": 713.87,
                    "end": 713.93,
                    "word": ",",
                    "probability": 0.501953125
                },
                {
                    "start": 713.93,
                    "end": 714.05,
                    "word": " what",
                    "probability": 0.833984375
                },
                {
                    "start": 714.05,
                    "end": 714.31,
                    "word": " action",
                    "probability": 0.70458984375
                },
                {
                    "start": 714.31,
                    "end": 714.51,
                    "word": " should",
                    "probability": 0.90185546875
                },
                {
                    "start": 714.51,
                    "end": 714.63,
                    "word": " the",
                    "probability": 0.7890625
                },
                {
                    "start": 714.63,
                    "end": 714.79,
                    "word": " agent",
                    "probability": 0.80029296875
                },
                {
                    "start": 714.79,
                    "end": 715.05,
                    "word": " take",
                    "probability": 0.81298828125
                },
                {
                    "start": 715.05,
                    "end": 715.17,
                    "word": " at",
                    "probability": 0.83544921875
                },
                {
                    "start": 715.17,
                    "end": 715.41,
                    "word": " that",
                    "probability": 0.8701171875
                },
                {
                    "start": 715.41,
                    "end": 715.93,
                    "word": " state",
                    "probability": 0.84423828125
                },
                {
                    "start": 715.93,
                    "end": 716.23,
                    "word": ",",
                    "probability": 0.59912109375
                },
                {
                    "start": 716.23,
                    "end": 716.67,
                    "word": " which",
                    "probability": 0.89990234375
                },
                {
                    "start": 716.67,
                    "end": 717.05,
                    "word": " takes",
                    "probability": 0.71728515625
                },
                {
                    "start": 717.05,
                    "end": 717.15,
                    "word": " it",
                    "probability": 0.869140625
                },
                {
                    "start": 717.15,
                    "end": 717.23,
                    "word": " to",
                    "probability": 0.90185546875
                },
                {
                    "start": 717.23,
                    "end": 717.47,
                    "word": " another",
                    "probability": 0.82275390625
                },
                {
                    "start": 717.47,
                    "end": 717.85,
                    "word": " state",
                    "probability": 0.888671875
                },
                {
                    "start": 717.85,
                    "end": 718.09,
                    "word": ",",
                    "probability": 0.483642578125
                }
            ]
        },
        {
            "id": 106,
            "seek": 73710,
            "start": 718.09,
            "end": 722.07,
            "text": " and then the policy is used to figure out the action to take in that state, and this goes on",
            "tokens": [
                50600,
                293,
                550,
                264,
                3897,
                307,
                1143,
                281,
                2573,
                484,
                264,
                3069,
                281,
                747,
                294,
                300,
                1785,
                11,
                293,
                341,
                1709,
                322,
                50800
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2182348858524155,
            "compression_ratio": 1.7892156862745099,
            "no_speech_prob": 0.0156707763671875,
            "words": [
                {
                    "start": 718.09,
                    "end": 718.27,
                    "word": " and",
                    "probability": 0.8779296875
                },
                {
                    "start": 718.27,
                    "end": 718.39,
                    "word": " then",
                    "probability": 0.71337890625
                },
                {
                    "start": 718.39,
                    "end": 718.55,
                    "word": " the",
                    "probability": 0.7998046875
                },
                {
                    "start": 718.55,
                    "end": 718.79,
                    "word": " policy",
                    "probability": 0.658203125
                },
                {
                    "start": 718.79,
                    "end": 718.93,
                    "word": " is",
                    "probability": 0.8330078125
                },
                {
                    "start": 718.93,
                    "end": 719.13,
                    "word": " used",
                    "probability": 0.79638671875
                },
                {
                    "start": 719.13,
                    "end": 719.55,
                    "word": " to",
                    "probability": 0.89697265625
                },
                {
                    "start": 719.55,
                    "end": 719.97,
                    "word": " figure",
                    "probability": 0.9091796875
                },
                {
                    "start": 719.97,
                    "end": 720.13,
                    "word": " out",
                    "probability": 0.81787109375
                },
                {
                    "start": 720.13,
                    "end": 720.23,
                    "word": " the",
                    "probability": 0.80322265625
                },
                {
                    "start": 720.23,
                    "end": 720.45,
                    "word": " action",
                    "probability": 0.75634765625
                },
                {
                    "start": 720.45,
                    "end": 720.57,
                    "word": " to",
                    "probability": 0.84375
                },
                {
                    "start": 720.57,
                    "end": 720.69,
                    "word": " take",
                    "probability": 0.8115234375
                },
                {
                    "start": 720.69,
                    "end": 720.77,
                    "word": " in",
                    "probability": 0.65380859375
                },
                {
                    "start": 720.77,
                    "end": 720.89,
                    "word": " that",
                    "probability": 0.87841796875
                },
                {
                    "start": 720.89,
                    "end": 721.09,
                    "word": " state",
                    "probability": 0.8798828125
                },
                {
                    "start": 721.09,
                    "end": 721.25,
                    "word": ",",
                    "probability": 0.489501953125
                },
                {
                    "start": 721.25,
                    "end": 721.35,
                    "word": " and",
                    "probability": 0.8935546875
                },
                {
                    "start": 721.35,
                    "end": 721.67,
                    "word": " this",
                    "probability": 0.86572265625
                },
                {
                    "start": 721.67,
                    "end": 721.91,
                    "word": " goes",
                    "probability": 0.86572265625
                },
                {
                    "start": 721.91,
                    "end": 722.07,
                    "word": " on",
                    "probability": 0.90869140625
                }
            ]
        },
        {
            "id": 107,
            "seek": 73710,
            "start": 722.07,
            "end": 727.65,
            "text": " until the termination of the episode or the horizon. Now that we are familiar with the value",
            "tokens": [
                50800,
                1826,
                264,
                1433,
                2486,
                295,
                264,
                3500,
                420,
                264,
                18046,
                13,
                823,
                300,
                321,
                366,
                4963,
                365,
                264,
                2158,
                51080
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2182348858524155,
            "compression_ratio": 1.7892156862745099,
            "no_speech_prob": 0.0156707763671875,
            "words": [
                {
                    "start": 722.07,
                    "end": 722.41,
                    "word": " until",
                    "probability": 0.79443359375
                },
                {
                    "start": 722.41,
                    "end": 723.13,
                    "word": " the",
                    "probability": 0.783203125
                },
                {
                    "start": 723.13,
                    "end": 723.61,
                    "word": " termination",
                    "probability": 0.886474609375
                },
                {
                    "start": 723.61,
                    "end": 724.21,
                    "word": " of",
                    "probability": 0.904296875
                },
                {
                    "start": 724.21,
                    "end": 724.45,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 724.45,
                    "end": 724.81,
                    "word": " episode",
                    "probability": 0.62939453125
                },
                {
                    "start": 724.81,
                    "end": 725.07,
                    "word": " or",
                    "probability": 0.81787109375
                },
                {
                    "start": 725.07,
                    "end": 725.25,
                    "word": " the",
                    "probability": 0.79833984375
                },
                {
                    "start": 725.25,
                    "end": 725.67,
                    "word": " horizon",
                    "probability": 0.7568359375
                },
                {
                    "start": 725.67,
                    "end": 726.15,
                    "word": ".",
                    "probability": 0.986328125
                },
                {
                    "start": 726.35,
                    "end": 726.63,
                    "word": " Now",
                    "probability": 0.88427734375
                },
                {
                    "start": 726.63,
                    "end": 726.73,
                    "word": " that",
                    "probability": 0.86572265625
                },
                {
                    "start": 726.73,
                    "end": 726.81,
                    "word": " we",
                    "probability": 0.68798828125
                },
                {
                    "start": 726.81,
                    "end": 726.87,
                    "word": " are",
                    "probability": 0.51611328125
                },
                {
                    "start": 726.87,
                    "end": 727.13,
                    "word": " familiar",
                    "probability": 0.76416015625
                },
                {
                    "start": 727.13,
                    "end": 727.31,
                    "word": " with",
                    "probability": 0.81201171875
                },
                {
                    "start": 727.31,
                    "end": 727.45,
                    "word": " the",
                    "probability": 0.78466796875
                },
                {
                    "start": 727.45,
                    "end": 727.65,
                    "word": " value",
                    "probability": 0.8505859375
                }
            ]
        },
        {
            "id": 108,
            "seek": 73710,
            "start": 727.65,
            "end": 743.23,
            "text": " function and the Q function, let's go a bit deeper. The policy pi, the value function,",
            "tokens": [
                51080,
                2445,
                293,
                264,
                1249,
                2445,
                11,
                718,
                311,
                352,
                257,
                857,
                7731,
                13,
                440,
                3897,
                3895,
                11,
                264,
                2158,
                2445,
                11,
                51512
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2182348858524155,
            "compression_ratio": 1.7892156862745099,
            "no_speech_prob": 0.0156707763671875,
            "words": [
                {
                    "start": 727.65,
                    "end": 727.93,
                    "word": " function",
                    "probability": 0.85498046875
                },
                {
                    "start": 727.93,
                    "end": 728.29,
                    "word": " and",
                    "probability": 0.8857421875
                },
                {
                    "start": 728.29,
                    "end": 728.59,
                    "word": " the",
                    "probability": 0.68505859375
                },
                {
                    "start": 728.59,
                    "end": 728.75,
                    "word": " Q",
                    "probability": 0.62451171875
                },
                {
                    "start": 728.75,
                    "end": 729.13,
                    "word": " function",
                    "probability": 0.8271484375
                },
                {
                    "start": 729.13,
                    "end": 729.93,
                    "word": ",",
                    "probability": 0.94287109375
                },
                {
                    "start": 729.93,
                    "end": 730.51,
                    "word": " let's",
                    "probability": 0.944580078125
                },
                {
                    "start": 730.51,
                    "end": 731.87,
                    "word": " go",
                    "probability": 0.892578125
                },
                {
                    "start": 732.01,
                    "end": 732.13,
                    "word": " a bit",
                    "probability": 0.90771484375
                },
                {
                    "start": 732.13,
                    "end": 732.61,
                    "word": " deeper",
                    "probability": 0.83642578125
                },
                {
                    "start": 732.61,
                    "end": 733.09,
                    "word": ".",
                    "probability": 0.81982421875
                },
                {
                    "start": 733.63,
                    "end": 733.85,
                    "word": " The",
                    "probability": 0.80810546875
                },
                {
                    "start": 733.85,
                    "end": 734.75,
                    "word": " policy",
                    "probability": 0.78076171875
                },
                {
                    "start": 734.75,
                    "end": 735.19,
                    "word": " pi",
                    "probability": 0.476806640625
                },
                {
                    "start": 735.19,
                    "end": 735.35,
                    "word": ",",
                    "probability": 0.8876953125
                },
                {
                    "start": 735.35,
                    "end": 735.51,
                    "word": " the",
                    "probability": 0.82080078125
                },
                {
                    "start": 735.51,
                    "end": 735.85,
                    "word": " value",
                    "probability": 0.93115234375
                },
                {
                    "start": 735.85,
                    "end": 736.37,
                    "word": " function",
                    "probability": 0.87890625
                },
                {
                    "start": 736.37,
                    "end": 743.23,
                    "word": ",",
                    "probability": 0.6904296875
                }
            ]
        },
        {
            "id": 109,
            "seek": 76150,
            "start": 743.23,
            "end": 749.07,
            "text": " or the reward model are all functions. We want to learn at least one of these from experience.",
            "tokens": [
                50372,
                420,
                264,
                7782,
                2316,
                366,
                439,
                6828,
                13,
                492,
                528,
                281,
                1466,
                412,
                1935,
                472,
                295,
                613,
                490,
                1752,
                13,
                50656
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23619186479684917,
            "compression_ratio": 1.6849315068493151,
            "no_speech_prob": 0.32421875,
            "words": [
                {
                    "start": 743.23,
                    "end": 743.55,
                    "word": " or",
                    "probability": 0.294677734375
                },
                {
                    "start": 743.55,
                    "end": 743.69,
                    "word": " the",
                    "probability": 0.796875
                },
                {
                    "start": 743.69,
                    "end": 743.97,
                    "word": " reward",
                    "probability": 0.83349609375
                },
                {
                    "start": 743.97,
                    "end": 744.31,
                    "word": " model",
                    "probability": 0.8291015625
                },
                {
                    "start": 744.31,
                    "end": 744.67,
                    "word": " are",
                    "probability": 0.59375
                },
                {
                    "start": 744.67,
                    "end": 744.97,
                    "word": " all",
                    "probability": 0.8876953125
                },
                {
                    "start": 744.97,
                    "end": 745.45,
                    "word": " functions",
                    "probability": 0.86572265625
                },
                {
                    "start": 745.45,
                    "end": 746.01,
                    "word": ".",
                    "probability": 0.9482421875
                },
                {
                    "start": 746.21,
                    "end": 746.39,
                    "word": " We",
                    "probability": 0.90185546875
                },
                {
                    "start": 746.39,
                    "end": 746.53,
                    "word": " want",
                    "probability": 0.81884765625
                },
                {
                    "start": 746.53,
                    "end": 746.65,
                    "word": " to",
                    "probability": 0.9033203125
                },
                {
                    "start": 746.65,
                    "end": 746.85,
                    "word": " learn",
                    "probability": 0.9150390625
                },
                {
                    "start": 746.85,
                    "end": 747.05,
                    "word": " at",
                    "probability": 0.9052734375
                },
                {
                    "start": 747.05,
                    "end": 747.25,
                    "word": " least",
                    "probability": 0.859375
                },
                {
                    "start": 747.25,
                    "end": 747.59,
                    "word": " one",
                    "probability": 0.818359375
                },
                {
                    "start": 747.59,
                    "end": 747.71,
                    "word": " of",
                    "probability": 0.89013671875
                },
                {
                    "start": 747.71,
                    "end": 747.89,
                    "word": " these",
                    "probability": 0.70556640625
                },
                {
                    "start": 747.89,
                    "end": 748.09,
                    "word": " from",
                    "probability": 0.82568359375
                },
                {
                    "start": 748.09,
                    "end": 748.69,
                    "word": " experience",
                    "probability": 0.814453125
                },
                {
                    "start": 748.69,
                    "end": 749.07,
                    "word": ".",
                    "probability": 0.923828125
                }
            ]
        },
        {
            "id": 110,
            "seek": 76150,
            "start": 749.07,
            "end": 755.51,
            "text": " This is the essence of RL. If there are too many states, however, we cannot just",
            "tokens": [
                50656,
                639,
                307,
                264,
                12801,
                295,
                497,
                43,
                13,
                759,
                456,
                366,
                886,
                867,
                4368,
                11,
                4461,
                11,
                321,
                2644,
                445,
                50968
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23619186479684917,
            "compression_ratio": 1.6849315068493151,
            "no_speech_prob": 0.32421875,
            "words": [
                {
                    "start": 749.07,
                    "end": 749.37,
                    "word": " This",
                    "probability": 0.81640625
                },
                {
                    "start": 749.37,
                    "end": 749.71,
                    "word": " is",
                    "probability": 0.90380859375
                },
                {
                    "start": 749.71,
                    "end": 750.05,
                    "word": " the",
                    "probability": 0.8173828125
                },
                {
                    "start": 750.05,
                    "end": 750.35,
                    "word": " essence",
                    "probability": 0.78125
                },
                {
                    "start": 750.35,
                    "end": 750.59,
                    "word": " of",
                    "probability": 0.89990234375
                },
                {
                    "start": 750.59,
                    "end": 750.95,
                    "word": " RL",
                    "probability": 0.5621337890625
                },
                {
                    "start": 750.95,
                    "end": 751.39,
                    "word": ".",
                    "probability": 0.98046875
                },
                {
                    "start": 751.39,
                    "end": 751.87,
                    "word": " If",
                    "probability": 0.89794921875
                },
                {
                    "start": 751.87,
                    "end": 751.99,
                    "word": " there",
                    "probability": 0.853515625
                },
                {
                    "start": 751.99,
                    "end": 752.15,
                    "word": " are",
                    "probability": 0.892578125
                },
                {
                    "start": 752.15,
                    "end": 752.49,
                    "word": " too",
                    "probability": 0.8056640625
                },
                {
                    "start": 752.49,
                    "end": 752.73,
                    "word": " many",
                    "probability": 0.80712890625
                },
                {
                    "start": 752.73,
                    "end": 753.31,
                    "word": " states",
                    "probability": 0.84326171875
                },
                {
                    "start": 753.31,
                    "end": 753.65,
                    "word": ",",
                    "probability": 0.865234375
                },
                {
                    "start": 753.65,
                    "end": 754.03,
                    "word": " however",
                    "probability": 0.890625
                },
                {
                    "start": 754.03,
                    "end": 754.63,
                    "word": ",",
                    "probability": 0.95751953125
                },
                {
                    "start": 754.63,
                    "end": 754.87,
                    "word": " we",
                    "probability": 0.90478515625
                },
                {
                    "start": 754.87,
                    "end": 755.05,
                    "word": " cannot",
                    "probability": 0.79345703125
                },
                {
                    "start": 755.05,
                    "end": 755.51,
                    "word": " just",
                    "probability": 0.8935546875
                }
            ]
        },
        {
            "id": 111,
            "seek": 76150,
            "start": 756.67,
            "end": 762.83,
            "text": " tabulate everything and try to experience every state and the corresponding result from that state",
            "tokens": [
                51036,
                4421,
                5256,
                1203,
                293,
                853,
                281,
                1752,
                633,
                1785,
                293,
                264,
                11760,
                1874,
                490,
                300,
                1785,
                51340
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23619186479684917,
            "compression_ratio": 1.6849315068493151,
            "no_speech_prob": 0.32421875,
            "words": [
                {
                    "start": 756.67,
                    "end": 757.23,
                    "word": " tabulate",
                    "probability": 0.937255859375
                },
                {
                    "start": 757.23,
                    "end": 757.59,
                    "word": " everything",
                    "probability": 0.88623046875
                },
                {
                    "start": 757.59,
                    "end": 758.01,
                    "word": " and",
                    "probability": 0.8701171875
                },
                {
                    "start": 758.01,
                    "end": 758.29,
                    "word": " try",
                    "probability": 0.81884765625
                },
                {
                    "start": 758.29,
                    "end": 758.51,
                    "word": " to",
                    "probability": 0.9091796875
                },
                {
                    "start": 758.51,
                    "end": 758.91,
                    "word": " experience",
                    "probability": 0.8251953125
                },
                {
                    "start": 758.91,
                    "end": 759.39,
                    "word": " every",
                    "probability": 0.7451171875
                },
                {
                    "start": 759.39,
                    "end": 759.89,
                    "word": " state",
                    "probability": 0.82958984375
                },
                {
                    "start": 759.89,
                    "end": 760.65,
                    "word": " and",
                    "probability": 0.8330078125
                },
                {
                    "start": 760.65,
                    "end": 760.91,
                    "word": " the",
                    "probability": 0.80126953125
                },
                {
                    "start": 760.91,
                    "end": 761.29,
                    "word": " corresponding",
                    "probability": 0.76708984375
                },
                {
                    "start": 761.29,
                    "end": 762.15,
                    "word": " result",
                    "probability": 0.7998046875
                },
                {
                    "start": 762.15,
                    "end": 762.35,
                    "word": " from",
                    "probability": 0.814453125
                },
                {
                    "start": 762.35,
                    "end": 762.55,
                    "word": " that",
                    "probability": 0.87890625
                },
                {
                    "start": 762.55,
                    "end": 762.83,
                    "word": " state",
                    "probability": 0.8681640625
                }
            ]
        },
        {
            "id": 112,
            "seek": 76150,
            "start": 762.83,
            "end": 767.63,
            "text": " or that state and any action performed, because as the size of the problem, as the state space",
            "tokens": [
                51340,
                420,
                300,
                1785,
                293,
                604,
                3069,
                10332,
                11,
                570,
                382,
                264,
                2744,
                295,
                264,
                1154,
                11,
                382,
                264,
                1785,
                1901,
                51592
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23619186479684917,
            "compression_ratio": 1.6849315068493151,
            "no_speech_prob": 0.32421875,
            "words": [
                {
                    "start": 762.83,
                    "end": 763.01,
                    "word": " or",
                    "probability": 0.67529296875
                },
                {
                    "start": 763.01,
                    "end": 763.27,
                    "word": " that",
                    "probability": 0.87109375
                },
                {
                    "start": 763.27,
                    "end": 763.65,
                    "word": " state",
                    "probability": 0.86962890625
                },
                {
                    "start": 763.65,
                    "end": 763.93,
                    "word": " and",
                    "probability": 0.8505859375
                },
                {
                    "start": 763.93,
                    "end": 764.19,
                    "word": " any",
                    "probability": 0.79345703125
                },
                {
                    "start": 764.19,
                    "end": 764.45,
                    "word": " action",
                    "probability": 0.80859375
                },
                {
                    "start": 764.45,
                    "end": 764.75,
                    "word": " performed",
                    "probability": 0.76953125
                },
                {
                    "start": 764.75,
                    "end": 764.85,
                    "word": ",",
                    "probability": 0.24951171875
                },
                {
                    "start": 764.85,
                    "end": 765.07,
                    "word": " because",
                    "probability": 0.8310546875
                },
                {
                    "start": 765.07,
                    "end": 765.69,
                    "word": " as",
                    "probability": 0.84619140625
                },
                {
                    "start": 765.69,
                    "end": 765.81,
                    "word": " the",
                    "probability": 0.67724609375
                },
                {
                    "start": 765.81,
                    "end": 766.05,
                    "word": " size",
                    "probability": 0.82373046875
                },
                {
                    "start": 766.05,
                    "end": 766.17,
                    "word": " of",
                    "probability": 0.9013671875
                },
                {
                    "start": 766.17,
                    "end": 766.29,
                    "word": " the",
                    "probability": 0.81103515625
                },
                {
                    "start": 766.29,
                    "end": 766.59,
                    "word": " problem",
                    "probability": 0.7734375
                },
                {
                    "start": 766.59,
                    "end": 766.73,
                    "word": ",",
                    "probability": 0.7939453125
                },
                {
                    "start": 766.73,
                    "end": 766.83,
                    "word": " as",
                    "probability": 0.9013671875
                },
                {
                    "start": 766.83,
                    "end": 766.97,
                    "word": " the",
                    "probability": 0.60205078125
                },
                {
                    "start": 766.97,
                    "end": 767.23,
                    "word": " state",
                    "probability": 0.80322265625
                },
                {
                    "start": 767.23,
                    "end": 767.63,
                    "word": " space",
                    "probability": 0.7744140625
                }
            ]
        },
        {
            "id": 113,
            "seek": 79148,
            "start": 767.63,
            "end": 774.07,
            "text": " and the action space increase in settings like, let's say, playing GTA 5 or a self-driving",
            "tokens": [
                50372,
                293,
                264,
                3069,
                1901,
                3488,
                294,
                6257,
                411,
                11,
                718,
                311,
                584,
                11,
                2433,
                35575,
                1025,
                420,
                257,
                2698,
                12,
                47094,
                50684
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22197165440038308,
            "compression_ratio": 1.6370656370656371,
            "no_speech_prob": 0.003765106201171875,
            "words": [
                {
                    "start": 767.63,
                    "end": 768.05,
                    "word": " and",
                    "probability": 0.7021484375
                },
                {
                    "start": 768.05,
                    "end": 768.21,
                    "word": " the",
                    "probability": 0.80322265625
                },
                {
                    "start": 768.21,
                    "end": 768.43,
                    "word": " action",
                    "probability": 0.6953125
                },
                {
                    "start": 768.43,
                    "end": 768.67,
                    "word": " space",
                    "probability": 0.74951171875
                },
                {
                    "start": 768.67,
                    "end": 769.23,
                    "word": " increase",
                    "probability": 0.7177734375
                },
                {
                    "start": 769.23,
                    "end": 769.57,
                    "word": " in",
                    "probability": 0.802734375
                },
                {
                    "start": 769.57,
                    "end": 769.87,
                    "word": " settings",
                    "probability": 0.82763671875
                },
                {
                    "start": 769.87,
                    "end": 770.15,
                    "word": " like",
                    "probability": 0.80224609375
                },
                {
                    "start": 770.15,
                    "end": 770.23,
                    "word": ",",
                    "probability": 0.5126953125
                },
                {
                    "start": 770.23,
                    "end": 770.41,
                    "word": " let's",
                    "probability": 0.932861328125
                },
                {
                    "start": 770.41,
                    "end": 770.51,
                    "word": " say",
                    "probability": 0.89697265625
                },
                {
                    "start": 770.51,
                    "end": 770.55,
                    "word": ",",
                    "probability": 0.6728515625
                },
                {
                    "start": 770.55,
                    "end": 770.83,
                    "word": " playing",
                    "probability": 0.810546875
                },
                {
                    "start": 770.83,
                    "end": 771.63,
                    "word": " GTA",
                    "probability": 0.9267578125
                },
                {
                    "start": 771.63,
                    "end": 772.07,
                    "word": " 5",
                    "probability": 0.40283203125
                },
                {
                    "start": 772.07,
                    "end": 772.49,
                    "word": " or",
                    "probability": 0.6875
                },
                {
                    "start": 773.25,
                    "end": 773.63,
                    "word": " a self",
                    "probability": 0.80078125
                },
                {
                    "start": 773.63,
                    "end": 774.07,
                    "word": "-driving",
                    "probability": 0.867431640625
                }
            ]
        },
        {
            "id": 114,
            "seek": 79148,
            "start": 774.83,
            "end": 781.09,
            "text": " navigation policy in the real world, these state spaces just explode. Dimensionality",
            "tokens": [
                50724,
                17346,
                3897,
                294,
                264,
                957,
                1002,
                11,
                613,
                1785,
                7673,
                445,
                21411,
                13,
                20975,
                3378,
                1860,
                51040
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22197165440038308,
            "compression_ratio": 1.6370656370656371,
            "no_speech_prob": 0.003765106201171875,
            "words": [
                {
                    "start": 774.83,
                    "end": 775.23,
                    "word": " navigation",
                    "probability": 0.7333984375
                },
                {
                    "start": 775.23,
                    "end": 775.89,
                    "word": " policy",
                    "probability": 0.81396484375
                },
                {
                    "start": 775.89,
                    "end": 776.23,
                    "word": " in",
                    "probability": 0.9013671875
                },
                {
                    "start": 776.23,
                    "end": 776.35,
                    "word": " the",
                    "probability": 0.78076171875
                },
                {
                    "start": 776.35,
                    "end": 776.49,
                    "word": " real",
                    "probability": 0.87841796875
                },
                {
                    "start": 776.49,
                    "end": 776.85,
                    "word": " world",
                    "probability": 0.87353515625
                },
                {
                    "start": 776.85,
                    "end": 777.33,
                    "word": ",",
                    "probability": 0.6484375
                },
                {
                    "start": 777.33,
                    "end": 777.53,
                    "word": " these",
                    "probability": 0.6455078125
                },
                {
                    "start": 777.53,
                    "end": 777.97,
                    "word": " state",
                    "probability": 0.56298828125
                },
                {
                    "start": 777.97,
                    "end": 779.17,
                    "word": " spaces",
                    "probability": 0.7890625
                },
                {
                    "start": 779.17,
                    "end": 779.41,
                    "word": " just",
                    "probability": 0.88427734375
                },
                {
                    "start": 779.41,
                    "end": 780.05,
                    "word": " explode",
                    "probability": 0.87646484375
                },
                {
                    "start": 780.05,
                    "end": 780.33,
                    "word": ".",
                    "probability": 0.83349609375
                },
                {
                    "start": 780.45,
                    "end": 781.09,
                    "word": " Dimensionality",
                    "probability": 0.7400716145833334
                }
            ]
        },
        {
            "id": 115,
            "seek": 79148,
            "start": 781.63,
            "end": 786.71,
            "text": " is too high and it's just not feasible to store every possible state. In those cases,",
            "tokens": [
                51068,
                307,
                886,
                1090,
                293,
                309,
                311,
                445,
                406,
                26648,
                281,
                3531,
                633,
                1944,
                1785,
                13,
                682,
                729,
                3331,
                11,
                51312
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22197165440038308,
            "compression_ratio": 1.6370656370656371,
            "no_speech_prob": 0.003765106201171875,
            "words": [
                {
                    "start": 781.63,
                    "end": 781.95,
                    "word": " is",
                    "probability": 0.90087890625
                },
                {
                    "start": 781.95,
                    "end": 782.21,
                    "word": " too",
                    "probability": 0.8056640625
                },
                {
                    "start": 782.21,
                    "end": 782.55,
                    "word": " high",
                    "probability": 0.857421875
                },
                {
                    "start": 782.55,
                    "end": 782.95,
                    "word": " and",
                    "probability": 0.6337890625
                },
                {
                    "start": 782.95,
                    "end": 783.55,
                    "word": " it's",
                    "probability": 0.9423828125
                },
                {
                    "start": 783.55,
                    "end": 784.05,
                    "word": " just",
                    "probability": 0.87841796875
                },
                {
                    "start": 784.05,
                    "end": 784.25,
                    "word": " not",
                    "probability": 0.90576171875
                },
                {
                    "start": 784.25,
                    "end": 784.55,
                    "word": " feasible",
                    "probability": 0.794921875
                },
                {
                    "start": 784.55,
                    "end": 784.87,
                    "word": " to",
                    "probability": 0.9013671875
                },
                {
                    "start": 784.87,
                    "end": 785.13,
                    "word": " store",
                    "probability": 0.81298828125
                },
                {
                    "start": 785.13,
                    "end": 785.43,
                    "word": " every",
                    "probability": 0.74853515625
                },
                {
                    "start": 785.43,
                    "end": 785.71,
                    "word": " possible",
                    "probability": 0.8642578125
                },
                {
                    "start": 785.71,
                    "end": 786.01,
                    "word": " state",
                    "probability": 0.8681640625
                },
                {
                    "start": 786.01,
                    "end": 786.07,
                    "word": ".",
                    "probability": 0.9765625
                },
                {
                    "start": 786.07,
                    "end": 786.15,
                    "word": " In",
                    "probability": 0.8974609375
                },
                {
                    "start": 786.15,
                    "end": 786.41,
                    "word": " those",
                    "probability": 0.8427734375
                },
                {
                    "start": 786.41,
                    "end": 786.69,
                    "word": " cases",
                    "probability": 0.810546875
                },
                {
                    "start": 786.69,
                    "end": 786.71,
                    "word": ",",
                    "probability": 0.8623046875
                }
            ]
        },
        {
            "id": 116,
            "seek": 79148,
            "start": 786.71,
            "end": 791.19,
            "text": " we need to approximate. In general and traditionally, this is called RL with",
            "tokens": [
                51312,
                321,
                643,
                281,
                30874,
                13,
                682,
                2674,
                293,
                19067,
                11,
                341,
                307,
                1219,
                497,
                43,
                365,
                51540
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22197165440038308,
            "compression_ratio": 1.6370656370656371,
            "no_speech_prob": 0.003765106201171875,
            "words": [
                {
                    "start": 786.71,
                    "end": 786.83,
                    "word": " we",
                    "probability": 0.90380859375
                },
                {
                    "start": 786.83,
                    "end": 786.95,
                    "word": " need",
                    "probability": 0.81787109375
                },
                {
                    "start": 786.95,
                    "end": 787.05,
                    "word": " to",
                    "probability": 0.90625
                },
                {
                    "start": 787.05,
                    "end": 787.41,
                    "word": " approximate",
                    "probability": 0.80078125
                },
                {
                    "start": 787.41,
                    "end": 788.05,
                    "word": ".",
                    "probability": 0.9716796875
                },
                {
                    "start": 788.19,
                    "end": 788.35,
                    "word": " In",
                    "probability": 0.91796875
                },
                {
                    "start": 788.35,
                    "end": 788.89,
                    "word": " general",
                    "probability": 0.802734375
                },
                {
                    "start": 788.89,
                    "end": 789.13,
                    "word": " and",
                    "probability": 0.640625
                },
                {
                    "start": 789.13,
                    "end": 789.61,
                    "word": " traditionally",
                    "probability": 0.70166015625
                },
                {
                    "start": 789.61,
                    "end": 789.79,
                    "word": ",",
                    "probability": 0.94384765625
                },
                {
                    "start": 789.79,
                    "end": 789.95,
                    "word": " this",
                    "probability": 0.88232421875
                },
                {
                    "start": 789.95,
                    "end": 790.11,
                    "word": " is",
                    "probability": 0.9091796875
                },
                {
                    "start": 790.11,
                    "end": 790.41,
                    "word": " called",
                    "probability": 0.79638671875
                },
                {
                    "start": 790.41,
                    "end": 790.93,
                    "word": " RL",
                    "probability": 0.944580078125
                },
                {
                    "start": 790.93,
                    "end": 791.19,
                    "word": " with",
                    "probability": 0.72216796875
                }
            ]
        },
        {
            "id": 117,
            "seek": 79148,
            "start": 791.19,
            "end": 797.61,
            "text": " function approximation. If function approximation is done using deep neural networks,",
            "tokens": [
                51540,
                2445,
                28023,
                13,
                759,
                2445,
                28023,
                307,
                1096,
                1228,
                2452,
                18161,
                9590,
                11,
                51828
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22197165440038308,
            "compression_ratio": 1.6370656370656371,
            "no_speech_prob": 0.003765106201171875,
            "words": [
                {
                    "start": 791.19,
                    "end": 791.55,
                    "word": " function",
                    "probability": 0.8486328125
                },
                {
                    "start": 791.55,
                    "end": 792.03,
                    "word": " approximation",
                    "probability": 0.80859375
                },
                {
                    "start": 792.03,
                    "end": 792.67,
                    "word": ".",
                    "probability": 0.99658203125
                },
                {
                    "start": 792.93,
                    "end": 793.27,
                    "word": " If",
                    "probability": 0.900390625
                },
                {
                    "start": 793.27,
                    "end": 793.79,
                    "word": " function",
                    "probability": 0.876953125
                },
                {
                    "start": 793.79,
                    "end": 794.21,
                    "word": " approximation",
                    "probability": 0.81689453125
                },
                {
                    "start": 794.21,
                    "end": 794.97,
                    "word": " is",
                    "probability": 0.90673828125
                },
                {
                    "start": 794.97,
                    "end": 795.45,
                    "word": " done",
                    "probability": 0.8095703125
                },
                {
                    "start": 795.45,
                    "end": 795.77,
                    "word": " using",
                    "probability": 0.82666015625
                },
                {
                    "start": 795.77,
                    "end": 796.19,
                    "word": " deep",
                    "probability": 0.92626953125
                },
                {
                    "start": 796.19,
                    "end": 796.41,
                    "word": " neural",
                    "probability": 0.85546875
                },
                {
                    "start": 796.41,
                    "end": 796.99,
                    "word": " networks",
                    "probability": 0.537109375
                },
                {
                    "start": 796.99,
                    "end": 797.61,
                    "word": ",",
                    "probability": 0.99462890625
                }
            ]
        },
        {
            "id": 118,
            "seek": 81750,
            "start": 797.61,
            "end": 804.45,
            "text": " then we call the training setting or the RL setting deep reinforcement learning. The term",
            "tokens": [
                50364,
                550,
                321,
                818,
                264,
                3097,
                3287,
                420,
                264,
                497,
                43,
                3287,
                2452,
                29280,
                2539,
                13,
                440,
                1433,
                50704
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20694443931182224,
            "compression_ratio": 1.5654008438818565,
            "no_speech_prob": 0.00020992755889892578,
            "words": [
                {
                    "start": 797.61,
                    "end": 797.93,
                    "word": " then",
                    "probability": 0.0210113525390625
                },
                {
                    "start": 797.93,
                    "end": 798.47,
                    "word": " we",
                    "probability": 0.693359375
                },
                {
                    "start": 798.47,
                    "end": 798.83,
                    "word": " call",
                    "probability": 0.7626953125
                },
                {
                    "start": 798.83,
                    "end": 799.57,
                    "word": " the",
                    "probability": 0.78271484375
                },
                {
                    "start": 799.57,
                    "end": 800.39,
                    "word": " training",
                    "probability": 0.82373046875
                },
                {
                    "start": 800.39,
                    "end": 801.19,
                    "word": " setting",
                    "probability": 0.8037109375
                },
                {
                    "start": 801.19,
                    "end": 801.43,
                    "word": " or",
                    "probability": 0.52783203125
                },
                {
                    "start": 801.43,
                    "end": 801.79,
                    "word": " the",
                    "probability": 0.79736328125
                },
                {
                    "start": 801.79,
                    "end": 802.19,
                    "word": " RL",
                    "probability": 0.90576171875
                },
                {
                    "start": 802.19,
                    "end": 802.53,
                    "word": " setting",
                    "probability": 0.82861328125
                },
                {
                    "start": 802.53,
                    "end": 802.93,
                    "word": " deep",
                    "probability": 0.283935546875
                },
                {
                    "start": 802.93,
                    "end": 803.39,
                    "word": " reinforcement",
                    "probability": 0.88720703125
                },
                {
                    "start": 803.39,
                    "end": 803.73,
                    "word": " learning",
                    "probability": 0.80224609375
                },
                {
                    "start": 803.73,
                    "end": 803.95,
                    "word": ".",
                    "probability": 0.72900390625
                },
                {
                    "start": 803.95,
                    "end": 804.27,
                    "word": " The",
                    "probability": 0.74072265625
                },
                {
                    "start": 804.27,
                    "end": 804.45,
                    "word": " term",
                    "probability": 0.78125
                }
            ]
        },
        {
            "id": 119,
            "seek": 81750,
            "start": 804.45,
            "end": 811.45,
            "text": " is relatively new. It was, I believe, introduced in late 2013, early 2014 by me and David Silver",
            "tokens": [
                50704,
                307,
                7226,
                777,
                13,
                467,
                390,
                11,
                286,
                1697,
                11,
                7268,
                294,
                3469,
                9012,
                11,
                2440,
                8227,
                538,
                385,
                293,
                4389,
                15861,
                51060
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20694443931182224,
            "compression_ratio": 1.5654008438818565,
            "no_speech_prob": 0.00020992755889892578,
            "words": [
                {
                    "start": 804.45,
                    "end": 804.59,
                    "word": " is",
                    "probability": 0.89501953125
                },
                {
                    "start": 804.59,
                    "end": 805.05,
                    "word": " relatively",
                    "probability": 0.79345703125
                },
                {
                    "start": 805.05,
                    "end": 805.41,
                    "word": " new",
                    "probability": 0.810546875
                },
                {
                    "start": 805.41,
                    "end": 805.55,
                    "word": ".",
                    "probability": 0.31201171875
                },
                {
                    "start": 805.55,
                    "end": 805.57,
                    "word": " It",
                    "probability": 0.8720703125
                },
                {
                    "start": 805.57,
                    "end": 805.89,
                    "word": " was",
                    "probability": 0.892578125
                },
                {
                    "start": 805.89,
                    "end": 806.11,
                    "word": ",",
                    "probability": 0.476806640625
                },
                {
                    "start": 806.11,
                    "end": 806.15,
                    "word": " I",
                    "probability": 0.99072265625
                },
                {
                    "start": 806.15,
                    "end": 806.33,
                    "word": " believe",
                    "probability": 0.7041015625
                },
                {
                    "start": 806.33,
                    "end": 806.37,
                    "word": ",",
                    "probability": 0.90380859375
                },
                {
                    "start": 806.37,
                    "end": 806.71,
                    "word": " introduced",
                    "probability": 0.75439453125
                },
                {
                    "start": 806.71,
                    "end": 806.87,
                    "word": " in",
                    "probability": 0.892578125
                },
                {
                    "start": 806.87,
                    "end": 807.01,
                    "word": " late",
                    "probability": 0.86865234375
                },
                {
                    "start": 807.01,
                    "end": 807.87,
                    "word": " 2013",
                    "probability": 0.73876953125
                },
                {
                    "start": 807.87,
                    "end": 808.03,
                    "word": ",",
                    "probability": 0.56982421875
                },
                {
                    "start": 808.03,
                    "end": 808.19,
                    "word": " early",
                    "probability": 0.9365234375
                },
                {
                    "start": 808.19,
                    "end": 808.85,
                    "word": " 2014",
                    "probability": 0.74560546875
                },
                {
                    "start": 808.85,
                    "end": 809.41,
                    "word": " by",
                    "probability": 0.64208984375
                },
                {
                    "start": 809.41,
                    "end": 810.55,
                    "word": " me",
                    "probability": 0.78857421875
                },
                {
                    "start": 810.55,
                    "end": 810.91,
                    "word": " and",
                    "probability": 0.79736328125
                },
                {
                    "start": 810.91,
                    "end": 811.13,
                    "word": " David",
                    "probability": 0.82177734375
                },
                {
                    "start": 811.13,
                    "end": 811.45,
                    "word": " Silver",
                    "probability": 0.666015625
                }
            ]
        },
        {
            "id": 120,
            "seek": 81750,
            "start": 811.45,
            "end": 817.19,
            "text": " and others at DeepMind in their DeepQ learning paper. However, the concept is not very new.",
            "tokens": [
                51060,
                293,
                2357,
                412,
                14895,
                44,
                471,
                294,
                641,
                14895,
                48,
                2539,
                3035,
                13,
                2908,
                11,
                264,
                3410,
                307,
                406,
                588,
                777,
                13,
                51336
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20694443931182224,
            "compression_ratio": 1.5654008438818565,
            "no_speech_prob": 0.00020992755889892578,
            "words": [
                {
                    "start": 811.45,
                    "end": 811.77,
                    "word": " and",
                    "probability": 0.80615234375
                },
                {
                    "start": 811.77,
                    "end": 812.09,
                    "word": " others",
                    "probability": 0.73876953125
                },
                {
                    "start": 812.09,
                    "end": 812.33,
                    "word": " at",
                    "probability": 0.8798828125
                },
                {
                    "start": 812.33,
                    "end": 812.93,
                    "word": " DeepMind",
                    "probability": 0.8741861979166666
                },
                {
                    "start": 812.93,
                    "end": 813.21,
                    "word": " in",
                    "probability": 0.73779296875
                },
                {
                    "start": 813.21,
                    "end": 813.37,
                    "word": " their",
                    "probability": 0.833984375
                },
                {
                    "start": 813.37,
                    "end": 814.11,
                    "word": " DeepQ",
                    "probability": 0.322021484375
                },
                {
                    "start": 814.11,
                    "end": 814.35,
                    "word": " learning",
                    "probability": 0.58935546875
                },
                {
                    "start": 814.35,
                    "end": 814.77,
                    "word": " paper",
                    "probability": 0.81591796875
                },
                {
                    "start": 814.77,
                    "end": 815.17,
                    "word": ".",
                    "probability": 0.91845703125
                },
                {
                    "start": 815.17,
                    "end": 815.47,
                    "word": " However",
                    "probability": 0.818359375
                },
                {
                    "start": 815.47,
                    "end": 815.55,
                    "word": ",",
                    "probability": 0.94970703125
                },
                {
                    "start": 815.55,
                    "end": 815.71,
                    "word": " the",
                    "probability": 0.8173828125
                },
                {
                    "start": 815.71,
                    "end": 815.97,
                    "word": " concept",
                    "probability": 0.80615234375
                },
                {
                    "start": 815.97,
                    "end": 816.19,
                    "word": " is",
                    "probability": 0.9052734375
                },
                {
                    "start": 816.19,
                    "end": 816.35,
                    "word": " not",
                    "probability": 0.89697265625
                },
                {
                    "start": 816.35,
                    "end": 816.63,
                    "word": " very",
                    "probability": 0.74072265625
                },
                {
                    "start": 816.63,
                    "end": 816.93,
                    "word": " new",
                    "probability": 0.798828125
                },
                {
                    "start": 816.93,
                    "end": 817.19,
                    "word": ".",
                    "probability": 0.9443359375
                }
            ]
        },
        {
            "id": 121,
            "seek": 81750,
            "start": 817.19,
            "end": 823.63,
            "text": " That said, there have been fantastic and even mind-blowing advances in this area in the past",
            "tokens": [
                51336,
                663,
                848,
                11,
                456,
                362,
                668,
                5456,
                293,
                754,
                1575,
                12,
                43788,
                25297,
                294,
                341,
                1859,
                294,
                264,
                1791,
                51668
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20694443931182224,
            "compression_ratio": 1.5654008438818565,
            "no_speech_prob": 0.00020992755889892578,
            "words": [
                {
                    "start": 817.19,
                    "end": 817.35,
                    "word": " That",
                    "probability": 0.826171875
                },
                {
                    "start": 817.35,
                    "end": 817.55,
                    "word": " said",
                    "probability": 0.88818359375
                },
                {
                    "start": 817.55,
                    "end": 817.59,
                    "word": ",",
                    "probability": 0.89892578125
                },
                {
                    "start": 817.59,
                    "end": 817.69,
                    "word": " there",
                    "probability": 0.85302734375
                },
                {
                    "start": 817.69,
                    "end": 817.91,
                    "word": " have",
                    "probability": 0.79296875
                },
                {
                    "start": 817.91,
                    "end": 818.25,
                    "word": " been",
                    "probability": 0.8271484375
                },
                {
                    "start": 818.25,
                    "end": 819.69,
                    "word": " fantastic",
                    "probability": 0.8701171875
                },
                {
                    "start": 819.69,
                    "end": 820.45,
                    "word": " and",
                    "probability": 0.8662109375
                },
                {
                    "start": 820.45,
                    "end": 821.23,
                    "word": " even",
                    "probability": 0.75
                },
                {
                    "start": 821.23,
                    "end": 821.59,
                    "word": " mind",
                    "probability": 0.7958984375
                },
                {
                    "start": 821.59,
                    "end": 821.83,
                    "word": "-blowing",
                    "probability": 0.771728515625
                },
                {
                    "start": 821.83,
                    "end": 822.31,
                    "word": " advances",
                    "probability": 0.90283203125
                },
                {
                    "start": 822.31,
                    "end": 822.73,
                    "word": " in",
                    "probability": 0.904296875
                },
                {
                    "start": 822.73,
                    "end": 822.89,
                    "word": " this",
                    "probability": 0.86669921875
                },
                {
                    "start": 822.89,
                    "end": 823.09,
                    "word": " area",
                    "probability": 0.8212890625
                },
                {
                    "start": 823.09,
                    "end": 823.25,
                    "word": " in",
                    "probability": 0.89013671875
                },
                {
                    "start": 823.25,
                    "end": 823.37,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 823.37,
                    "end": 823.63,
                    "word": " past",
                    "probability": 0.81494140625
                }
            ]
        },
        {
            "id": 122,
            "seek": 84486,
            "start": 823.63,
            "end": 828.69,
            "text": " three or four years. We'll talk about some of those as we move forward. A quick overview of",
            "tokens": [
                50364,
                1045,
                420,
                1451,
                924,
                13,
                492,
                603,
                751,
                466,
                512,
                295,
                729,
                382,
                321,
                1286,
                2128,
                13,
                316,
                1702,
                12492,
                295,
                50612
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2167553188318902,
            "compression_ratio": 1.5909090909090908,
            "no_speech_prob": 0.0182647705078125,
            "words": [
                {
                    "start": 823.63,
                    "end": 823.91,
                    "word": " three",
                    "probability": 0.2218017578125
                },
                {
                    "start": 823.91,
                    "end": 824.05,
                    "word": " or",
                    "probability": 0.89599609375
                },
                {
                    "start": 824.05,
                    "end": 824.23,
                    "word": " four",
                    "probability": 0.88671875
                },
                {
                    "start": 824.23,
                    "end": 824.51,
                    "word": " years",
                    "probability": 0.87744140625
                },
                {
                    "start": 824.51,
                    "end": 824.61,
                    "word": ".",
                    "probability": 0.5380859375
                },
                {
                    "start": 824.61,
                    "end": 824.65,
                    "word": " We'll",
                    "probability": 0.6937255859375
                },
                {
                    "start": 824.65,
                    "end": 824.83,
                    "word": " talk",
                    "probability": 0.8232421875
                },
                {
                    "start": 824.83,
                    "end": 824.99,
                    "word": " about",
                    "probability": 0.82421875
                },
                {
                    "start": 824.99,
                    "end": 825.15,
                    "word": " some",
                    "probability": 0.78857421875
                },
                {
                    "start": 825.15,
                    "end": 825.25,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 825.25,
                    "end": 825.39,
                    "word": " those",
                    "probability": 0.84375
                },
                {
                    "start": 825.39,
                    "end": 825.53,
                    "word": " as",
                    "probability": 0.884765625
                },
                {
                    "start": 825.53,
                    "end": 825.73,
                    "word": " we",
                    "probability": 0.90380859375
                },
                {
                    "start": 825.73,
                    "end": 826.47,
                    "word": " move",
                    "probability": 0.873046875
                },
                {
                    "start": 826.47,
                    "end": 826.81,
                    "word": " forward",
                    "probability": 0.82470703125
                },
                {
                    "start": 826.81,
                    "end": 827.25,
                    "word": ".",
                    "probability": 0.9736328125
                },
                {
                    "start": 827.25,
                    "end": 827.51,
                    "word": " A",
                    "probability": 0.97998046875
                },
                {
                    "start": 827.51,
                    "end": 827.97,
                    "word": " quick",
                    "probability": 0.9150390625
                },
                {
                    "start": 827.97,
                    "end": 828.53,
                    "word": " overview",
                    "probability": 0.79736328125
                },
                {
                    "start": 828.53,
                    "end": 828.69,
                    "word": " of",
                    "probability": 0.89990234375
                }
            ]
        },
        {
            "id": 123,
            "seek": 84486,
            "start": 828.69,
            "end": 834.25,
            "text": " the taxonomy of different RL approaches and RL agents. Remember, we have value, function, policy,",
            "tokens": [
                50612,
                264,
                3366,
                23423,
                295,
                819,
                497,
                43,
                11587,
                293,
                497,
                43,
                12554,
                13,
                5459,
                11,
                321,
                362,
                2158,
                11,
                2445,
                11,
                3897,
                11,
                50892
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2167553188318902,
            "compression_ratio": 1.5909090909090908,
            "no_speech_prob": 0.0182647705078125,
            "words": [
                {
                    "start": 828.69,
                    "end": 828.79,
                    "word": " the",
                    "probability": 0.7998046875
                },
                {
                    "start": 828.79,
                    "end": 829.29,
                    "word": " taxonomy",
                    "probability": 0.84326171875
                },
                {
                    "start": 829.29,
                    "end": 829.53,
                    "word": " of",
                    "probability": 0.90087890625
                },
                {
                    "start": 829.53,
                    "end": 829.95,
                    "word": " different",
                    "probability": 0.66796875
                },
                {
                    "start": 829.95,
                    "end": 830.25,
                    "word": " RL",
                    "probability": 0.88427734375
                },
                {
                    "start": 830.25,
                    "end": 830.57,
                    "word": " approaches",
                    "probability": 0.6142578125
                },
                {
                    "start": 830.57,
                    "end": 830.81,
                    "word": " and",
                    "probability": 0.88427734375
                },
                {
                    "start": 830.81,
                    "end": 831.11,
                    "word": " RL",
                    "probability": 0.9150390625
                },
                {
                    "start": 831.11,
                    "end": 831.45,
                    "word": " agents",
                    "probability": 0.7236328125
                },
                {
                    "start": 831.45,
                    "end": 831.85,
                    "word": ".",
                    "probability": 0.91943359375
                },
                {
                    "start": 831.85,
                    "end": 832.21,
                    "word": " Remember",
                    "probability": 0.9091796875
                },
                {
                    "start": 832.21,
                    "end": 832.37,
                    "word": ",",
                    "probability": 0.4794921875
                },
                {
                    "start": 832.37,
                    "end": 832.49,
                    "word": " we",
                    "probability": 0.90380859375
                },
                {
                    "start": 832.49,
                    "end": 832.83,
                    "word": " have",
                    "probability": 0.88037109375
                },
                {
                    "start": 832.83,
                    "end": 833.21,
                    "word": " value",
                    "probability": 0.83203125
                },
                {
                    "start": 833.21,
                    "end": 833.33,
                    "word": ",",
                    "probability": 0.3798828125
                },
                {
                    "start": 833.33,
                    "end": 833.63,
                    "word": " function",
                    "probability": 0.88330078125
                },
                {
                    "start": 833.63,
                    "end": 833.79,
                    "word": ",",
                    "probability": 0.9765625
                },
                {
                    "start": 833.79,
                    "end": 834.15,
                    "word": " policy",
                    "probability": 0.82568359375
                },
                {
                    "start": 834.15,
                    "end": 834.25,
                    "word": ",",
                    "probability": 0.65576171875
                }
            ]
        },
        {
            "id": 124,
            "seek": 84486,
            "start": 834.25,
            "end": 839.69,
            "text": " and model. There are different approaches to solving the RL problem. Sometimes we just want to",
            "tokens": [
                50892,
                293,
                2316,
                13,
                821,
                366,
                819,
                11587,
                281,
                12606,
                264,
                497,
                43,
                1154,
                13,
                4803,
                321,
                445,
                528,
                281,
                51148
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2167553188318902,
            "compression_ratio": 1.5909090909090908,
            "no_speech_prob": 0.0182647705078125,
            "words": [
                {
                    "start": 834.25,
                    "end": 834.41,
                    "word": " and",
                    "probability": 0.89990234375
                },
                {
                    "start": 834.41,
                    "end": 834.75,
                    "word": " model",
                    "probability": 0.8037109375
                },
                {
                    "start": 834.75,
                    "end": 835.35,
                    "word": ".",
                    "probability": 0.9619140625
                },
                {
                    "start": 835.35,
                    "end": 835.49,
                    "word": " There",
                    "probability": 0.78857421875
                },
                {
                    "start": 835.49,
                    "end": 835.65,
                    "word": " are",
                    "probability": 0.8876953125
                },
                {
                    "start": 835.65,
                    "end": 836.19,
                    "word": " different",
                    "probability": 0.705078125
                },
                {
                    "start": 836.19,
                    "end": 836.77,
                    "word": " approaches",
                    "probability": 0.7119140625
                },
                {
                    "start": 836.77,
                    "end": 836.99,
                    "word": " to",
                    "probability": 0.90625
                },
                {
                    "start": 836.99,
                    "end": 837.21,
                    "word": " solving",
                    "probability": 0.83837890625
                },
                {
                    "start": 837.21,
                    "end": 837.37,
                    "word": " the",
                    "probability": 0.81787109375
                },
                {
                    "start": 837.37,
                    "end": 837.61,
                    "word": " RL",
                    "probability": 0.955322265625
                },
                {
                    "start": 837.61,
                    "end": 837.93,
                    "word": " problem",
                    "probability": 0.48876953125
                },
                {
                    "start": 837.93,
                    "end": 838.07,
                    "word": ".",
                    "probability": 0.9677734375
                },
                {
                    "start": 838.07,
                    "end": 838.31,
                    "word": " Sometimes",
                    "probability": 0.88916015625
                },
                {
                    "start": 838.31,
                    "end": 838.79,
                    "word": " we",
                    "probability": 0.64501953125
                },
                {
                    "start": 838.79,
                    "end": 838.99,
                    "word": " just",
                    "probability": 0.8857421875
                },
                {
                    "start": 838.99,
                    "end": 839.27,
                    "word": " want",
                    "probability": 0.81494140625
                },
                {
                    "start": 839.27,
                    "end": 839.69,
                    "word": " to",
                    "probability": 0.904296875
                }
            ]
        },
        {
            "id": 125,
            "seek": 84486,
            "start": 840.43,
            "end": 850.99,
            "text": " find a policy directly. The approaches that respond or satisfy that need are called policy-based RL.",
            "tokens": [
                51216,
                915,
                257,
                3897,
                3838,
                13,
                440,
                11587,
                300,
                4196,
                420,
                19319,
                300,
                643,
                366,
                1219,
                3897,
                12,
                6032,
                497,
                43,
                13,
                51700
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2167553188318902,
            "compression_ratio": 1.5909090909090908,
            "no_speech_prob": 0.0182647705078125,
            "words": [
                {
                    "start": 840.43,
                    "end": 841.03,
                    "word": " find",
                    "probability": 0.767578125
                },
                {
                    "start": 841.19,
                    "end": 841.53,
                    "word": " a policy",
                    "probability": 0.8173828125
                },
                {
                    "start": 841.53,
                    "end": 842.03,
                    "word": " directly",
                    "probability": 0.73828125
                },
                {
                    "start": 842.03,
                    "end": 842.63,
                    "word": ".",
                    "probability": 0.97509765625
                },
                {
                    "start": 843.43,
                    "end": 844.03,
                    "word": " The",
                    "probability": 0.477294921875
                },
                {
                    "start": 844.03,
                    "end": 844.39,
                    "word": " approaches",
                    "probability": 0.62353515625
                },
                {
                    "start": 844.39,
                    "end": 844.91,
                    "word": " that",
                    "probability": 0.8798828125
                },
                {
                    "start": 844.91,
                    "end": 846.03,
                    "word": " respond",
                    "probability": 0.734375
                },
                {
                    "start": 846.03,
                    "end": 846.37,
                    "word": " or",
                    "probability": 0.64208984375
                },
                {
                    "start": 846.37,
                    "end": 847.97,
                    "word": " satisfy",
                    "probability": 0.78173828125
                },
                {
                    "start": 847.97,
                    "end": 848.35,
                    "word": " that",
                    "probability": 0.86669921875
                },
                {
                    "start": 848.35,
                    "end": 848.65,
                    "word": " need",
                    "probability": 0.8388671875
                },
                {
                    "start": 848.65,
                    "end": 848.85,
                    "word": " are",
                    "probability": 0.88818359375
                },
                {
                    "start": 848.85,
                    "end": 849.07,
                    "word": " called",
                    "probability": 0.7841796875
                },
                {
                    "start": 849.07,
                    "end": 849.57,
                    "word": " policy",
                    "probability": 0.7421875
                },
                {
                    "start": 849.57,
                    "end": 849.95,
                    "word": "-based",
                    "probability": 0.833984375
                },
                {
                    "start": 849.95,
                    "end": 850.39,
                    "word": " RL",
                    "probability": 0.95556640625
                },
                {
                    "start": 850.39,
                    "end": 850.99,
                    "word": ".",
                    "probability": 0.9736328125
                }
            ]
        },
        {
            "id": 126,
            "seek": 87102,
            "start": 850.99,
            "end": 856.39,
            "text": " Sometimes we want to find a Q function or a V function first and then derive the policy",
            "tokens": [
                50372,
                4803,
                321,
                528,
                281,
                915,
                257,
                1249,
                2445,
                420,
                257,
                691,
                2445,
                700,
                293,
                550,
                28446,
                264,
                3897,
                50648
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19589119928854484,
            "compression_ratio": 2.061946902654867,
            "no_speech_prob": 5.227327346801758e-05,
            "words": [
                {
                    "start": 850.99,
                    "end": 851.55,
                    "word": " Sometimes",
                    "probability": 0.6572265625
                },
                {
                    "start": 851.55,
                    "end": 851.97,
                    "word": " we",
                    "probability": 0.8095703125
                },
                {
                    "start": 851.97,
                    "end": 852.41,
                    "word": " want",
                    "probability": 0.81201171875
                },
                {
                    "start": 852.41,
                    "end": 852.55,
                    "word": " to",
                    "probability": 0.904296875
                },
                {
                    "start": 852.55,
                    "end": 852.85,
                    "word": " find",
                    "probability": 0.80322265625
                },
                {
                    "start": 853.21,
                    "end": 853.59,
                    "word": " a Q",
                    "probability": 0.80078125
                },
                {
                    "start": 853.59,
                    "end": 853.91,
                    "word": " function",
                    "probability": 0.8115234375
                },
                {
                    "start": 853.91,
                    "end": 854.07,
                    "word": " or",
                    "probability": 0.81201171875
                },
                {
                    "start": 854.17,
                    "end": 854.31,
                    "word": " a V",
                    "probability": 0.9453125
                },
                {
                    "start": 854.31,
                    "end": 854.61,
                    "word": " function",
                    "probability": 0.8896484375
                },
                {
                    "start": 854.61,
                    "end": 854.97,
                    "word": " first",
                    "probability": 0.8046875
                },
                {
                    "start": 854.97,
                    "end": 855.25,
                    "word": " and",
                    "probability": 0.6259765625
                },
                {
                    "start": 855.25,
                    "end": 855.43,
                    "word": " then",
                    "probability": 0.7421875
                },
                {
                    "start": 855.43,
                    "end": 855.81,
                    "word": " derive",
                    "probability": 0.81201171875
                },
                {
                    "start": 855.81,
                    "end": 855.95,
                    "word": " the",
                    "probability": 0.599609375
                },
                {
                    "start": 855.95,
                    "end": 856.39,
                    "word": " policy",
                    "probability": 0.7724609375
                }
            ]
        },
        {
            "id": 127,
            "seek": 87102,
            "start": 856.39,
            "end": 862.43,
            "text": " from those functions. These approaches are called value-based. Sometimes we want to first learn the",
            "tokens": [
                50648,
                490,
                729,
                6828,
                13,
                1981,
                11587,
                366,
                1219,
                2158,
                12,
                6032,
                13,
                4803,
                321,
                528,
                281,
                700,
                1466,
                264,
                50932
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19589119928854484,
            "compression_ratio": 2.061946902654867,
            "no_speech_prob": 5.227327346801758e-05,
            "words": [
                {
                    "start": 856.39,
                    "end": 857.05,
                    "word": " from",
                    "probability": 0.81640625
                },
                {
                    "start": 857.05,
                    "end": 857.39,
                    "word": " those",
                    "probability": 0.84619140625
                },
                {
                    "start": 857.39,
                    "end": 857.83,
                    "word": " functions",
                    "probability": 0.8466796875
                },
                {
                    "start": 857.83,
                    "end": 858.37,
                    "word": ".",
                    "probability": 0.93994140625
                },
                {
                    "start": 858.37,
                    "end": 858.65,
                    "word": " These",
                    "probability": 0.9033203125
                },
                {
                    "start": 858.65,
                    "end": 858.99,
                    "word": " approaches",
                    "probability": 0.67529296875
                },
                {
                    "start": 858.99,
                    "end": 859.19,
                    "word": " are",
                    "probability": 0.90087890625
                },
                {
                    "start": 859.19,
                    "end": 859.51,
                    "word": " called",
                    "probability": 0.787109375
                },
                {
                    "start": 859.51,
                    "end": 860.07,
                    "word": " value",
                    "probability": 0.76220703125
                },
                {
                    "start": 860.07,
                    "end": 860.33,
                    "word": "-based",
                    "probability": 0.742431640625
                },
                {
                    "start": 860.33,
                    "end": 860.49,
                    "word": ".",
                    "probability": 0.927734375
                },
                {
                    "start": 860.49,
                    "end": 860.85,
                    "word": " Sometimes",
                    "probability": 0.8857421875
                },
                {
                    "start": 860.85,
                    "end": 861.47,
                    "word": " we",
                    "probability": 0.767578125
                },
                {
                    "start": 861.47,
                    "end": 861.61,
                    "word": " want",
                    "probability": 0.8115234375
                },
                {
                    "start": 861.61,
                    "end": 861.73,
                    "word": " to",
                    "probability": 0.90234375
                },
                {
                    "start": 861.73,
                    "end": 861.91,
                    "word": " first",
                    "probability": 0.8134765625
                },
                {
                    "start": 861.91,
                    "end": 862.25,
                    "word": " learn",
                    "probability": 0.90283203125
                },
                {
                    "start": 862.25,
                    "end": 862.43,
                    "word": " the",
                    "probability": 0.441650390625
                }
            ]
        },
        {
            "id": 128,
            "seek": 87102,
            "start": 862.43,
            "end": 867.23,
            "text": " model of the environment and then solve the problem. These are called model-based approaches.",
            "tokens": [
                50932,
                2316,
                295,
                264,
                2823,
                293,
                550,
                5039,
                264,
                1154,
                13,
                1981,
                366,
                1219,
                2316,
                12,
                6032,
                11587,
                13,
                51168
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19589119928854484,
            "compression_ratio": 2.061946902654867,
            "no_speech_prob": 5.227327346801758e-05,
            "words": [
                {
                    "start": 862.43,
                    "end": 862.79,
                    "word": " model",
                    "probability": 0.8193359375
                },
                {
                    "start": 862.79,
                    "end": 863.07,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 863.07,
                    "end": 863.19,
                    "word": " the",
                    "probability": 0.80322265625
                },
                {
                    "start": 863.19,
                    "end": 863.63,
                    "word": " environment",
                    "probability": 0.76220703125
                },
                {
                    "start": 863.63,
                    "end": 864.11,
                    "word": " and",
                    "probability": 0.8173828125
                },
                {
                    "start": 864.11,
                    "end": 864.39,
                    "word": " then",
                    "probability": 0.73583984375
                },
                {
                    "start": 864.39,
                    "end": 865.07,
                    "word": " solve",
                    "probability": 0.8544921875
                },
                {
                    "start": 865.07,
                    "end": 865.21,
                    "word": " the",
                    "probability": 0.8076171875
                },
                {
                    "start": 865.21,
                    "end": 865.43,
                    "word": " problem",
                    "probability": 0.771484375
                },
                {
                    "start": 865.43,
                    "end": 865.49,
                    "word": ".",
                    "probability": 0.94140625
                },
                {
                    "start": 865.49,
                    "end": 865.61,
                    "word": " These",
                    "probability": 0.900390625
                },
                {
                    "start": 865.61,
                    "end": 865.71,
                    "word": " are",
                    "probability": 0.8984375
                },
                {
                    "start": 865.71,
                    "end": 865.89,
                    "word": " called",
                    "probability": 0.75244140625
                },
                {
                    "start": 865.89,
                    "end": 866.29,
                    "word": " model",
                    "probability": 0.7744140625
                },
                {
                    "start": 866.29,
                    "end": 866.63,
                    "word": "-based",
                    "probability": 0.88134765625
                },
                {
                    "start": 866.63,
                    "end": 867.09,
                    "word": " approaches",
                    "probability": 0.67919921875
                },
                {
                    "start": 867.09,
                    "end": 867.23,
                    "word": ".",
                    "probability": 0.9697265625
                }
            ]
        },
        {
            "id": 129,
            "seek": 87102,
            "start": 867.23,
            "end": 870.62,
            "text": " Sometimes we don't have the model and we don't want to learn the model explicitly,",
            "tokens": [
                51168,
                4803,
                321,
                500,
                380,
                362,
                264,
                2316,
                293,
                321,
                500,
                380,
                528,
                281,
                1466,
                264,
                2316,
                20803,
                11,
                51336
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19589119928854484,
            "compression_ratio": 2.061946902654867,
            "no_speech_prob": 5.227327346801758e-05,
            "words": [
                {
                    "start": 867.23,
                    "end": 867.45,
                    "word": " Sometimes",
                    "probability": 0.89501953125
                },
                {
                    "start": 867.45,
                    "end": 867.75,
                    "word": " we",
                    "probability": 0.8828125
                },
                {
                    "start": 867.75,
                    "end": 867.97,
                    "word": " don't",
                    "probability": 0.9423828125
                },
                {
                    "start": 867.97,
                    "end": 868.31,
                    "word": " have",
                    "probability": 0.8779296875
                },
                {
                    "start": 868.31,
                    "end": 868.47,
                    "word": " the",
                    "probability": 0.80029296875
                },
                {
                    "start": 868.47,
                    "end": 868.63,
                    "word": " model",
                    "probability": 0.80078125
                },
                {
                    "start": 868.63,
                    "end": 868.75,
                    "word": " and",
                    "probability": 0.79150390625
                },
                {
                    "start": 868.75,
                    "end": 868.85,
                    "word": " we",
                    "probability": 0.88623046875
                },
                {
                    "start": 868.85,
                    "end": 869.01,
                    "word": " don't",
                    "probability": 0.946533203125
                },
                {
                    "start": 869.01,
                    "end": 869.23,
                    "word": " want",
                    "probability": 0.822265625
                },
                {
                    "start": 869.23,
                    "end": 869.33,
                    "word": " to",
                    "probability": 0.90234375
                },
                {
                    "start": 869.33,
                    "end": 869.47,
                    "word": " learn",
                    "probability": 0.90771484375
                },
                {
                    "start": 869.47,
                    "end": 869.59,
                    "word": " the",
                    "probability": 0.80859375
                },
                {
                    "start": 869.59,
                    "end": 869.77,
                    "word": " model",
                    "probability": 0.81298828125
                },
                {
                    "start": 869.77,
                    "end": 870.35,
                    "word": " explicitly",
                    "probability": 0.68896484375
                },
                {
                    "start": 870.35,
                    "end": 870.62,
                    "word": ",",
                    "probability": 0.395751953125
                }
            ]
        },
        {
            "id": 130,
            "seek": 87102,
            "start": 871.25,
            "end": 877.15,
            "text": " and those are called model-free. When we're dealing with both value, function, and policy at the same",
            "tokens": [
                51368,
                293,
                729,
                366,
                1219,
                2316,
                12,
                10792,
                13,
                1133,
                321,
                434,
                6260,
                365,
                1293,
                2158,
                11,
                2445,
                11,
                293,
                3897,
                412,
                264,
                912,
                51672
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19589119928854484,
            "compression_ratio": 2.061946902654867,
            "no_speech_prob": 5.227327346801758e-05,
            "words": [
                {
                    "start": 871.25,
                    "end": 871.39,
                    "word": " and",
                    "probability": 0.88720703125
                },
                {
                    "start": 871.39,
                    "end": 871.55,
                    "word": " those",
                    "probability": 0.84423828125
                },
                {
                    "start": 871.55,
                    "end": 871.69,
                    "word": " are",
                    "probability": 0.8984375
                },
                {
                    "start": 871.69,
                    "end": 872.05,
                    "word": " called",
                    "probability": 0.7890625
                },
                {
                    "start": 872.05,
                    "end": 872.45,
                    "word": " model",
                    "probability": 0.76025390625
                },
                {
                    "start": 872.45,
                    "end": 872.77,
                    "word": "-free",
                    "probability": 0.8662109375
                },
                {
                    "start": 872.77,
                    "end": 873.31,
                    "word": ".",
                    "probability": 0.98828125
                },
                {
                    "start": 873.37,
                    "end": 873.71,
                    "word": " When",
                    "probability": 0.81787109375
                },
                {
                    "start": 873.71,
                    "end": 873.95,
                    "word": " we're",
                    "probability": 0.775634765625
                },
                {
                    "start": 873.95,
                    "end": 874.23,
                    "word": " dealing",
                    "probability": 0.79833984375
                },
                {
                    "start": 874.23,
                    "end": 874.79,
                    "word": " with",
                    "probability": 0.81201171875
                },
                {
                    "start": 874.79,
                    "end": 875.49,
                    "word": " both",
                    "probability": 0.81982421875
                },
                {
                    "start": 875.49,
                    "end": 875.91,
                    "word": " value",
                    "probability": 0.88134765625
                },
                {
                    "start": 875.91,
                    "end": 875.95,
                    "word": ",",
                    "probability": 0.00374603271484375
                },
                {
                    "start": 875.95,
                    "end": 876.21,
                    "word": " function",
                    "probability": 0.896484375
                },
                {
                    "start": 876.21,
                    "end": 876.25,
                    "word": ",",
                    "probability": 0.595703125
                },
                {
                    "start": 876.25,
                    "end": 876.37,
                    "word": " and",
                    "probability": 0.892578125
                },
                {
                    "start": 876.37,
                    "end": 876.63,
                    "word": " policy",
                    "probability": 0.8134765625
                },
                {
                    "start": 876.63,
                    "end": 876.79,
                    "word": " at",
                    "probability": 0.89208984375
                },
                {
                    "start": 876.79,
                    "end": 876.93,
                    "word": " the",
                    "probability": 0.80908203125
                },
                {
                    "start": 876.93,
                    "end": 877.15,
                    "word": " same",
                    "probability": 0.8115234375
                }
            ]
        },
        {
            "id": 131,
            "seek": 89578,
            "start": 877.15,
            "end": 881.73,
            "text": " time, in other words, we have one agent learning the value function, another learning the policy,",
            "tokens": [
                50364,
                565,
                11,
                294,
                661,
                2283,
                11,
                321,
                362,
                472,
                9461,
                2539,
                264,
                2158,
                2445,
                11,
                1071,
                2539,
                264,
                3897,
                11,
                50584
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20312500631679659,
            "compression_ratio": 1.6652360515021458,
            "no_speech_prob": 0.0010986328125,
            "words": [
                {
                    "start": 877.15,
                    "end": 877.47,
                    "word": " time",
                    "probability": 0.12017822265625
                },
                {
                    "start": 877.47,
                    "end": 877.85,
                    "word": ",",
                    "probability": 0.2568359375
                },
                {
                    "start": 877.85,
                    "end": 877.97,
                    "word": " in",
                    "probability": 0.8369140625
                },
                {
                    "start": 877.97,
                    "end": 878.13,
                    "word": " other",
                    "probability": 0.78515625
                },
                {
                    "start": 878.13,
                    "end": 878.33,
                    "word": " words",
                    "probability": 0.73388671875
                },
                {
                    "start": 878.33,
                    "end": 878.41,
                    "word": ",",
                    "probability": 0.6484375
                },
                {
                    "start": 878.41,
                    "end": 878.49,
                    "word": " we",
                    "probability": 0.873046875
                },
                {
                    "start": 878.49,
                    "end": 878.69,
                    "word": " have",
                    "probability": 0.880859375
                },
                {
                    "start": 878.69,
                    "end": 879.05,
                    "word": " one",
                    "probability": 0.80908203125
                },
                {
                    "start": 879.05,
                    "end": 879.51,
                    "word": " agent",
                    "probability": 0.78564453125
                },
                {
                    "start": 879.51,
                    "end": 879.95,
                    "word": " learning",
                    "probability": 0.83203125
                },
                {
                    "start": 879.95,
                    "end": 880.13,
                    "word": " the",
                    "probability": 0.78076171875
                },
                {
                    "start": 880.13,
                    "end": 880.37,
                    "word": " value",
                    "probability": 0.87451171875
                },
                {
                    "start": 880.37,
                    "end": 880.63,
                    "word": " function",
                    "probability": 0.8359375
                },
                {
                    "start": 880.63,
                    "end": 880.71,
                    "word": ",",
                    "probability": 0.60888671875
                },
                {
                    "start": 880.71,
                    "end": 880.85,
                    "word": " another",
                    "probability": 0.76513671875
                },
                {
                    "start": 880.85,
                    "end": 881.09,
                    "word": " learning",
                    "probability": 0.7509765625
                },
                {
                    "start": 881.09,
                    "end": 881.25,
                    "word": " the",
                    "probability": 0.8017578125
                },
                {
                    "start": 881.25,
                    "end": 881.55,
                    "word": " policy",
                    "probability": 0.8251953125
                },
                {
                    "start": 881.55,
                    "end": 881.73,
                    "word": ",",
                    "probability": 0.4716796875
                }
            ]
        },
        {
            "id": 132,
            "seek": 89578,
            "start": 881.73,
            "end": 890.91,
            "text": " and then contrasting those with each other in a kind of zero-sum setting. We call those actor-critic",
            "tokens": [
                50584,
                293,
                550,
                8712,
                278,
                729,
                365,
                1184,
                661,
                294,
                257,
                733,
                295,
                4018,
                12,
                82,
                449,
                3287,
                13,
                492,
                818,
                729,
                8747,
                12,
                32255,
                299,
                51052
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20312500631679659,
            "compression_ratio": 1.6652360515021458,
            "no_speech_prob": 0.0010986328125,
            "words": [
                {
                    "start": 881.73,
                    "end": 881.75,
                    "word": " and",
                    "probability": 0.88720703125
                },
                {
                    "start": 881.75,
                    "end": 882.23,
                    "word": " then",
                    "probability": 0.72119140625
                },
                {
                    "start": 882.23,
                    "end": 883.41,
                    "word": " contrasting",
                    "probability": 0.818115234375
                },
                {
                    "start": 883.41,
                    "end": 883.61,
                    "word": " those",
                    "probability": 0.84716796875
                },
                {
                    "start": 883.61,
                    "end": 884.27,
                    "word": " with",
                    "probability": 0.685546875
                },
                {
                    "start": 884.27,
                    "end": 884.45,
                    "word": " each",
                    "probability": 0.65283203125
                },
                {
                    "start": 884.45,
                    "end": 884.67,
                    "word": " other",
                    "probability": 0.79052734375
                },
                {
                    "start": 884.67,
                    "end": 885.67,
                    "word": " in",
                    "probability": 0.87255859375
                },
                {
                    "start": 886.15,
                    "end": 888.65,
                    "word": " a kind",
                    "probability": 0.454833984375
                },
                {
                    "start": 888.65,
                    "end": 888.77,
                    "word": " of",
                    "probability": 0.90478515625
                },
                {
                    "start": 888.77,
                    "end": 889.03,
                    "word": " zero",
                    "probability": 0.7880859375
                },
                {
                    "start": 889.03,
                    "end": 889.27,
                    "word": "-sum",
                    "probability": 0.8675130208333334
                },
                {
                    "start": 889.27,
                    "end": 889.61,
                    "word": " setting",
                    "probability": 0.84619140625
                },
                {
                    "start": 889.61,
                    "end": 889.77,
                    "word": ".",
                    "probability": 0.7109375
                },
                {
                    "start": 889.77,
                    "end": 889.79,
                    "word": " We",
                    "probability": 0.8916015625
                },
                {
                    "start": 889.79,
                    "end": 889.97,
                    "word": " call",
                    "probability": 0.7880859375
                },
                {
                    "start": 889.97,
                    "end": 890.21,
                    "word": " those",
                    "probability": 0.8564453125
                },
                {
                    "start": 890.21,
                    "end": 890.51,
                    "word": " actor",
                    "probability": 0.60791015625
                },
                {
                    "start": 890.51,
                    "end": 890.91,
                    "word": "-critic",
                    "probability": 0.8603515625
                }
            ]
        },
        {
            "id": 133,
            "seek": 89578,
            "start": 891.97,
            "end": 896.43,
            "text": " RL approaches. As you can see, there are different approaches to different settings and different",
            "tokens": [
                51104,
                497,
                43,
                11587,
                13,
                1018,
                291,
                393,
                536,
                11,
                456,
                366,
                819,
                11587,
                281,
                819,
                6257,
                293,
                819,
                51328
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20312500631679659,
            "compression_ratio": 1.6652360515021458,
            "no_speech_prob": 0.0010986328125,
            "words": [
                {
                    "start": 891.97,
                    "end": 892.45,
                    "word": " RL",
                    "probability": 0.889404296875
                },
                {
                    "start": 892.45,
                    "end": 892.93,
                    "word": " approaches",
                    "probability": 0.61474609375
                },
                {
                    "start": 892.93,
                    "end": 893.11,
                    "word": ".",
                    "probability": 0.98486328125
                },
                {
                    "start": 893.11,
                    "end": 893.21,
                    "word": " As",
                    "probability": 0.9052734375
                },
                {
                    "start": 893.21,
                    "end": 893.29,
                    "word": " you",
                    "probability": 0.88037109375
                },
                {
                    "start": 893.29,
                    "end": 893.39,
                    "word": " can",
                    "probability": 0.89794921875
                },
                {
                    "start": 893.39,
                    "end": 893.65,
                    "word": " see",
                    "probability": 0.8154296875
                },
                {
                    "start": 893.65,
                    "end": 893.85,
                    "word": ",",
                    "probability": 0.931640625
                },
                {
                    "start": 893.85,
                    "end": 893.95,
                    "word": " there",
                    "probability": 0.8447265625
                },
                {
                    "start": 893.95,
                    "end": 894.13,
                    "word": " are",
                    "probability": 0.8876953125
                },
                {
                    "start": 894.13,
                    "end": 894.51,
                    "word": " different",
                    "probability": 0.70068359375
                },
                {
                    "start": 894.51,
                    "end": 894.85,
                    "word": " approaches",
                    "probability": 0.703125
                },
                {
                    "start": 894.85,
                    "end": 895.61,
                    "word": " to",
                    "probability": 0.88671875
                },
                {
                    "start": 895.61,
                    "end": 895.85,
                    "word": " different",
                    "probability": 0.6767578125
                },
                {
                    "start": 895.85,
                    "end": 896.11,
                    "word": " settings",
                    "probability": 0.8212890625
                },
                {
                    "start": 896.11,
                    "end": 896.23,
                    "word": " and",
                    "probability": 0.86962890625
                },
                {
                    "start": 896.23,
                    "end": 896.43,
                    "word": " different",
                    "probability": 0.6728515625
                }
            ]
        },
        {
            "id": 134,
            "seek": 89578,
            "start": 896.43,
            "end": 901.91,
            "text": " problems. However, all of those can still be grounded on top of the Markov decision process",
            "tokens": [
                51328,
                2740,
                13,
                2908,
                11,
                439,
                295,
                729,
                393,
                920,
                312,
                23535,
                322,
                1192,
                295,
                264,
                3934,
                5179,
                3537,
                1399,
                51600
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20312500631679659,
            "compression_ratio": 1.6652360515021458,
            "no_speech_prob": 0.0010986328125,
            "words": [
                {
                    "start": 896.43,
                    "end": 896.89,
                    "word": " problems",
                    "probability": 0.703125
                },
                {
                    "start": 896.89,
                    "end": 896.95,
                    "word": ".",
                    "probability": 0.6513671875
                },
                {
                    "start": 896.95,
                    "end": 897.17,
                    "word": " However",
                    "probability": 0.8232421875
                },
                {
                    "start": 897.17,
                    "end": 897.25,
                    "word": ",",
                    "probability": 0.98681640625
                },
                {
                    "start": 897.25,
                    "end": 897.43,
                    "word": " all",
                    "probability": 0.8935546875
                },
                {
                    "start": 897.43,
                    "end": 897.53,
                    "word": " of",
                    "probability": 0.88671875
                },
                {
                    "start": 897.53,
                    "end": 897.81,
                    "word": " those",
                    "probability": 0.8515625
                },
                {
                    "start": 897.81,
                    "end": 898.09,
                    "word": " can",
                    "probability": 0.9033203125
                },
                {
                    "start": 898.09,
                    "end": 898.41,
                    "word": " still",
                    "probability": 0.90234375
                },
                {
                    "start": 898.41,
                    "end": 899.25,
                    "word": " be",
                    "probability": 0.8984375
                },
                {
                    "start": 899.25,
                    "end": 899.99,
                    "word": " grounded",
                    "probability": 0.81396484375
                },
                {
                    "start": 899.99,
                    "end": 900.27,
                    "word": " on",
                    "probability": 0.89990234375
                },
                {
                    "start": 900.27,
                    "end": 900.47,
                    "word": " top",
                    "probability": 0.822265625
                },
                {
                    "start": 900.47,
                    "end": 900.69,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 900.69,
                    "end": 900.87,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 900.87,
                    "end": 901.21,
                    "word": " Markov",
                    "probability": 0.92041015625
                },
                {
                    "start": 901.21,
                    "end": 901.49,
                    "word": " decision",
                    "probability": 0.64013671875
                },
                {
                    "start": 901.49,
                    "end": 901.91,
                    "word": " process",
                    "probability": 0.78369140625
                }
            ]
        },
        {
            "id": 135,
            "seek": 91604,
            "start": 901.91,
            "end": 907.83,
            "text": " framework and the general solution approach we looked at before. One of the better-known",
            "tokens": [
                50364,
                8388,
                293,
                264,
                2674,
                3827,
                3109,
                321,
                2956,
                412,
                949,
                13,
                1485,
                295,
                264,
                1101,
                12,
                6861,
                50656
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23147322152342115,
            "compression_ratio": 1.574585635359116,
            "no_speech_prob": 0.0022869110107421875,
            "words": [
                {
                    "start": 901.91,
                    "end": 902.27,
                    "word": " framework",
                    "probability": 0.61279296875
                },
                {
                    "start": 902.27,
                    "end": 902.99,
                    "word": " and",
                    "probability": 0.74462890625
                },
                {
                    "start": 902.99,
                    "end": 903.21,
                    "word": " the",
                    "probability": 0.802734375
                },
                {
                    "start": 903.21,
                    "end": 903.63,
                    "word": " general",
                    "probability": 0.78369140625
                },
                {
                    "start": 903.63,
                    "end": 904.47,
                    "word": " solution",
                    "probability": 0.81005859375
                },
                {
                    "start": 904.47,
                    "end": 904.77,
                    "word": " approach",
                    "probability": 0.6953125
                },
                {
                    "start": 904.77,
                    "end": 904.87,
                    "word": " we",
                    "probability": 0.86865234375
                },
                {
                    "start": 904.87,
                    "end": 905.01,
                    "word": " looked",
                    "probability": 0.70458984375
                },
                {
                    "start": 905.01,
                    "end": 905.19,
                    "word": " at",
                    "probability": 0.9130859375
                },
                {
                    "start": 905.19,
                    "end": 906.03,
                    "word": " before",
                    "probability": 0.79150390625
                },
                {
                    "start": 906.03,
                    "end": 906.33,
                    "word": ".",
                    "probability": 0.94970703125
                },
                {
                    "start": 906.35,
                    "end": 906.95,
                    "word": " One",
                    "probability": 0.8857421875
                },
                {
                    "start": 906.95,
                    "end": 907.07,
                    "word": " of",
                    "probability": 0.90087890625
                },
                {
                    "start": 907.07,
                    "end": 907.23,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 907.23,
                    "end": 907.51,
                    "word": " better",
                    "probability": 0.8291015625
                },
                {
                    "start": 907.51,
                    "end": 907.83,
                    "word": "-known",
                    "probability": 0.52490234375
                }
            ]
        },
        {
            "id": 136,
            "seek": 91604,
            "start": 907.83,
            "end": 915.09,
            "text": " approaches to RL, which falls under the value-based approaches, is Q-learning. The objective in Q-",
            "tokens": [
                50684,
                11587,
                281,
                497,
                43,
                11,
                597,
                8804,
                833,
                264,
                2158,
                12,
                6032,
                11587,
                11,
                307,
                1249,
                12,
                47204,
                13,
                440,
                10024,
                294,
                1249,
                12,
                51012
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23147322152342115,
            "compression_ratio": 1.574585635359116,
            "no_speech_prob": 0.0022869110107421875,
            "words": [
                {
                    "start": 907.83,
                    "end": 908.79,
                    "word": " approaches",
                    "probability": 0.7080078125
                },
                {
                    "start": 908.79,
                    "end": 909.19,
                    "word": " to",
                    "probability": 0.90771484375
                },
                {
                    "start": 909.19,
                    "end": 909.63,
                    "word": " RL",
                    "probability": 0.93994140625
                },
                {
                    "start": 909.63,
                    "end": 909.77,
                    "word": ",",
                    "probability": 0.50146484375
                },
                {
                    "start": 909.77,
                    "end": 909.85,
                    "word": " which",
                    "probability": 0.91845703125
                },
                {
                    "start": 909.85,
                    "end": 910.07,
                    "word": " falls",
                    "probability": 0.6708984375
                },
                {
                    "start": 910.07,
                    "end": 910.23,
                    "word": " under",
                    "probability": 0.82177734375
                },
                {
                    "start": 910.23,
                    "end": 910.83,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 910.83,
                    "end": 911.37,
                    "word": " value",
                    "probability": 0.90625
                },
                {
                    "start": 911.37,
                    "end": 911.63,
                    "word": "-based",
                    "probability": 0.869140625
                },
                {
                    "start": 911.63,
                    "end": 912.11,
                    "word": " approaches",
                    "probability": 0.6455078125
                },
                {
                    "start": 912.11,
                    "end": 912.29,
                    "word": ",",
                    "probability": 0.8486328125
                },
                {
                    "start": 912.29,
                    "end": 912.45,
                    "word": " is",
                    "probability": 0.90283203125
                },
                {
                    "start": 912.45,
                    "end": 912.67,
                    "word": " Q",
                    "probability": 0.71435546875
                },
                {
                    "start": 912.67,
                    "end": 912.91,
                    "word": "-learning",
                    "probability": 0.765869140625
                },
                {
                    "start": 912.91,
                    "end": 913.47,
                    "word": ".",
                    "probability": 0.984375
                },
                {
                    "start": 913.47,
                    "end": 914.03,
                    "word": " The",
                    "probability": 0.79052734375
                },
                {
                    "start": 914.03,
                    "end": 914.61,
                    "word": " objective",
                    "probability": 0.81298828125
                },
                {
                    "start": 914.61,
                    "end": 914.79,
                    "word": " in",
                    "probability": 0.8876953125
                },
                {
                    "start": 914.79,
                    "end": 914.93,
                    "word": " Q",
                    "probability": 0.98974609375
                },
                {
                    "start": 914.93,
                    "end": 915.09,
                    "word": "-",
                    "probability": 0.984375
                }
            ]
        },
        {
            "id": 137,
            "seek": 91604,
            "start": 915.09,
            "end": 922.17,
            "text": " learning is to derive the optimal policy pi star based on optimal Q. So the optimal Q is one that",
            "tokens": [
                51012,
                2539,
                307,
                281,
                28446,
                264,
                16252,
                3897,
                3895,
                3543,
                2361,
                322,
                16252,
                1249,
                13,
                407,
                264,
                16252,
                1249,
                307,
                472,
                300,
                51384
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23147322152342115,
            "compression_ratio": 1.574585635359116,
            "no_speech_prob": 0.0022869110107421875,
            "words": [
                {
                    "start": 915.09,
                    "end": 915.09,
                    "word": " learning",
                    "probability": 0.0013790130615234375
                },
                {
                    "start": 915.09,
                    "end": 915.29,
                    "word": " is",
                    "probability": 0.89892578125
                },
                {
                    "start": 915.29,
                    "end": 915.41,
                    "word": " to",
                    "probability": 0.90576171875
                },
                {
                    "start": 915.41,
                    "end": 915.73,
                    "word": " derive",
                    "probability": 0.77685546875
                },
                {
                    "start": 915.73,
                    "end": 915.87,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 915.87,
                    "end": 916.17,
                    "word": " optimal",
                    "probability": 0.73828125
                },
                {
                    "start": 916.17,
                    "end": 916.53,
                    "word": " policy",
                    "probability": 0.83935546875
                },
                {
                    "start": 916.53,
                    "end": 916.75,
                    "word": " pi",
                    "probability": 0.7490234375
                },
                {
                    "start": 916.75,
                    "end": 916.99,
                    "word": " star",
                    "probability": 0.3359375
                },
                {
                    "start": 916.99,
                    "end": 917.43,
                    "word": " based",
                    "probability": 0.77783203125
                },
                {
                    "start": 917.43,
                    "end": 917.75,
                    "word": " on",
                    "probability": 0.90380859375
                },
                {
                    "start": 917.75,
                    "end": 918.33,
                    "word": " optimal",
                    "probability": 0.72705078125
                },
                {
                    "start": 918.33,
                    "end": 918.91,
                    "word": " Q",
                    "probability": 0.9423828125
                },
                {
                    "start": 918.91,
                    "end": 919.23,
                    "word": ".",
                    "probability": 0.8916015625
                },
                {
                    "start": 919.23,
                    "end": 919.49,
                    "word": " So",
                    "probability": 0.822265625
                },
                {
                    "start": 919.49,
                    "end": 920.19,
                    "word": " the",
                    "probability": 0.5791015625
                },
                {
                    "start": 920.19,
                    "end": 920.51,
                    "word": " optimal",
                    "probability": 0.73828125
                },
                {
                    "start": 920.51,
                    "end": 921.09,
                    "word": " Q",
                    "probability": 0.9873046875
                },
                {
                    "start": 921.09,
                    "end": 921.77,
                    "word": " is",
                    "probability": 0.9013671875
                },
                {
                    "start": 921.77,
                    "end": 921.97,
                    "word": " one",
                    "probability": 0.82861328125
                },
                {
                    "start": 921.97,
                    "end": 922.17,
                    "word": " that",
                    "probability": 0.884765625
                }
            ]
        },
        {
            "id": 138,
            "seek": 94466,
            "start": 922.17,
            "end": 935.17,
            "text": " maximizes the value of each state and action. And from that point onward, with an iterative",
            "tokens": [
                50368,
                5138,
                5660,
                264,
                2158,
                295,
                1184,
                1785,
                293,
                3069,
                13,
                400,
                490,
                300,
                935,
                322,
                1007,
                11,
                365,
                364,
                17138,
                1166,
                51012
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20965484896702552,
            "compression_ratio": 1.4642857142857142,
            "no_speech_prob": 0.00047278404235839844,
            "words": [
                {
                    "start": 922.17,
                    "end": 923.33,
                    "word": " maximizes",
                    "probability": 0.8671875
                },
                {
                    "start": 923.33,
                    "end": 924.49,
                    "word": " the",
                    "probability": 0.80810546875
                },
                {
                    "start": 924.49,
                    "end": 928.27,
                    "word": " value",
                    "probability": 0.541015625
                },
                {
                    "start": 928.27,
                    "end": 928.65,
                    "word": " of",
                    "probability": 0.90576171875
                },
                {
                    "start": 928.65,
                    "end": 928.89,
                    "word": " each",
                    "probability": 0.85498046875
                },
                {
                    "start": 928.89,
                    "end": 929.21,
                    "word": " state",
                    "probability": 0.89111328125
                },
                {
                    "start": 929.21,
                    "end": 929.45,
                    "word": " and",
                    "probability": 0.86376953125
                },
                {
                    "start": 929.45,
                    "end": 929.93,
                    "word": " action",
                    "probability": 0.79443359375
                },
                {
                    "start": 929.93,
                    "end": 930.69,
                    "word": ".",
                    "probability": 0.54443359375
                },
                {
                    "start": 932.11,
                    "end": 932.23,
                    "word": " And",
                    "probability": 0.763671875
                },
                {
                    "start": 932.23,
                    "end": 933.07,
                    "word": " from",
                    "probability": 0.77978515625
                },
                {
                    "start": 933.07,
                    "end": 933.35,
                    "word": " that",
                    "probability": 0.8818359375
                },
                {
                    "start": 933.35,
                    "end": 933.63,
                    "word": " point",
                    "probability": 0.8515625
                },
                {
                    "start": 933.63,
                    "end": 934.17,
                    "word": " onward",
                    "probability": 0.802734375
                },
                {
                    "start": 934.17,
                    "end": 934.39,
                    "word": ",",
                    "probability": 0.6103515625
                },
                {
                    "start": 934.39,
                    "end": 934.61,
                    "word": " with",
                    "probability": 0.798828125
                },
                {
                    "start": 934.61,
                    "end": 934.79,
                    "word": " an",
                    "probability": 0.912109375
                },
                {
                    "start": 934.79,
                    "end": 935.17,
                    "word": " iterative",
                    "probability": 0.897705078125
                }
            ]
        },
        {
            "id": 139,
            "seek": 94466,
            "start": 935.17,
            "end": 941.89,
            "text": " formulation based on Bellman equations or dynamic programming, it becomes possible to find this",
            "tokens": [
                51012,
                37642,
                2361,
                322,
                11485,
                1601,
                11787,
                420,
                8546,
                9410,
                11,
                309,
                3643,
                1944,
                281,
                915,
                341,
                51360
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20965484896702552,
            "compression_ratio": 1.4642857142857142,
            "no_speech_prob": 0.00047278404235839844,
            "words": [
                {
                    "start": 935.17,
                    "end": 935.61,
                    "word": " formulation",
                    "probability": 0.88818359375
                },
                {
                    "start": 935.61,
                    "end": 935.91,
                    "word": " based",
                    "probability": 0.77490234375
                },
                {
                    "start": 935.91,
                    "end": 936.13,
                    "word": " on",
                    "probability": 0.9072265625
                },
                {
                    "start": 936.13,
                    "end": 936.51,
                    "word": " Bellman",
                    "probability": 0.86181640625
                },
                {
                    "start": 936.51,
                    "end": 937.63,
                    "word": " equations",
                    "probability": 0.217041015625
                },
                {
                    "start": 937.63,
                    "end": 937.95,
                    "word": " or",
                    "probability": 0.77001953125
                },
                {
                    "start": 937.95,
                    "end": 938.81,
                    "word": " dynamic",
                    "probability": 0.67333984375
                },
                {
                    "start": 938.81,
                    "end": 939.31,
                    "word": " programming",
                    "probability": 0.71044921875
                },
                {
                    "start": 939.31,
                    "end": 939.69,
                    "word": ",",
                    "probability": 0.95361328125
                },
                {
                    "start": 939.69,
                    "end": 940.07,
                    "word": " it",
                    "probability": 0.89404296875
                },
                {
                    "start": 940.07,
                    "end": 940.29,
                    "word": " becomes",
                    "probability": 0.82568359375
                },
                {
                    "start": 940.29,
                    "end": 940.81,
                    "word": " possible",
                    "probability": 0.84521484375
                },
                {
                    "start": 940.81,
                    "end": 941.07,
                    "word": " to",
                    "probability": 0.9091796875
                },
                {
                    "start": 941.07,
                    "end": 941.47,
                    "word": " find",
                    "probability": 0.81640625
                },
                {
                    "start": 941.47,
                    "end": 941.89,
                    "word": " this",
                    "probability": 0.83740234375
                }
            ]
        },
        {
            "id": 140,
            "seek": 94466,
            "start": 941.89,
            "end": 950.79,
            "text": " through bootstrapping and iterative re-estimation of the Q value. There are different approaches to",
            "tokens": [
                51360,
                807,
                11450,
                19639,
                3759,
                293,
                17138,
                1166,
                319,
                12,
                377,
                332,
                399,
                295,
                264,
                1249,
                2158,
                13,
                821,
                366,
                819,
                11587,
                281,
                51796
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20965484896702552,
            "compression_ratio": 1.4642857142857142,
            "no_speech_prob": 0.00047278404235839844,
            "words": [
                {
                    "start": 941.89,
                    "end": 942.41,
                    "word": " through",
                    "probability": 0.7705078125
                },
                {
                    "start": 942.41,
                    "end": 943.31,
                    "word": " bootstrapping",
                    "probability": 0.84814453125
                },
                {
                    "start": 943.31,
                    "end": 943.93,
                    "word": " and",
                    "probability": 0.65380859375
                },
                {
                    "start": 943.93,
                    "end": 944.67,
                    "word": " iterative",
                    "probability": 0.90185546875
                },
                {
                    "start": 944.67,
                    "end": 946.59,
                    "word": " re",
                    "probability": 0.9296875
                },
                {
                    "start": 946.59,
                    "end": 947.19,
                    "word": "-estimation",
                    "probability": 0.889892578125
                },
                {
                    "start": 947.19,
                    "end": 947.93,
                    "word": " of",
                    "probability": 0.9013671875
                },
                {
                    "start": 947.93,
                    "end": 948.11,
                    "word": " the",
                    "probability": 0.80712890625
                },
                {
                    "start": 948.11,
                    "end": 948.27,
                    "word": " Q",
                    "probability": 0.88623046875
                },
                {
                    "start": 948.27,
                    "end": 948.63,
                    "word": " value",
                    "probability": 0.49658203125
                },
                {
                    "start": 948.63,
                    "end": 949.39,
                    "word": ".",
                    "probability": 0.99365234375
                },
                {
                    "start": 949.49,
                    "end": 949.71,
                    "word": " There",
                    "probability": 0.791015625
                },
                {
                    "start": 949.71,
                    "end": 949.81,
                    "word": " are",
                    "probability": 0.87890625
                },
                {
                    "start": 949.81,
                    "end": 950.03,
                    "word": " different",
                    "probability": 0.70361328125
                },
                {
                    "start": 950.03,
                    "end": 950.37,
                    "word": " approaches",
                    "probability": 0.70751953125
                },
                {
                    "start": 950.37,
                    "end": 950.79,
                    "word": " to",
                    "probability": 0.9052734375
                }
            ]
        },
        {
            "id": 141,
            "seek": 96904,
            "start": 950.79,
            "end": 958.18,
            "text": " Q-learning in large state spaces or large action spaces where function approximation is required.",
            "tokens": [
                50364,
                1249,
                12,
                47204,
                294,
                2416,
                1785,
                7673,
                420,
                2416,
                3069,
                7673,
                689,
                2445,
                28023,
                307,
                4739,
                13,
                50716
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2899038461538462,
            "compression_ratio": 1.622754491017964,
            "no_speech_prob": 0.0009851455688476562,
            "words": [
                {
                    "start": 950.79,
                    "end": 951.09,
                    "word": " Q",
                    "probability": 0.272705078125
                },
                {
                    "start": 951.09,
                    "end": 951.45,
                    "word": "-learning",
                    "probability": 0.665771484375
                },
                {
                    "start": 951.45,
                    "end": 952.25,
                    "word": " in",
                    "probability": 0.8916015625
                },
                {
                    "start": 952.25,
                    "end": 953.95,
                    "word": " large",
                    "probability": 0.88525390625
                },
                {
                    "start": 953.95,
                    "end": 954.27,
                    "word": " state",
                    "probability": 0.76611328125
                },
                {
                    "start": 954.27,
                    "end": 954.55,
                    "word": " spaces",
                    "probability": 0.84912109375
                },
                {
                    "start": 954.55,
                    "end": 954.73,
                    "word": " or",
                    "probability": 0.82080078125
                },
                {
                    "start": 954.73,
                    "end": 954.97,
                    "word": " large",
                    "probability": 0.94482421875
                },
                {
                    "start": 954.97,
                    "end": 955.29,
                    "word": " action",
                    "probability": 0.7802734375
                },
                {
                    "start": 955.29,
                    "end": 955.73,
                    "word": " spaces",
                    "probability": 0.857421875
                },
                {
                    "start": 955.73,
                    "end": 956.27,
                    "word": " where",
                    "probability": 0.74560546875
                },
                {
                    "start": 956.27,
                    "end": 956.67,
                    "word": " function",
                    "probability": 0.7744140625
                },
                {
                    "start": 956.67,
                    "end": 957.09,
                    "word": " approximation",
                    "probability": 0.78173828125
                },
                {
                    "start": 957.09,
                    "end": 957.37,
                    "word": " is",
                    "probability": 0.90625
                },
                {
                    "start": 957.37,
                    "end": 957.81,
                    "word": " required",
                    "probability": 0.73779296875
                },
                {
                    "start": 957.81,
                    "end": 958.18,
                    "word": ".",
                    "probability": 0.845703125
                }
            ]
        },
        {
            "id": 142,
            "seek": 96904,
            "start": 958.69,
            "end": 965.78,
            "text": " One such approach is to parametrize Q as s, a, and an estimation parameter theta.",
            "tokens": [
                50760,
                1485,
                1270,
                3109,
                307,
                281,
                6220,
                302,
                470,
                1381,
                1249,
                382,
                262,
                11,
                257,
                11,
                293,
                364,
                35701,
                13075,
                9725,
                13,
                51096
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2899038461538462,
            "compression_ratio": 1.622754491017964,
            "no_speech_prob": 0.0009851455688476562,
            "words": [
                {
                    "start": 958.69,
                    "end": 959.09,
                    "word": " One",
                    "probability": 0.88623046875
                },
                {
                    "start": 959.09,
                    "end": 960.07,
                    "word": " such",
                    "probability": 0.6513671875
                },
                {
                    "start": 960.07,
                    "end": 960.35,
                    "word": " approach",
                    "probability": 0.7626953125
                },
                {
                    "start": 960.35,
                    "end": 960.51,
                    "word": " is",
                    "probability": 0.90380859375
                },
                {
                    "start": 960.51,
                    "end": 960.75,
                    "word": " to",
                    "probability": 0.89990234375
                },
                {
                    "start": 960.75,
                    "end": 961.75,
                    "word": " parametrize",
                    "probability": 0.79071044921875
                },
                {
                    "start": 961.75,
                    "end": 962.35,
                    "word": " Q",
                    "probability": 0.966796875
                },
                {
                    "start": 962.35,
                    "end": 962.71,
                    "word": " as",
                    "probability": 0.890625
                },
                {
                    "start": 962.71,
                    "end": 963.03,
                    "word": " s",
                    "probability": 0.1630859375
                },
                {
                    "start": 963.03,
                    "end": 963.07,
                    "word": ",",
                    "probability": 0.6787109375
                },
                {
                    "start": 963.29,
                    "end": 963.43,
                    "word": " a,",
                    "probability": 0.7080078125
                },
                {
                    "start": 963.43,
                    "end": 963.71,
                    "word": " and",
                    "probability": 0.8095703125
                },
                {
                    "start": 963.71,
                    "end": 963.97,
                    "word": " an",
                    "probability": 0.8603515625
                },
                {
                    "start": 963.97,
                    "end": 964.31,
                    "word": " estimation",
                    "probability": 0.767578125
                },
                {
                    "start": 964.31,
                    "end": 964.81,
                    "word": " parameter",
                    "probability": 0.890625
                },
                {
                    "start": 964.81,
                    "end": 965.41,
                    "word": " theta",
                    "probability": 0.42138671875
                },
                {
                    "start": 965.41,
                    "end": 965.78,
                    "word": ".",
                    "probability": 0.4951171875
                }
            ]
        },
        {
            "id": 143,
            "seek": 96904,
            "start": 967.07,
            "end": 975.17,
            "text": " If we solve this parametrized Q function by neural networks, where theta corresponds to the",
            "tokens": [
                51232,
                759,
                321,
                5039,
                341,
                6220,
                302,
                470,
                11312,
                1249,
                2445,
                538,
                18161,
                9590,
                11,
                689,
                9725,
                23249,
                281,
                264,
                51576
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2899038461538462,
            "compression_ratio": 1.622754491017964,
            "no_speech_prob": 0.0009851455688476562,
            "words": [
                {
                    "start": 967.07,
                    "end": 967.81,
                    "word": " If",
                    "probability": 0.28857421875
                },
                {
                    "start": 967.81,
                    "end": 968.55,
                    "word": " we",
                    "probability": 0.623046875
                },
                {
                    "start": 968.55,
                    "end": 968.83,
                    "word": " solve",
                    "probability": 0.791015625
                },
                {
                    "start": 968.83,
                    "end": 969.17,
                    "word": " this",
                    "probability": 0.875
                },
                {
                    "start": 969.17,
                    "end": 970.67,
                    "word": " parametrized",
                    "probability": 0.8743896484375
                },
                {
                    "start": 970.67,
                    "end": 971.21,
                    "word": " Q",
                    "probability": 0.9794921875
                },
                {
                    "start": 971.21,
                    "end": 971.63,
                    "word": " function",
                    "probability": 0.77099609375
                },
                {
                    "start": 971.63,
                    "end": 971.97,
                    "word": " by",
                    "probability": 0.62646484375
                },
                {
                    "start": 971.97,
                    "end": 972.31,
                    "word": " neural",
                    "probability": 0.91943359375
                },
                {
                    "start": 972.31,
                    "end": 972.83,
                    "word": " networks",
                    "probability": 0.66943359375
                },
                {
                    "start": 972.83,
                    "end": 973.53,
                    "word": ",",
                    "probability": 0.432373046875
                },
                {
                    "start": 973.53,
                    "end": 973.95,
                    "word": " where",
                    "probability": 0.89501953125
                },
                {
                    "start": 973.95,
                    "end": 974.55,
                    "word": " theta",
                    "probability": 0.68359375
                },
                {
                    "start": 974.55,
                    "end": 974.93,
                    "word": " corresponds",
                    "probability": 0.81494140625
                },
                {
                    "start": 974.93,
                    "end": 975.11,
                    "word": " to",
                    "probability": 0.8876953125
                },
                {
                    "start": 975.11,
                    "end": 975.17,
                    "word": " the",
                    "probability": 0.77685546875
                }
            ]
        },
        {
            "id": 144,
            "seek": 99412,
            "start": 975.17,
            "end": 983.45,
            "text": " neural network, the approach is called Q-networks. This solution approach was proposed",
            "tokens": [
                50384,
                18161,
                3209,
                11,
                264,
                3109,
                307,
                1219,
                1249,
                12,
                7129,
                18357,
                13,
                639,
                3827,
                3109,
                390,
                10348,
                50772
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2709821296589715,
            "compression_ratio": 1.4719101123595506,
            "no_speech_prob": 0.1688232421875,
            "words": [
                {
                    "start": 975.17,
                    "end": 975.69,
                    "word": " neural",
                    "probability": 0.003368377685546875
                },
                {
                    "start": 975.69,
                    "end": 976.23,
                    "word": " network",
                    "probability": 0.62255859375
                },
                {
                    "start": 976.23,
                    "end": 976.31,
                    "word": ",",
                    "probability": 0.51025390625
                },
                {
                    "start": 976.31,
                    "end": 976.87,
                    "word": " the",
                    "probability": 0.8037109375
                },
                {
                    "start": 976.87,
                    "end": 977.33,
                    "word": " approach",
                    "probability": 0.794921875
                },
                {
                    "start": 977.33,
                    "end": 978.03,
                    "word": " is",
                    "probability": 0.90087890625
                },
                {
                    "start": 978.03,
                    "end": 978.47,
                    "word": " called",
                    "probability": 0.82275390625
                },
                {
                    "start": 978.47,
                    "end": 978.91,
                    "word": " Q",
                    "probability": 0.489013671875
                },
                {
                    "start": 978.91,
                    "end": 979.41,
                    "word": "-networks",
                    "probability": 0.7659505208333334
                },
                {
                    "start": 979.41,
                    "end": 979.85,
                    "word": ".",
                    "probability": 0.96826171875
                },
                {
                    "start": 979.85,
                    "end": 980.57,
                    "word": " This",
                    "probability": 0.7958984375
                },
                {
                    "start": 980.57,
                    "end": 981.69,
                    "word": " solution",
                    "probability": 0.849609375
                },
                {
                    "start": 981.69,
                    "end": 981.95,
                    "word": " approach",
                    "probability": 0.759765625
                },
                {
                    "start": 981.95,
                    "end": 982.29,
                    "word": " was",
                    "probability": 0.9072265625
                },
                {
                    "start": 982.29,
                    "end": 983.45,
                    "word": " proposed",
                    "probability": 0.7861328125
                }
            ]
        },
        {
            "id": 145,
            "seek": 99412,
            "start": 984.95,
            "end": 992.77,
            "text": " very early in 2000s, even earlier, but was perfected in some sense in 2013-2014 by",
            "tokens": [
                50860,
                588,
                2440,
                294,
                8132,
                82,
                11,
                754,
                3071,
                11,
                457,
                390,
                2176,
                292,
                294,
                512,
                2020,
                294,
                9012,
                12,
                2009,
                7271,
                538,
                51256
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2709821296589715,
            "compression_ratio": 1.4719101123595506,
            "no_speech_prob": 0.1688232421875,
            "words": [
                {
                    "start": 984.95,
                    "end": 985.39,
                    "word": " very",
                    "probability": 0.7001953125
                },
                {
                    "start": 985.39,
                    "end": 985.77,
                    "word": " early",
                    "probability": 0.93359375
                },
                {
                    "start": 985.77,
                    "end": 986.11,
                    "word": " in",
                    "probability": 0.86669921875
                },
                {
                    "start": 986.11,
                    "end": 987.03,
                    "word": " 2000s",
                    "probability": 0.69775390625
                },
                {
                    "start": 987.03,
                    "end": 987.03,
                    "word": ",",
                    "probability": 0.71044921875
                },
                {
                    "start": 987.03,
                    "end": 987.15,
                    "word": " even",
                    "probability": 0.7578125
                },
                {
                    "start": 987.15,
                    "end": 987.65,
                    "word": " earlier",
                    "probability": 0.9208984375
                },
                {
                    "start": 987.65,
                    "end": 988.15,
                    "word": ",",
                    "probability": 0.87939453125
                },
                {
                    "start": 988.15,
                    "end": 988.43,
                    "word": " but",
                    "probability": 0.89208984375
                },
                {
                    "start": 988.43,
                    "end": 988.67,
                    "word": " was",
                    "probability": 0.8876953125
                },
                {
                    "start": 988.67,
                    "end": 989.67,
                    "word": " perfected",
                    "probability": 0.924072265625
                },
                {
                    "start": 989.67,
                    "end": 989.95,
                    "word": " in",
                    "probability": 0.7470703125
                },
                {
                    "start": 989.95,
                    "end": 990.21,
                    "word": " some",
                    "probability": 0.77587890625
                },
                {
                    "start": 990.21,
                    "end": 990.53,
                    "word": " sense",
                    "probability": 0.732421875
                },
                {
                    "start": 990.53,
                    "end": 990.81,
                    "word": " in",
                    "probability": 0.75927734375
                },
                {
                    "start": 990.81,
                    "end": 991.59,
                    "word": " 2013",
                    "probability": 0.7783203125
                },
                {
                    "start": 991.59,
                    "end": 992.35,
                    "word": "-2014",
                    "probability": 0.7425130208333334
                },
                {
                    "start": 992.35,
                    "end": 992.77,
                    "word": " by",
                    "probability": 0.8369140625
                }
            ]
        },
        {
            "id": 146,
            "seek": 99412,
            "start": 992.77,
            "end": 1000.25,
            "text": " DeepMind, by David Silver and his team, which resulted in the proposal of deep Q-networks or",
            "tokens": [
                51256,
                14895,
                44,
                471,
                11,
                538,
                4389,
                15861,
                293,
                702,
                1469,
                11,
                597,
                18753,
                294,
                264,
                11494,
                295,
                2452,
                1249,
                12,
                7129,
                18357,
                420,
                51600
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2709821296589715,
            "compression_ratio": 1.4719101123595506,
            "no_speech_prob": 0.1688232421875,
            "words": [
                {
                    "start": 992.77,
                    "end": 993.65,
                    "word": " DeepMind",
                    "probability": 0.87841796875
                },
                {
                    "start": 993.65,
                    "end": 993.79,
                    "word": ",",
                    "probability": 0.466796875
                },
                {
                    "start": 993.79,
                    "end": 994.03,
                    "word": " by",
                    "probability": 0.86083984375
                },
                {
                    "start": 994.03,
                    "end": 994.85,
                    "word": " David",
                    "probability": 0.84228515625
                },
                {
                    "start": 994.85,
                    "end": 995.07,
                    "word": " Silver",
                    "probability": 0.6826171875
                },
                {
                    "start": 995.07,
                    "end": 995.25,
                    "word": " and",
                    "probability": 0.751953125
                },
                {
                    "start": 995.25,
                    "end": 995.39,
                    "word": " his",
                    "probability": 0.8701171875
                },
                {
                    "start": 995.39,
                    "end": 995.67,
                    "word": " team",
                    "probability": 0.91552734375
                },
                {
                    "start": 995.67,
                    "end": 996.35,
                    "word": ",",
                    "probability": 0.86474609375
                },
                {
                    "start": 996.35,
                    "end": 996.65,
                    "word": " which",
                    "probability": 0.91552734375
                },
                {
                    "start": 996.65,
                    "end": 997.05,
                    "word": " resulted",
                    "probability": 0.82421875
                },
                {
                    "start": 997.05,
                    "end": 997.65,
                    "word": " in",
                    "probability": 0.9013671875
                },
                {
                    "start": 997.65,
                    "end": 998.11,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 998.11,
                    "end": 998.51,
                    "word": " proposal",
                    "probability": 0.67822265625
                },
                {
                    "start": 998.51,
                    "end": 999.05,
                    "word": " of",
                    "probability": 0.9052734375
                },
                {
                    "start": 999.05,
                    "end": 999.27,
                    "word": " deep",
                    "probability": 0.296630859375
                },
                {
                    "start": 999.27,
                    "end": 999.49,
                    "word": " Q",
                    "probability": 0.908203125
                },
                {
                    "start": 999.49,
                    "end": 999.81,
                    "word": "-networks",
                    "probability": 0.8958333333333334
                },
                {
                    "start": 999.81,
                    "end": 1000.25,
                    "word": " or",
                    "probability": 0.6669921875
                }
            ]
        },
        {
            "id": 147,
            "seek": 102160,
            "start": 1000.25,
            "end": 1007.27,
            "text": " QNs. So we're talking about deep networks. What does that mean? It means that these networks use",
            "tokens": [
                50364,
                1249,
                45,
                82,
                13,
                407,
                321,
                434,
                1417,
                466,
                2452,
                9590,
                13,
                708,
                775,
                300,
                914,
                30,
                467,
                1355,
                300,
                613,
                9590,
                764,
                50732
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25405844310661413,
            "compression_ratio": 1.5767441860465117,
            "no_speech_prob": 0.07806396484375,
            "words": [
                {
                    "start": 1000.25,
                    "end": 1000.77,
                    "word": " QNs",
                    "probability": 0.5712788899739584
                },
                {
                    "start": 1000.77,
                    "end": 1001.35,
                    "word": ".",
                    "probability": 0.79931640625
                },
                {
                    "start": 1001.35,
                    "end": 1001.69,
                    "word": " So",
                    "probability": 0.615234375
                },
                {
                    "start": 1001.69,
                    "end": 1002.19,
                    "word": " we're",
                    "probability": 0.5408935546875
                },
                {
                    "start": 1002.19,
                    "end": 1002.43,
                    "word": " talking",
                    "probability": 0.75537109375
                },
                {
                    "start": 1002.43,
                    "end": 1002.87,
                    "word": " about",
                    "probability": 0.822265625
                },
                {
                    "start": 1002.87,
                    "end": 1003.47,
                    "word": " deep",
                    "probability": 0.75537109375
                },
                {
                    "start": 1003.47,
                    "end": 1004.03,
                    "word": " networks",
                    "probability": 0.6845703125
                },
                {
                    "start": 1004.03,
                    "end": 1004.39,
                    "word": ".",
                    "probability": 0.79931640625
                },
                {
                    "start": 1004.39,
                    "end": 1004.49,
                    "word": " What",
                    "probability": 0.7880859375
                },
                {
                    "start": 1004.49,
                    "end": 1004.61,
                    "word": " does",
                    "probability": 0.92236328125
                },
                {
                    "start": 1004.61,
                    "end": 1004.79,
                    "word": " that",
                    "probability": 0.8828125
                },
                {
                    "start": 1004.79,
                    "end": 1004.91,
                    "word": " mean",
                    "probability": 0.92724609375
                },
                {
                    "start": 1004.91,
                    "end": 1004.99,
                    "word": "?",
                    "probability": 0.9697265625
                },
                {
                    "start": 1004.99,
                    "end": 1005.07,
                    "word": " It",
                    "probability": 0.8740234375
                },
                {
                    "start": 1005.07,
                    "end": 1005.25,
                    "word": " means",
                    "probability": 0.876953125
                },
                {
                    "start": 1005.25,
                    "end": 1005.45,
                    "word": " that",
                    "probability": 0.86572265625
                },
                {
                    "start": 1005.45,
                    "end": 1005.77,
                    "word": " these",
                    "probability": 0.73193359375
                },
                {
                    "start": 1005.77,
                    "end": 1006.39,
                    "word": " networks",
                    "probability": 0.66455078125
                },
                {
                    "start": 1006.39,
                    "end": 1007.27,
                    "word": " use",
                    "probability": 0.78857421875
                }
            ]
        },
        {
            "id": 148,
            "seek": 102160,
            "start": 1007.27,
            "end": 1013.01,
            "text": " deep neural networks, like CNNs, which help with both function approximation",
            "tokens": [
                50732,
                2452,
                18161,
                9590,
                11,
                411,
                24859,
                82,
                11,
                597,
                854,
                365,
                1293,
                2445,
                28023,
                51012
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25405844310661413,
            "compression_ratio": 1.5767441860465117,
            "no_speech_prob": 0.07806396484375,
            "words": [
                {
                    "start": 1007.27,
                    "end": 1007.99,
                    "word": " deep",
                    "probability": 0.92822265625
                },
                {
                    "start": 1007.99,
                    "end": 1008.53,
                    "word": " neural",
                    "probability": 0.8779296875
                },
                {
                    "start": 1008.53,
                    "end": 1009.07,
                    "word": " networks",
                    "probability": 0.66748046875
                },
                {
                    "start": 1009.07,
                    "end": 1009.55,
                    "word": ",",
                    "probability": 0.27001953125
                },
                {
                    "start": 1009.55,
                    "end": 1009.83,
                    "word": " like",
                    "probability": 0.8818359375
                },
                {
                    "start": 1009.83,
                    "end": 1010.65,
                    "word": " CNNs",
                    "probability": 0.943603515625
                },
                {
                    "start": 1010.65,
                    "end": 1010.79,
                    "word": ",",
                    "probability": 0.8759765625
                },
                {
                    "start": 1010.79,
                    "end": 1011.09,
                    "word": " which",
                    "probability": 0.9111328125
                },
                {
                    "start": 1011.09,
                    "end": 1011.41,
                    "word": " help",
                    "probability": 0.8056640625
                },
                {
                    "start": 1011.41,
                    "end": 1011.69,
                    "word": " with",
                    "probability": 0.8212890625
                },
                {
                    "start": 1011.69,
                    "end": 1012.05,
                    "word": " both",
                    "probability": 0.82958984375
                },
                {
                    "start": 1012.05,
                    "end": 1012.49,
                    "word": " function",
                    "probability": 0.8583984375
                },
                {
                    "start": 1012.49,
                    "end": 1013.01,
                    "word": " approximation",
                    "probability": 0.7626953125
                }
            ]
        },
        {
            "id": 149,
            "seek": 102160,
            "start": 1013.69,
            "end": 1021.81,
            "text": " and also end-to-end feature learning. One of the advantages of deep learning is its",
            "tokens": [
                51040,
                293,
                611,
                917,
                12,
                1353,
                12,
                521,
                4111,
                2539,
                13,
                1485,
                295,
                264,
                14906,
                295,
                2452,
                2539,
                307,
                1080,
                51456
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25405844310661413,
            "compression_ratio": 1.5767441860465117,
            "no_speech_prob": 0.07806396484375,
            "words": [
                {
                    "start": 1013.69,
                    "end": 1014.05,
                    "word": " and",
                    "probability": 0.73193359375
                },
                {
                    "start": 1014.05,
                    "end": 1014.45,
                    "word": " also",
                    "probability": 0.7919921875
                },
                {
                    "start": 1014.45,
                    "end": 1014.75,
                    "word": " end",
                    "probability": 0.8017578125
                },
                {
                    "start": 1014.75,
                    "end": 1014.89,
                    "word": "-to",
                    "probability": 0.89697265625
                },
                {
                    "start": 1014.89,
                    "end": 1015.01,
                    "word": "-end",
                    "probability": 0.9541015625
                },
                {
                    "start": 1015.01,
                    "end": 1015.27,
                    "word": " feature",
                    "probability": 0.8525390625
                },
                {
                    "start": 1015.27,
                    "end": 1015.49,
                    "word": " learning",
                    "probability": 0.8369140625
                },
                {
                    "start": 1015.49,
                    "end": 1015.61,
                    "word": ".",
                    "probability": 0.654296875
                },
                {
                    "start": 1015.61,
                    "end": 1015.65,
                    "word": " One",
                    "probability": 0.8876953125
                },
                {
                    "start": 1015.65,
                    "end": 1015.75,
                    "word": " of",
                    "probability": 0.8974609375
                },
                {
                    "start": 1015.75,
                    "end": 1015.83,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 1015.83,
                    "end": 1016.29,
                    "word": " advantages",
                    "probability": 0.859375
                },
                {
                    "start": 1016.29,
                    "end": 1017.59,
                    "word": " of",
                    "probability": 0.9013671875
                },
                {
                    "start": 1017.59,
                    "end": 1019.87,
                    "word": " deep",
                    "probability": 0.63525390625
                },
                {
                    "start": 1019.87,
                    "end": 1020.45,
                    "word": " learning",
                    "probability": 0.83349609375
                },
                {
                    "start": 1020.45,
                    "end": 1021.01,
                    "word": " is",
                    "probability": 0.89599609375
                },
                {
                    "start": 1021.01,
                    "end": 1021.81,
                    "word": " its",
                    "probability": 0.7568359375
                }
            ]
        },
        {
            "id": 150,
            "seek": 102160,
            "start": 1021.81,
            "end": 1027.73,
            "text": " superior performance in learning feature representations, especially from images.",
            "tokens": [
                51456,
                13028,
                3389,
                294,
                2539,
                4111,
                33358,
                11,
                2318,
                490,
                5267,
                13,
                51724
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25405844310661413,
            "compression_ratio": 1.5767441860465117,
            "no_speech_prob": 0.07806396484375,
            "words": [
                {
                    "start": 1021.81,
                    "end": 1022.55,
                    "word": " superior",
                    "probability": 0.919921875
                },
                {
                    "start": 1022.55,
                    "end": 1023.07,
                    "word": " performance",
                    "probability": 0.82568359375
                },
                {
                    "start": 1023.07,
                    "end": 1023.57,
                    "word": " in",
                    "probability": 0.90185546875
                },
                {
                    "start": 1023.57,
                    "end": 1025.41,
                    "word": " learning",
                    "probability": 0.8701171875
                },
                {
                    "start": 1025.41,
                    "end": 1025.71,
                    "word": " feature",
                    "probability": 0.767578125
                },
                {
                    "start": 1025.71,
                    "end": 1026.41,
                    "word": " representations",
                    "probability": 0.448486328125
                },
                {
                    "start": 1026.41,
                    "end": 1026.59,
                    "word": ",",
                    "probability": 0.9091796875
                },
                {
                    "start": 1026.59,
                    "end": 1026.79,
                    "word": " especially",
                    "probability": 0.7001953125
                },
                {
                    "start": 1026.79,
                    "end": 1027.01,
                    "word": " from",
                    "probability": 0.82421875
                },
                {
                    "start": 1027.01,
                    "end": 1027.37,
                    "word": " images",
                    "probability": 0.845703125
                },
                {
                    "start": 1027.37,
                    "end": 1027.73,
                    "word": ".",
                    "probability": 0.998046875
                }
            ]
        },
        {
            "id": 151,
            "seek": 105150,
            "start": 1028.63,
            "end": 1035.43,
            "text": " Now, for those of you who are more statistically oriented, you may have seen the term IID.",
            "tokens": [
                50412,
                823,
                11,
                337,
                729,
                295,
                291,
                567,
                366,
                544,
                36478,
                21841,
                11,
                291,
                815,
                362,
                1612,
                264,
                1433,
                286,
                2777,
                13,
                50712
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22716346555031264,
            "compression_ratio": 1.6307692307692307,
            "no_speech_prob": 0.0001767873764038086,
            "words": [
                {
                    "start": 1028.63,
                    "end": 1028.99,
                    "word": " Now",
                    "probability": 0.453369140625
                },
                {
                    "start": 1028.99,
                    "end": 1029.61,
                    "word": ",",
                    "probability": 0.708984375
                },
                {
                    "start": 1029.61,
                    "end": 1030.07,
                    "word": " for",
                    "probability": 0.87451171875
                },
                {
                    "start": 1030.07,
                    "end": 1030.27,
                    "word": " those",
                    "probability": 0.8642578125
                },
                {
                    "start": 1030.27,
                    "end": 1030.37,
                    "word": " of",
                    "probability": 0.89892578125
                },
                {
                    "start": 1030.37,
                    "end": 1030.47,
                    "word": " you",
                    "probability": 0.89453125
                },
                {
                    "start": 1030.47,
                    "end": 1030.63,
                    "word": " who",
                    "probability": 0.81103515625
                },
                {
                    "start": 1030.63,
                    "end": 1030.83,
                    "word": " are",
                    "probability": 0.9013671875
                },
                {
                    "start": 1030.83,
                    "end": 1031.19,
                    "word": " more",
                    "probability": 0.8740234375
                },
                {
                    "start": 1031.19,
                    "end": 1031.65,
                    "word": " statistically",
                    "probability": 0.73095703125
                },
                {
                    "start": 1031.65,
                    "end": 1032.19,
                    "word": " oriented",
                    "probability": 0.61279296875
                },
                {
                    "start": 1032.19,
                    "end": 1032.41,
                    "word": ",",
                    "probability": 0.81298828125
                },
                {
                    "start": 1032.41,
                    "end": 1032.47,
                    "word": " you",
                    "probability": 0.892578125
                },
                {
                    "start": 1032.47,
                    "end": 1032.85,
                    "word": " may",
                    "probability": 0.8876953125
                },
                {
                    "start": 1032.85,
                    "end": 1033.19,
                    "word": " have",
                    "probability": 0.88720703125
                },
                {
                    "start": 1033.19,
                    "end": 1033.61,
                    "word": " seen",
                    "probability": 0.8798828125
                },
                {
                    "start": 1033.61,
                    "end": 1033.79,
                    "word": " the",
                    "probability": 0.8193359375
                },
                {
                    "start": 1033.79,
                    "end": 1034.05,
                    "word": " term",
                    "probability": 0.81201171875
                },
                {
                    "start": 1034.05,
                    "end": 1034.71,
                    "word": " IID",
                    "probability": 0.787353515625
                },
                {
                    "start": 1034.71,
                    "end": 1035.43,
                    "word": ".",
                    "probability": 0.8505859375
                }
            ]
        },
        {
            "id": 152,
            "seek": 105150,
            "start": 1035.73,
            "end": 1042.07,
            "text": " It means that the data are independent and identically",
            "tokens": [
                50752,
                467,
                1355,
                300,
                264,
                1412,
                366,
                6695,
                293,
                2473,
                984,
                51084
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22716346555031264,
            "compression_ratio": 1.6307692307692307,
            "no_speech_prob": 0.0001767873764038086,
            "words": [
                {
                    "start": 1035.73,
                    "end": 1036.11,
                    "word": " It",
                    "probability": 0.892578125
                },
                {
                    "start": 1036.11,
                    "end": 1036.29,
                    "word": " means",
                    "probability": 0.87451171875
                },
                {
                    "start": 1036.29,
                    "end": 1036.49,
                    "word": " that",
                    "probability": 0.8720703125
                },
                {
                    "start": 1036.49,
                    "end": 1036.81,
                    "word": " the",
                    "probability": 0.80029296875
                },
                {
                    "start": 1036.81,
                    "end": 1037.45,
                    "word": " data",
                    "probability": 0.8125
                },
                {
                    "start": 1037.45,
                    "end": 1038.33,
                    "word": " are",
                    "probability": 0.8525390625
                },
                {
                    "start": 1038.33,
                    "end": 1039.75,
                    "word": " independent",
                    "probability": 0.044525146484375
                },
                {
                    "start": 1039.75,
                    "end": 1041.59,
                    "word": " and",
                    "probability": 0.86474609375
                },
                {
                    "start": 1041.59,
                    "end": 1042.07,
                    "word": " identically",
                    "probability": 0.91845703125
                }
            ]
        },
        {
            "id": 153,
            "seek": 105150,
            "start": 1042.07,
            "end": 1050.77,
            "text": " distributed. It means that one data point does not depend on another data point, and also",
            "tokens": [
                51084,
                12631,
                13,
                467,
                1355,
                300,
                472,
                1412,
                935,
                775,
                406,
                5672,
                322,
                1071,
                1412,
                935,
                11,
                293,
                611,
                51512
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22716346555031264,
            "compression_ratio": 1.6307692307692307,
            "no_speech_prob": 0.0001767873764038086,
            "words": [
                {
                    "start": 1042.07,
                    "end": 1042.73,
                    "word": " distributed",
                    "probability": 0.8359375
                },
                {
                    "start": 1042.73,
                    "end": 1042.93,
                    "word": ".",
                    "probability": 0.947265625
                },
                {
                    "start": 1042.93,
                    "end": 1042.97,
                    "word": " It",
                    "probability": 0.88330078125
                },
                {
                    "start": 1042.97,
                    "end": 1043.17,
                    "word": " means",
                    "probability": 0.8828125
                },
                {
                    "start": 1043.17,
                    "end": 1043.99,
                    "word": " that",
                    "probability": 0.87646484375
                },
                {
                    "start": 1043.99,
                    "end": 1045.69,
                    "word": " one",
                    "probability": 0.806640625
                },
                {
                    "start": 1045.69,
                    "end": 1045.97,
                    "word": " data",
                    "probability": 0.78759765625
                },
                {
                    "start": 1045.97,
                    "end": 1046.39,
                    "word": " point",
                    "probability": 0.76611328125
                },
                {
                    "start": 1046.39,
                    "end": 1047.29,
                    "word": " does",
                    "probability": 0.9287109375
                },
                {
                    "start": 1047.29,
                    "end": 1047.63,
                    "word": " not",
                    "probability": 0.89794921875
                },
                {
                    "start": 1047.63,
                    "end": 1048.13,
                    "word": " depend",
                    "probability": 0.78759765625
                },
                {
                    "start": 1048.13,
                    "end": 1048.61,
                    "word": " on",
                    "probability": 0.91064453125
                },
                {
                    "start": 1048.61,
                    "end": 1048.97,
                    "word": " another",
                    "probability": 0.81787109375
                },
                {
                    "start": 1048.97,
                    "end": 1049.23,
                    "word": " data",
                    "probability": 0.7998046875
                },
                {
                    "start": 1049.23,
                    "end": 1049.57,
                    "word": " point",
                    "probability": 0.8525390625
                },
                {
                    "start": 1049.57,
                    "end": 1049.85,
                    "word": ",",
                    "probability": 0.372314453125
                },
                {
                    "start": 1049.85,
                    "end": 1050.39,
                    "word": " and",
                    "probability": 0.890625
                },
                {
                    "start": 1050.39,
                    "end": 1050.77,
                    "word": " also",
                    "probability": 0.78125
                }
            ]
        },
        {
            "id": 154,
            "seek": 105150,
            "start": 1050.77,
            "end": 1057.63,
            "text": " each data point is equally likely to occur. These do not hold for the RL settings.",
            "tokens": [
                51540,
                1184,
                1412,
                935,
                307,
                12309,
                3700,
                281,
                5160,
                13,
                1981,
                360,
                406,
                1797,
                337,
                264,
                497,
                43,
                6257,
                13,
                51852
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22716346555031264,
            "compression_ratio": 1.6307692307692307,
            "no_speech_prob": 0.0001767873764038086,
            "words": [
                {
                    "start": 1050.77,
                    "end": 1051.59,
                    "word": " each",
                    "probability": 0.55908203125
                },
                {
                    "start": 1051.59,
                    "end": 1051.89,
                    "word": " data",
                    "probability": 0.80517578125
                },
                {
                    "start": 1051.89,
                    "end": 1052.13,
                    "word": " point",
                    "probability": 0.8544921875
                },
                {
                    "start": 1052.13,
                    "end": 1052.41,
                    "word": " is",
                    "probability": 0.90625
                },
                {
                    "start": 1052.41,
                    "end": 1052.77,
                    "word": " equally",
                    "probability": 0.8359375
                },
                {
                    "start": 1052.77,
                    "end": 1053.15,
                    "word": " likely",
                    "probability": 0.74853515625
                },
                {
                    "start": 1053.15,
                    "end": 1053.89,
                    "word": " to",
                    "probability": 0.90966796875
                },
                {
                    "start": 1053.89,
                    "end": 1054.19,
                    "word": " occur",
                    "probability": 0.82666015625
                },
                {
                    "start": 1054.19,
                    "end": 1054.39,
                    "word": ".",
                    "probability": 0.9951171875
                },
                {
                    "start": 1054.39,
                    "end": 1054.83,
                    "word": " These",
                    "probability": 0.9052734375
                },
                {
                    "start": 1054.83,
                    "end": 1055.49,
                    "word": " do",
                    "probability": 0.83447265625
                },
                {
                    "start": 1055.49,
                    "end": 1055.65,
                    "word": " not",
                    "probability": 0.88916015625
                },
                {
                    "start": 1055.65,
                    "end": 1055.97,
                    "word": " hold",
                    "probability": 0.8330078125
                },
                {
                    "start": 1055.97,
                    "end": 1056.79,
                    "word": " for",
                    "probability": 0.90283203125
                },
                {
                    "start": 1056.79,
                    "end": 1057.03,
                    "word": " the",
                    "probability": 0.822265625
                },
                {
                    "start": 1057.03,
                    "end": 1057.27,
                    "word": " RL",
                    "probability": 0.890869140625
                },
                {
                    "start": 1057.27,
                    "end": 1057.59,
                    "word": " settings",
                    "probability": 0.7431640625
                },
                {
                    "start": 1057.59,
                    "end": 1057.63,
                    "word": ".",
                    "probability": 0.76708984375
                }
            ]
        },
        {
            "id": 155,
            "seek": 108072,
            "start": 1057.63,
            "end": 1064.17,
            "text": " A particular state, for example, is highly correlated with its previous state and action.",
            "tokens": [
                50400,
                316,
                1729,
                1785,
                11,
                337,
                1365,
                11,
                307,
                5405,
                38574,
                365,
                1080,
                3894,
                1785,
                293,
                3069,
                13,
                50664
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21122685185185186,
            "compression_ratio": 1.5605381165919283,
            "no_speech_prob": 4.26173210144043e-05,
            "words": [
                {
                    "start": 1057.63,
                    "end": 1058.05,
                    "word": " A",
                    "probability": 0.00128936767578125
                },
                {
                    "start": 1058.05,
                    "end": 1058.97,
                    "word": " particular",
                    "probability": 0.83349609375
                },
                {
                    "start": 1058.97,
                    "end": 1059.55,
                    "word": " state",
                    "probability": 0.86181640625
                },
                {
                    "start": 1059.55,
                    "end": 1059.61,
                    "word": ",",
                    "probability": 0.57958984375
                },
                {
                    "start": 1059.61,
                    "end": 1059.71,
                    "word": " for",
                    "probability": 0.88330078125
                },
                {
                    "start": 1059.71,
                    "end": 1060.07,
                    "word": " example",
                    "probability": 0.9326171875
                },
                {
                    "start": 1060.07,
                    "end": 1060.61,
                    "word": ",",
                    "probability": 0.919921875
                },
                {
                    "start": 1060.61,
                    "end": 1060.73,
                    "word": " is",
                    "probability": 0.884765625
                },
                {
                    "start": 1060.73,
                    "end": 1060.99,
                    "word": " highly",
                    "probability": 0.80712890625
                },
                {
                    "start": 1060.99,
                    "end": 1061.41,
                    "word": " correlated",
                    "probability": 0.7939453125
                },
                {
                    "start": 1061.41,
                    "end": 1061.83,
                    "word": " with",
                    "probability": 0.791015625
                },
                {
                    "start": 1061.83,
                    "end": 1062.31,
                    "word": " its",
                    "probability": 0.77197265625
                },
                {
                    "start": 1062.31,
                    "end": 1062.73,
                    "word": " previous",
                    "probability": 0.779296875
                },
                {
                    "start": 1062.73,
                    "end": 1063.17,
                    "word": " state",
                    "probability": 0.8935546875
                },
                {
                    "start": 1063.17,
                    "end": 1063.33,
                    "word": " and",
                    "probability": 0.77392578125
                },
                {
                    "start": 1063.33,
                    "end": 1063.69,
                    "word": " action",
                    "probability": 0.80029296875
                },
                {
                    "start": 1063.69,
                    "end": 1064.17,
                    "word": ".",
                    "probability": 0.94970703125
                }
            ]
        },
        {
            "id": 156,
            "seek": 108072,
            "start": 1064.17,
            "end": 1069.73,
            "text": " Why does this matter? Well, a lot of our supervised and deep learning approaches",
            "tokens": [
                50692,
                1545,
                775,
                341,
                1871,
                30,
                1042,
                11,
                257,
                688,
                295,
                527,
                46533,
                293,
                2452,
                2539,
                11587,
                50968
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21122685185185186,
            "compression_ratio": 1.5605381165919283,
            "no_speech_prob": 4.26173210144043e-05,
            "words": [
                {
                    "start": 1064.17,
                    "end": 1064.47,
                    "word": " Why",
                    "probability": 0.80224609375
                },
                {
                    "start": 1064.47,
                    "end": 1064.59,
                    "word": " does",
                    "probability": 0.9072265625
                },
                {
                    "start": 1064.59,
                    "end": 1064.75,
                    "word": " this",
                    "probability": 0.86669921875
                },
                {
                    "start": 1064.75,
                    "end": 1064.97,
                    "word": " matter",
                    "probability": 0.8271484375
                },
                {
                    "start": 1064.97,
                    "end": 1065.15,
                    "word": "?",
                    "probability": 0.99169921875
                },
                {
                    "start": 1065.15,
                    "end": 1065.27,
                    "word": " Well",
                    "probability": 0.451904296875
                },
                {
                    "start": 1065.27,
                    "end": 1065.37,
                    "word": ",",
                    "probability": 0.9072265625
                },
                {
                    "start": 1065.49,
                    "end": 1065.69,
                    "word": " a lot",
                    "probability": 0.8857421875
                },
                {
                    "start": 1065.69,
                    "end": 1065.99,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 1065.99,
                    "end": 1066.33,
                    "word": " our",
                    "probability": 0.77734375
                },
                {
                    "start": 1066.33,
                    "end": 1066.97,
                    "word": " supervised",
                    "probability": 0.7587890625
                },
                {
                    "start": 1066.97,
                    "end": 1067.81,
                    "word": " and",
                    "probability": 0.86328125
                },
                {
                    "start": 1067.81,
                    "end": 1068.99,
                    "word": " deep",
                    "probability": 0.9033203125
                },
                {
                    "start": 1068.99,
                    "end": 1069.25,
                    "word": " learning",
                    "probability": 0.779296875
                },
                {
                    "start": 1069.25,
                    "end": 1069.73,
                    "word": " approaches",
                    "probability": 0.666015625
                }
            ]
        },
        {
            "id": 157,
            "seek": 108072,
            "start": 1070.57,
            "end": 1078.55,
            "text": " are based on the assumption that the training data is IID. When this is not possible,",
            "tokens": [
                51016,
                366,
                2361,
                322,
                264,
                15302,
                300,
                264,
                3097,
                1412,
                307,
                286,
                2777,
                13,
                1133,
                341,
                307,
                406,
                1944,
                11,
                51400
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21122685185185186,
            "compression_ratio": 1.5605381165919283,
            "no_speech_prob": 4.26173210144043e-05,
            "words": [
                {
                    "start": 1070.57,
                    "end": 1070.95,
                    "word": " are",
                    "probability": 0.8583984375
                },
                {
                    "start": 1070.95,
                    "end": 1071.27,
                    "word": " based",
                    "probability": 0.83251953125
                },
                {
                    "start": 1071.27,
                    "end": 1071.41,
                    "word": " on",
                    "probability": 0.908203125
                },
                {
                    "start": 1071.41,
                    "end": 1071.53,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 1071.53,
                    "end": 1071.83,
                    "word": " assumption",
                    "probability": 0.89453125
                },
                {
                    "start": 1071.83,
                    "end": 1072.09,
                    "word": " that",
                    "probability": 0.875
                },
                {
                    "start": 1072.09,
                    "end": 1072.25,
                    "word": " the",
                    "probability": 0.802734375
                },
                {
                    "start": 1072.25,
                    "end": 1073.19,
                    "word": " training",
                    "probability": 0.388427734375
                },
                {
                    "start": 1073.19,
                    "end": 1073.63,
                    "word": " data",
                    "probability": 0.81201171875
                },
                {
                    "start": 1073.63,
                    "end": 1074.29,
                    "word": " is",
                    "probability": 0.896484375
                },
                {
                    "start": 1074.29,
                    "end": 1074.83,
                    "word": " IID",
                    "probability": 0.841552734375
                },
                {
                    "start": 1074.83,
                    "end": 1075.33,
                    "word": ".",
                    "probability": 0.9658203125
                },
                {
                    "start": 1075.33,
                    "end": 1075.45,
                    "word": " When",
                    "probability": 0.79248046875
                },
                {
                    "start": 1075.45,
                    "end": 1075.63,
                    "word": " this",
                    "probability": 0.87451171875
                },
                {
                    "start": 1075.63,
                    "end": 1075.77,
                    "word": " is",
                    "probability": 0.9052734375
                },
                {
                    "start": 1075.77,
                    "end": 1076.23,
                    "word": " not",
                    "probability": 0.900390625
                },
                {
                    "start": 1076.23,
                    "end": 1077.91,
                    "word": " possible",
                    "probability": 0.84423828125
                },
                {
                    "start": 1077.91,
                    "end": 1078.55,
                    "word": ",",
                    "probability": 0.90869140625
                }
            ]
        },
        {
            "id": 158,
            "seek": 108072,
            "start": 1078.55,
            "end": 1086.85,
            "text": " what happens is, in response to this problem, what happened was the DQN approach introduced",
            "tokens": [
                51400,
                437,
                2314,
                307,
                11,
                294,
                4134,
                281,
                341,
                1154,
                11,
                437,
                2011,
                390,
                264,
                413,
                48,
                45,
                3109,
                7268,
                51832
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21122685185185186,
            "compression_ratio": 1.5605381165919283,
            "no_speech_prob": 4.26173210144043e-05,
            "words": [
                {
                    "start": 1078.55,
                    "end": 1078.65,
                    "word": " what",
                    "probability": 0.75390625
                },
                {
                    "start": 1078.65,
                    "end": 1079.05,
                    "word": " happens",
                    "probability": 0.904296875
                },
                {
                    "start": 1079.05,
                    "end": 1079.69,
                    "word": " is",
                    "probability": 0.8359375
                },
                {
                    "start": 1079.69,
                    "end": 1080.57,
                    "word": ",",
                    "probability": 0.4072265625
                },
                {
                    "start": 1080.57,
                    "end": 1081.73,
                    "word": " in",
                    "probability": 0.89111328125
                },
                {
                    "start": 1081.73,
                    "end": 1082.15,
                    "word": " response",
                    "probability": 0.89453125
                },
                {
                    "start": 1082.15,
                    "end": 1082.35,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 1082.35,
                    "end": 1082.53,
                    "word": " this",
                    "probability": 0.88232421875
                },
                {
                    "start": 1082.53,
                    "end": 1083.01,
                    "word": " problem",
                    "probability": 0.77197265625
                },
                {
                    "start": 1083.01,
                    "end": 1083.75,
                    "word": ",",
                    "probability": 0.916015625
                },
                {
                    "start": 1083.75,
                    "end": 1083.93,
                    "word": " what",
                    "probability": 0.73876953125
                },
                {
                    "start": 1083.93,
                    "end": 1084.27,
                    "word": " happened",
                    "probability": 0.69287109375
                },
                {
                    "start": 1084.27,
                    "end": 1084.71,
                    "word": " was",
                    "probability": 0.88818359375
                },
                {
                    "start": 1084.71,
                    "end": 1085.49,
                    "word": " the",
                    "probability": 0.71875
                },
                {
                    "start": 1085.49,
                    "end": 1086.07,
                    "word": " DQN",
                    "probability": 0.8958333333333334
                },
                {
                    "start": 1086.07,
                    "end": 1086.51,
                    "word": " approach",
                    "probability": 0.78466796875
                },
                {
                    "start": 1086.51,
                    "end": 1086.85,
                    "word": " introduced",
                    "probability": 0.7333984375
                }
            ]
        },
        {
            "id": 159,
            "seek": 110932,
            "start": 1086.85,
            "end": 1092.81,
            "text": " experience replay. It's like a bag of all data which is randomly sampled in each training",
            "tokens": [
                50368,
                1752,
                23836,
                13,
                467,
                311,
                411,
                257,
                3411,
                295,
                439,
                1412,
                597,
                307,
                16979,
                3247,
                15551,
                294,
                1184,
                3097,
                50664
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20998475973199054,
            "compression_ratio": 1.6486486486486487,
            "no_speech_prob": 0.003482818603515625,
            "words": [
                {
                    "start": 1086.85,
                    "end": 1087.39,
                    "word": " experience",
                    "probability": 0.2425537109375
                },
                {
                    "start": 1087.39,
                    "end": 1087.89,
                    "word": " replay",
                    "probability": 0.7724609375
                },
                {
                    "start": 1087.89,
                    "end": 1088.01,
                    "word": ".",
                    "probability": 0.58642578125
                },
                {
                    "start": 1088.01,
                    "end": 1088.05,
                    "word": " It's",
                    "probability": 0.91796875
                },
                {
                    "start": 1088.05,
                    "end": 1088.19,
                    "word": " like",
                    "probability": 0.86328125
                },
                {
                    "start": 1088.37,
                    "end": 1088.69,
                    "word": " a bag",
                    "probability": 0.91015625
                },
                {
                    "start": 1088.69,
                    "end": 1089.15,
                    "word": " of",
                    "probability": 0.90185546875
                },
                {
                    "start": 1089.15,
                    "end": 1089.45,
                    "word": " all",
                    "probability": 0.490478515625
                },
                {
                    "start": 1089.45,
                    "end": 1089.83,
                    "word": " data",
                    "probability": 0.83984375
                },
                {
                    "start": 1089.83,
                    "end": 1090.25,
                    "word": " which",
                    "probability": 0.5771484375
                },
                {
                    "start": 1090.25,
                    "end": 1090.47,
                    "word": " is",
                    "probability": 0.90234375
                },
                {
                    "start": 1090.47,
                    "end": 1090.91,
                    "word": " randomly",
                    "probability": 0.7001953125
                },
                {
                    "start": 1090.91,
                    "end": 1091.45,
                    "word": " sampled",
                    "probability": 0.906982421875
                },
                {
                    "start": 1091.45,
                    "end": 1092.05,
                    "word": " in",
                    "probability": 0.8896484375
                },
                {
                    "start": 1092.05,
                    "end": 1092.33,
                    "word": " each",
                    "probability": 0.87451171875
                },
                {
                    "start": 1092.33,
                    "end": 1092.81,
                    "word": " training",
                    "probability": 0.79931640625
                }
            ]
        },
        {
            "id": 160,
            "seek": 110932,
            "start": 1092.81,
            "end": 1098.53,
            "text": " iteration to reduce the correlation, the sequential correlation, the temporal correlation",
            "tokens": [
                50664,
                24784,
                281,
                5407,
                264,
                20009,
                11,
                264,
                42881,
                20009,
                11,
                264,
                30881,
                20009,
                50952
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20998475973199054,
            "compression_ratio": 1.6486486486486487,
            "no_speech_prob": 0.003482818603515625,
            "words": [
                {
                    "start": 1092.81,
                    "end": 1093.39,
                    "word": " iteration",
                    "probability": 0.75439453125
                },
                {
                    "start": 1093.39,
                    "end": 1094.09,
                    "word": " to",
                    "probability": 0.8232421875
                },
                {
                    "start": 1094.09,
                    "end": 1094.73,
                    "word": " reduce",
                    "probability": 0.75634765625
                },
                {
                    "start": 1094.73,
                    "end": 1095.41,
                    "word": " the",
                    "probability": 0.826171875
                },
                {
                    "start": 1095.41,
                    "end": 1096.67,
                    "word": " correlation",
                    "probability": 0.30322265625
                },
                {
                    "start": 1096.67,
                    "end": 1096.85,
                    "word": ",",
                    "probability": 0.7265625
                },
                {
                    "start": 1096.85,
                    "end": 1096.91,
                    "word": " the",
                    "probability": 0.80810546875
                },
                {
                    "start": 1096.91,
                    "end": 1097.23,
                    "word": " sequential",
                    "probability": 0.87451171875
                },
                {
                    "start": 1097.23,
                    "end": 1097.63,
                    "word": " correlation",
                    "probability": 0.7861328125
                },
                {
                    "start": 1097.63,
                    "end": 1097.75,
                    "word": ",",
                    "probability": 0.90625
                },
                {
                    "start": 1097.75,
                    "end": 1097.83,
                    "word": " the",
                    "probability": 0.79541015625
                },
                {
                    "start": 1097.83,
                    "end": 1098.07,
                    "word": " temporal",
                    "probability": 0.8349609375
                },
                {
                    "start": 1098.07,
                    "end": 1098.53,
                    "word": " correlation",
                    "probability": 0.7724609375
                }
            ]
        },
        {
            "id": 161,
            "seek": 110932,
            "start": 1098.53,
            "end": 1106.83,
            "text": " of data points, and also make it more likely for data to be evenly distributed. Also, to",
            "tokens": [
                50952,
                295,
                1412,
                2793,
                11,
                293,
                611,
                652,
                309,
                544,
                3700,
                337,
                1412,
                281,
                312,
                17658,
                12631,
                13,
                2743,
                11,
                281,
                51352
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20998475973199054,
            "compression_ratio": 1.6486486486486487,
            "no_speech_prob": 0.003482818603515625,
            "words": [
                {
                    "start": 1098.53,
                    "end": 1098.91,
                    "word": " of",
                    "probability": 0.80859375
                },
                {
                    "start": 1098.91,
                    "end": 1099.17,
                    "word": " data",
                    "probability": 0.80859375
                },
                {
                    "start": 1099.17,
                    "end": 1099.49,
                    "word": " points",
                    "probability": 0.7685546875
                },
                {
                    "start": 1099.49,
                    "end": 1099.55,
                    "word": ",",
                    "probability": 0.302978515625
                },
                {
                    "start": 1099.55,
                    "end": 1099.71,
                    "word": " and",
                    "probability": 0.890625
                },
                {
                    "start": 1099.71,
                    "end": 1100.07,
                    "word": " also",
                    "probability": 0.82177734375
                },
                {
                    "start": 1100.07,
                    "end": 1100.71,
                    "word": " make",
                    "probability": 0.85205078125
                },
                {
                    "start": 1100.71,
                    "end": 1100.83,
                    "word": " it",
                    "probability": 0.89990234375
                },
                {
                    "start": 1100.83,
                    "end": 1101.01,
                    "word": " more",
                    "probability": 0.88916015625
                },
                {
                    "start": 1101.01,
                    "end": 1101.33,
                    "word": " likely",
                    "probability": 0.77734375
                },
                {
                    "start": 1101.33,
                    "end": 1101.79,
                    "word": " for",
                    "probability": 0.89453125
                },
                {
                    "start": 1101.79,
                    "end": 1102.15,
                    "word": " data",
                    "probability": 0.81201171875
                },
                {
                    "start": 1102.15,
                    "end": 1102.33,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 1102.33,
                    "end": 1103.19,
                    "word": " be",
                    "probability": 0.90185546875
                },
                {
                    "start": 1103.19,
                    "end": 1104.79,
                    "word": " evenly",
                    "probability": 0.70849609375
                },
                {
                    "start": 1104.79,
                    "end": 1105.43,
                    "word": " distributed",
                    "probability": 0.85400390625
                },
                {
                    "start": 1105.43,
                    "end": 1105.99,
                    "word": ".",
                    "probability": 0.95751953125
                },
                {
                    "start": 1105.99,
                    "end": 1106.43,
                    "word": " Also",
                    "probability": 0.92041015625
                },
                {
                    "start": 1106.43,
                    "end": 1106.57,
                    "word": ",",
                    "probability": 0.2861328125
                },
                {
                    "start": 1106.57,
                    "end": 1106.83,
                    "word": " to",
                    "probability": 0.90234375
                }
            ]
        },
        {
            "id": 162,
            "seek": 110932,
            "start": 1107.49,
            "end": 1115.45,
            "text": " reduce the effect of oscillation during training, DQN uses fixed parameters for a target network.",
            "tokens": [
                51396,
                5407,
                264,
                1802,
                295,
                18225,
                399,
                1830,
                3097,
                11,
                413,
                48,
                45,
                4960,
                6806,
                9834,
                337,
                257,
                3779,
                3209,
                13,
                51788
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20998475973199054,
            "compression_ratio": 1.6486486486486487,
            "no_speech_prob": 0.003482818603515625,
            "words": [
                {
                    "start": 1107.49,
                    "end": 1107.95,
                    "word": " reduce",
                    "probability": 0.73486328125
                },
                {
                    "start": 1107.95,
                    "end": 1108.11,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 1108.11,
                    "end": 1108.33,
                    "word": " effect",
                    "probability": 0.82470703125
                },
                {
                    "start": 1108.33,
                    "end": 1108.55,
                    "word": " of",
                    "probability": 0.9033203125
                },
                {
                    "start": 1108.55,
                    "end": 1109.25,
                    "word": " oscillation",
                    "probability": 0.873046875
                },
                {
                    "start": 1109.25,
                    "end": 1109.79,
                    "word": " during",
                    "probability": 0.76806640625
                },
                {
                    "start": 1109.79,
                    "end": 1110.29,
                    "word": " training",
                    "probability": 0.80419921875
                },
                {
                    "start": 1110.29,
                    "end": 1111.35,
                    "word": ",",
                    "probability": 0.828125
                },
                {
                    "start": 1111.35,
                    "end": 1111.85,
                    "word": " DQN",
                    "probability": 0.8424479166666666
                },
                {
                    "start": 1111.85,
                    "end": 1112.73,
                    "word": " uses",
                    "probability": 0.763671875
                },
                {
                    "start": 1112.73,
                    "end": 1113.47,
                    "word": " fixed",
                    "probability": 0.841796875
                },
                {
                    "start": 1113.47,
                    "end": 1113.95,
                    "word": " parameters",
                    "probability": 0.8798828125
                },
                {
                    "start": 1113.95,
                    "end": 1114.31,
                    "word": " for",
                    "probability": 0.89453125
                },
                {
                    "start": 1114.51,
                    "end": 1114.81,
                    "word": " a target",
                    "probability": 0.8564453125
                },
                {
                    "start": 1114.81,
                    "end": 1115.35,
                    "word": " network",
                    "probability": 0.84716796875
                },
                {
                    "start": 1115.35,
                    "end": 1115.45,
                    "word": ".",
                    "probability": 0.51171875
                }
            ]
        },
        {
            "id": 163,
            "seek": 113450,
            "start": 1115.45,
            "end": 1122.79,
            "text": " The target of optimization is fixed and is updated every few thousand iterations,",
            "tokens": [
                50364,
                440,
                3779,
                295,
                19618,
                307,
                6806,
                293,
                307,
                10588,
                633,
                1326,
                4714,
                36540,
                11,
                50740
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22937499284744262,
            "compression_ratio": 1.5953488372093023,
            "no_speech_prob": 0.0002415180206298828,
            "words": [
                {
                    "start": 1115.45,
                    "end": 1115.59,
                    "word": " The",
                    "probability": 0.054290771484375
                },
                {
                    "start": 1115.59,
                    "end": 1115.97,
                    "word": " target",
                    "probability": 0.72509765625
                },
                {
                    "start": 1115.97,
                    "end": 1116.31,
                    "word": " of",
                    "probability": 0.8544921875
                },
                {
                    "start": 1116.31,
                    "end": 1116.79,
                    "word": " optimization",
                    "probability": 0.69921875
                },
                {
                    "start": 1116.79,
                    "end": 1117.11,
                    "word": " is",
                    "probability": 0.89892578125
                },
                {
                    "start": 1117.11,
                    "end": 1117.57,
                    "word": " fixed",
                    "probability": 0.8232421875
                },
                {
                    "start": 1117.57,
                    "end": 1118.41,
                    "word": " and",
                    "probability": 0.7392578125
                },
                {
                    "start": 1118.41,
                    "end": 1118.59,
                    "word": " is",
                    "probability": 0.83984375
                },
                {
                    "start": 1118.59,
                    "end": 1119.05,
                    "word": " updated",
                    "probability": 0.8359375
                },
                {
                    "start": 1119.05,
                    "end": 1119.97,
                    "word": " every",
                    "probability": 0.6982421875
                },
                {
                    "start": 1119.97,
                    "end": 1121.11,
                    "word": " few",
                    "probability": 0.7685546875
                },
                {
                    "start": 1121.11,
                    "end": 1121.55,
                    "word": " thousand",
                    "probability": 0.7001953125
                },
                {
                    "start": 1121.55,
                    "end": 1122.33,
                    "word": " iterations",
                    "probability": 0.7255859375
                },
                {
                    "start": 1122.33,
                    "end": 1122.79,
                    "word": ",",
                    "probability": 0.1895751953125
                }
            ]
        },
        {
            "id": 164,
            "seek": 113450,
            "start": 1122.79,
            "end": 1126.27,
            "text": " so that reduces the oscillation problem. And, of course, the rewards are normalized to",
            "tokens": [
                50740,
                370,
                300,
                18081,
                264,
                18225,
                399,
                1154,
                13,
                400,
                11,
                295,
                1164,
                11,
                264,
                17203,
                366,
                48704,
                281,
                50912
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22937499284744262,
            "compression_ratio": 1.5953488372093023,
            "no_speech_prob": 0.0002415180206298828,
            "words": [
                {
                    "start": 1122.79,
                    "end": 1123.21,
                    "word": " so",
                    "probability": 0.86767578125
                },
                {
                    "start": 1123.21,
                    "end": 1123.37,
                    "word": " that",
                    "probability": 0.87646484375
                },
                {
                    "start": 1123.37,
                    "end": 1123.65,
                    "word": " reduces",
                    "probability": 0.89111328125
                },
                {
                    "start": 1123.65,
                    "end": 1123.85,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 1123.85,
                    "end": 1124.19,
                    "word": " oscillation",
                    "probability": 0.661865234375
                },
                {
                    "start": 1124.19,
                    "end": 1124.49,
                    "word": " problem",
                    "probability": 0.763671875
                },
                {
                    "start": 1124.49,
                    "end": 1124.61,
                    "word": ".",
                    "probability": 0.7783203125
                },
                {
                    "start": 1124.61,
                    "end": 1124.63,
                    "word": " And",
                    "probability": 0.72265625
                },
                {
                    "start": 1124.63,
                    "end": 1124.75,
                    "word": ",",
                    "probability": 0.154541015625
                },
                {
                    "start": 1124.75,
                    "end": 1124.75,
                    "word": " of",
                    "probability": 0.9111328125
                },
                {
                    "start": 1124.75,
                    "end": 1124.87,
                    "word": " course",
                    "probability": 0.85693359375
                },
                {
                    "start": 1124.87,
                    "end": 1124.91,
                    "word": ",",
                    "probability": 0.9833984375
                },
                {
                    "start": 1124.91,
                    "end": 1124.99,
                    "word": " the",
                    "probability": 0.71728515625
                },
                {
                    "start": 1124.99,
                    "end": 1125.25,
                    "word": " rewards",
                    "probability": 0.70361328125
                },
                {
                    "start": 1125.25,
                    "end": 1125.53,
                    "word": " are",
                    "probability": 0.904296875
                },
                {
                    "start": 1125.53,
                    "end": 1125.93,
                    "word": " normalized",
                    "probability": 0.904296875
                },
                {
                    "start": 1125.93,
                    "end": 1126.27,
                    "word": " to",
                    "probability": 0.89599609375
                }
            ]
        },
        {
            "id": 165,
            "seek": 113450,
            "start": 1126.27,
            "end": 1133.77,
            "text": " minus one to one to make sure the performance, the reward signals are bounded.",
            "tokens": [
                50912,
                3175,
                472,
                281,
                472,
                281,
                652,
                988,
                264,
                3389,
                11,
                264,
                7782,
                12354,
                366,
                37498,
                13,
                51268
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22937499284744262,
            "compression_ratio": 1.5953488372093023,
            "no_speech_prob": 0.0002415180206298828,
            "words": [
                {
                    "start": 1126.27,
                    "end": 1126.71,
                    "word": " minus",
                    "probability": 0.4169921875
                },
                {
                    "start": 1126.71,
                    "end": 1126.95,
                    "word": " one",
                    "probability": 0.489990234375
                },
                {
                    "start": 1126.95,
                    "end": 1127.07,
                    "word": " to",
                    "probability": 0.85498046875
                },
                {
                    "start": 1127.07,
                    "end": 1127.31,
                    "word": " one",
                    "probability": 0.80078125
                },
                {
                    "start": 1127.31,
                    "end": 1127.61,
                    "word": " to",
                    "probability": 0.83251953125
                },
                {
                    "start": 1127.61,
                    "end": 1128.39,
                    "word": " make",
                    "probability": 0.8798828125
                },
                {
                    "start": 1128.39,
                    "end": 1128.73,
                    "word": " sure",
                    "probability": 0.80322265625
                },
                {
                    "start": 1128.73,
                    "end": 1129.77,
                    "word": " the",
                    "probability": 0.79931640625
                },
                {
                    "start": 1129.77,
                    "end": 1131.59,
                    "word": " performance",
                    "probability": 0.7822265625
                },
                {
                    "start": 1131.59,
                    "end": 1131.77,
                    "word": ",",
                    "probability": 0.611328125
                },
                {
                    "start": 1131.77,
                    "end": 1131.85,
                    "word": " the",
                    "probability": 0.802734375
                },
                {
                    "start": 1131.85,
                    "end": 1132.35,
                    "word": " reward",
                    "probability": 0.814453125
                },
                {
                    "start": 1132.35,
                    "end": 1132.79,
                    "word": " signals",
                    "probability": 0.9033203125
                },
                {
                    "start": 1132.79,
                    "end": 1133.05,
                    "word": " are",
                    "probability": 0.7265625
                },
                {
                    "start": 1133.05,
                    "end": 1133.45,
                    "word": " bounded",
                    "probability": 0.818359375
                },
                {
                    "start": 1133.45,
                    "end": 1133.77,
                    "word": ".",
                    "probability": 0.99072265625
                }
            ]
        },
        {
            "id": 166,
            "seek": 113450,
            "start": 1134.93,
            "end": 1140.63,
            "text": " Now, now that we have a preliminary understanding of deep reinforcement learning and one of its",
            "tokens": [
                51336,
                823,
                11,
                586,
                300,
                321,
                362,
                257,
                28817,
                3701,
                295,
                2452,
                29280,
                2539,
                293,
                472,
                295,
                1080,
                51620
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22937499284744262,
            "compression_ratio": 1.5953488372093023,
            "no_speech_prob": 0.0002415180206298828,
            "words": [
                {
                    "start": 1134.93,
                    "end": 1135.21,
                    "word": " Now",
                    "probability": 0.89990234375
                },
                {
                    "start": 1135.21,
                    "end": 1135.53,
                    "word": ",",
                    "probability": 0.34716796875
                },
                {
                    "start": 1135.53,
                    "end": 1136.31,
                    "word": " now",
                    "probability": 0.62353515625
                },
                {
                    "start": 1136.31,
                    "end": 1136.45,
                    "word": " that",
                    "probability": 0.8759765625
                },
                {
                    "start": 1136.45,
                    "end": 1136.57,
                    "word": " we",
                    "probability": 0.90087890625
                },
                {
                    "start": 1136.57,
                    "end": 1136.89,
                    "word": " have",
                    "probability": 0.88037109375
                },
                {
                    "start": 1137.29,
                    "end": 1137.65,
                    "word": " a preliminary",
                    "probability": 0.8916015625
                },
                {
                    "start": 1137.65,
                    "end": 1138.21,
                    "word": " understanding",
                    "probability": 0.7099609375
                },
                {
                    "start": 1138.21,
                    "end": 1138.85,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 1138.85,
                    "end": 1139.25,
                    "word": " deep",
                    "probability": 0.59521484375
                },
                {
                    "start": 1139.25,
                    "end": 1139.65,
                    "word": " reinforcement",
                    "probability": 0.87109375
                },
                {
                    "start": 1139.65,
                    "end": 1139.95,
                    "word": " learning",
                    "probability": 0.83251953125
                },
                {
                    "start": 1139.95,
                    "end": 1140.23,
                    "word": " and",
                    "probability": 0.85546875
                },
                {
                    "start": 1140.23,
                    "end": 1140.43,
                    "word": " one",
                    "probability": 0.80908203125
                },
                {
                    "start": 1140.43,
                    "end": 1140.53,
                    "word": " of",
                    "probability": 0.9033203125
                },
                {
                    "start": 1140.53,
                    "end": 1140.63,
                    "word": " its",
                    "probability": 0.81103515625
                }
            ]
        },
        {
            "id": 167,
            "seek": 116092,
            "start": 1140.63,
            "end": 1146.13,
            "text": " implementations, one of the parallel approaches called DQN, let's take a quick look at adversarial",
            "tokens": [
                50364,
                4445,
                763,
                11,
                472,
                295,
                264,
                8952,
                11587,
                1219,
                413,
                48,
                45,
                11,
                718,
                311,
                747,
                257,
                1702,
                574,
                412,
                17641,
                44745,
                50640
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21189692407323604,
            "compression_ratio": 1.7435897435897436,
            "no_speech_prob": 0.00555419921875,
            "words": [
                {
                    "start": 1140.63,
                    "end": 1141.35,
                    "word": " implementations",
                    "probability": 0.76123046875
                },
                {
                    "start": 1141.35,
                    "end": 1141.43,
                    "word": ",",
                    "probability": 0.7001953125
                },
                {
                    "start": 1141.43,
                    "end": 1141.51,
                    "word": " one",
                    "probability": 0.7998046875
                },
                {
                    "start": 1141.51,
                    "end": 1141.83,
                    "word": " of",
                    "probability": 0.89501953125
                },
                {
                    "start": 1141.83,
                    "end": 1142.45,
                    "word": " the",
                    "probability": 0.59814453125
                },
                {
                    "start": 1142.45,
                    "end": 1142.71,
                    "word": " parallel",
                    "probability": 0.7421875
                },
                {
                    "start": 1142.71,
                    "end": 1143.09,
                    "word": " approaches",
                    "probability": 0.67041015625
                },
                {
                    "start": 1143.09,
                    "end": 1143.33,
                    "word": " called",
                    "probability": 0.72021484375
                },
                {
                    "start": 1143.33,
                    "end": 1143.93,
                    "word": " DQN",
                    "probability": 0.84765625
                },
                {
                    "start": 1143.93,
                    "end": 1144.53,
                    "word": ",",
                    "probability": 0.25244140625
                },
                {
                    "start": 1144.53,
                    "end": 1144.85,
                    "word": " let's",
                    "probability": 0.931884765625
                },
                {
                    "start": 1144.85,
                    "end": 1144.99,
                    "word": " take",
                    "probability": 0.81201171875
                },
                {
                    "start": 1145.09,
                    "end": 1145.25,
                    "word": " a quick",
                    "probability": 0.90625
                },
                {
                    "start": 1145.25,
                    "end": 1145.47,
                    "word": " look",
                    "probability": 0.9130859375
                },
                {
                    "start": 1145.47,
                    "end": 1145.69,
                    "word": " at",
                    "probability": 0.90380859375
                },
                {
                    "start": 1145.69,
                    "end": 1146.13,
                    "word": " adversarial",
                    "probability": 0.9248046875
                }
            ]
        },
        {
            "id": 168,
            "seek": 116092,
            "start": 1146.13,
            "end": 1150.09,
            "text": " machine learning. And I assume that by now, those of you who are not familiar with adversarial",
            "tokens": [
                50640,
                3479,
                2539,
                13,
                400,
                286,
                6552,
                300,
                538,
                586,
                11,
                729,
                295,
                291,
                567,
                366,
                406,
                4963,
                365,
                17641,
                44745,
                50840
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21189692407323604,
            "compression_ratio": 1.7435897435897436,
            "no_speech_prob": 0.00555419921875,
            "words": [
                {
                    "start": 1146.13,
                    "end": 1146.43,
                    "word": " machine",
                    "probability": 0.662109375
                },
                {
                    "start": 1146.43,
                    "end": 1146.71,
                    "word": " learning",
                    "probability": 0.83984375
                },
                {
                    "start": 1146.71,
                    "end": 1146.79,
                    "word": ".",
                    "probability": 0.463134765625
                },
                {
                    "start": 1146.79,
                    "end": 1146.85,
                    "word": " And",
                    "probability": 0.376708984375
                },
                {
                    "start": 1146.85,
                    "end": 1146.93,
                    "word": " I",
                    "probability": 0.5673828125
                },
                {
                    "start": 1146.93,
                    "end": 1147.09,
                    "word": " assume",
                    "probability": 0.83837890625
                },
                {
                    "start": 1147.09,
                    "end": 1147.33,
                    "word": " that",
                    "probability": 0.87890625
                },
                {
                    "start": 1147.33,
                    "end": 1147.75,
                    "word": " by",
                    "probability": 0.7861328125
                },
                {
                    "start": 1147.75,
                    "end": 1148.19,
                    "word": " now",
                    "probability": 0.8955078125
                },
                {
                    "start": 1148.19,
                    "end": 1148.19,
                    "word": ",",
                    "probability": 0.54541015625
                },
                {
                    "start": 1148.19,
                    "end": 1148.41,
                    "word": " those",
                    "probability": 0.8427734375
                },
                {
                    "start": 1148.41,
                    "end": 1148.53,
                    "word": " of",
                    "probability": 0.9013671875
                },
                {
                    "start": 1148.53,
                    "end": 1148.65,
                    "word": " you",
                    "probability": 0.89111328125
                },
                {
                    "start": 1148.65,
                    "end": 1148.75,
                    "word": " who",
                    "probability": 0.81005859375
                },
                {
                    "start": 1148.75,
                    "end": 1148.85,
                    "word": " are",
                    "probability": 0.890625
                },
                {
                    "start": 1148.85,
                    "end": 1149.11,
                    "word": " not",
                    "probability": 0.89013671875
                },
                {
                    "start": 1149.11,
                    "end": 1149.43,
                    "word": " familiar",
                    "probability": 0.77587890625
                },
                {
                    "start": 1149.43,
                    "end": 1149.67,
                    "word": " with",
                    "probability": 0.81103515625
                },
                {
                    "start": 1149.67,
                    "end": 1150.09,
                    "word": " adversarial",
                    "probability": 0.9306640625
                }
            ]
        },
        {
            "id": 169,
            "seek": 116092,
            "start": 1150.09,
            "end": 1155.81,
            "text": " examples have become, have been introduced to this concept. The idea here in adversarial examples,",
            "tokens": [
                50840,
                5110,
                362,
                1813,
                11,
                362,
                668,
                7268,
                281,
                341,
                3410,
                13,
                440,
                1558,
                510,
                294,
                17641,
                44745,
                5110,
                11,
                51128
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21189692407323604,
            "compression_ratio": 1.7435897435897436,
            "no_speech_prob": 0.00555419921875,
            "words": [
                {
                    "start": 1150.09,
                    "end": 1150.51,
                    "word": " examples",
                    "probability": 0.796875
                },
                {
                    "start": 1150.51,
                    "end": 1150.91,
                    "word": " have",
                    "probability": 0.81591796875
                },
                {
                    "start": 1150.91,
                    "end": 1151.27,
                    "word": " become",
                    "probability": 0.72998046875
                },
                {
                    "start": 1151.27,
                    "end": 1151.93,
                    "word": ",",
                    "probability": 0.454345703125
                },
                {
                    "start": 1151.93,
                    "end": 1152.57,
                    "word": " have",
                    "probability": 0.7890625
                },
                {
                    "start": 1152.57,
                    "end": 1152.71,
                    "word": " been",
                    "probability": 0.8173828125
                },
                {
                    "start": 1152.71,
                    "end": 1153.09,
                    "word": " introduced",
                    "probability": 0.75048828125
                },
                {
                    "start": 1153.09,
                    "end": 1153.27,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 1153.27,
                    "end": 1153.43,
                    "word": " this",
                    "probability": 0.88037109375
                },
                {
                    "start": 1153.43,
                    "end": 1153.67,
                    "word": " concept",
                    "probability": 0.80517578125
                },
                {
                    "start": 1153.67,
                    "end": 1153.67,
                    "word": ".",
                    "probability": 0.83154296875
                },
                {
                    "start": 1153.67,
                    "end": 1153.87,
                    "word": " The",
                    "probability": 0.80810546875
                },
                {
                    "start": 1153.87,
                    "end": 1154.25,
                    "word": " idea",
                    "probability": 0.783203125
                },
                {
                    "start": 1154.25,
                    "end": 1154.67,
                    "word": " here",
                    "probability": 0.81298828125
                },
                {
                    "start": 1154.67,
                    "end": 1154.95,
                    "word": " in",
                    "probability": 0.78369140625
                },
                {
                    "start": 1154.95,
                    "end": 1155.37,
                    "word": " adversarial",
                    "probability": 0.924072265625
                },
                {
                    "start": 1155.37,
                    "end": 1155.81,
                    "word": " examples",
                    "probability": 0.79150390625
                },
                {
                    "start": 1155.81,
                    "end": 1155.81,
                    "word": ",",
                    "probability": 0.85546875
                }
            ]
        },
        {
            "id": 170,
            "seek": 116092,
            "start": 1155.81,
            "end": 1160.59,
            "text": " we are now talking in the realm of, we are speaking in the context of image classifier",
            "tokens": [
                51128,
                321,
                366,
                586,
                1417,
                294,
                264,
                15355,
                295,
                11,
                321,
                366,
                4124,
                294,
                264,
                4319,
                295,
                3256,
                1508,
                9902,
                51360
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21189692407323604,
            "compression_ratio": 1.7435897435897436,
            "no_speech_prob": 0.00555419921875,
            "words": [
                {
                    "start": 1155.81,
                    "end": 1156.07,
                    "word": " we",
                    "probability": 0.9013671875
                },
                {
                    "start": 1156.07,
                    "end": 1156.17,
                    "word": " are",
                    "probability": 0.88671875
                },
                {
                    "start": 1156.17,
                    "end": 1156.35,
                    "word": " now",
                    "probability": 0.89453125
                },
                {
                    "start": 1156.35,
                    "end": 1156.69,
                    "word": " talking",
                    "probability": 0.76171875
                },
                {
                    "start": 1156.69,
                    "end": 1157.65,
                    "word": " in",
                    "probability": 0.8193359375
                },
                {
                    "start": 1157.65,
                    "end": 1157.81,
                    "word": " the",
                    "probability": 0.7900390625
                },
                {
                    "start": 1157.81,
                    "end": 1158.03,
                    "word": " realm",
                    "probability": 0.8330078125
                },
                {
                    "start": 1158.03,
                    "end": 1158.15,
                    "word": " of",
                    "probability": 0.8212890625
                },
                {
                    "start": 1158.15,
                    "end": 1158.15,
                    "word": ",",
                    "probability": 0.341552734375
                },
                {
                    "start": 1158.15,
                    "end": 1158.23,
                    "word": " we",
                    "probability": 0.69140625
                },
                {
                    "start": 1158.23,
                    "end": 1158.27,
                    "word": " are",
                    "probability": 0.38525390625
                },
                {
                    "start": 1158.27,
                    "end": 1158.57,
                    "word": " speaking",
                    "probability": 0.814453125
                },
                {
                    "start": 1158.57,
                    "end": 1159.21,
                    "word": " in",
                    "probability": 0.8740234375
                },
                {
                    "start": 1159.21,
                    "end": 1159.37,
                    "word": " the",
                    "probability": 0.81787109375
                },
                {
                    "start": 1159.37,
                    "end": 1159.69,
                    "word": " context",
                    "probability": 0.7734375
                },
                {
                    "start": 1159.69,
                    "end": 1159.93,
                    "word": " of",
                    "probability": 0.9111328125
                },
                {
                    "start": 1159.93,
                    "end": 1160.17,
                    "word": " image",
                    "probability": 0.8388671875
                },
                {
                    "start": 1160.17,
                    "end": 1160.59,
                    "word": " classifier",
                    "probability": 0.86083984375
                }
            ]
        },
        {
            "id": 171,
            "seek": 116092,
            "start": 1160.59,
            "end": 1167.05,
            "text": " supervised learning. Let's say we have an image classifier trained on a set of images of pandas,",
            "tokens": [
                51360,
                46533,
                2539,
                13,
                961,
                311,
                584,
                321,
                362,
                364,
                3256,
                1508,
                9902,
                8895,
                322,
                257,
                992,
                295,
                5267,
                295,
                4565,
                296,
                11,
                51676
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21189692407323604,
            "compression_ratio": 1.7435897435897436,
            "no_speech_prob": 0.00555419921875,
            "words": [
                {
                    "start": 1160.59,
                    "end": 1160.91,
                    "word": " supervised",
                    "probability": 0.58251953125
                },
                {
                    "start": 1160.91,
                    "end": 1161.29,
                    "word": " learning",
                    "probability": 0.80908203125
                },
                {
                    "start": 1161.29,
                    "end": 1161.43,
                    "word": ".",
                    "probability": 0.85498046875
                },
                {
                    "start": 1161.43,
                    "end": 1161.55,
                    "word": " Let's",
                    "probability": 0.94189453125
                },
                {
                    "start": 1161.55,
                    "end": 1161.63,
                    "word": " say",
                    "probability": 0.89453125
                },
                {
                    "start": 1161.63,
                    "end": 1161.73,
                    "word": " we",
                    "probability": 0.89013671875
                },
                {
                    "start": 1161.73,
                    "end": 1161.85,
                    "word": " have",
                    "probability": 0.87548828125
                },
                {
                    "start": 1161.85,
                    "end": 1161.95,
                    "word": " an",
                    "probability": 0.89990234375
                },
                {
                    "start": 1161.95,
                    "end": 1162.13,
                    "word": " image",
                    "probability": 0.8466796875
                },
                {
                    "start": 1162.13,
                    "end": 1162.71,
                    "word": " classifier",
                    "probability": 0.964599609375
                },
                {
                    "start": 1162.71,
                    "end": 1164.19,
                    "word": " trained",
                    "probability": 0.58447265625
                },
                {
                    "start": 1164.19,
                    "end": 1164.67,
                    "word": " on",
                    "probability": 0.9091796875
                },
                {
                    "start": 1165.09,
                    "end": 1165.39,
                    "word": " a set",
                    "probability": 0.880859375
                },
                {
                    "start": 1165.39,
                    "end": 1165.67,
                    "word": " of",
                    "probability": 0.91162109375
                },
                {
                    "start": 1165.67,
                    "end": 1166.07,
                    "word": " images",
                    "probability": 0.857421875
                },
                {
                    "start": 1166.07,
                    "end": 1166.33,
                    "word": " of",
                    "probability": 0.87060546875
                },
                {
                    "start": 1166.33,
                    "end": 1166.79,
                    "word": " pandas",
                    "probability": 0.89453125
                },
                {
                    "start": 1166.79,
                    "end": 1167.05,
                    "word": ",",
                    "probability": 0.96728515625
                }
            ]
        },
        {
            "id": 172,
            "seek": 118490,
            "start": 1167.05,
            "end": 1173.29,
            "text": " cats, and others to identify the object in those images. In the beginning, we pass the image of a",
            "tokens": [
                50372,
                11111,
                11,
                293,
                2357,
                281,
                5876,
                264,
                2657,
                294,
                729,
                5267,
                13,
                682,
                264,
                2863,
                11,
                321,
                1320,
                264,
                3256,
                295,
                257,
                50656
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22837936618300372,
            "compression_ratio": 1.7632850241545894,
            "no_speech_prob": 0.00970458984375,
            "words": [
                {
                    "start": 1167.05,
                    "end": 1167.25,
                    "word": " cats",
                    "probability": 0.000202178955078125
                },
                {
                    "start": 1167.25,
                    "end": 1167.39,
                    "word": ",",
                    "probability": 0.32470703125
                },
                {
                    "start": 1167.39,
                    "end": 1167.57,
                    "word": " and",
                    "probability": 0.7548828125
                },
                {
                    "start": 1167.57,
                    "end": 1168.05,
                    "word": " others",
                    "probability": 0.7099609375
                },
                {
                    "start": 1168.05,
                    "end": 1169.49,
                    "word": " to",
                    "probability": 0.66455078125
                },
                {
                    "start": 1169.49,
                    "end": 1169.91,
                    "word": " identify",
                    "probability": 0.8779296875
                },
                {
                    "start": 1169.91,
                    "end": 1170.09,
                    "word": " the",
                    "probability": 0.8037109375
                },
                {
                    "start": 1170.09,
                    "end": 1170.33,
                    "word": " object",
                    "probability": 0.78125
                },
                {
                    "start": 1170.33,
                    "end": 1170.47,
                    "word": " in",
                    "probability": 0.82421875
                },
                {
                    "start": 1170.47,
                    "end": 1170.63,
                    "word": " those",
                    "probability": 0.81689453125
                },
                {
                    "start": 1170.63,
                    "end": 1170.93,
                    "word": " images",
                    "probability": 0.83837890625
                },
                {
                    "start": 1170.93,
                    "end": 1171.27,
                    "word": ".",
                    "probability": 0.9560546875
                },
                {
                    "start": 1171.27,
                    "end": 1171.37,
                    "word": " In",
                    "probability": 0.8818359375
                },
                {
                    "start": 1171.37,
                    "end": 1171.51,
                    "word": " the",
                    "probability": 0.81787109375
                },
                {
                    "start": 1171.51,
                    "end": 1171.79,
                    "word": " beginning",
                    "probability": 0.78759765625
                },
                {
                    "start": 1171.79,
                    "end": 1171.91,
                    "word": ",",
                    "probability": 0.69482421875
                },
                {
                    "start": 1171.91,
                    "end": 1171.99,
                    "word": " we",
                    "probability": 0.892578125
                },
                {
                    "start": 1171.99,
                    "end": 1172.25,
                    "word": " pass",
                    "probability": 0.6240234375
                },
                {
                    "start": 1172.25,
                    "end": 1172.37,
                    "word": " the",
                    "probability": 0.7685546875
                },
                {
                    "start": 1172.37,
                    "end": 1172.55,
                    "word": " image",
                    "probability": 0.7958984375
                },
                {
                    "start": 1172.55,
                    "end": 1172.79,
                    "word": " of",
                    "probability": 0.9072265625
                },
                {
                    "start": 1172.99,
                    "end": 1173.29,
                    "word": " a panda",
                    "probability": 0.73193359375
                }
            ]
        },
        {
            "id": 173,
            "seek": 118490,
            "start": 1173.29,
            "end": 1179.59,
            "text": " panda to the classifier, and it correctly classifies it as panda. Now, it's been demonstrated,",
            "tokens": [
                50656,
                46685,
                281,
                264,
                1508,
                9902,
                11,
                293,
                309,
                8944,
                1508,
                11221,
                309,
                382,
                46685,
                13,
                823,
                11,
                309,
                311,
                668,
                18772,
                11,
                50976
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22837936618300372,
            "compression_ratio": 1.7632850241545894,
            "no_speech_prob": 0.00970458984375,
            "words": [
                {
                    "start": 1173.29,
                    "end": 1173.71,
                    "word": " to",
                    "probability": 0.87841796875
                },
                {
                    "start": 1173.71,
                    "end": 1174.39,
                    "word": " the",
                    "probability": 0.783203125
                },
                {
                    "start": 1174.39,
                    "end": 1174.91,
                    "word": " classifier",
                    "probability": 0.95654296875
                },
                {
                    "start": 1174.91,
                    "end": 1175.11,
                    "word": ",",
                    "probability": 0.4140625
                },
                {
                    "start": 1175.11,
                    "end": 1175.17,
                    "word": " and",
                    "probability": 0.88134765625
                },
                {
                    "start": 1175.17,
                    "end": 1175.31,
                    "word": " it",
                    "probability": 0.89990234375
                },
                {
                    "start": 1175.31,
                    "end": 1175.65,
                    "word": " correctly",
                    "probability": 0.81640625
                },
                {
                    "start": 1175.65,
                    "end": 1176.13,
                    "word": " classifies",
                    "probability": 0.96337890625
                },
                {
                    "start": 1176.13,
                    "end": 1176.29,
                    "word": " it",
                    "probability": 0.89306640625
                },
                {
                    "start": 1176.29,
                    "end": 1176.49,
                    "word": " as",
                    "probability": 0.91943359375
                },
                {
                    "start": 1176.49,
                    "end": 1176.77,
                    "word": " panda",
                    "probability": 0.434326171875
                },
                {
                    "start": 1176.77,
                    "end": 1177.27,
                    "word": ".",
                    "probability": 0.98974609375
                },
                {
                    "start": 1177.45,
                    "end": 1177.69,
                    "word": " Now",
                    "probability": 0.7724609375
                },
                {
                    "start": 1177.69,
                    "end": 1177.83,
                    "word": ",",
                    "probability": 0.75
                },
                {
                    "start": 1177.83,
                    "end": 1178.01,
                    "word": " it's",
                    "probability": 0.920166015625
                },
                {
                    "start": 1178.01,
                    "end": 1178.21,
                    "word": " been",
                    "probability": 0.8232421875
                },
                {
                    "start": 1178.21,
                    "end": 1178.87,
                    "word": " demonstrated",
                    "probability": 0.865234375
                },
                {
                    "start": 1178.87,
                    "end": 1179.39,
                    "word": ",",
                    "probability": 0.646484375
                },
                {
                    "start": 1179.39,
                    "end": 1179.59,
                    "word": " it's",
                    "probability": 0.929931640625
                }
            ]
        },
        {
            "id": 174,
            "seek": 118490,
            "start": 1179.59,
            "end": 1187.49,
            "text": " it's been actually established by now, that it is possible to make the classifier,",
            "tokens": [
                50976,
                309,
                311,
                668,
                767,
                7545,
                538,
                586,
                11,
                300,
                309,
                307,
                1944,
                281,
                652,
                264,
                1508,
                9902,
                11,
                51268
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22837936618300372,
            "compression_ratio": 1.7632850241545894,
            "no_speech_prob": 0.00970458984375,
            "words": [
                {
                    "start": 1179.59,
                    "end": 1179.73,
                    "word": " been",
                    "probability": 0.8173828125
                },
                {
                    "start": 1179.73,
                    "end": 1180.03,
                    "word": " actually",
                    "probability": 0.705078125
                },
                {
                    "start": 1180.03,
                    "end": 1180.45,
                    "word": " established",
                    "probability": 0.841796875
                },
                {
                    "start": 1180.45,
                    "end": 1180.73,
                    "word": " by",
                    "probability": 0.8896484375
                },
                {
                    "start": 1180.73,
                    "end": 1180.99,
                    "word": " now",
                    "probability": 0.87109375
                },
                {
                    "start": 1180.99,
                    "end": 1181.35,
                    "word": ",",
                    "probability": 0.748046875
                },
                {
                    "start": 1181.35,
                    "end": 1181.71,
                    "word": " that",
                    "probability": 0.8701171875
                },
                {
                    "start": 1181.71,
                    "end": 1182.29,
                    "word": " it",
                    "probability": 0.890625
                },
                {
                    "start": 1182.29,
                    "end": 1182.47,
                    "word": " is",
                    "probability": 0.88671875
                },
                {
                    "start": 1182.47,
                    "end": 1183.03,
                    "word": " possible",
                    "probability": 0.84228515625
                },
                {
                    "start": 1183.03,
                    "end": 1183.85,
                    "word": " to",
                    "probability": 0.9033203125
                },
                {
                    "start": 1183.85,
                    "end": 1184.23,
                    "word": " make",
                    "probability": 0.61572265625
                },
                {
                    "start": 1184.23,
                    "end": 1184.41,
                    "word": " the",
                    "probability": 0.8095703125
                },
                {
                    "start": 1184.41,
                    "end": 1185.19,
                    "word": " classifier",
                    "probability": 0.953857421875
                },
                {
                    "start": 1185.19,
                    "end": 1186.01,
                    "word": ",",
                    "probability": 0.2255859375
                },
                {
                    "start": 1186.01,
                    "end": 1187.05,
                    "word": " to",
                    "probability": 0.87109375
                },
                {
                    "start": 1187.05,
                    "end": 1187.49,
                    "word": " induce",
                    "probability": 0.71923828125
                }
            ]
        },
        {
            "id": 175,
            "seek": 118490,
            "start": 1187.49,
            "end": 1191.03,
            "text": " to induce incorrect classifications in deep learning models or machine learning models in",
            "tokens": [
                51352,
                281,
                41263,
                18424,
                1508,
                7833,
                294,
                2452,
                2539,
                5245,
                420,
                3479,
                2539,
                5245,
                294,
                51560
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22837936618300372,
            "compression_ratio": 1.7632850241545894,
            "no_speech_prob": 0.00970458984375,
            "words": [
                {
                    "start": 1187.49,
                    "end": 1188.11,
                    "word": " incorrect",
                    "probability": 0.92333984375
                },
                {
                    "start": 1188.11,
                    "end": 1189.07,
                    "word": " classifications",
                    "probability": 0.94287109375
                },
                {
                    "start": 1189.07,
                    "end": 1189.41,
                    "word": " in",
                    "probability": 0.873046875
                },
                {
                    "start": 1189.41,
                    "end": 1189.59,
                    "word": " deep",
                    "probability": 0.85888671875
                },
                {
                    "start": 1189.59,
                    "end": 1189.79,
                    "word": " learning",
                    "probability": 0.7890625
                },
                {
                    "start": 1189.79,
                    "end": 1190.07,
                    "word": " models",
                    "probability": 0.767578125
                },
                {
                    "start": 1190.07,
                    "end": 1190.23,
                    "word": " or",
                    "probability": 0.5712890625
                },
                {
                    "start": 1190.23,
                    "end": 1190.45,
                    "word": " machine",
                    "probability": 0.67138671875
                },
                {
                    "start": 1190.45,
                    "end": 1190.65,
                    "word": " learning",
                    "probability": 0.80712890625
                },
                {
                    "start": 1190.65,
                    "end": 1190.87,
                    "word": " models",
                    "probability": 0.7998046875
                },
                {
                    "start": 1190.87,
                    "end": 1191.03,
                    "word": " in",
                    "probability": 0.861328125
                }
            ]
        },
        {
            "id": 176,
            "seek": 121404,
            "start": 1191.03,
            "end": 1198.59,
            "text": " general by adding minute, minimal perturbations to the original image. As you can see,",
            "tokens": [
                50364,
                2674,
                538,
                5127,
                3456,
                11,
                13206,
                40468,
                763,
                281,
                264,
                3380,
                3256,
                13,
                1018,
                291,
                393,
                536,
                11,
                50708
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2292968787252903,
            "compression_ratio": 1.64,
            "no_speech_prob": 0.00555419921875,
            "words": [
                {
                    "start": 1191.03,
                    "end": 1191.29,
                    "word": " general",
                    "probability": 0.4189453125
                },
                {
                    "start": 1191.29,
                    "end": 1191.55,
                    "word": " by",
                    "probability": 0.60791015625
                },
                {
                    "start": 1191.55,
                    "end": 1192.21,
                    "word": " adding",
                    "probability": 0.80322265625
                },
                {
                    "start": 1192.21,
                    "end": 1194.95,
                    "word": " minute",
                    "probability": 0.45166015625
                },
                {
                    "start": 1194.95,
                    "end": 1195.19,
                    "word": ",",
                    "probability": 0.432861328125
                },
                {
                    "start": 1195.19,
                    "end": 1195.67,
                    "word": " minimal",
                    "probability": 0.82861328125
                },
                {
                    "start": 1195.67,
                    "end": 1196.55,
                    "word": " perturbations",
                    "probability": 0.86669921875
                },
                {
                    "start": 1196.55,
                    "end": 1196.69,
                    "word": " to",
                    "probability": 0.88232421875
                },
                {
                    "start": 1196.69,
                    "end": 1196.77,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 1196.77,
                    "end": 1197.05,
                    "word": " original",
                    "probability": 0.83154296875
                },
                {
                    "start": 1197.05,
                    "end": 1197.33,
                    "word": " image",
                    "probability": 0.837890625
                },
                {
                    "start": 1197.33,
                    "end": 1197.39,
                    "word": ".",
                    "probability": 0.89013671875
                },
                {
                    "start": 1197.39,
                    "end": 1197.49,
                    "word": " As",
                    "probability": 0.8828125
                },
                {
                    "start": 1197.49,
                    "end": 1197.57,
                    "word": " you",
                    "probability": 0.89599609375
                },
                {
                    "start": 1197.57,
                    "end": 1197.71,
                    "word": " can",
                    "probability": 0.89990234375
                },
                {
                    "start": 1197.71,
                    "end": 1198.03,
                    "word": " see",
                    "probability": 0.81103515625
                },
                {
                    "start": 1198.03,
                    "end": 1198.59,
                    "word": ",",
                    "probability": 0.8232421875
                }
            ]
        },
        {
            "id": 177,
            "seek": 121404,
            "start": 1198.59,
            "end": 1206.71,
            "text": " it's almost impossible to detect or see any changes in this final image. These perturbations,",
            "tokens": [
                50740,
                309,
                311,
                1920,
                6243,
                281,
                5531,
                420,
                536,
                604,
                2962,
                294,
                341,
                2572,
                3256,
                13,
                1981,
                40468,
                763,
                11,
                51140
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2292968787252903,
            "compression_ratio": 1.64,
            "no_speech_prob": 0.00555419921875,
            "words": [
                {
                    "start": 1198.59,
                    "end": 1198.85,
                    "word": " it's",
                    "probability": 0.9296875
                },
                {
                    "start": 1198.85,
                    "end": 1199.19,
                    "word": " almost",
                    "probability": 0.65966796875
                },
                {
                    "start": 1199.19,
                    "end": 1199.79,
                    "word": " impossible",
                    "probability": 0.78369140625
                },
                {
                    "start": 1199.79,
                    "end": 1200.39,
                    "word": " to",
                    "probability": 0.90576171875
                },
                {
                    "start": 1200.39,
                    "end": 1201.83,
                    "word": " detect",
                    "probability": 0.9267578125
                },
                {
                    "start": 1201.83,
                    "end": 1202.37,
                    "word": " or",
                    "probability": 0.85986328125
                },
                {
                    "start": 1202.37,
                    "end": 1203.59,
                    "word": " see",
                    "probability": 0.7939453125
                },
                {
                    "start": 1203.59,
                    "end": 1204.23,
                    "word": " any",
                    "probability": 0.81396484375
                },
                {
                    "start": 1204.23,
                    "end": 1204.63,
                    "word": " changes",
                    "probability": 0.81298828125
                },
                {
                    "start": 1204.63,
                    "end": 1204.79,
                    "word": " in",
                    "probability": 0.90283203125
                },
                {
                    "start": 1204.79,
                    "end": 1204.91,
                    "word": " this",
                    "probability": 0.84912109375
                },
                {
                    "start": 1204.91,
                    "end": 1205.21,
                    "word": " final",
                    "probability": 0.8828125
                },
                {
                    "start": 1205.21,
                    "end": 1205.63,
                    "word": " image",
                    "probability": 0.8349609375
                },
                {
                    "start": 1205.63,
                    "end": 1205.85,
                    "word": ".",
                    "probability": 0.880859375
                },
                {
                    "start": 1205.85,
                    "end": 1206.05,
                    "word": " These",
                    "probability": 0.9091796875
                },
                {
                    "start": 1206.05,
                    "end": 1206.61,
                    "word": " perturbations",
                    "probability": 0.57806396484375
                },
                {
                    "start": 1206.61,
                    "end": 1206.71,
                    "word": ",",
                    "probability": 0.8466796875
                }
            ]
        },
        {
            "id": 178,
            "seek": 121404,
            "start": 1206.71,
            "end": 1215.89,
            "text": " these pixel perturbations are very small. Now, this is one example of different attacks or",
            "tokens": [
                51140,
                613,
                19261,
                40468,
                763,
                366,
                588,
                1359,
                13,
                823,
                11,
                341,
                307,
                472,
                1365,
                295,
                819,
                8122,
                420,
                51604
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2292968787252903,
            "compression_ratio": 1.64,
            "no_speech_prob": 0.00555419921875,
            "words": [
                {
                    "start": 1206.71,
                    "end": 1206.83,
                    "word": " these",
                    "probability": 0.71875
                },
                {
                    "start": 1206.83,
                    "end": 1207.17,
                    "word": " pixel",
                    "probability": 0.79833984375
                },
                {
                    "start": 1207.17,
                    "end": 1207.73,
                    "word": " perturbations",
                    "probability": 0.8720703125
                },
                {
                    "start": 1207.73,
                    "end": 1207.97,
                    "word": " are",
                    "probability": 0.57275390625
                },
                {
                    "start": 1207.97,
                    "end": 1208.19,
                    "word": " very",
                    "probability": 0.7392578125
                },
                {
                    "start": 1208.19,
                    "end": 1208.49,
                    "word": " small",
                    "probability": 0.84912109375
                },
                {
                    "start": 1208.49,
                    "end": 1209.13,
                    "word": ".",
                    "probability": 0.98828125
                },
                {
                    "start": 1209.25,
                    "end": 1209.47,
                    "word": " Now",
                    "probability": 0.83251953125
                },
                {
                    "start": 1209.47,
                    "end": 1210.17,
                    "word": ",",
                    "probability": 0.728515625
                },
                {
                    "start": 1210.17,
                    "end": 1211.25,
                    "word": " this",
                    "probability": 0.87548828125
                },
                {
                    "start": 1211.25,
                    "end": 1211.43,
                    "word": " is",
                    "probability": 0.90869140625
                },
                {
                    "start": 1211.43,
                    "end": 1211.89,
                    "word": " one",
                    "probability": 0.8203125
                },
                {
                    "start": 1211.89,
                    "end": 1212.47,
                    "word": " example",
                    "probability": 0.9296875
                },
                {
                    "start": 1212.47,
                    "end": 1213.39,
                    "word": " of",
                    "probability": 0.90234375
                },
                {
                    "start": 1213.39,
                    "end": 1215.05,
                    "word": " different",
                    "probability": 0.67822265625
                },
                {
                    "start": 1215.05,
                    "end": 1215.53,
                    "word": " attacks",
                    "probability": 0.69921875
                },
                {
                    "start": 1215.53,
                    "end": 1215.89,
                    "word": " or",
                    "probability": 0.818359375
                }
            ]
        },
        {
            "id": 179,
            "seek": 121404,
            "start": 1215.89,
            "end": 1220.17,
            "text": " different vulnerabilities in the classical realm of adversarial machine learning, which is mostly",
            "tokens": [
                51604,
                819,
                37633,
                294,
                264,
                13735,
                15355,
                295,
                17641,
                44745,
                3479,
                2539,
                11,
                597,
                307,
                5240,
                51820
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2292968787252903,
            "compression_ratio": 1.64,
            "no_speech_prob": 0.00555419921875,
            "words": [
                {
                    "start": 1215.89,
                    "end": 1216.13,
                    "word": " different",
                    "probability": 0.7021484375
                },
                {
                    "start": 1216.13,
                    "end": 1216.65,
                    "word": " vulnerabilities",
                    "probability": 0.82666015625
                },
                {
                    "start": 1216.65,
                    "end": 1217.49,
                    "word": " in",
                    "probability": 0.884765625
                },
                {
                    "start": 1217.49,
                    "end": 1217.59,
                    "word": " the",
                    "probability": 0.8046875
                },
                {
                    "start": 1217.59,
                    "end": 1217.95,
                    "word": " classical",
                    "probability": 0.90185546875
                },
                {
                    "start": 1217.95,
                    "end": 1218.25,
                    "word": " realm",
                    "probability": 0.82470703125
                },
                {
                    "start": 1218.25,
                    "end": 1218.37,
                    "word": " of",
                    "probability": 0.87646484375
                },
                {
                    "start": 1218.37,
                    "end": 1218.81,
                    "word": " adversarial",
                    "probability": 0.923583984375
                },
                {
                    "start": 1218.81,
                    "end": 1219.13,
                    "word": " machine",
                    "probability": 0.60302734375
                },
                {
                    "start": 1219.13,
                    "end": 1219.45,
                    "word": " learning",
                    "probability": 0.8359375
                },
                {
                    "start": 1219.45,
                    "end": 1219.55,
                    "word": ",",
                    "probability": 0.8916015625
                },
                {
                    "start": 1219.55,
                    "end": 1219.81,
                    "word": " which",
                    "probability": 0.90869140625
                },
                {
                    "start": 1219.81,
                    "end": 1219.95,
                    "word": " is",
                    "probability": 0.9052734375
                },
                {
                    "start": 1219.95,
                    "end": 1220.17,
                    "word": " mostly",
                    "probability": 0.8408203125
                }
            ]
        },
        {
            "id": 180,
            "seek": 123954,
            "start": 1220.17,
            "end": 1225.87,
            "text": " concerned with supervised learning and sometimes unsupervised learning. In general, the adversarial",
            "tokens": [
                50364,
                5922,
                365,
                46533,
                2539,
                293,
                2171,
                2693,
                12879,
                24420,
                2539,
                13,
                682,
                2674,
                11,
                264,
                17641,
                44745,
                50652
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20192307558579323,
            "compression_ratio": 1.7302325581395348,
            "no_speech_prob": 0.0003249645233154297,
            "words": [
                {
                    "start": 1220.17,
                    "end": 1220.53,
                    "word": " concerned",
                    "probability": 0.1458740234375
                },
                {
                    "start": 1220.53,
                    "end": 1220.93,
                    "word": " with",
                    "probability": 0.8154296875
                },
                {
                    "start": 1220.93,
                    "end": 1221.41,
                    "word": " supervised",
                    "probability": 0.7763671875
                },
                {
                    "start": 1221.41,
                    "end": 1221.77,
                    "word": " learning",
                    "probability": 0.82421875
                },
                {
                    "start": 1221.77,
                    "end": 1222.71,
                    "word": " and",
                    "probability": 0.6865234375
                },
                {
                    "start": 1222.71,
                    "end": 1223.21,
                    "word": " sometimes",
                    "probability": 0.86181640625
                },
                {
                    "start": 1223.21,
                    "end": 1223.85,
                    "word": " unsupervised",
                    "probability": 0.8977864583333334
                },
                {
                    "start": 1223.85,
                    "end": 1224.19,
                    "word": " learning",
                    "probability": 0.83740234375
                },
                {
                    "start": 1224.19,
                    "end": 1224.69,
                    "word": ".",
                    "probability": 0.9208984375
                },
                {
                    "start": 1224.69,
                    "end": 1224.77,
                    "word": " In",
                    "probability": 0.91552734375
                },
                {
                    "start": 1224.77,
                    "end": 1225.23,
                    "word": " general",
                    "probability": 0.8076171875
                },
                {
                    "start": 1225.23,
                    "end": 1225.33,
                    "word": ",",
                    "probability": 0.8359375
                },
                {
                    "start": 1225.33,
                    "end": 1225.43,
                    "word": " the",
                    "probability": 0.79931640625
                },
                {
                    "start": 1225.43,
                    "end": 1225.87,
                    "word": " adversarial",
                    "probability": 0.894287109375
                }
            ]
        },
        {
            "id": 181,
            "seek": 123954,
            "start": 1225.87,
            "end": 1234.35,
            "text": " objectives in AML or adversarial machine learning can be classified under the traditional CIA",
            "tokens": [
                50652,
                15961,
                294,
                6475,
                43,
                420,
                17641,
                44745,
                3479,
                2539,
                393,
                312,
                20627,
                833,
                264,
                5164,
                25143,
                51076
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20192307558579323,
            "compression_ratio": 1.7302325581395348,
            "no_speech_prob": 0.0003249645233154297,
            "words": [
                {
                    "start": 1225.87,
                    "end": 1226.31,
                    "word": " objectives",
                    "probability": 0.755859375
                },
                {
                    "start": 1226.31,
                    "end": 1226.81,
                    "word": " in",
                    "probability": 0.89990234375
                },
                {
                    "start": 1226.81,
                    "end": 1228.35,
                    "word": " AML",
                    "probability": 0.8896484375
                },
                {
                    "start": 1228.35,
                    "end": 1228.49,
                    "word": " or",
                    "probability": 0.5048828125
                },
                {
                    "start": 1228.49,
                    "end": 1228.81,
                    "word": " adversarial",
                    "probability": 0.8798828125
                },
                {
                    "start": 1228.81,
                    "end": 1229.15,
                    "word": " machine",
                    "probability": 0.67724609375
                },
                {
                    "start": 1229.15,
                    "end": 1229.49,
                    "word": " learning",
                    "probability": 0.83447265625
                },
                {
                    "start": 1229.49,
                    "end": 1230.15,
                    "word": " can",
                    "probability": 0.82177734375
                },
                {
                    "start": 1230.15,
                    "end": 1230.51,
                    "word": " be",
                    "probability": 0.9013671875
                },
                {
                    "start": 1230.51,
                    "end": 1231.81,
                    "word": " classified",
                    "probability": 0.93798828125
                },
                {
                    "start": 1231.81,
                    "end": 1232.19,
                    "word": " under",
                    "probability": 0.814453125
                },
                {
                    "start": 1232.19,
                    "end": 1232.59,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 1232.59,
                    "end": 1233.51,
                    "word": " traditional",
                    "probability": 0.71337890625
                },
                {
                    "start": 1233.51,
                    "end": 1234.35,
                    "word": " CIA",
                    "probability": 0.75390625
                }
            ]
        },
        {
            "id": 182,
            "seek": 123954,
            "start": 1234.97,
            "end": 1239.01,
            "text": " triad, confidentiality, integrity, and availability. So, with respect to confidentiality",
            "tokens": [
                51104,
                1376,
                345,
                11,
                27054,
                507,
                11,
                16000,
                11,
                293,
                17945,
                13,
                407,
                11,
                365,
                3104,
                281,
                27054,
                507,
                51328
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20192307558579323,
            "compression_ratio": 1.7302325581395348,
            "no_speech_prob": 0.0003249645233154297,
            "words": [
                {
                    "start": 1234.97,
                    "end": 1235.45,
                    "word": " triad",
                    "probability": 0.84521484375
                },
                {
                    "start": 1235.45,
                    "end": 1235.57,
                    "word": ",",
                    "probability": 0.40087890625
                },
                {
                    "start": 1235.57,
                    "end": 1236.19,
                    "word": " confidentiality",
                    "probability": 0.8173828125
                },
                {
                    "start": 1236.19,
                    "end": 1236.27,
                    "word": ",",
                    "probability": 0.921875
                },
                {
                    "start": 1236.27,
                    "end": 1236.51,
                    "word": " integrity",
                    "probability": 0.71435546875
                },
                {
                    "start": 1236.51,
                    "end": 1236.53,
                    "word": ",",
                    "probability": 0.52783203125
                },
                {
                    "start": 1236.53,
                    "end": 1236.71,
                    "word": " and",
                    "probability": 0.89111328125
                },
                {
                    "start": 1236.71,
                    "end": 1237.09,
                    "word": " availability",
                    "probability": 0.92822265625
                },
                {
                    "start": 1237.09,
                    "end": 1237.53,
                    "word": ".",
                    "probability": 0.95849609375
                },
                {
                    "start": 1237.53,
                    "end": 1237.57,
                    "word": " So",
                    "probability": 0.82080078125
                },
                {
                    "start": 1237.57,
                    "end": 1237.65,
                    "word": ",",
                    "probability": 0.166259765625
                },
                {
                    "start": 1237.65,
                    "end": 1237.69,
                    "word": " with",
                    "probability": 0.81103515625
                },
                {
                    "start": 1237.69,
                    "end": 1237.93,
                    "word": " respect",
                    "probability": 0.77197265625
                },
                {
                    "start": 1237.93,
                    "end": 1238.11,
                    "word": " to",
                    "probability": 0.91162109375
                },
                {
                    "start": 1238.11,
                    "end": 1239.01,
                    "word": " confidentiality",
                    "probability": 0.924560546875
                }
            ]
        },
        {
            "id": 183,
            "seek": 123954,
            "start": 1239.01,
            "end": 1245.67,
            "text": " and privacy, an adversary may wish to target the confidentiality of the model parameters,",
            "tokens": [
                51328,
                293,
                11427,
                11,
                364,
                48222,
                815,
                3172,
                281,
                3779,
                264,
                27054,
                507,
                295,
                264,
                2316,
                9834,
                11,
                51632
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20192307558579323,
            "compression_ratio": 1.7302325581395348,
            "no_speech_prob": 0.0003249645233154297,
            "words": [
                {
                    "start": 1239.01,
                    "end": 1239.65,
                    "word": " and",
                    "probability": 0.8505859375
                },
                {
                    "start": 1239.65,
                    "end": 1240.15,
                    "word": " privacy",
                    "probability": 0.7802734375
                },
                {
                    "start": 1240.15,
                    "end": 1240.79,
                    "word": ",",
                    "probability": 0.92333984375
                },
                {
                    "start": 1240.79,
                    "end": 1240.89,
                    "word": " an",
                    "probability": 0.89990234375
                },
                {
                    "start": 1240.89,
                    "end": 1241.33,
                    "word": " adversary",
                    "probability": 0.8408203125
                },
                {
                    "start": 1241.33,
                    "end": 1242.09,
                    "word": " may",
                    "probability": 0.8759765625
                },
                {
                    "start": 1242.09,
                    "end": 1242.45,
                    "word": " wish",
                    "probability": 0.81201171875
                },
                {
                    "start": 1242.45,
                    "end": 1242.79,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 1242.79,
                    "end": 1243.41,
                    "word": " target",
                    "probability": 0.86474609375
                },
                {
                    "start": 1243.41,
                    "end": 1243.57,
                    "word": " the",
                    "probability": 0.8076171875
                },
                {
                    "start": 1243.57,
                    "end": 1244.29,
                    "word": " confidentiality",
                    "probability": 0.927734375
                },
                {
                    "start": 1244.29,
                    "end": 1244.53,
                    "word": " of",
                    "probability": 0.91064453125
                },
                {
                    "start": 1244.53,
                    "end": 1244.71,
                    "word": " the",
                    "probability": 0.81640625
                },
                {
                    "start": 1244.71,
                    "end": 1244.93,
                    "word": " model",
                    "probability": 0.8154296875
                },
                {
                    "start": 1244.93,
                    "end": 1245.53,
                    "word": " parameters",
                    "probability": 0.86865234375
                },
                {
                    "start": 1245.53,
                    "end": 1245.67,
                    "word": ",",
                    "probability": 0.8896484375
                }
            ]
        },
        {
            "id": 184,
            "seek": 126486,
            "start": 1245.67,
            "end": 1253.65,
            "text": " architecture, no intellectual property theft is an issue in larger models, or it may target the",
            "tokens": [
                50372,
                9482,
                11,
                572,
                12576,
                4707,
                28508,
                307,
                364,
                2734,
                294,
                4833,
                5245,
                11,
                420,
                309,
                815,
                3779,
                264,
                50760
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2505580435196559,
            "compression_ratio": 1.7566371681415929,
            "no_speech_prob": 0.033599853515625,
            "words": [
                {
                    "start": 1245.67,
                    "end": 1246.19,
                    "word": " architecture",
                    "probability": 0.1358642578125
                },
                {
                    "start": 1246.19,
                    "end": 1246.95,
                    "word": ",",
                    "probability": 0.54150390625
                },
                {
                    "start": 1246.95,
                    "end": 1247.15,
                    "word": " no",
                    "probability": 0.82958984375
                },
                {
                    "start": 1247.15,
                    "end": 1247.97,
                    "word": " intellectual",
                    "probability": 0.87841796875
                },
                {
                    "start": 1247.97,
                    "end": 1248.45,
                    "word": " property",
                    "probability": 0.7802734375
                },
                {
                    "start": 1248.45,
                    "end": 1248.79,
                    "word": " theft",
                    "probability": 0.8037109375
                },
                {
                    "start": 1248.79,
                    "end": 1249.09,
                    "word": " is",
                    "probability": 0.8125
                },
                {
                    "start": 1249.09,
                    "end": 1249.35,
                    "word": " an",
                    "probability": 0.91357421875
                },
                {
                    "start": 1249.35,
                    "end": 1249.69,
                    "word": " issue",
                    "probability": 0.89111328125
                },
                {
                    "start": 1249.69,
                    "end": 1250.93,
                    "word": " in",
                    "probability": 0.74951171875
                },
                {
                    "start": 1250.93,
                    "end": 1251.25,
                    "word": " larger",
                    "probability": 0.8115234375
                },
                {
                    "start": 1251.25,
                    "end": 1251.65,
                    "word": " models",
                    "probability": 0.80517578125
                },
                {
                    "start": 1251.65,
                    "end": 1251.85,
                    "word": ",",
                    "probability": 0.458251953125
                },
                {
                    "start": 1251.85,
                    "end": 1252.13,
                    "word": " or",
                    "probability": 0.90478515625
                },
                {
                    "start": 1252.13,
                    "end": 1252.77,
                    "word": " it",
                    "probability": 0.8291015625
                },
                {
                    "start": 1252.77,
                    "end": 1253.03,
                    "word": " may",
                    "probability": 0.8818359375
                },
                {
                    "start": 1253.03,
                    "end": 1253.43,
                    "word": " target",
                    "probability": 0.89697265625
                },
                {
                    "start": 1253.43,
                    "end": 1253.65,
                    "word": " the",
                    "probability": 0.8046875
                }
            ]
        },
        {
            "id": 185,
            "seek": 126486,
            "start": 1253.65,
            "end": 1259.31,
            "text": " privacy, the adversary may target the privacy of the training and test data. If medical, for example,",
            "tokens": [
                50760,
                11427,
                11,
                264,
                48222,
                815,
                3779,
                264,
                11427,
                295,
                264,
                3097,
                293,
                1500,
                1412,
                13,
                759,
                4625,
                11,
                337,
                1365,
                11,
                51040
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2505580435196559,
            "compression_ratio": 1.7566371681415929,
            "no_speech_prob": 0.033599853515625,
            "words": [
                {
                    "start": 1253.65,
                    "end": 1254.07,
                    "word": " privacy",
                    "probability": 0.78125
                },
                {
                    "start": 1254.07,
                    "end": 1254.19,
                    "word": ",",
                    "probability": 0.36572265625
                },
                {
                    "start": 1254.19,
                    "end": 1254.27,
                    "word": " the",
                    "probability": 0.79248046875
                },
                {
                    "start": 1254.27,
                    "end": 1254.53,
                    "word": " adversary",
                    "probability": 0.82470703125
                },
                {
                    "start": 1254.53,
                    "end": 1254.75,
                    "word": " may",
                    "probability": 0.8701171875
                },
                {
                    "start": 1254.75,
                    "end": 1255.03,
                    "word": " target",
                    "probability": 0.8740234375
                },
                {
                    "start": 1255.03,
                    "end": 1255.19,
                    "word": " the",
                    "probability": 0.80859375
                },
                {
                    "start": 1255.19,
                    "end": 1255.69,
                    "word": " privacy",
                    "probability": 0.8076171875
                },
                {
                    "start": 1255.69,
                    "end": 1256.03,
                    "word": " of",
                    "probability": 0.88623046875
                },
                {
                    "start": 1256.03,
                    "end": 1256.15,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 1256.15,
                    "end": 1256.49,
                    "word": " training",
                    "probability": 0.81591796875
                },
                {
                    "start": 1256.49,
                    "end": 1256.81,
                    "word": " and",
                    "probability": 0.8974609375
                },
                {
                    "start": 1256.81,
                    "end": 1257.11,
                    "word": " test",
                    "probability": 0.810546875
                },
                {
                    "start": 1257.11,
                    "end": 1257.51,
                    "word": " data",
                    "probability": 0.8115234375
                },
                {
                    "start": 1257.51,
                    "end": 1257.73,
                    "word": ".",
                    "probability": 0.93310546875
                },
                {
                    "start": 1257.85,
                    "end": 1258.41,
                    "word": " If",
                    "probability": 0.8017578125
                },
                {
                    "start": 1258.41,
                    "end": 1258.73,
                    "word": " medical",
                    "probability": 0.90087890625
                },
                {
                    "start": 1258.73,
                    "end": 1258.95,
                    "word": ",",
                    "probability": 0.51611328125
                },
                {
                    "start": 1258.95,
                    "end": 1259.07,
                    "word": " for",
                    "probability": 0.8955078125
                },
                {
                    "start": 1259.07,
                    "end": 1259.27,
                    "word": " example",
                    "probability": 0.9267578125
                },
                {
                    "start": 1259.27,
                    "end": 1259.31,
                    "word": ",",
                    "probability": 0.890625
                }
            ]
        },
        {
            "id": 186,
            "seek": 126486,
            "start": 1259.31,
            "end": 1264.45,
            "text": " if medical records were used in the training data, there have been proof of concept attacks showing",
            "tokens": [
                51040,
                498,
                4625,
                7724,
                645,
                1143,
                294,
                264,
                3097,
                1412,
                11,
                456,
                362,
                668,
                8177,
                295,
                3410,
                8122,
                4099,
                51304
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2505580435196559,
            "compression_ratio": 1.7566371681415929,
            "no_speech_prob": 0.033599853515625,
            "words": [
                {
                    "start": 1259.31,
                    "end": 1259.41,
                    "word": " if",
                    "probability": 0.66796875
                },
                {
                    "start": 1259.41,
                    "end": 1259.59,
                    "word": " medical",
                    "probability": 0.89892578125
                },
                {
                    "start": 1259.59,
                    "end": 1259.93,
                    "word": " records",
                    "probability": 0.76904296875
                },
                {
                    "start": 1259.93,
                    "end": 1260.17,
                    "word": " were",
                    "probability": 0.86669921875
                },
                {
                    "start": 1260.17,
                    "end": 1260.53,
                    "word": " used",
                    "probability": 0.80322265625
                },
                {
                    "start": 1260.53,
                    "end": 1260.65,
                    "word": " in",
                    "probability": 0.908203125
                },
                {
                    "start": 1260.65,
                    "end": 1260.73,
                    "word": " the",
                    "probability": 0.78173828125
                },
                {
                    "start": 1260.73,
                    "end": 1260.95,
                    "word": " training",
                    "probability": 0.8173828125
                },
                {
                    "start": 1260.95,
                    "end": 1261.21,
                    "word": " data",
                    "probability": 0.79296875
                },
                {
                    "start": 1261.21,
                    "end": 1261.43,
                    "word": ",",
                    "probability": 0.92578125
                },
                {
                    "start": 1261.43,
                    "end": 1261.71,
                    "word": " there",
                    "probability": 0.84521484375
                },
                {
                    "start": 1261.71,
                    "end": 1261.97,
                    "word": " have",
                    "probability": 0.73974609375
                },
                {
                    "start": 1261.97,
                    "end": 1262.63,
                    "word": " been",
                    "probability": 0.80859375
                },
                {
                    "start": 1262.63,
                    "end": 1263.49,
                    "word": " proof",
                    "probability": 0.78369140625
                },
                {
                    "start": 1263.49,
                    "end": 1263.61,
                    "word": " of",
                    "probability": 0.84716796875
                },
                {
                    "start": 1263.61,
                    "end": 1263.83,
                    "word": " concept",
                    "probability": 0.80419921875
                },
                {
                    "start": 1263.83,
                    "end": 1264.21,
                    "word": " attacks",
                    "probability": 0.55908203125
                },
                {
                    "start": 1264.21,
                    "end": 1264.45,
                    "word": " showing",
                    "probability": 0.7255859375
                }
            ]
        },
        {
            "id": 187,
            "seek": 126486,
            "start": 1264.45,
            "end": 1270.99,
            "text": " that it is possible to infer whether a particular patient or the records pertaining to a particular",
            "tokens": [
                51304,
                300,
                309,
                307,
                1944,
                281,
                13596,
                1968,
                257,
                1729,
                4537,
                420,
                264,
                7724,
                49582,
                281,
                257,
                1729,
                51632
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2505580435196559,
            "compression_ratio": 1.7566371681415929,
            "no_speech_prob": 0.033599853515625,
            "words": [
                {
                    "start": 1264.45,
                    "end": 1264.79,
                    "word": " that",
                    "probability": 0.87890625
                },
                {
                    "start": 1264.79,
                    "end": 1265.17,
                    "word": " it",
                    "probability": 0.89404296875
                },
                {
                    "start": 1265.17,
                    "end": 1265.33,
                    "word": " is",
                    "probability": 0.89892578125
                },
                {
                    "start": 1265.33,
                    "end": 1265.71,
                    "word": " possible",
                    "probability": 0.84765625
                },
                {
                    "start": 1265.71,
                    "end": 1265.97,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 1265.97,
                    "end": 1266.21,
                    "word": " infer",
                    "probability": 0.87060546875
                },
                {
                    "start": 1266.21,
                    "end": 1266.51,
                    "word": " whether",
                    "probability": 0.85791015625
                },
                {
                    "start": 1266.77,
                    "end": 1267.21,
                    "word": " a particular",
                    "probability": 0.818359375
                },
                {
                    "start": 1267.21,
                    "end": 1268.75,
                    "word": " patient",
                    "probability": 0.64794921875
                },
                {
                    "start": 1268.75,
                    "end": 1269.39,
                    "word": " or",
                    "probability": 0.68896484375
                },
                {
                    "start": 1269.39,
                    "end": 1269.59,
                    "word": " the",
                    "probability": 0.7998046875
                },
                {
                    "start": 1269.59,
                    "end": 1269.95,
                    "word": " records",
                    "probability": 0.77197265625
                },
                {
                    "start": 1269.95,
                    "end": 1270.51,
                    "word": " pertaining",
                    "probability": 0.7666015625
                },
                {
                    "start": 1270.51,
                    "end": 1270.65,
                    "word": " to",
                    "probability": 0.9013671875
                },
                {
                    "start": 1270.75,
                    "end": 1270.99,
                    "word": " a particular",
                    "probability": 0.8115234375
                }
            ]
        },
        {
            "id": 188,
            "seek": 129268,
            "start": 1270.99,
            "end": 1276.29,
            "text": " patient were used in the training data or not, or sometimes it's possible to reconstruct the",
            "tokens": [
                50364,
                4537,
                645,
                1143,
                294,
                264,
                3097,
                1412,
                420,
                406,
                11,
                420,
                2171,
                309,
                311,
                1944,
                281,
                31499,
                264,
                50628
            ],
            "temperature": 0.0,
            "avg_logprob": -0.18979778827405444,
            "compression_ratio": 1.718146718146718,
            "no_speech_prob": 0.0010318756103515625,
            "words": [
                {
                    "start": 1270.99,
                    "end": 1271.35,
                    "word": " patient",
                    "probability": 0.0217132568359375
                },
                {
                    "start": 1271.35,
                    "end": 1271.97,
                    "word": " were",
                    "probability": 0.113525390625
                },
                {
                    "start": 1271.97,
                    "end": 1272.43,
                    "word": " used",
                    "probability": 0.806640625
                },
                {
                    "start": 1272.43,
                    "end": 1272.67,
                    "word": " in",
                    "probability": 0.9052734375
                },
                {
                    "start": 1272.67,
                    "end": 1272.79,
                    "word": " the",
                    "probability": 0.80322265625
                },
                {
                    "start": 1272.79,
                    "end": 1272.99,
                    "word": " training",
                    "probability": 0.80517578125
                },
                {
                    "start": 1272.99,
                    "end": 1273.23,
                    "word": " data",
                    "probability": 0.798828125
                },
                {
                    "start": 1273.23,
                    "end": 1273.39,
                    "word": " or",
                    "probability": 0.88232421875
                },
                {
                    "start": 1273.39,
                    "end": 1273.67,
                    "word": " not",
                    "probability": 0.8994140625
                },
                {
                    "start": 1273.67,
                    "end": 1274.13,
                    "word": ",",
                    "probability": 0.322998046875
                },
                {
                    "start": 1274.13,
                    "end": 1274.23,
                    "word": " or",
                    "probability": 0.88525390625
                },
                {
                    "start": 1274.23,
                    "end": 1274.57,
                    "word": " sometimes",
                    "probability": 0.9091796875
                },
                {
                    "start": 1274.57,
                    "end": 1274.77,
                    "word": " it's",
                    "probability": 0.93701171875
                },
                {
                    "start": 1274.77,
                    "end": 1275.03,
                    "word": " possible",
                    "probability": 0.84619140625
                },
                {
                    "start": 1275.03,
                    "end": 1275.23,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 1275.23,
                    "end": 1275.75,
                    "word": " reconstruct",
                    "probability": 0.8232421875
                },
                {
                    "start": 1275.75,
                    "end": 1276.29,
                    "word": " the",
                    "probability": 0.806640625
                }
            ]
        },
        {
            "id": 189,
            "seek": 129268,
            "start": 1276.29,
            "end": 1281.59,
            "text": " training data set by just having access to the model itself. And that is a major HIPAA violation.",
            "tokens": [
                50628,
                3097,
                1412,
                992,
                538,
                445,
                1419,
                2105,
                281,
                264,
                2316,
                2564,
                13,
                400,
                300,
                307,
                257,
                2563,
                389,
                9139,
                5265,
                22840,
                13,
                50872
            ],
            "temperature": 0.0,
            "avg_logprob": -0.18979778827405444,
            "compression_ratio": 1.718146718146718,
            "no_speech_prob": 0.0010318756103515625,
            "words": [
                {
                    "start": 1276.29,
                    "end": 1276.67,
                    "word": " training",
                    "probability": 0.83349609375
                },
                {
                    "start": 1276.67,
                    "end": 1276.95,
                    "word": " data",
                    "probability": 0.468505859375
                },
                {
                    "start": 1276.95,
                    "end": 1277.23,
                    "word": " set",
                    "probability": 0.705078125
                },
                {
                    "start": 1277.23,
                    "end": 1277.53,
                    "word": " by",
                    "probability": 0.888671875
                },
                {
                    "start": 1277.53,
                    "end": 1277.75,
                    "word": " just",
                    "probability": 0.87939453125
                },
                {
                    "start": 1277.75,
                    "end": 1277.99,
                    "word": " having",
                    "probability": 0.79052734375
                },
                {
                    "start": 1277.99,
                    "end": 1278.29,
                    "word": " access",
                    "probability": 0.783203125
                },
                {
                    "start": 1278.29,
                    "end": 1278.41,
                    "word": " to",
                    "probability": 0.90283203125
                },
                {
                    "start": 1278.41,
                    "end": 1278.51,
                    "word": " the",
                    "probability": 0.8173828125
                },
                {
                    "start": 1278.51,
                    "end": 1278.65,
                    "word": " model",
                    "probability": 0.822265625
                },
                {
                    "start": 1278.65,
                    "end": 1279.01,
                    "word": " itself",
                    "probability": 0.78955078125
                },
                {
                    "start": 1279.01,
                    "end": 1279.13,
                    "word": ".",
                    "probability": 0.473876953125
                },
                {
                    "start": 1279.13,
                    "end": 1279.23,
                    "word": " And",
                    "probability": 0.8076171875
                },
                {
                    "start": 1279.23,
                    "end": 1279.41,
                    "word": " that",
                    "probability": 0.87890625
                },
                {
                    "start": 1279.41,
                    "end": 1279.71,
                    "word": " is",
                    "probability": 0.90478515625
                },
                {
                    "start": 1279.91,
                    "end": 1280.15,
                    "word": " a major",
                    "probability": 0.79345703125
                },
                {
                    "start": 1280.15,
                    "end": 1280.45,
                    "word": " HIPAA",
                    "probability": 0.9703776041666666
                },
                {
                    "start": 1280.45,
                    "end": 1281.07,
                    "word": " violation",
                    "probability": 0.81494140625
                },
                {
                    "start": 1281.07,
                    "end": 1281.59,
                    "word": ".",
                    "probability": 0.71044921875
                }
            ]
        },
        {
            "id": 190,
            "seek": 129268,
            "start": 1282.07,
            "end": 1288.75,
            "text": " In essence, it's a privacy violation. Also, with regards to integrity and availability,",
            "tokens": [
                50932,
                682,
                12801,
                11,
                309,
                311,
                257,
                11427,
                22840,
                13,
                2743,
                11,
                365,
                14258,
                281,
                16000,
                293,
                17945,
                11,
                51240
            ],
            "temperature": 0.0,
            "avg_logprob": -0.18979778827405444,
            "compression_ratio": 1.718146718146718,
            "no_speech_prob": 0.0010318756103515625,
            "words": [
                {
                    "start": 1282.07,
                    "end": 1282.59,
                    "word": " In",
                    "probability": 0.87890625
                },
                {
                    "start": 1282.59,
                    "end": 1282.97,
                    "word": " essence",
                    "probability": 0.828125
                },
                {
                    "start": 1282.97,
                    "end": 1283.09,
                    "word": ",",
                    "probability": 0.87646484375
                },
                {
                    "start": 1283.09,
                    "end": 1283.21,
                    "word": " it's",
                    "probability": 0.951171875
                },
                {
                    "start": 1283.33,
                    "end": 1283.65,
                    "word": " a privacy",
                    "probability": 0.80517578125
                },
                {
                    "start": 1283.65,
                    "end": 1284.13,
                    "word": " violation",
                    "probability": 0.81640625
                },
                {
                    "start": 1284.13,
                    "end": 1284.65,
                    "word": ".",
                    "probability": 0.9599609375
                },
                {
                    "start": 1284.99,
                    "end": 1285.31,
                    "word": " Also",
                    "probability": 0.91552734375
                },
                {
                    "start": 1285.31,
                    "end": 1285.81,
                    "word": ",",
                    "probability": 0.81103515625
                },
                {
                    "start": 1285.81,
                    "end": 1286.75,
                    "word": " with",
                    "probability": 0.7880859375
                },
                {
                    "start": 1286.75,
                    "end": 1286.97,
                    "word": " regards",
                    "probability": 0.7490234375
                },
                {
                    "start": 1286.97,
                    "end": 1287.11,
                    "word": " to",
                    "probability": 0.87255859375
                },
                {
                    "start": 1287.11,
                    "end": 1287.39,
                    "word": " integrity",
                    "probability": 0.72802734375
                },
                {
                    "start": 1287.39,
                    "end": 1287.57,
                    "word": " and",
                    "probability": 0.888671875
                },
                {
                    "start": 1287.57,
                    "end": 1288.09,
                    "word": " availability",
                    "probability": 0.93359375
                },
                {
                    "start": 1288.09,
                    "end": 1288.75,
                    "word": ",",
                    "probability": 0.921875
                }
            ]
        },
        {
            "id": 191,
            "seek": 129268,
            "start": 1288.75,
            "end": 1293.49,
            "text": " the attacker may target the integrity of predictions or the outcome of the model,",
            "tokens": [
                51240,
                264,
                35871,
                815,
                3779,
                264,
                16000,
                295,
                21264,
                420,
                264,
                9700,
                295,
                264,
                2316,
                11,
                51484
            ],
            "temperature": 0.0,
            "avg_logprob": -0.18979778827405444,
            "compression_ratio": 1.718146718146718,
            "no_speech_prob": 0.0010318756103515625,
            "words": [
                {
                    "start": 1288.75,
                    "end": 1288.85,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 1288.85,
                    "end": 1289.07,
                    "word": " attacker",
                    "probability": 0.62744140625
                },
                {
                    "start": 1289.07,
                    "end": 1289.25,
                    "word": " may",
                    "probability": 0.87060546875
                },
                {
                    "start": 1289.25,
                    "end": 1289.59,
                    "word": " target",
                    "probability": 0.84619140625
                },
                {
                    "start": 1289.59,
                    "end": 1289.77,
                    "word": " the",
                    "probability": 0.7724609375
                },
                {
                    "start": 1289.77,
                    "end": 1290.11,
                    "word": " integrity",
                    "probability": 0.783203125
                },
                {
                    "start": 1290.11,
                    "end": 1290.29,
                    "word": " of",
                    "probability": 0.74853515625
                },
                {
                    "start": 1290.29,
                    "end": 1290.77,
                    "word": " predictions",
                    "probability": 0.7431640625
                },
                {
                    "start": 1290.77,
                    "end": 1292.17,
                    "word": " or",
                    "probability": 0.54931640625
                },
                {
                    "start": 1292.17,
                    "end": 1292.37,
                    "word": " the",
                    "probability": 0.81787109375
                },
                {
                    "start": 1292.37,
                    "end": 1292.79,
                    "word": " outcome",
                    "probability": 0.8134765625
                },
                {
                    "start": 1292.79,
                    "end": 1293.03,
                    "word": " of",
                    "probability": 0.90625
                },
                {
                    "start": 1293.03,
                    "end": 1293.15,
                    "word": " the",
                    "probability": 0.82080078125
                },
                {
                    "start": 1293.15,
                    "end": 1293.39,
                    "word": " model",
                    "probability": 0.82568359375
                },
                {
                    "start": 1293.39,
                    "end": 1293.49,
                    "word": ",",
                    "probability": 0.8486328125
                }
            ]
        },
        {
            "id": 192,
            "seek": 129268,
            "start": 1293.49,
            "end": 1298.81,
            "text": " the performance of the model. For example, adversarial examples are an attack on the",
            "tokens": [
                51484,
                264,
                3389,
                295,
                264,
                2316,
                13,
                1171,
                1365,
                11,
                17641,
                44745,
                5110,
                366,
                364,
                2690,
                322,
                264,
                51748
            ],
            "temperature": 0.0,
            "avg_logprob": -0.18979778827405444,
            "compression_ratio": 1.718146718146718,
            "no_speech_prob": 0.0010318756103515625,
            "words": [
                {
                    "start": 1293.49,
                    "end": 1293.65,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 1293.65,
                    "end": 1294.07,
                    "word": " performance",
                    "probability": 0.83251953125
                },
                {
                    "start": 1294.07,
                    "end": 1294.29,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 1294.29,
                    "end": 1294.39,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 1294.39,
                    "end": 1294.63,
                    "word": " model",
                    "probability": 0.8232421875
                },
                {
                    "start": 1294.63,
                    "end": 1295.15,
                    "word": ".",
                    "probability": 0.85546875
                },
                {
                    "start": 1295.33,
                    "end": 1295.83,
                    "word": " For",
                    "probability": 0.90087890625
                },
                {
                    "start": 1295.83,
                    "end": 1296.07,
                    "word": " example",
                    "probability": 0.93212890625
                },
                {
                    "start": 1296.07,
                    "end": 1296.15,
                    "word": ",",
                    "probability": 0.96142578125
                },
                {
                    "start": 1296.15,
                    "end": 1296.53,
                    "word": " adversarial",
                    "probability": 0.92578125
                },
                {
                    "start": 1296.53,
                    "end": 1297.17,
                    "word": " examples",
                    "probability": 0.78662109375
                },
                {
                    "start": 1297.17,
                    "end": 1297.67,
                    "word": " are",
                    "probability": 0.81689453125
                },
                {
                    "start": 1297.67,
                    "end": 1298.17,
                    "word": " an",
                    "probability": 0.89990234375
                },
                {
                    "start": 1298.17,
                    "end": 1298.47,
                    "word": " attack",
                    "probability": 0.8447265625
                },
                {
                    "start": 1298.47,
                    "end": 1298.67,
                    "word": " on",
                    "probability": 0.9072265625
                },
                {
                    "start": 1298.67,
                    "end": 1298.81,
                    "word": " the",
                    "probability": 0.7939453125
                }
            ]
        },
        {
            "id": 193,
            "seek": 132112,
            "start": 1298.81,
            "end": 1304.09,
            "text": " integrity of image classifiers or supervised machine learning models. And also an adversary",
            "tokens": [
                50364,
                16000,
                295,
                3256,
                1508,
                23463,
                420,
                46533,
                3479,
                2539,
                5245,
                13,
                400,
                611,
                364,
                48222,
                50632
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2083333295652236,
            "compression_ratio": 1.9043478260869566,
            "no_speech_prob": 0.00075531005859375,
            "words": [
                {
                    "start": 1298.81,
                    "end": 1299.13,
                    "word": " integrity",
                    "probability": 0.09130859375
                },
                {
                    "start": 1299.13,
                    "end": 1299.81,
                    "word": " of",
                    "probability": 0.90234375
                },
                {
                    "start": 1299.81,
                    "end": 1300.35,
                    "word": " image",
                    "probability": 0.7470703125
                },
                {
                    "start": 1300.35,
                    "end": 1300.89,
                    "word": " classifiers",
                    "probability": 0.916748046875
                },
                {
                    "start": 1300.89,
                    "end": 1301.15,
                    "word": " or",
                    "probability": 0.82275390625
                },
                {
                    "start": 1301.15,
                    "end": 1301.57,
                    "word": " supervised",
                    "probability": 0.6962890625
                },
                {
                    "start": 1301.57,
                    "end": 1301.81,
                    "word": " machine",
                    "probability": 0.6318359375
                },
                {
                    "start": 1301.81,
                    "end": 1302.05,
                    "word": " learning",
                    "probability": 0.7998046875
                },
                {
                    "start": 1302.05,
                    "end": 1302.41,
                    "word": " models",
                    "probability": 0.7939453125
                },
                {
                    "start": 1302.41,
                    "end": 1302.93,
                    "word": ".",
                    "probability": 0.463623046875
                },
                {
                    "start": 1302.93,
                    "end": 1302.99,
                    "word": " And",
                    "probability": 0.6591796875
                },
                {
                    "start": 1302.99,
                    "end": 1303.33,
                    "word": " also",
                    "probability": 0.8046875
                },
                {
                    "start": 1303.33,
                    "end": 1303.81,
                    "word": " an",
                    "probability": 0.59814453125
                },
                {
                    "start": 1303.81,
                    "end": 1304.09,
                    "word": " adversary",
                    "probability": 0.861328125
                }
            ]
        },
        {
            "id": 194,
            "seek": 132112,
            "start": 1304.09,
            "end": 1308.79,
            "text": " may target the availability of the system that is deploying machine learning, for example,",
            "tokens": [
                50632,
                815,
                3779,
                264,
                17945,
                295,
                264,
                1185,
                300,
                307,
                34198,
                3479,
                2539,
                11,
                337,
                1365,
                11,
                50856
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2083333295652236,
            "compression_ratio": 1.9043478260869566,
            "no_speech_prob": 0.00075531005859375,
            "words": [
                {
                    "start": 1304.09,
                    "end": 1304.35,
                    "word": " may",
                    "probability": 0.87255859375
                },
                {
                    "start": 1304.35,
                    "end": 1304.59,
                    "word": " target",
                    "probability": 0.8525390625
                },
                {
                    "start": 1304.59,
                    "end": 1304.73,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 1304.73,
                    "end": 1305.15,
                    "word": " availability",
                    "probability": 0.9248046875
                },
                {
                    "start": 1305.15,
                    "end": 1305.55,
                    "word": " of",
                    "probability": 0.90576171875
                },
                {
                    "start": 1305.55,
                    "end": 1305.69,
                    "word": " the",
                    "probability": 0.80712890625
                },
                {
                    "start": 1305.69,
                    "end": 1306.15,
                    "word": " system",
                    "probability": 0.87060546875
                },
                {
                    "start": 1306.15,
                    "end": 1306.73,
                    "word": " that",
                    "probability": 0.84130859375
                },
                {
                    "start": 1306.73,
                    "end": 1306.87,
                    "word": " is",
                    "probability": 0.9013671875
                },
                {
                    "start": 1306.87,
                    "end": 1307.33,
                    "word": " deploying",
                    "probability": 0.72900390625
                },
                {
                    "start": 1307.33,
                    "end": 1307.89,
                    "word": " machine",
                    "probability": 0.6953125
                },
                {
                    "start": 1307.89,
                    "end": 1308.19,
                    "word": " learning",
                    "probability": 0.83837890625
                },
                {
                    "start": 1308.19,
                    "end": 1308.31,
                    "word": ",",
                    "probability": 0.57666015625
                },
                {
                    "start": 1308.31,
                    "end": 1308.39,
                    "word": " for",
                    "probability": 0.892578125
                },
                {
                    "start": 1308.39,
                    "end": 1308.69,
                    "word": " example",
                    "probability": 0.9296875
                },
                {
                    "start": 1308.69,
                    "end": 1308.79,
                    "word": ",",
                    "probability": 0.82666015625
                }
            ]
        },
        {
            "id": 195,
            "seek": 132112,
            "start": 1308.87,
            "end": 1315.17,
            "text": " a facial recognition system or an autonomous navigation system in a driverless car. Now,",
            "tokens": [
                50856,
                257,
                15642,
                11150,
                1185,
                420,
                364,
                23797,
                17346,
                1185,
                294,
                257,
                6787,
                1832,
                1032,
                13,
                823,
                11,
                51144
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2083333295652236,
            "compression_ratio": 1.9043478260869566,
            "no_speech_prob": 0.00075531005859375,
            "words": [
                {
                    "start": 1308.87,
                    "end": 1309.05,
                    "word": " a facial",
                    "probability": 0.7978515625
                },
                {
                    "start": 1309.05,
                    "end": 1309.39,
                    "word": " recognition",
                    "probability": 0.84033203125
                },
                {
                    "start": 1309.39,
                    "end": 1309.73,
                    "word": " system",
                    "probability": 0.85693359375
                },
                {
                    "start": 1309.73,
                    "end": 1309.83,
                    "word": " or",
                    "probability": 0.7021484375
                },
                {
                    "start": 1309.83,
                    "end": 1309.97,
                    "word": " an",
                    "probability": 0.869140625
                },
                {
                    "start": 1309.97,
                    "end": 1310.21,
                    "word": " autonomous",
                    "probability": 0.8359375
                },
                {
                    "start": 1310.21,
                    "end": 1310.57,
                    "word": " navigation",
                    "probability": 0.77392578125
                },
                {
                    "start": 1310.57,
                    "end": 1311.11,
                    "word": " system",
                    "probability": 0.87744140625
                },
                {
                    "start": 1311.11,
                    "end": 1312.31,
                    "word": " in",
                    "probability": 0.82568359375
                },
                {
                    "start": 1312.61,
                    "end": 1313.05,
                    "word": " a driverless",
                    "probability": 0.885498046875
                },
                {
                    "start": 1313.05,
                    "end": 1313.45,
                    "word": " car",
                    "probability": 0.9013671875
                },
                {
                    "start": 1313.45,
                    "end": 1314.13,
                    "word": ".",
                    "probability": 0.99169921875
                },
                {
                    "start": 1314.45,
                    "end": 1314.47,
                    "word": " Now",
                    "probability": 0.89892578125
                },
                {
                    "start": 1314.47,
                    "end": 1315.17,
                    "word": ",",
                    "probability": 0.419189453125
                }
            ]
        },
        {
            "id": 196,
            "seek": 132112,
            "start": 1316.13,
            "end": 1323.23,
            "text": " now that we know so much about adversarial machine learning and security vulnerabilities",
            "tokens": [
                51252,
                586,
                300,
                321,
                458,
                370,
                709,
                466,
                17641,
                44745,
                3479,
                2539,
                293,
                3825,
                37633,
                51600
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2083333295652236,
            "compression_ratio": 1.9043478260869566,
            "no_speech_prob": 0.00075531005859375,
            "words": [
                {
                    "start": 1316.13,
                    "end": 1316.81,
                    "word": " now",
                    "probability": 0.6669921875
                },
                {
                    "start": 1316.81,
                    "end": 1317.03,
                    "word": " that",
                    "probability": 0.86767578125
                },
                {
                    "start": 1317.03,
                    "end": 1318.73,
                    "word": " we",
                    "probability": 0.8447265625
                },
                {
                    "start": 1318.73,
                    "end": 1319.03,
                    "word": " know",
                    "probability": 0.81298828125
                },
                {
                    "start": 1319.03,
                    "end": 1319.31,
                    "word": " so",
                    "probability": 0.8984375
                },
                {
                    "start": 1319.31,
                    "end": 1319.75,
                    "word": " much",
                    "probability": 0.8115234375
                },
                {
                    "start": 1319.75,
                    "end": 1320.43,
                    "word": " about",
                    "probability": 0.8291015625
                },
                {
                    "start": 1320.43,
                    "end": 1321.39,
                    "word": " adversarial",
                    "probability": 0.89892578125
                },
                {
                    "start": 1321.39,
                    "end": 1321.63,
                    "word": " machine",
                    "probability": 0.673828125
                },
                {
                    "start": 1321.63,
                    "end": 1321.87,
                    "word": " learning",
                    "probability": 0.828125
                },
                {
                    "start": 1321.87,
                    "end": 1322.19,
                    "word": " and",
                    "probability": 0.89306640625
                },
                {
                    "start": 1322.19,
                    "end": 1322.67,
                    "word": " security",
                    "probability": 0.91357421875
                },
                {
                    "start": 1322.67,
                    "end": 1323.23,
                    "word": " vulnerabilities",
                    "probability": 0.830078125
                }
            ]
        },
        {
            "id": 197,
            "seek": 132112,
            "start": 1323.23,
            "end": 1327.25,
            "text": " of classical machine learning, supervised learning and unsupervised learning,",
            "tokens": [
                51600,
                295,
                13735,
                3479,
                2539,
                11,
                46533,
                2539,
                293,
                2693,
                12879,
                24420,
                2539,
                11,
                51776
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2083333295652236,
            "compression_ratio": 1.9043478260869566,
            "no_speech_prob": 0.00075531005859375,
            "words": [
                {
                    "start": 1323.23,
                    "end": 1323.71,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 1323.71,
                    "end": 1324.23,
                    "word": " classical",
                    "probability": 0.912109375
                },
                {
                    "start": 1324.23,
                    "end": 1324.57,
                    "word": " machine",
                    "probability": 0.68701171875
                },
                {
                    "start": 1324.57,
                    "end": 1324.93,
                    "word": " learning",
                    "probability": 0.83056640625
                },
                {
                    "start": 1324.93,
                    "end": 1325.33,
                    "word": ",",
                    "probability": 0.9130859375
                },
                {
                    "start": 1325.33,
                    "end": 1325.55,
                    "word": " supervised",
                    "probability": 0.71923828125
                },
                {
                    "start": 1325.55,
                    "end": 1325.95,
                    "word": " learning",
                    "probability": 0.798828125
                },
                {
                    "start": 1325.95,
                    "end": 1326.11,
                    "word": " and",
                    "probability": 0.8134765625
                },
                {
                    "start": 1326.11,
                    "end": 1326.65,
                    "word": " unsupervised",
                    "probability": 0.9041341145833334
                },
                {
                    "start": 1326.65,
                    "end": 1326.93,
                    "word": " learning",
                    "probability": 0.830078125
                },
                {
                    "start": 1326.93,
                    "end": 1327.25,
                    "word": ",",
                    "probability": 0.7607421875
                }
            ]
        },
        {
            "id": 198,
            "seek": 134932,
            "start": 1328.05,
            "end": 1334.91,
            "text": " there is a major question. And it's whether deep RL is immune to those attacks.",
            "tokens": [
                50408,
                456,
                307,
                257,
                2563,
                1168,
                13,
                400,
                309,
                311,
                1968,
                2452,
                497,
                43,
                307,
                11992,
                281,
                729,
                8122,
                13,
                50708
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20644530635327102,
            "compression_ratio": 1.5701357466063348,
            "no_speech_prob": 5.561113357543945e-05,
            "words": [
                {
                    "start": 1328.05,
                    "end": 1328.41,
                    "word": " there",
                    "probability": 0.1434326171875
                },
                {
                    "start": 1328.41,
                    "end": 1328.89,
                    "word": " is",
                    "probability": 0.8984375
                },
                {
                    "start": 1329.51,
                    "end": 1329.87,
                    "word": " a major",
                    "probability": 0.7939453125
                },
                {
                    "start": 1329.87,
                    "end": 1330.35,
                    "word": " question",
                    "probability": 0.79833984375
                },
                {
                    "start": 1330.35,
                    "end": 1330.81,
                    "word": ".",
                    "probability": 0.132568359375
                },
                {
                    "start": 1330.81,
                    "end": 1330.85,
                    "word": " And",
                    "probability": 0.796875
                },
                {
                    "start": 1330.85,
                    "end": 1331.11,
                    "word": " it's",
                    "probability": 0.94140625
                },
                {
                    "start": 1331.11,
                    "end": 1331.51,
                    "word": " whether",
                    "probability": 0.8271484375
                },
                {
                    "start": 1331.51,
                    "end": 1331.99,
                    "word": " deep",
                    "probability": 0.5244140625
                },
                {
                    "start": 1331.99,
                    "end": 1332.49,
                    "word": " RL",
                    "probability": 0.93359375
                },
                {
                    "start": 1332.49,
                    "end": 1332.95,
                    "word": " is",
                    "probability": 0.89404296875
                },
                {
                    "start": 1332.95,
                    "end": 1333.41,
                    "word": " immune",
                    "probability": 0.8466796875
                },
                {
                    "start": 1333.41,
                    "end": 1333.57,
                    "word": " to",
                    "probability": 0.9052734375
                },
                {
                    "start": 1333.57,
                    "end": 1333.77,
                    "word": " those",
                    "probability": 0.84326171875
                },
                {
                    "start": 1333.77,
                    "end": 1334.19,
                    "word": " attacks",
                    "probability": 0.6669921875
                },
                {
                    "start": 1334.19,
                    "end": 1334.91,
                    "word": ".",
                    "probability": 0.88623046875
                }
            ]
        },
        {
            "id": 199,
            "seek": 134932,
            "start": 1334.91,
            "end": 1340.59,
            "text": " Back in 2016, late 2016, when the research community was just beginning to start,",
            "tokens": [
                50748,
                5833,
                294,
                6549,
                11,
                3469,
                6549,
                11,
                562,
                264,
                2132,
                1768,
                390,
                445,
                2863,
                281,
                722,
                11,
                51024
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20644530635327102,
            "compression_ratio": 1.5701357466063348,
            "no_speech_prob": 5.561113357543945e-05,
            "words": [
                {
                    "start": 1334.91,
                    "end": 1335.27,
                    "word": " Back",
                    "probability": 0.865234375
                },
                {
                    "start": 1335.27,
                    "end": 1335.41,
                    "word": " in",
                    "probability": 0.9072265625
                },
                {
                    "start": 1335.41,
                    "end": 1336.09,
                    "word": " 2016",
                    "probability": 0.623046875
                },
                {
                    "start": 1336.09,
                    "end": 1336.21,
                    "word": ",",
                    "probability": 0.91552734375
                },
                {
                    "start": 1336.21,
                    "end": 1336.33,
                    "word": " late",
                    "probability": 0.8388671875
                },
                {
                    "start": 1336.33,
                    "end": 1337.01,
                    "word": " 2016",
                    "probability": 0.68505859375
                },
                {
                    "start": 1337.01,
                    "end": 1337.03,
                    "word": ",",
                    "probability": 0.70654296875
                },
                {
                    "start": 1337.03,
                    "end": 1337.43,
                    "word": " when",
                    "probability": 0.87939453125
                },
                {
                    "start": 1337.43,
                    "end": 1337.83,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 1337.83,
                    "end": 1338.05,
                    "word": " research",
                    "probability": 0.7490234375
                },
                {
                    "start": 1338.05,
                    "end": 1338.53,
                    "word": " community",
                    "probability": 0.802734375
                },
                {
                    "start": 1338.53,
                    "end": 1338.95,
                    "word": " was",
                    "probability": 0.8486328125
                },
                {
                    "start": 1338.95,
                    "end": 1339.25,
                    "word": " just",
                    "probability": 0.8876953125
                },
                {
                    "start": 1339.25,
                    "end": 1339.63,
                    "word": " beginning",
                    "probability": 0.82275390625
                },
                {
                    "start": 1339.63,
                    "end": 1339.87,
                    "word": " to",
                    "probability": 0.89501953125
                },
                {
                    "start": 1339.87,
                    "end": 1340.17,
                    "word": " start",
                    "probability": 0.60791015625
                },
                {
                    "start": 1340.17,
                    "end": 1340.59,
                    "word": ",",
                    "probability": 0.65234375
                }
            ]
        },
        {
            "id": 200,
            "seek": 134932,
            "start": 1340.59,
            "end": 1345.79,
            "text": " was just beginning to pay attention to both deep reinforcement learning and the issue of",
            "tokens": [
                51024,
                390,
                445,
                2863,
                281,
                1689,
                3202,
                281,
                1293,
                2452,
                29280,
                2539,
                293,
                264,
                2734,
                295,
                51284
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20644530635327102,
            "compression_ratio": 1.5701357466063348,
            "no_speech_prob": 5.561113357543945e-05,
            "words": [
                {
                    "start": 1340.59,
                    "end": 1340.79,
                    "word": " was",
                    "probability": 0.4248046875
                },
                {
                    "start": 1340.79,
                    "end": 1340.95,
                    "word": " just",
                    "probability": 0.884765625
                },
                {
                    "start": 1340.95,
                    "end": 1341.15,
                    "word": " beginning",
                    "probability": 0.78857421875
                },
                {
                    "start": 1341.15,
                    "end": 1341.35,
                    "word": " to",
                    "probability": 0.90625
                },
                {
                    "start": 1341.35,
                    "end": 1341.47,
                    "word": " pay",
                    "probability": 0.8828125
                },
                {
                    "start": 1341.47,
                    "end": 1341.85,
                    "word": " attention",
                    "probability": 0.86328125
                },
                {
                    "start": 1341.85,
                    "end": 1342.47,
                    "word": " to",
                    "probability": 0.8955078125
                },
                {
                    "start": 1342.47,
                    "end": 1344.03,
                    "word": " both",
                    "probability": 0.76611328125
                },
                {
                    "start": 1344.03,
                    "end": 1344.29,
                    "word": " deep",
                    "probability": 0.8681640625
                },
                {
                    "start": 1344.29,
                    "end": 1344.65,
                    "word": " reinforcement",
                    "probability": 0.806640625
                },
                {
                    "start": 1344.65,
                    "end": 1344.95,
                    "word": " learning",
                    "probability": 0.8193359375
                },
                {
                    "start": 1344.95,
                    "end": 1345.31,
                    "word": " and",
                    "probability": 0.8701171875
                },
                {
                    "start": 1345.31,
                    "end": 1345.49,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 1345.49,
                    "end": 1345.65,
                    "word": " issue",
                    "probability": 0.88818359375
                },
                {
                    "start": 1345.65,
                    "end": 1345.79,
                    "word": " of",
                    "probability": 0.88525390625
                }
            ]
        },
        {
            "id": 201,
            "seek": 134932,
            "start": 1345.79,
            "end": 1355.45,
            "text": " adversarial examples. I came up with this question and decided to experiment with it a little to",
            "tokens": [
                51284,
                17641,
                44745,
                5110,
                13,
                286,
                1361,
                493,
                365,
                341,
                1168,
                293,
                3047,
                281,
                5120,
                365,
                309,
                257,
                707,
                281,
                51768
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20644530635327102,
            "compression_ratio": 1.5701357466063348,
            "no_speech_prob": 5.561113357543945e-05,
            "words": [
                {
                    "start": 1345.79,
                    "end": 1346.19,
                    "word": " adversarial",
                    "probability": 0.92236328125
                },
                {
                    "start": 1346.19,
                    "end": 1346.75,
                    "word": " examples",
                    "probability": 0.7978515625
                },
                {
                    "start": 1346.75,
                    "end": 1347.03,
                    "word": ".",
                    "probability": 0.52734375
                },
                {
                    "start": 1348.11,
                    "end": 1348.83,
                    "word": " I",
                    "probability": 0.9775390625
                },
                {
                    "start": 1348.83,
                    "end": 1349.07,
                    "word": " came",
                    "probability": 0.8037109375
                },
                {
                    "start": 1349.07,
                    "end": 1349.19,
                    "word": " up",
                    "probability": 0.90380859375
                },
                {
                    "start": 1349.19,
                    "end": 1349.29,
                    "word": " with",
                    "probability": 0.8173828125
                },
                {
                    "start": 1349.29,
                    "end": 1349.53,
                    "word": " this",
                    "probability": 0.88134765625
                },
                {
                    "start": 1349.53,
                    "end": 1350.15,
                    "word": " question",
                    "probability": 0.794921875
                },
                {
                    "start": 1350.15,
                    "end": 1351.49,
                    "word": " and",
                    "probability": 0.78125
                },
                {
                    "start": 1351.49,
                    "end": 1352.35,
                    "word": " decided",
                    "probability": 0.88232421875
                },
                {
                    "start": 1352.35,
                    "end": 1353.01,
                    "word": " to",
                    "probability": 0.9013671875
                },
                {
                    "start": 1353.01,
                    "end": 1354.59,
                    "word": " experiment",
                    "probability": 0.85693359375
                },
                {
                    "start": 1354.59,
                    "end": 1354.79,
                    "word": " with",
                    "probability": 0.794921875
                },
                {
                    "start": 1354.79,
                    "end": 1354.87,
                    "word": " it",
                    "probability": 0.88623046875
                },
                {
                    "start": 1354.97,
                    "end": 1355.11,
                    "word": " a little",
                    "probability": 0.810546875
                },
                {
                    "start": 1355.11,
                    "end": 1355.45,
                    "word": " to",
                    "probability": 0.490234375
                }
            ]
        },
        {
            "id": 202,
            "seek": 137932,
            "start": 1355.45,
            "end": 1362.21,
            "text": " find out whether deep RL can also be vulnerable to such attacks. I started from a simple observation.",
            "tokens": [
                50364,
                915,
                484,
                1968,
                2452,
                497,
                43,
                393,
                611,
                312,
                10955,
                281,
                1270,
                8122,
                13,
                286,
                1409,
                490,
                257,
                2199,
                14816,
                13,
                50696
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22873263323196658,
            "compression_ratio": 1.7915057915057915,
            "no_speech_prob": 0.0005192756652832031,
            "words": [
                {
                    "start": 1355.45,
                    "end": 1355.63,
                    "word": " find",
                    "probability": 0.290283203125
                },
                {
                    "start": 1355.63,
                    "end": 1355.75,
                    "word": " out",
                    "probability": 0.81201171875
                },
                {
                    "start": 1355.75,
                    "end": 1355.89,
                    "word": " whether",
                    "probability": 0.8251953125
                },
                {
                    "start": 1355.89,
                    "end": 1356.17,
                    "word": " deep",
                    "probability": 0.65234375
                },
                {
                    "start": 1356.17,
                    "end": 1356.43,
                    "word": " RL",
                    "probability": 0.907958984375
                },
                {
                    "start": 1356.43,
                    "end": 1356.69,
                    "word": " can",
                    "probability": 0.703125
                },
                {
                    "start": 1356.69,
                    "end": 1357.43,
                    "word": " also",
                    "probability": 0.7978515625
                },
                {
                    "start": 1357.43,
                    "end": 1357.59,
                    "word": " be",
                    "probability": 0.90869140625
                },
                {
                    "start": 1357.59,
                    "end": 1357.91,
                    "word": " vulnerable",
                    "probability": 0.8115234375
                },
                {
                    "start": 1357.91,
                    "end": 1358.41,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 1358.41,
                    "end": 1358.63,
                    "word": " such",
                    "probability": 0.8740234375
                },
                {
                    "start": 1358.63,
                    "end": 1359.01,
                    "word": " attacks",
                    "probability": 0.7099609375
                },
                {
                    "start": 1359.01,
                    "end": 1359.65,
                    "word": ".",
                    "probability": 0.94189453125
                },
                {
                    "start": 1359.87,
                    "end": 1360.51,
                    "word": " I",
                    "probability": 0.99072265625
                },
                {
                    "start": 1360.51,
                    "end": 1360.75,
                    "word": " started",
                    "probability": 0.79150390625
                },
                {
                    "start": 1360.75,
                    "end": 1360.95,
                    "word": " from",
                    "probability": 0.82568359375
                },
                {
                    "start": 1361.05,
                    "end": 1361.23,
                    "word": " a simple",
                    "probability": 0.9033203125
                },
                {
                    "start": 1361.23,
                    "end": 1361.71,
                    "word": " observation",
                    "probability": 0.79931640625
                },
                {
                    "start": 1361.71,
                    "end": 1362.21,
                    "word": ".",
                    "probability": 0.6181640625
                }
            ]
        },
        {
            "id": 203,
            "seek": 137932,
            "start": 1362.21,
            "end": 1366.79,
            "text": " The deep neural networks in DQN models and classifiers are both function approximators,",
            "tokens": [
                50696,
                440,
                2452,
                18161,
                9590,
                294,
                413,
                48,
                45,
                5245,
                293,
                1508,
                23463,
                366,
                1293,
                2445,
                8542,
                3391,
                11,
                50928
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22873263323196658,
            "compression_ratio": 1.7915057915057915,
            "no_speech_prob": 0.0005192756652832031,
            "words": [
                {
                    "start": 1362.21,
                    "end": 1362.29,
                    "word": " The",
                    "probability": 0.79345703125
                },
                {
                    "start": 1362.29,
                    "end": 1362.53,
                    "word": " deep",
                    "probability": 0.90283203125
                },
                {
                    "start": 1362.53,
                    "end": 1362.73,
                    "word": " neural",
                    "probability": 0.91015625
                },
                {
                    "start": 1362.73,
                    "end": 1363.21,
                    "word": " networks",
                    "probability": 0.66015625
                },
                {
                    "start": 1363.21,
                    "end": 1363.47,
                    "word": " in",
                    "probability": 0.85498046875
                },
                {
                    "start": 1363.47,
                    "end": 1363.85,
                    "word": " DQN",
                    "probability": 0.896484375
                },
                {
                    "start": 1363.85,
                    "end": 1364.21,
                    "word": " models",
                    "probability": 0.8115234375
                },
                {
                    "start": 1364.21,
                    "end": 1364.51,
                    "word": " and",
                    "probability": 0.88818359375
                },
                {
                    "start": 1364.51,
                    "end": 1364.95,
                    "word": " classifiers",
                    "probability": 0.928466796875
                },
                {
                    "start": 1364.95,
                    "end": 1365.17,
                    "word": " are",
                    "probability": 0.87646484375
                },
                {
                    "start": 1365.17,
                    "end": 1365.57,
                    "word": " both",
                    "probability": 0.80859375
                },
                {
                    "start": 1365.57,
                    "end": 1366.07,
                    "word": " function",
                    "probability": 0.82080078125
                },
                {
                    "start": 1366.07,
                    "end": 1366.65,
                    "word": " approximators",
                    "probability": 0.885498046875
                },
                {
                    "start": 1366.65,
                    "end": 1366.79,
                    "word": ",",
                    "probability": 0.348876953125
                }
            ]
        },
        {
            "id": 204,
            "seek": 137932,
            "start": 1366.79,
            "end": 1371.01,
            "text": " are both at training time, there are function approximators at test times or just functions.",
            "tokens": [
                50928,
                366,
                1293,
                412,
                3097,
                565,
                11,
                456,
                366,
                2445,
                8542,
                3391,
                412,
                1500,
                1413,
                420,
                445,
                6828,
                13,
                51128
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22873263323196658,
            "compression_ratio": 1.7915057915057915,
            "no_speech_prob": 0.0005192756652832031,
            "words": [
                {
                    "start": 1366.79,
                    "end": 1366.85,
                    "word": " are",
                    "probability": 0.26123046875
                },
                {
                    "start": 1366.85,
                    "end": 1367.31,
                    "word": " both",
                    "probability": 0.796875
                },
                {
                    "start": 1367.31,
                    "end": 1367.93,
                    "word": " at",
                    "probability": 0.2252197265625
                },
                {
                    "start": 1367.93,
                    "end": 1368.19,
                    "word": " training",
                    "probability": 0.8095703125
                },
                {
                    "start": 1368.19,
                    "end": 1368.41,
                    "word": " time",
                    "probability": 0.775390625
                },
                {
                    "start": 1368.41,
                    "end": 1368.47,
                    "word": ",",
                    "probability": 0.3935546875
                },
                {
                    "start": 1368.47,
                    "end": 1368.53,
                    "word": " there",
                    "probability": 0.399658203125
                },
                {
                    "start": 1368.53,
                    "end": 1368.65,
                    "word": " are",
                    "probability": 0.87890625
                },
                {
                    "start": 1368.65,
                    "end": 1368.95,
                    "word": " function",
                    "probability": 0.7734375
                },
                {
                    "start": 1368.95,
                    "end": 1369.41,
                    "word": " approximators",
                    "probability": 0.893798828125
                },
                {
                    "start": 1369.41,
                    "end": 1369.59,
                    "word": " at",
                    "probability": 0.73828125
                },
                {
                    "start": 1369.59,
                    "end": 1369.79,
                    "word": " test",
                    "probability": 0.7880859375
                },
                {
                    "start": 1369.79,
                    "end": 1369.97,
                    "word": " times",
                    "probability": 0.8095703125
                },
                {
                    "start": 1369.97,
                    "end": 1370.11,
                    "word": " or",
                    "probability": 0.350830078125
                },
                {
                    "start": 1370.11,
                    "end": 1370.29,
                    "word": " just",
                    "probability": 0.9013671875
                },
                {
                    "start": 1370.29,
                    "end": 1370.69,
                    "word": " functions",
                    "probability": 0.8681640625
                },
                {
                    "start": 1370.69,
                    "end": 1371.01,
                    "word": ".",
                    "probability": 0.89599609375
                }
            ]
        },
        {
            "id": 205,
            "seek": 137932,
            "start": 1371.81,
            "end": 1378.23,
            "text": " And I came up with this hypothesis. If classifiers are vulnerable to adversarial examples,",
            "tokens": [
                51180,
                400,
                286,
                1361,
                493,
                365,
                341,
                17291,
                13,
                759,
                1508,
                23463,
                366,
                10955,
                281,
                17641,
                44745,
                5110,
                11,
                51496
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22873263323196658,
            "compression_ratio": 1.7915057915057915,
            "no_speech_prob": 0.0005192756652832031,
            "words": [
                {
                    "start": 1371.81,
                    "end": 1372.23,
                    "word": " And",
                    "probability": 0.84228515625
                },
                {
                    "start": 1372.23,
                    "end": 1372.75,
                    "word": " I",
                    "probability": 0.93310546875
                },
                {
                    "start": 1372.75,
                    "end": 1373.35,
                    "word": " came",
                    "probability": 0.69384765625
                },
                {
                    "start": 1373.35,
                    "end": 1373.45,
                    "word": " up",
                    "probability": 0.9033203125
                },
                {
                    "start": 1373.45,
                    "end": 1373.55,
                    "word": " with",
                    "probability": 0.81298828125
                },
                {
                    "start": 1373.55,
                    "end": 1373.71,
                    "word": " this",
                    "probability": 0.88037109375
                },
                {
                    "start": 1373.71,
                    "end": 1374.27,
                    "word": " hypothesis",
                    "probability": 0.767578125
                },
                {
                    "start": 1374.27,
                    "end": 1374.65,
                    "word": ".",
                    "probability": 0.54052734375
                },
                {
                    "start": 1374.65,
                    "end": 1374.85,
                    "word": " If",
                    "probability": 0.89013671875
                },
                {
                    "start": 1374.85,
                    "end": 1375.49,
                    "word": " classifiers",
                    "probability": 0.923828125
                },
                {
                    "start": 1375.49,
                    "end": 1375.83,
                    "word": " are",
                    "probability": 0.892578125
                },
                {
                    "start": 1375.83,
                    "end": 1376.35,
                    "word": " vulnerable",
                    "probability": 0.83984375
                },
                {
                    "start": 1376.35,
                    "end": 1376.61,
                    "word": " to",
                    "probability": 0.90234375
                },
                {
                    "start": 1376.61,
                    "end": 1377.17,
                    "word": " adversarial",
                    "probability": 0.9267578125
                },
                {
                    "start": 1377.17,
                    "end": 1377.69,
                    "word": " examples",
                    "probability": 0.77978515625
                },
                {
                    "start": 1377.69,
                    "end": 1378.23,
                    "word": ",",
                    "probability": 0.95849609375
                }
            ]
        },
        {
            "id": 206,
            "seek": 137932,
            "start": 1378.23,
            "end": 1385.25,
            "text": " then action value approximators of DQNs may also be vulnerable. All right. So I started to",
            "tokens": [
                51496,
                550,
                3069,
                2158,
                8542,
                3391,
                295,
                413,
                48,
                45,
                82,
                815,
                611,
                312,
                10955,
                13,
                1057,
                558,
                13,
                407,
                286,
                1409,
                281,
                51848
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22873263323196658,
            "compression_ratio": 1.7915057915057915,
            "no_speech_prob": 0.0005192756652832031,
            "words": [
                {
                    "start": 1378.23,
                    "end": 1378.39,
                    "word": " then",
                    "probability": 0.73828125
                },
                {
                    "start": 1378.39,
                    "end": 1378.99,
                    "word": " action",
                    "probability": 0.791015625
                },
                {
                    "start": 1378.99,
                    "end": 1379.39,
                    "word": " value",
                    "probability": 0.8671875
                },
                {
                    "start": 1379.39,
                    "end": 1380.05,
                    "word": " approximators",
                    "probability": 0.90966796875
                },
                {
                    "start": 1380.05,
                    "end": 1380.37,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 1380.37,
                    "end": 1381.11,
                    "word": " DQNs",
                    "probability": 0.9278564453125
                },
                {
                    "start": 1381.11,
                    "end": 1381.25,
                    "word": " may",
                    "probability": 0.86962890625
                },
                {
                    "start": 1381.25,
                    "end": 1381.59,
                    "word": " also",
                    "probability": 0.8232421875
                },
                {
                    "start": 1381.59,
                    "end": 1381.77,
                    "word": " be",
                    "probability": 0.9091796875
                },
                {
                    "start": 1381.77,
                    "end": 1382.11,
                    "word": " vulnerable",
                    "probability": 0.81591796875
                },
                {
                    "start": 1382.11,
                    "end": 1382.75,
                    "word": ".",
                    "probability": 0.96337890625
                },
                {
                    "start": 1382.79,
                    "end": 1382.85,
                    "word": " All",
                    "probability": 0.603515625
                },
                {
                    "start": 1382.85,
                    "end": 1383.05,
                    "word": " right",
                    "probability": 0.88232421875
                },
                {
                    "start": 1383.05,
                    "end": 1383.69,
                    "word": ".",
                    "probability": 0.787109375
                },
                {
                    "start": 1383.75,
                    "end": 1384.17,
                    "word": " So",
                    "probability": 0.90625
                },
                {
                    "start": 1384.17,
                    "end": 1384.37,
                    "word": " I",
                    "probability": 0.95751953125
                },
                {
                    "start": 1384.37,
                    "end": 1384.89,
                    "word": " started",
                    "probability": 0.81396484375
                },
                {
                    "start": 1384.89,
                    "end": 1385.25,
                    "word": " to",
                    "probability": 0.8994140625
                }
            ]
        },
        {
            "id": 207,
            "seek": 140768,
            "start": 1385.45,
            "end": 1395.23,
            "text": " set up an experiment where the aim was to perform adversarial attacks. The adversary's goal",
            "tokens": [
                50364,
                992,
                493,
                364,
                5120,
                689,
                264,
                5939,
                390,
                281,
                2042,
                17641,
                44745,
                8122,
                13,
                440,
                48222,
                311,
                3387,
                50848
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2411859027850322,
            "compression_ratio": 1.7246376811594204,
            "no_speech_prob": 0.0006265640258789062,
            "words": [
                {
                    "start": 1385.45,
                    "end": 1385.73,
                    "word": " set",
                    "probability": 0.65771484375
                },
                {
                    "start": 1385.73,
                    "end": 1385.83,
                    "word": " up",
                    "probability": 0.89453125
                },
                {
                    "start": 1385.83,
                    "end": 1385.93,
                    "word": " an",
                    "probability": 0.9033203125
                },
                {
                    "start": 1385.93,
                    "end": 1386.33,
                    "word": " experiment",
                    "probability": 0.87060546875
                },
                {
                    "start": 1386.33,
                    "end": 1387.09,
                    "word": " where",
                    "probability": 0.81103515625
                },
                {
                    "start": 1387.09,
                    "end": 1389.05,
                    "word": " the",
                    "probability": 0.7763671875
                },
                {
                    "start": 1389.05,
                    "end": 1391.57,
                    "word": " aim",
                    "probability": 0.79345703125
                },
                {
                    "start": 1391.57,
                    "end": 1391.93,
                    "word": " was",
                    "probability": 0.89453125
                },
                {
                    "start": 1391.93,
                    "end": 1392.07,
                    "word": " to",
                    "probability": 0.9033203125
                },
                {
                    "start": 1392.07,
                    "end": 1392.37,
                    "word": " perform",
                    "probability": 0.7724609375
                },
                {
                    "start": 1392.37,
                    "end": 1393.13,
                    "word": " adversarial",
                    "probability": 0.93505859375
                },
                {
                    "start": 1393.13,
                    "end": 1393.97,
                    "word": " attacks",
                    "probability": 0.71875
                },
                {
                    "start": 1393.97,
                    "end": 1394.17,
                    "word": ".",
                    "probability": 0.787109375
                },
                {
                    "start": 1394.17,
                    "end": 1394.27,
                    "word": " The",
                    "probability": 0.79052734375
                },
                {
                    "start": 1394.27,
                    "end": 1394.85,
                    "word": " adversary's",
                    "probability": 0.844970703125
                },
                {
                    "start": 1394.85,
                    "end": 1395.23,
                    "word": " goal",
                    "probability": 0.90380859375
                }
            ]
        },
        {
            "id": 208,
            "seek": 140768,
            "start": 1395.85,
            "end": 1401.11,
            "text": " was twofold. Test time attack to perturb the performance of target's learned policy. So",
            "tokens": [
                50884,
                390,
                732,
                18353,
                13,
                9279,
                565,
                2690,
                281,
                40468,
                264,
                3389,
                295,
                3779,
                311,
                3264,
                3897,
                13,
                407,
                51156
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2411859027850322,
            "compression_ratio": 1.7246376811594204,
            "no_speech_prob": 0.0006265640258789062,
            "words": [
                {
                    "start": 1395.85,
                    "end": 1396.37,
                    "word": " was",
                    "probability": 0.88916015625
                },
                {
                    "start": 1396.37,
                    "end": 1397.29,
                    "word": " twofold",
                    "probability": 0.71533203125
                },
                {
                    "start": 1397.29,
                    "end": 1397.49,
                    "word": ".",
                    "probability": 0.36669921875
                },
                {
                    "start": 1397.49,
                    "end": 1397.69,
                    "word": " Test",
                    "probability": 0.861328125
                },
                {
                    "start": 1397.69,
                    "end": 1397.89,
                    "word": " time",
                    "probability": 0.47607421875
                },
                {
                    "start": 1397.89,
                    "end": 1398.25,
                    "word": " attack",
                    "probability": 0.82958984375
                },
                {
                    "start": 1398.25,
                    "end": 1398.47,
                    "word": " to",
                    "probability": 0.65869140625
                },
                {
                    "start": 1398.47,
                    "end": 1398.91,
                    "word": " perturb",
                    "probability": 0.8359375
                },
                {
                    "start": 1398.91,
                    "end": 1399.15,
                    "word": " the",
                    "probability": 0.80615234375
                },
                {
                    "start": 1399.15,
                    "end": 1399.57,
                    "word": " performance",
                    "probability": 0.8330078125
                },
                {
                    "start": 1399.57,
                    "end": 1399.77,
                    "word": " of",
                    "probability": 0.904296875
                },
                {
                    "start": 1399.77,
                    "end": 1400.17,
                    "word": " target's",
                    "probability": 0.629150390625
                },
                {
                    "start": 1400.17,
                    "end": 1400.41,
                    "word": " learned",
                    "probability": 0.466064453125
                },
                {
                    "start": 1400.41,
                    "end": 1400.81,
                    "word": " policy",
                    "probability": 0.763671875
                },
                {
                    "start": 1400.81,
                    "end": 1400.97,
                    "word": ".",
                    "probability": 0.66259765625
                },
                {
                    "start": 1400.97,
                    "end": 1401.11,
                    "word": " So",
                    "probability": 0.861328125
                }
            ]
        },
        {
            "id": 209,
            "seek": 140768,
            "start": 1401.11,
            "end": 1405.67,
            "text": " the target at this point is fully trained and is deployed in the environment and the adversary",
            "tokens": [
                51156,
                264,
                3779,
                412,
                341,
                935,
                307,
                4498,
                8895,
                293,
                307,
                17826,
                294,
                264,
                2823,
                293,
                264,
                48222,
                51380
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2411859027850322,
            "compression_ratio": 1.7246376811594204,
            "no_speech_prob": 0.0006265640258789062,
            "words": [
                {
                    "start": 1401.11,
                    "end": 1401.55,
                    "word": " the",
                    "probability": 0.68701171875
                },
                {
                    "start": 1401.55,
                    "end": 1401.83,
                    "word": " target",
                    "probability": 0.79541015625
                },
                {
                    "start": 1401.83,
                    "end": 1402.11,
                    "word": " at",
                    "probability": 0.82763671875
                },
                {
                    "start": 1402.11,
                    "end": 1402.29,
                    "word": " this",
                    "probability": 0.8818359375
                },
                {
                    "start": 1402.29,
                    "end": 1402.51,
                    "word": " point",
                    "probability": 0.86865234375
                },
                {
                    "start": 1402.51,
                    "end": 1402.73,
                    "word": " is",
                    "probability": 0.875
                },
                {
                    "start": 1402.73,
                    "end": 1402.93,
                    "word": " fully",
                    "probability": 0.740234375
                },
                {
                    "start": 1402.93,
                    "end": 1403.29,
                    "word": " trained",
                    "probability": 0.8134765625
                },
                {
                    "start": 1403.29,
                    "end": 1403.47,
                    "word": " and",
                    "probability": 0.86865234375
                },
                {
                    "start": 1403.47,
                    "end": 1403.65,
                    "word": " is",
                    "probability": 0.732421875
                },
                {
                    "start": 1403.65,
                    "end": 1404.03,
                    "word": " deployed",
                    "probability": 0.61962890625
                },
                {
                    "start": 1404.03,
                    "end": 1404.19,
                    "word": " in",
                    "probability": 0.90478515625
                },
                {
                    "start": 1404.19,
                    "end": 1404.27,
                    "word": " the",
                    "probability": 0.7900390625
                },
                {
                    "start": 1404.27,
                    "end": 1404.65,
                    "word": " environment",
                    "probability": 0.744140625
                },
                {
                    "start": 1404.65,
                    "end": 1405.25,
                    "word": " and",
                    "probability": 0.372802734375
                },
                {
                    "start": 1405.25,
                    "end": 1405.35,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 1405.35,
                    "end": 1405.67,
                    "word": " adversary",
                    "probability": 0.82861328125
                }
            ]
        },
        {
            "id": 210,
            "seek": 140768,
            "start": 1405.67,
            "end": 1413.81,
            "text": " wants to somehow cause the agent to perform incorrectly, to manipulate its policy.",
            "tokens": [
                51380,
                2738,
                281,
                6063,
                3082,
                264,
                9461,
                281,
                2042,
                42892,
                11,
                281,
                20459,
                1080,
                3897,
                13,
                51768
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2411859027850322,
            "compression_ratio": 1.7246376811594204,
            "no_speech_prob": 0.0006265640258789062,
            "words": [
                {
                    "start": 1405.67,
                    "end": 1405.99,
                    "word": " wants",
                    "probability": 0.74609375
                },
                {
                    "start": 1405.99,
                    "end": 1406.61,
                    "word": " to",
                    "probability": 0.90625
                },
                {
                    "start": 1406.61,
                    "end": 1407.37,
                    "word": " somehow",
                    "probability": 0.87060546875
                },
                {
                    "start": 1407.37,
                    "end": 1408.89,
                    "word": " cause",
                    "probability": 0.7919921875
                },
                {
                    "start": 1408.89,
                    "end": 1409.27,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 1409.27,
                    "end": 1409.73,
                    "word": " agent",
                    "probability": 0.828125
                },
                {
                    "start": 1409.73,
                    "end": 1410.05,
                    "word": " to",
                    "probability": 0.9091796875
                },
                {
                    "start": 1410.05,
                    "end": 1410.65,
                    "word": " perform",
                    "probability": 0.77587890625
                },
                {
                    "start": 1410.65,
                    "end": 1411.99,
                    "word": " incorrectly",
                    "probability": 0.87109375
                },
                {
                    "start": 1411.99,
                    "end": 1412.23,
                    "word": ",",
                    "probability": 0.417724609375
                },
                {
                    "start": 1412.23,
                    "end": 1412.49,
                    "word": " to",
                    "probability": 0.91015625
                },
                {
                    "start": 1412.49,
                    "end": 1412.85,
                    "word": " manipulate",
                    "probability": 0.9091796875
                },
                {
                    "start": 1412.85,
                    "end": 1413.07,
                    "word": " its",
                    "probability": 0.7041015625
                },
                {
                    "start": 1413.07,
                    "end": 1413.49,
                    "word": " policy",
                    "probability": 0.81787109375
                },
                {
                    "start": 1413.49,
                    "end": 1413.81,
                    "word": ".",
                    "probability": 0.99609375
                }
            ]
        },
        {
            "id": 211,
            "seek": 143676,
            "start": 1414.89,
            "end": 1419.45,
            "text": " What does the adversary know about the target? It knows the type of input to the target. For",
            "tokens": [
                50396,
                708,
                775,
                264,
                48222,
                458,
                466,
                264,
                3779,
                30,
                467,
                3255,
                264,
                2010,
                295,
                4846,
                281,
                264,
                3779,
                13,
                1171,
                50640
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1997023752757481,
            "compression_ratio": 1.7826086956521738,
            "no_speech_prob": 0.0020198822021484375,
            "words": [
                {
                    "start": 1414.89,
                    "end": 1415.41,
                    "word": " What",
                    "probability": 0.36328125
                },
                {
                    "start": 1415.41,
                    "end": 1415.59,
                    "word": " does",
                    "probability": 0.87158203125
                },
                {
                    "start": 1415.59,
                    "end": 1415.71,
                    "word": " the",
                    "probability": 0.80615234375
                },
                {
                    "start": 1415.71,
                    "end": 1416.03,
                    "word": " adversary",
                    "probability": 0.7744140625
                },
                {
                    "start": 1416.03,
                    "end": 1416.45,
                    "word": " know",
                    "probability": 0.8017578125
                },
                {
                    "start": 1416.45,
                    "end": 1416.71,
                    "word": " about",
                    "probability": 0.83349609375
                },
                {
                    "start": 1416.71,
                    "end": 1416.89,
                    "word": " the",
                    "probability": 0.80078125
                },
                {
                    "start": 1416.89,
                    "end": 1417.11,
                    "word": " target",
                    "probability": 0.8564453125
                },
                {
                    "start": 1417.11,
                    "end": 1417.49,
                    "word": "?",
                    "probability": 0.95654296875
                },
                {
                    "start": 1417.49,
                    "end": 1417.75,
                    "word": " It",
                    "probability": 0.86572265625
                },
                {
                    "start": 1417.75,
                    "end": 1417.97,
                    "word": " knows",
                    "probability": 0.72265625
                },
                {
                    "start": 1417.97,
                    "end": 1418.09,
                    "word": " the",
                    "probability": 0.78759765625
                },
                {
                    "start": 1418.09,
                    "end": 1418.33,
                    "word": " type",
                    "probability": 0.91796875
                },
                {
                    "start": 1418.33,
                    "end": 1418.45,
                    "word": " of",
                    "probability": 0.90478515625
                },
                {
                    "start": 1418.45,
                    "end": 1418.71,
                    "word": " input",
                    "probability": 0.85205078125
                },
                {
                    "start": 1418.71,
                    "end": 1418.85,
                    "word": " to",
                    "probability": 0.875
                },
                {
                    "start": 1418.85,
                    "end": 1418.95,
                    "word": " the",
                    "probability": 0.80517578125
                },
                {
                    "start": 1418.95,
                    "end": 1419.19,
                    "word": " target",
                    "probability": 0.8662109375
                },
                {
                    "start": 1419.19,
                    "end": 1419.37,
                    "word": ".",
                    "probability": 0.6005859375
                },
                {
                    "start": 1419.37,
                    "end": 1419.45,
                    "word": " For",
                    "probability": 0.849609375
                }
            ]
        },
        {
            "id": 212,
            "seek": 143676,
            "start": 1419.45,
            "end": 1426.93,
            "text": " example, it knows whether the target policy is looking at image data, text, audio, and such.",
            "tokens": [
                50640,
                1365,
                11,
                309,
                3255,
                1968,
                264,
                3779,
                3897,
                307,
                1237,
                412,
                3256,
                1412,
                11,
                2487,
                11,
                6278,
                11,
                293,
                1270,
                13,
                51012
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1997023752757481,
            "compression_ratio": 1.7826086956521738,
            "no_speech_prob": 0.0020198822021484375,
            "words": [
                {
                    "start": 1419.45,
                    "end": 1419.69,
                    "word": " example",
                    "probability": 0.92626953125
                },
                {
                    "start": 1419.69,
                    "end": 1419.75,
                    "word": ",",
                    "probability": 0.85693359375
                },
                {
                    "start": 1419.75,
                    "end": 1419.89,
                    "word": " it",
                    "probability": 0.8876953125
                },
                {
                    "start": 1419.89,
                    "end": 1420.21,
                    "word": " knows",
                    "probability": 0.73046875
                },
                {
                    "start": 1420.21,
                    "end": 1420.49,
                    "word": " whether",
                    "probability": 0.8369140625
                },
                {
                    "start": 1420.49,
                    "end": 1420.93,
                    "word": " the",
                    "probability": 0.7978515625
                },
                {
                    "start": 1420.93,
                    "end": 1421.37,
                    "word": " target",
                    "probability": 0.8369140625
                },
                {
                    "start": 1421.37,
                    "end": 1421.83,
                    "word": " policy",
                    "probability": 0.79931640625
                },
                {
                    "start": 1421.83,
                    "end": 1423.23,
                    "word": " is",
                    "probability": 0.88623046875
                },
                {
                    "start": 1423.23,
                    "end": 1424.03,
                    "word": " looking",
                    "probability": 0.7919921875
                },
                {
                    "start": 1424.03,
                    "end": 1424.19,
                    "word": " at",
                    "probability": 0.91552734375
                },
                {
                    "start": 1424.19,
                    "end": 1424.49,
                    "word": " image",
                    "probability": 0.83349609375
                },
                {
                    "start": 1424.49,
                    "end": 1424.87,
                    "word": " data",
                    "probability": 0.7880859375
                },
                {
                    "start": 1424.87,
                    "end": 1425.37,
                    "word": ",",
                    "probability": 0.93310546875
                },
                {
                    "start": 1425.37,
                    "end": 1425.75,
                    "word": " text",
                    "probability": 0.9169921875
                },
                {
                    "start": 1425.75,
                    "end": 1425.91,
                    "word": ",",
                    "probability": 0.9716796875
                },
                {
                    "start": 1425.91,
                    "end": 1426.17,
                    "word": " audio",
                    "probability": 0.89990234375
                },
                {
                    "start": 1426.17,
                    "end": 1426.29,
                    "word": ",",
                    "probability": 0.432373046875
                },
                {
                    "start": 1426.29,
                    "end": 1426.37,
                    "word": " and",
                    "probability": 0.88671875
                },
                {
                    "start": 1426.37,
                    "end": 1426.63,
                    "word": " such",
                    "probability": 0.88818359375
                },
                {
                    "start": 1426.63,
                    "end": 1426.93,
                    "word": ".",
                    "probability": 0.9638671875
                }
            ]
        },
        {
            "id": 213,
            "seek": 143676,
            "start": 1426.93,
            "end": 1431.05,
            "text": " Why? Because it helps with estimating the architecture. For example, if it's images,",
            "tokens": [
                51012,
                1545,
                30,
                1436,
                309,
                3665,
                365,
                8017,
                990,
                264,
                9482,
                13,
                1171,
                1365,
                11,
                498,
                309,
                311,
                5267,
                11,
                51216
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1997023752757481,
            "compression_ratio": 1.7826086956521738,
            "no_speech_prob": 0.0020198822021484375,
            "words": [
                {
                    "start": 1426.93,
                    "end": 1427.09,
                    "word": " Why",
                    "probability": 0.8134765625
                },
                {
                    "start": 1427.09,
                    "end": 1427.19,
                    "word": "?",
                    "probability": 0.95947265625
                },
                {
                    "start": 1427.19,
                    "end": 1427.37,
                    "word": " Because",
                    "probability": 0.89990234375
                },
                {
                    "start": 1427.37,
                    "end": 1427.49,
                    "word": " it",
                    "probability": 0.900390625
                },
                {
                    "start": 1427.49,
                    "end": 1427.71,
                    "word": " helps",
                    "probability": 0.716796875
                },
                {
                    "start": 1427.71,
                    "end": 1428.11,
                    "word": " with",
                    "probability": 0.80224609375
                },
                {
                    "start": 1428.11,
                    "end": 1429.11,
                    "word": " estimating",
                    "probability": 0.875244140625
                },
                {
                    "start": 1429.11,
                    "end": 1429.31,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 1429.31,
                    "end": 1429.69,
                    "word": " architecture",
                    "probability": 0.8369140625
                },
                {
                    "start": 1429.69,
                    "end": 1429.79,
                    "word": ".",
                    "probability": 0.52490234375
                },
                {
                    "start": 1429.79,
                    "end": 1429.89,
                    "word": " For",
                    "probability": 0.89501953125
                },
                {
                    "start": 1429.89,
                    "end": 1430.11,
                    "word": " example",
                    "probability": 0.93115234375
                },
                {
                    "start": 1430.11,
                    "end": 1430.17,
                    "word": ",",
                    "probability": 0.95361328125
                },
                {
                    "start": 1430.17,
                    "end": 1430.25,
                    "word": " if",
                    "probability": 0.615234375
                },
                {
                    "start": 1430.25,
                    "end": 1430.41,
                    "word": " it's",
                    "probability": 0.905029296875
                },
                {
                    "start": 1430.41,
                    "end": 1430.77,
                    "word": " images",
                    "probability": 0.84423828125
                },
                {
                    "start": 1430.77,
                    "end": 1431.05,
                    "word": ",",
                    "probability": 0.8701171875
                }
            ]
        },
        {
            "id": 214,
            "seek": 143676,
            "start": 1431.05,
            "end": 1438.69,
            "text": " the adversary can come up with a good guess that the target architecture includes CNNs",
            "tokens": [
                51216,
                264,
                48222,
                393,
                808,
                493,
                365,
                257,
                665,
                2041,
                300,
                264,
                3779,
                9482,
                5974,
                24859,
                82,
                51600
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1997023752757481,
            "compression_ratio": 1.7826086956521738,
            "no_speech_prob": 0.0020198822021484375,
            "words": [
                {
                    "start": 1431.05,
                    "end": 1431.13,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 1431.13,
                    "end": 1431.47,
                    "word": " adversary",
                    "probability": 0.74365234375
                },
                {
                    "start": 1431.47,
                    "end": 1431.83,
                    "word": " can",
                    "probability": 0.900390625
                },
                {
                    "start": 1431.83,
                    "end": 1432.57,
                    "word": " come",
                    "probability": 0.76123046875
                },
                {
                    "start": 1432.57,
                    "end": 1432.71,
                    "word": " up",
                    "probability": 0.8828125
                },
                {
                    "start": 1432.71,
                    "end": 1433.55,
                    "word": " with",
                    "probability": 0.8056640625
                },
                {
                    "start": 1433.87,
                    "end": 1435.11,
                    "word": " a good",
                    "probability": 0.767578125
                },
                {
                    "start": 1435.11,
                    "end": 1435.55,
                    "word": " guess",
                    "probability": 0.87890625
                },
                {
                    "start": 1435.55,
                    "end": 1435.85,
                    "word": " that",
                    "probability": 0.869140625
                },
                {
                    "start": 1435.85,
                    "end": 1436.07,
                    "word": " the",
                    "probability": 0.80029296875
                },
                {
                    "start": 1436.07,
                    "end": 1437.41,
                    "word": " target",
                    "probability": 0.6484375
                },
                {
                    "start": 1437.41,
                    "end": 1437.77,
                    "word": " architecture",
                    "probability": 0.82763671875
                },
                {
                    "start": 1437.77,
                    "end": 1438.15,
                    "word": " includes",
                    "probability": 0.82958984375
                },
                {
                    "start": 1438.15,
                    "end": 1438.69,
                    "word": " CNNs",
                    "probability": 0.883544921875
                }
            ]
        },
        {
            "id": 215,
            "seek": 143676,
            "start": 1438.69,
            "end": 1442.89,
            "text": " or convolutional neural networks. I also assume that the adversary knows the reward function.",
            "tokens": [
                51600,
                420,
                45216,
                304,
                18161,
                9590,
                13,
                286,
                611,
                6552,
                300,
                264,
                48222,
                3255,
                264,
                7782,
                2445,
                13,
                51812
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1997023752757481,
            "compression_ratio": 1.7826086956521738,
            "no_speech_prob": 0.0020198822021484375,
            "words": [
                {
                    "start": 1438.69,
                    "end": 1438.79,
                    "word": " or",
                    "probability": 0.61572265625
                },
                {
                    "start": 1438.79,
                    "end": 1439.23,
                    "word": " convolutional",
                    "probability": 0.849365234375
                },
                {
                    "start": 1439.23,
                    "end": 1439.35,
                    "word": " neural",
                    "probability": 0.8388671875
                },
                {
                    "start": 1439.35,
                    "end": 1439.65,
                    "word": " networks",
                    "probability": 0.67333984375
                },
                {
                    "start": 1439.65,
                    "end": 1440.09,
                    "word": ".",
                    "probability": 0.99169921875
                },
                {
                    "start": 1440.09,
                    "end": 1440.23,
                    "word": " I",
                    "probability": 0.95947265625
                },
                {
                    "start": 1440.23,
                    "end": 1440.41,
                    "word": " also",
                    "probability": 0.81884765625
                },
                {
                    "start": 1440.41,
                    "end": 1440.63,
                    "word": " assume",
                    "probability": 0.75634765625
                },
                {
                    "start": 1440.63,
                    "end": 1440.75,
                    "word": " that",
                    "probability": 0.87109375
                },
                {
                    "start": 1440.75,
                    "end": 1440.89,
                    "word": " the",
                    "probability": 0.81201171875
                },
                {
                    "start": 1440.89,
                    "end": 1441.21,
                    "word": " adversary",
                    "probability": 0.82568359375
                },
                {
                    "start": 1441.21,
                    "end": 1441.61,
                    "word": " knows",
                    "probability": 0.71240234375
                },
                {
                    "start": 1441.61,
                    "end": 1441.77,
                    "word": " the",
                    "probability": 0.78515625
                },
                {
                    "start": 1441.77,
                    "end": 1442.11,
                    "word": " reward",
                    "probability": 0.9111328125
                },
                {
                    "start": 1442.11,
                    "end": 1442.51,
                    "word": " function",
                    "probability": 0.87841796875
                },
                {
                    "start": 1442.51,
                    "end": 1442.89,
                    "word": ".",
                    "probability": 0.6044921875
                }
            ]
        },
        {
            "id": 216,
            "seek": 146202,
            "start": 1442.89,
            "end": 1447.21,
            "text": " So the adversary may have access to the environment and if it's, for example, a game",
            "tokens": [
                50364,
                407,
                264,
                48222,
                815,
                362,
                2105,
                281,
                264,
                2823,
                293,
                498,
                309,
                311,
                11,
                337,
                1365,
                11,
                257,
                1216,
                50580
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2118252832442522,
            "compression_ratio": 1.7302325581395348,
            "no_speech_prob": 0.0003407001495361328,
            "words": [
                {
                    "start": 1442.89,
                    "end": 1443.09,
                    "word": " So",
                    "probability": 0.63525390625
                },
                {
                    "start": 1443.09,
                    "end": 1443.23,
                    "word": " the",
                    "probability": 0.76025390625
                },
                {
                    "start": 1443.23,
                    "end": 1443.55,
                    "word": " adversary",
                    "probability": 0.8349609375
                },
                {
                    "start": 1443.55,
                    "end": 1443.87,
                    "word": " may",
                    "probability": 0.8876953125
                },
                {
                    "start": 1443.87,
                    "end": 1444.07,
                    "word": " have",
                    "probability": 0.87744140625
                },
                {
                    "start": 1444.07,
                    "end": 1444.39,
                    "word": " access",
                    "probability": 0.80859375
                },
                {
                    "start": 1444.39,
                    "end": 1444.55,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 1444.55,
                    "end": 1444.67,
                    "word": " the",
                    "probability": 0.80615234375
                },
                {
                    "start": 1444.67,
                    "end": 1445.09,
                    "word": " environment",
                    "probability": 0.68603515625
                },
                {
                    "start": 1445.09,
                    "end": 1445.81,
                    "word": " and",
                    "probability": 0.410400390625
                },
                {
                    "start": 1445.81,
                    "end": 1446.39,
                    "word": " if",
                    "probability": 0.8359375
                },
                {
                    "start": 1446.39,
                    "end": 1446.55,
                    "word": " it's",
                    "probability": 0.932861328125
                },
                {
                    "start": 1446.55,
                    "end": 1446.67,
                    "word": ",",
                    "probability": 0.5166015625
                },
                {
                    "start": 1446.67,
                    "end": 1446.67,
                    "word": " for",
                    "probability": 0.888671875
                },
                {
                    "start": 1446.67,
                    "end": 1446.85,
                    "word": " example",
                    "probability": 0.93359375
                },
                {
                    "start": 1446.85,
                    "end": 1446.89,
                    "word": ",",
                    "probability": 0.9169921875
                },
                {
                    "start": 1447.01,
                    "end": 1447.21,
                    "word": " a game",
                    "probability": 0.6728515625
                }
            ]
        },
        {
            "id": 217,
            "seek": 146202,
            "start": 1447.21,
            "end": 1454.13,
            "text": " environment, it knows what the scores are, how the scores are generated. What is not known? The",
            "tokens": [
                50580,
                2823,
                11,
                309,
                3255,
                437,
                264,
                13444,
                366,
                11,
                577,
                264,
                13444,
                366,
                10833,
                13,
                708,
                307,
                406,
                2570,
                30,
                440,
                50924
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2118252832442522,
            "compression_ratio": 1.7302325581395348,
            "no_speech_prob": 0.0003407001495361328,
            "words": [
                {
                    "start": 1447.21,
                    "end": 1447.63,
                    "word": " environment",
                    "probability": 0.75830078125
                },
                {
                    "start": 1447.63,
                    "end": 1447.79,
                    "word": ",",
                    "probability": 0.765625
                },
                {
                    "start": 1447.79,
                    "end": 1447.91,
                    "word": " it",
                    "probability": 0.892578125
                },
                {
                    "start": 1447.91,
                    "end": 1448.29,
                    "word": " knows",
                    "probability": 0.75
                },
                {
                    "start": 1448.29,
                    "end": 1449.65,
                    "word": " what",
                    "probability": 0.84912109375
                },
                {
                    "start": 1449.65,
                    "end": 1450.11,
                    "word": " the",
                    "probability": 0.80859375
                },
                {
                    "start": 1450.11,
                    "end": 1450.75,
                    "word": " scores",
                    "probability": 0.67724609375
                },
                {
                    "start": 1450.75,
                    "end": 1450.99,
                    "word": " are",
                    "probability": 0.890625
                },
                {
                    "start": 1450.99,
                    "end": 1451.05,
                    "word": ",",
                    "probability": 0.90087890625
                },
                {
                    "start": 1451.05,
                    "end": 1451.17,
                    "word": " how",
                    "probability": 0.8896484375
                },
                {
                    "start": 1451.17,
                    "end": 1451.51,
                    "word": " the",
                    "probability": 0.80615234375
                },
                {
                    "start": 1451.51,
                    "end": 1452.01,
                    "word": " scores",
                    "probability": 0.71044921875
                },
                {
                    "start": 1452.01,
                    "end": 1452.19,
                    "word": " are",
                    "probability": 0.896484375
                },
                {
                    "start": 1452.19,
                    "end": 1452.57,
                    "word": " generated",
                    "probability": 0.7802734375
                },
                {
                    "start": 1452.57,
                    "end": 1452.81,
                    "word": ".",
                    "probability": 0.93505859375
                },
                {
                    "start": 1452.81,
                    "end": 1452.89,
                    "word": " What",
                    "probability": 0.810546875
                },
                {
                    "start": 1452.89,
                    "end": 1453.03,
                    "word": " is",
                    "probability": 0.90234375
                },
                {
                    "start": 1453.03,
                    "end": 1453.33,
                    "word": " not",
                    "probability": 0.88818359375
                },
                {
                    "start": 1453.33,
                    "end": 1453.61,
                    "word": " known",
                    "probability": 0.7255859375
                },
                {
                    "start": 1453.61,
                    "end": 1453.89,
                    "word": "?",
                    "probability": 0.90234375
                },
                {
                    "start": 1453.89,
                    "end": 1454.13,
                    "word": " The",
                    "probability": 0.80419921875
                }
            ]
        },
        {
            "id": 218,
            "seek": 146202,
            "start": 1454.13,
            "end": 1461.71,
            "text": " knowledge of target's neural network architecture is not known. So it's a black box attack. And also",
            "tokens": [
                50924,
                3601,
                295,
                3779,
                311,
                18161,
                3209,
                9482,
                307,
                406,
                2570,
                13,
                407,
                309,
                311,
                257,
                2211,
                2424,
                2690,
                13,
                400,
                611,
                51304
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2118252832442522,
            "compression_ratio": 1.7302325581395348,
            "no_speech_prob": 0.0003407001495361328,
            "words": [
                {
                    "start": 1454.13,
                    "end": 1454.49,
                    "word": " knowledge",
                    "probability": 0.83251953125
                },
                {
                    "start": 1454.49,
                    "end": 1454.69,
                    "word": " of",
                    "probability": 0.90869140625
                },
                {
                    "start": 1454.69,
                    "end": 1455.15,
                    "word": " target's",
                    "probability": 0.5928955078125
                },
                {
                    "start": 1455.15,
                    "end": 1455.33,
                    "word": " neural",
                    "probability": 0.923828125
                },
                {
                    "start": 1455.33,
                    "end": 1455.69,
                    "word": " network",
                    "probability": 0.853515625
                },
                {
                    "start": 1455.69,
                    "end": 1456.23,
                    "word": " architecture",
                    "probability": 0.84521484375
                },
                {
                    "start": 1456.23,
                    "end": 1457.45,
                    "word": " is",
                    "probability": 0.89453125
                },
                {
                    "start": 1457.45,
                    "end": 1457.89,
                    "word": " not",
                    "probability": 0.89794921875
                },
                {
                    "start": 1457.89,
                    "end": 1458.19,
                    "word": " known",
                    "probability": 0.72705078125
                },
                {
                    "start": 1458.19,
                    "end": 1458.33,
                    "word": ".",
                    "probability": 0.890625
                },
                {
                    "start": 1458.33,
                    "end": 1458.45,
                    "word": " So",
                    "probability": 0.89208984375
                },
                {
                    "start": 1458.45,
                    "end": 1458.67,
                    "word": " it's",
                    "probability": 0.945068359375
                },
                {
                    "start": 1458.81,
                    "end": 1459.05,
                    "word": " a black",
                    "probability": 0.85546875
                },
                {
                    "start": 1459.05,
                    "end": 1459.43,
                    "word": " box",
                    "probability": 0.76953125
                },
                {
                    "start": 1459.43,
                    "end": 1460.97,
                    "word": " attack",
                    "probability": 0.75341796875
                },
                {
                    "start": 1460.97,
                    "end": 1461.13,
                    "word": ".",
                    "probability": 0.7978515625
                },
                {
                    "start": 1461.13,
                    "end": 1461.57,
                    "word": " And",
                    "probability": 0.8759765625
                },
                {
                    "start": 1461.57,
                    "end": 1461.71,
                    "word": " also",
                    "probability": 0.81689453125
                }
            ]
        },
        {
            "id": 219,
            "seek": 146202,
            "start": 1461.71,
            "end": 1468.15,
            "text": " the initial parameters, the initialization of the target neural network is also not known.",
            "tokens": [
                51304,
                264,
                5883,
                9834,
                11,
                264,
                5883,
                2144,
                295,
                264,
                3779,
                18161,
                3209,
                307,
                611,
                406,
                2570,
                13,
                51632
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2118252832442522,
            "compression_ratio": 1.7302325581395348,
            "no_speech_prob": 0.0003407001495361328,
            "words": [
                {
                    "start": 1461.71,
                    "end": 1461.87,
                    "word": " the",
                    "probability": 0.75537109375
                },
                {
                    "start": 1461.87,
                    "end": 1462.13,
                    "word": " initial",
                    "probability": 0.86328125
                },
                {
                    "start": 1462.13,
                    "end": 1462.61,
                    "word": " parameters",
                    "probability": 0.783203125
                },
                {
                    "start": 1462.61,
                    "end": 1462.77,
                    "word": ",",
                    "probability": 0.89111328125
                },
                {
                    "start": 1462.77,
                    "end": 1462.87,
                    "word": " the",
                    "probability": 0.79541015625
                },
                {
                    "start": 1462.87,
                    "end": 1463.65,
                    "word": " initialization",
                    "probability": 0.894775390625
                },
                {
                    "start": 1463.65,
                    "end": 1464.25,
                    "word": " of",
                    "probability": 0.8984375
                },
                {
                    "start": 1464.25,
                    "end": 1464.97,
                    "word": " the",
                    "probability": 0.8046875
                },
                {
                    "start": 1464.97,
                    "end": 1466.53,
                    "word": " target",
                    "probability": 0.34033203125
                },
                {
                    "start": 1466.53,
                    "end": 1466.75,
                    "word": " neural",
                    "probability": 0.9228515625
                },
                {
                    "start": 1466.75,
                    "end": 1467.05,
                    "word": " network",
                    "probability": 0.8388671875
                },
                {
                    "start": 1467.05,
                    "end": 1467.49,
                    "word": " is",
                    "probability": 0.80810546875
                },
                {
                    "start": 1467.49,
                    "end": 1467.69,
                    "word": " also",
                    "probability": 0.81787109375
                },
                {
                    "start": 1467.69,
                    "end": 1467.87,
                    "word": " not",
                    "probability": 0.89599609375
                },
                {
                    "start": 1467.87,
                    "end": 1468.03,
                    "word": " known",
                    "probability": 0.75146484375
                },
                {
                    "start": 1468.03,
                    "end": 1468.15,
                    "word": ".",
                    "probability": 0.9931640625
                }
            ]
        },
        {
            "id": 220,
            "seek": 148724,
            "start": 1468.15,
            "end": 1471.91,
            "text": " What is available to adversary in terms of actions? The adversary may perturb",
            "tokens": [
                50368,
                708,
                307,
                2435,
                281,
                48222,
                294,
                2115,
                295,
                5909,
                30,
                440,
                48222,
                815,
                40468,
                50548
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21431326866149902,
            "compression_ratio": 1.5826086956521739,
            "no_speech_prob": 0.00394439697265625,
            "words": [
                {
                    "start": 1468.15,
                    "end": 1468.49,
                    "word": " What",
                    "probability": 0.0975341796875
                },
                {
                    "start": 1468.49,
                    "end": 1468.87,
                    "word": " is",
                    "probability": 0.89697265625
                },
                {
                    "start": 1468.87,
                    "end": 1469.33,
                    "word": " available",
                    "probability": 0.73876953125
                },
                {
                    "start": 1469.33,
                    "end": 1469.53,
                    "word": " to",
                    "probability": 0.90087890625
                },
                {
                    "start": 1469.53,
                    "end": 1469.95,
                    "word": " adversary",
                    "probability": 0.68115234375
                },
                {
                    "start": 1469.95,
                    "end": 1470.13,
                    "word": " in",
                    "probability": 0.88525390625
                },
                {
                    "start": 1470.13,
                    "end": 1470.27,
                    "word": " terms",
                    "probability": 0.7724609375
                },
                {
                    "start": 1470.27,
                    "end": 1470.37,
                    "word": " of",
                    "probability": 0.90625
                },
                {
                    "start": 1470.37,
                    "end": 1470.69,
                    "word": " actions",
                    "probability": 0.85009765625
                },
                {
                    "start": 1470.69,
                    "end": 1470.81,
                    "word": "?",
                    "probability": 0.630859375
                },
                {
                    "start": 1470.81,
                    "end": 1470.85,
                    "word": " The",
                    "probability": 0.783203125
                },
                {
                    "start": 1470.85,
                    "end": 1471.17,
                    "word": " adversary",
                    "probability": 0.8330078125
                },
                {
                    "start": 1471.17,
                    "end": 1471.41,
                    "word": " may",
                    "probability": 0.859375
                },
                {
                    "start": 1471.41,
                    "end": 1471.91,
                    "word": " perturb",
                    "probability": 0.81298828125
                }
            ]
        },
        {
            "id": 221,
            "seek": 148724,
            "start": 1472.55,
            "end": 1479.61,
            "text": " the environment where the target performs. For example, it can change pixel values in a game",
            "tokens": [
                50584,
                264,
                2823,
                689,
                264,
                3779,
                26213,
                13,
                1171,
                1365,
                11,
                309,
                393,
                1319,
                19261,
                4190,
                294,
                257,
                1216,
                50932
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21431326866149902,
            "compression_ratio": 1.5826086956521739,
            "no_speech_prob": 0.00394439697265625,
            "words": [
                {
                    "start": 1472.55,
                    "end": 1473.03,
                    "word": " the",
                    "probability": 0.79052734375
                },
                {
                    "start": 1473.03,
                    "end": 1474.85,
                    "word": " environment",
                    "probability": 0.7392578125
                },
                {
                    "start": 1474.85,
                    "end": 1475.19,
                    "word": " where",
                    "probability": 0.89306640625
                },
                {
                    "start": 1475.19,
                    "end": 1475.35,
                    "word": " the",
                    "probability": 0.80810546875
                },
                {
                    "start": 1475.35,
                    "end": 1475.79,
                    "word": " target",
                    "probability": 0.765625
                },
                {
                    "start": 1475.79,
                    "end": 1476.47,
                    "word": " performs",
                    "probability": 0.6572265625
                },
                {
                    "start": 1476.47,
                    "end": 1476.61,
                    "word": ".",
                    "probability": 0.58984375
                },
                {
                    "start": 1476.61,
                    "end": 1476.75,
                    "word": " For",
                    "probability": 0.8564453125
                },
                {
                    "start": 1476.75,
                    "end": 1477.05,
                    "word": " example",
                    "probability": 0.93017578125
                },
                {
                    "start": 1477.05,
                    "end": 1477.67,
                    "word": ",",
                    "probability": 0.91455078125
                },
                {
                    "start": 1477.67,
                    "end": 1477.83,
                    "word": " it",
                    "probability": 0.8994140625
                },
                {
                    "start": 1477.83,
                    "end": 1478.01,
                    "word": " can",
                    "probability": 0.8994140625
                },
                {
                    "start": 1478.01,
                    "end": 1478.29,
                    "word": " change",
                    "probability": 0.822265625
                },
                {
                    "start": 1478.29,
                    "end": 1478.63,
                    "word": " pixel",
                    "probability": 0.73828125
                },
                {
                    "start": 1478.63,
                    "end": 1479.15,
                    "word": " values",
                    "probability": 0.9306640625
                },
                {
                    "start": 1479.15,
                    "end": 1479.33,
                    "word": " in",
                    "probability": 0.90185546875
                },
                {
                    "start": 1479.43,
                    "end": 1479.61,
                    "word": " a game",
                    "probability": 0.76123046875
                }
            ]
        },
        {
            "id": 222,
            "seek": 148724,
            "start": 1479.61,
            "end": 1488.57,
            "text": " environment through a man-in-the-middle attack. I consider in this work two techniques for",
            "tokens": [
                50932,
                2823,
                807,
                257,
                587,
                12,
                259,
                12,
                3322,
                12,
                25394,
                2285,
                2690,
                13,
                286,
                1949,
                294,
                341,
                589,
                732,
                7512,
                337,
                51376
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21431326866149902,
            "compression_ratio": 1.5826086956521739,
            "no_speech_prob": 0.00394439697265625,
            "words": [
                {
                    "start": 1479.61,
                    "end": 1480.09,
                    "word": " environment",
                    "probability": 0.74755859375
                },
                {
                    "start": 1480.09,
                    "end": 1480.97,
                    "word": " through",
                    "probability": 0.73974609375
                },
                {
                    "start": 1481.45,
                    "end": 1481.75,
                    "word": " a man",
                    "probability": 0.779296875
                },
                {
                    "start": 1481.75,
                    "end": 1481.87,
                    "word": "-in",
                    "probability": 0.7861328125
                },
                {
                    "start": 1481.87,
                    "end": 1481.97,
                    "word": "-the",
                    "probability": 0.93994140625
                },
                {
                    "start": 1481.97,
                    "end": 1482.15,
                    "word": "-middle",
                    "probability": 0.93310546875
                },
                {
                    "start": 1482.15,
                    "end": 1482.59,
                    "word": " attack",
                    "probability": 0.837890625
                },
                {
                    "start": 1482.59,
                    "end": 1483.21,
                    "word": ".",
                    "probability": 0.98876953125
                },
                {
                    "start": 1483.79,
                    "end": 1484.33,
                    "word": " I",
                    "probability": 0.986328125
                },
                {
                    "start": 1484.33,
                    "end": 1484.81,
                    "word": " consider",
                    "probability": 0.7841796875
                },
                {
                    "start": 1484.81,
                    "end": 1485.57,
                    "word": " in",
                    "probability": 0.78515625
                },
                {
                    "start": 1485.57,
                    "end": 1485.75,
                    "word": " this",
                    "probability": 0.86328125
                },
                {
                    "start": 1485.75,
                    "end": 1485.99,
                    "word": " work",
                    "probability": 0.82958984375
                },
                {
                    "start": 1485.99,
                    "end": 1486.41,
                    "word": " two",
                    "probability": 0.6982421875
                },
                {
                    "start": 1486.41,
                    "end": 1488.13,
                    "word": " techniques",
                    "probability": 0.84326171875
                },
                {
                    "start": 1488.13,
                    "end": 1488.57,
                    "word": " for",
                    "probability": 0.89306640625
                }
            ]
        },
        {
            "id": 223,
            "seek": 148724,
            "start": 1488.57,
            "end": 1493.37,
            "text": " perturbing that environment. One is the classical fast gradient sign method for generating adversarial",
            "tokens": [
                51376,
                13269,
                374,
                4324,
                300,
                2823,
                13,
                1485,
                307,
                264,
                13735,
                2370,
                16235,
                1465,
                3170,
                337,
                17746,
                17641,
                44745,
                51624
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21431326866149902,
            "compression_ratio": 1.5826086956521739,
            "no_speech_prob": 0.00394439697265625,
            "words": [
                {
                    "start": 1488.57,
                    "end": 1489.05,
                    "word": " perturbing",
                    "probability": 0.8816731770833334
                },
                {
                    "start": 1489.05,
                    "end": 1489.25,
                    "word": " that",
                    "probability": 0.87939453125
                },
                {
                    "start": 1489.25,
                    "end": 1489.65,
                    "word": " environment",
                    "probability": 0.76171875
                },
                {
                    "start": 1489.65,
                    "end": 1489.89,
                    "word": ".",
                    "probability": 0.95068359375
                },
                {
                    "start": 1489.89,
                    "end": 1490.07,
                    "word": " One",
                    "probability": 0.89013671875
                },
                {
                    "start": 1490.07,
                    "end": 1490.19,
                    "word": " is",
                    "probability": 0.90625
                },
                {
                    "start": 1490.19,
                    "end": 1490.31,
                    "word": " the",
                    "probability": 0.78857421875
                },
                {
                    "start": 1490.31,
                    "end": 1490.65,
                    "word": " classical",
                    "probability": 0.78076171875
                },
                {
                    "start": 1490.65,
                    "end": 1490.93,
                    "word": " fast",
                    "probability": 0.51416015625
                },
                {
                    "start": 1490.93,
                    "end": 1491.17,
                    "word": " gradient",
                    "probability": 0.42919921875
                },
                {
                    "start": 1491.17,
                    "end": 1491.43,
                    "word": " sign",
                    "probability": 0.53466796875
                },
                {
                    "start": 1491.43,
                    "end": 1491.73,
                    "word": " method",
                    "probability": 0.88232421875
                },
                {
                    "start": 1491.73,
                    "end": 1492.47,
                    "word": " for",
                    "probability": 0.8564453125
                },
                {
                    "start": 1492.47,
                    "end": 1492.95,
                    "word": " generating",
                    "probability": 0.8408203125
                },
                {
                    "start": 1492.95,
                    "end": 1493.37,
                    "word": " adversarial",
                    "probability": 0.915283203125
                }
            ]
        },
        {
            "id": 224,
            "seek": 151722,
            "start": 1493.37,
            "end": 1497.83,
            "text": " examples and the other is the Jacobian-based saliency map attack or the JSMA approach",
            "tokens": [
                50364,
                5110,
                293,
                264,
                661,
                307,
                264,
                14117,
                952,
                12,
                6032,
                1845,
                7848,
                4471,
                2690,
                420,
                264,
                33063,
                9998,
                3109,
                50600
            ],
            "temperature": 0.0,
            "avg_logprob": -0.30283506137808575,
            "compression_ratio": 1.5082644628099173,
            "no_speech_prob": 0.0460205078125,
            "words": [
                {
                    "start": 1493.37,
                    "end": 1493.75,
                    "word": " examples",
                    "probability": 0.26904296875
                },
                {
                    "start": 1493.75,
                    "end": 1493.93,
                    "word": " and",
                    "probability": 0.399169921875
                },
                {
                    "start": 1493.93,
                    "end": 1494.07,
                    "word": " the",
                    "probability": 0.81640625
                },
                {
                    "start": 1494.07,
                    "end": 1494.27,
                    "word": " other",
                    "probability": 0.77490234375
                },
                {
                    "start": 1494.27,
                    "end": 1494.67,
                    "word": " is",
                    "probability": 0.9150390625
                },
                {
                    "start": 1494.67,
                    "end": 1494.85,
                    "word": " the",
                    "probability": 0.79345703125
                },
                {
                    "start": 1494.85,
                    "end": 1495.31,
                    "word": " Jacobian",
                    "probability": 0.79345703125
                },
                {
                    "start": 1495.31,
                    "end": 1495.57,
                    "word": "-based",
                    "probability": 0.63232421875
                },
                {
                    "start": 1495.57,
                    "end": 1495.95,
                    "word": " saliency",
                    "probability": 0.743896484375
                },
                {
                    "start": 1495.95,
                    "end": 1496.21,
                    "word": " map",
                    "probability": 0.7744140625
                },
                {
                    "start": 1496.21,
                    "end": 1496.47,
                    "word": " attack",
                    "probability": 0.763671875
                },
                {
                    "start": 1496.47,
                    "end": 1496.63,
                    "word": " or",
                    "probability": 0.68017578125
                },
                {
                    "start": 1496.63,
                    "end": 1496.85,
                    "word": " the",
                    "probability": 0.8037109375
                },
                {
                    "start": 1496.85,
                    "end": 1497.35,
                    "word": " JSMA",
                    "probability": 0.74609375
                },
                {
                    "start": 1497.35,
                    "end": 1497.83,
                    "word": " approach",
                    "probability": 0.78564453125
                }
            ]
        },
        {
            "id": 225,
            "seek": 151722,
            "start": 1497.83,
            "end": 1509.49,
            "text": " introduced by Paperno in 2015, I believe. Now, in that experiment, I used the classical DQN",
            "tokens": [
                50600,
                7268,
                538,
                15919,
                1248,
                78,
                294,
                7546,
                11,
                286,
                1697,
                13,
                823,
                11,
                294,
                300,
                5120,
                11,
                286,
                1143,
                264,
                13735,
                413,
                48,
                45,
                51188
            ],
            "temperature": 0.0,
            "avg_logprob": -0.30283506137808575,
            "compression_ratio": 1.5082644628099173,
            "no_speech_prob": 0.0460205078125,
            "words": [
                {
                    "start": 1497.83,
                    "end": 1498.49,
                    "word": " introduced",
                    "probability": 0.6513671875
                },
                {
                    "start": 1498.49,
                    "end": 1498.93,
                    "word": " by",
                    "probability": 0.90576171875
                },
                {
                    "start": 1498.93,
                    "end": 1500.33,
                    "word": " Paperno",
                    "probability": 0.5863444010416666
                },
                {
                    "start": 1500.33,
                    "end": 1500.75,
                    "word": " in",
                    "probability": 0.88525390625
                },
                {
                    "start": 1500.75,
                    "end": 1502.01,
                    "word": " 2015",
                    "probability": 0.73291015625
                },
                {
                    "start": 1502.01,
                    "end": 1502.23,
                    "word": ",",
                    "probability": 0.44384765625
                },
                {
                    "start": 1502.23,
                    "end": 1502.29,
                    "word": " I",
                    "probability": 0.9833984375
                },
                {
                    "start": 1502.29,
                    "end": 1502.63,
                    "word": " believe",
                    "probability": 0.69921875
                },
                {
                    "start": 1502.63,
                    "end": 1503.35,
                    "word": ".",
                    "probability": 0.9521484375
                },
                {
                    "start": 1504.15,
                    "end": 1504.55,
                    "word": " Now",
                    "probability": 0.77197265625
                },
                {
                    "start": 1504.55,
                    "end": 1504.99,
                    "word": ",",
                    "probability": 0.5712890625
                },
                {
                    "start": 1504.99,
                    "end": 1505.79,
                    "word": " in",
                    "probability": 0.87451171875
                },
                {
                    "start": 1505.79,
                    "end": 1505.91,
                    "word": " that",
                    "probability": 0.87841796875
                },
                {
                    "start": 1505.91,
                    "end": 1506.47,
                    "word": " experiment",
                    "probability": 0.5888671875
                },
                {
                    "start": 1506.47,
                    "end": 1507.51,
                    "word": ",",
                    "probability": 0.55517578125
                },
                {
                    "start": 1507.51,
                    "end": 1507.67,
                    "word": " I",
                    "probability": 0.99609375
                },
                {
                    "start": 1507.67,
                    "end": 1508.09,
                    "word": " used",
                    "probability": 0.80126953125
                },
                {
                    "start": 1508.09,
                    "end": 1508.45,
                    "word": " the",
                    "probability": 0.80126953125
                },
                {
                    "start": 1508.45,
                    "end": 1508.85,
                    "word": " classical",
                    "probability": 0.8818359375
                },
                {
                    "start": 1508.85,
                    "end": 1509.49,
                    "word": " DQN",
                    "probability": 0.9270833333333334
                }
            ]
        },
        {
            "id": 226,
            "seek": 151722,
            "start": 1509.49,
            "end": 1518.25,
            "text": " approach, the classical DQN architecture introduced by Nie in an Atari game, the game of Pong.",
            "tokens": [
                51188,
                3109,
                11,
                264,
                13735,
                413,
                48,
                45,
                9482,
                7268,
                538,
                12016,
                294,
                364,
                41381,
                1216,
                11,
                264,
                1216,
                295,
                430,
                556,
                13,
                51568
            ],
            "temperature": 0.0,
            "avg_logprob": -0.30283506137808575,
            "compression_ratio": 1.5082644628099173,
            "no_speech_prob": 0.0460205078125,
            "words": [
                {
                    "start": 1509.49,
                    "end": 1510.43,
                    "word": " approach",
                    "probability": 0.60009765625
                },
                {
                    "start": 1510.43,
                    "end": 1511.25,
                    "word": ",",
                    "probability": 0.7265625
                },
                {
                    "start": 1511.25,
                    "end": 1511.45,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 1511.45,
                    "end": 1511.77,
                    "word": " classical",
                    "probability": 0.89453125
                },
                {
                    "start": 1511.77,
                    "end": 1512.13,
                    "word": " DQN",
                    "probability": 0.9666341145833334
                },
                {
                    "start": 1512.13,
                    "end": 1512.55,
                    "word": " architecture",
                    "probability": 0.845703125
                },
                {
                    "start": 1512.55,
                    "end": 1513.15,
                    "word": " introduced",
                    "probability": 0.7392578125
                },
                {
                    "start": 1513.15,
                    "end": 1513.51,
                    "word": " by",
                    "probability": 0.90625
                },
                {
                    "start": 1513.51,
                    "end": 1513.89,
                    "word": " Nie",
                    "probability": 0.31494140625
                },
                {
                    "start": 1513.89,
                    "end": 1515.19,
                    "word": " in",
                    "probability": 0.64453125
                },
                {
                    "start": 1515.19,
                    "end": 1515.61,
                    "word": " an",
                    "probability": 0.89453125
                },
                {
                    "start": 1515.61,
                    "end": 1515.89,
                    "word": " Atari",
                    "probability": 0.923828125
                },
                {
                    "start": 1515.89,
                    "end": 1516.23,
                    "word": " game",
                    "probability": 0.818359375
                },
                {
                    "start": 1516.23,
                    "end": 1516.63,
                    "word": ",",
                    "probability": 0.8818359375
                },
                {
                    "start": 1516.63,
                    "end": 1516.83,
                    "word": " the",
                    "probability": 0.74755859375
                },
                {
                    "start": 1516.83,
                    "end": 1517.05,
                    "word": " game",
                    "probability": 0.50244140625
                },
                {
                    "start": 1517.05,
                    "end": 1517.19,
                    "word": " of",
                    "probability": 0.73388671875
                },
                {
                    "start": 1517.19,
                    "end": 1517.53,
                    "word": " Pong",
                    "probability": 0.904296875
                },
                {
                    "start": 1517.53,
                    "end": 1518.25,
                    "word": ".",
                    "probability": 0.662109375
                }
            ]
        },
        {
            "id": 227,
            "seek": 151722,
            "start": 1518.45,
            "end": 1523.35,
            "text": " And this was all implemented in OpenAI Gym with TensorFlow. Back then, PyTorch wasn't really",
            "tokens": [
                51612,
                400,
                341,
                390,
                439,
                12270,
                294,
                7238,
                48698,
                38635,
                365,
                37624,
                13,
                5833,
                550,
                11,
                9953,
                51,
                284,
                339,
                2067,
                380,
                534,
                51860
            ],
            "temperature": 0.0,
            "avg_logprob": -0.30283506137808575,
            "compression_ratio": 1.5082644628099173,
            "no_speech_prob": 0.0460205078125,
            "words": [
                {
                    "start": 1518.45,
                    "end": 1518.85,
                    "word": " And",
                    "probability": 0.66259765625
                },
                {
                    "start": 1518.85,
                    "end": 1519.51,
                    "word": " this",
                    "probability": 0.85791015625
                },
                {
                    "start": 1519.51,
                    "end": 1519.67,
                    "word": " was",
                    "probability": 0.8837890625
                },
                {
                    "start": 1519.67,
                    "end": 1519.87,
                    "word": " all",
                    "probability": 0.89990234375
                },
                {
                    "start": 1519.87,
                    "end": 1520.21,
                    "word": " implemented",
                    "probability": 0.68505859375
                },
                {
                    "start": 1520.21,
                    "end": 1520.37,
                    "word": " in",
                    "probability": 0.89794921875
                },
                {
                    "start": 1520.37,
                    "end": 1520.71,
                    "word": " OpenAI",
                    "probability": 0.865234375
                },
                {
                    "start": 1520.71,
                    "end": 1521.09,
                    "word": " Gym",
                    "probability": 0.412841796875
                },
                {
                    "start": 1521.09,
                    "end": 1521.57,
                    "word": " with",
                    "probability": 0.59375
                },
                {
                    "start": 1521.57,
                    "end": 1521.97,
                    "word": " TensorFlow",
                    "probability": 0.734375
                },
                {
                    "start": 1521.97,
                    "end": 1522.23,
                    "word": ".",
                    "probability": 0.51318359375
                },
                {
                    "start": 1522.23,
                    "end": 1522.33,
                    "word": " Back",
                    "probability": 0.9306640625
                },
                {
                    "start": 1522.33,
                    "end": 1522.51,
                    "word": " then",
                    "probability": 0.720703125
                },
                {
                    "start": 1522.51,
                    "end": 1522.55,
                    "word": ",",
                    "probability": 0.712890625
                },
                {
                    "start": 1522.55,
                    "end": 1522.95,
                    "word": " PyTorch",
                    "probability": 0.953125
                },
                {
                    "start": 1522.95,
                    "end": 1523.15,
                    "word": " wasn't",
                    "probability": 0.909912109375
                },
                {
                    "start": 1523.15,
                    "end": 1523.35,
                    "word": " really",
                    "probability": 0.75390625
                }
            ]
        },
        {
            "id": 228,
            "seek": 153874,
            "start": 1523.35,
            "end": 1530.73,
            "text": " and trained the agent against heuristic AI. And here's the initial proof of concept result.",
            "tokens": [
                50412,
                293,
                8895,
                264,
                9461,
                1970,
                415,
                374,
                3142,
                7318,
                13,
                400,
                510,
                311,
                264,
                5883,
                8177,
                295,
                3410,
                1874,
                13,
                50712
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21727195382118225,
            "compression_ratio": 1.4867724867724867,
            "no_speech_prob": 0.00017130374908447266,
            "words": [
                {
                    "start": 1523.35,
                    "end": 1524.17,
                    "word": " and",
                    "probability": 0.07403564453125
                },
                {
                    "start": 1524.17,
                    "end": 1525.25,
                    "word": " trained",
                    "probability": 0.1492919921875
                },
                {
                    "start": 1525.25,
                    "end": 1525.49,
                    "word": " the",
                    "probability": 0.818359375
                },
                {
                    "start": 1525.49,
                    "end": 1525.81,
                    "word": " agent",
                    "probability": 0.75048828125
                },
                {
                    "start": 1525.81,
                    "end": 1526.19,
                    "word": " against",
                    "probability": 0.861328125
                },
                {
                    "start": 1526.19,
                    "end": 1526.81,
                    "word": " heuristic",
                    "probability": 0.92724609375
                },
                {
                    "start": 1526.81,
                    "end": 1527.17,
                    "word": " AI",
                    "probability": 0.89501953125
                },
                {
                    "start": 1527.17,
                    "end": 1527.65,
                    "word": ".",
                    "probability": 0.52978515625
                },
                {
                    "start": 1527.93,
                    "end": 1528.03,
                    "word": " And",
                    "probability": 0.84765625
                },
                {
                    "start": 1528.03,
                    "end": 1528.29,
                    "word": " here's",
                    "probability": 0.80322265625
                },
                {
                    "start": 1528.29,
                    "end": 1528.49,
                    "word": " the",
                    "probability": 0.80859375
                },
                {
                    "start": 1528.49,
                    "end": 1529.03,
                    "word": " initial",
                    "probability": 0.8505859375
                },
                {
                    "start": 1529.03,
                    "end": 1529.57,
                    "word": " proof",
                    "probability": 0.82861328125
                },
                {
                    "start": 1529.57,
                    "end": 1529.69,
                    "word": " of",
                    "probability": 0.76025390625
                },
                {
                    "start": 1529.69,
                    "end": 1529.91,
                    "word": " concept",
                    "probability": 0.78271484375
                },
                {
                    "start": 1529.91,
                    "end": 1530.33,
                    "word": " result",
                    "probability": 0.78369140625
                },
                {
                    "start": 1530.33,
                    "end": 1530.73,
                    "word": ".",
                    "probability": 0.96728515625
                }
            ]
        },
        {
            "id": 229,
            "seek": 153874,
            "start": 1530.73,
            "end": 1533.93,
            "text": " This is for the white box attack. Of course, later on, I will show you the results for the",
            "tokens": [
                50740,
                639,
                307,
                337,
                264,
                2418,
                2424,
                2690,
                13,
                2720,
                1164,
                11,
                1780,
                322,
                11,
                286,
                486,
                855,
                291,
                264,
                3542,
                337,
                264,
                50888
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21727195382118225,
            "compression_ratio": 1.4867724867724867,
            "no_speech_prob": 0.00017130374908447266,
            "words": [
                {
                    "start": 1530.73,
                    "end": 1531.17,
                    "word": " This",
                    "probability": 0.8017578125
                },
                {
                    "start": 1531.17,
                    "end": 1531.27,
                    "word": " is",
                    "probability": 0.90234375
                },
                {
                    "start": 1531.27,
                    "end": 1531.39,
                    "word": " for",
                    "probability": 0.8876953125
                },
                {
                    "start": 1531.39,
                    "end": 1531.51,
                    "word": " the",
                    "probability": 0.8173828125
                },
                {
                    "start": 1531.51,
                    "end": 1531.73,
                    "word": " white",
                    "probability": 0.71826171875
                },
                {
                    "start": 1531.73,
                    "end": 1531.93,
                    "word": " box",
                    "probability": 0.6708984375
                },
                {
                    "start": 1531.93,
                    "end": 1532.15,
                    "word": " attack",
                    "probability": 0.7626953125
                },
                {
                    "start": 1532.15,
                    "end": 1532.27,
                    "word": ".",
                    "probability": 0.431640625
                },
                {
                    "start": 1532.27,
                    "end": 1532.33,
                    "word": " Of",
                    "probability": 0.87646484375
                },
                {
                    "start": 1532.33,
                    "end": 1532.47,
                    "word": " course",
                    "probability": 0.859375
                },
                {
                    "start": 1532.47,
                    "end": 1532.57,
                    "word": ",",
                    "probability": 0.85986328125
                },
                {
                    "start": 1532.57,
                    "end": 1532.73,
                    "word": " later",
                    "probability": 0.74658203125
                },
                {
                    "start": 1532.73,
                    "end": 1532.85,
                    "word": " on",
                    "probability": 0.8798828125
                },
                {
                    "start": 1532.85,
                    "end": 1532.91,
                    "word": ",",
                    "probability": 0.50830078125
                },
                {
                    "start": 1532.91,
                    "end": 1532.91,
                    "word": " I",
                    "probability": 0.98046875
                },
                {
                    "start": 1532.91,
                    "end": 1533.01,
                    "word": " will",
                    "probability": 0.75048828125
                },
                {
                    "start": 1533.01,
                    "end": 1533.17,
                    "word": " show",
                    "probability": 0.87939453125
                },
                {
                    "start": 1533.17,
                    "end": 1533.25,
                    "word": " you",
                    "probability": 0.88427734375
                },
                {
                    "start": 1533.25,
                    "end": 1533.33,
                    "word": " the",
                    "probability": 0.78857421875
                },
                {
                    "start": 1533.33,
                    "end": 1533.53,
                    "word": " results",
                    "probability": 0.66015625
                },
                {
                    "start": 1533.53,
                    "end": 1533.77,
                    "word": " for",
                    "probability": 0.888671875
                },
                {
                    "start": 1533.77,
                    "end": 1533.93,
                    "word": " the",
                    "probability": 0.8095703125
                }
            ]
        },
        {
            "id": 230,
            "seek": 153874,
            "start": 1533.93,
            "end": 1544.87,
            "text": " black box attack. And you can see that for FGSM and JSMA, regardless of how far along the training",
            "tokens": [
                50888,
                2211,
                2424,
                2690,
                13,
                400,
                291,
                393,
                536,
                300,
                337,
                479,
                24446,
                44,
                293,
                33063,
                9998,
                11,
                10060,
                295,
                577,
                1400,
                2051,
                264,
                3097,
                51464
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21727195382118225,
            "compression_ratio": 1.4867724867724867,
            "no_speech_prob": 0.00017130374908447266,
            "words": [
                {
                    "start": 1533.93,
                    "end": 1534.13,
                    "word": " black",
                    "probability": 0.81787109375
                },
                {
                    "start": 1534.13,
                    "end": 1534.33,
                    "word": " box",
                    "probability": 0.767578125
                },
                {
                    "start": 1534.33,
                    "end": 1534.71,
                    "word": " attack",
                    "probability": 0.849609375
                },
                {
                    "start": 1534.71,
                    "end": 1534.95,
                    "word": ".",
                    "probability": 0.80224609375
                },
                {
                    "start": 1534.95,
                    "end": 1535.19,
                    "word": " And",
                    "probability": 0.8916015625
                },
                {
                    "start": 1535.19,
                    "end": 1535.29,
                    "word": " you",
                    "probability": 0.89208984375
                },
                {
                    "start": 1535.29,
                    "end": 1535.43,
                    "word": " can",
                    "probability": 0.89599609375
                },
                {
                    "start": 1535.43,
                    "end": 1535.69,
                    "word": " see",
                    "probability": 0.8232421875
                },
                {
                    "start": 1535.69,
                    "end": 1536.03,
                    "word": " that",
                    "probability": 0.876953125
                },
                {
                    "start": 1536.03,
                    "end": 1536.79,
                    "word": " for",
                    "probability": 0.8125
                },
                {
                    "start": 1536.79,
                    "end": 1539.01,
                    "word": " FGSM",
                    "probability": 0.8759765625
                },
                {
                    "start": 1539.01,
                    "end": 1539.17,
                    "word": " and",
                    "probability": 0.86669921875
                },
                {
                    "start": 1539.17,
                    "end": 1539.87,
                    "word": " JSMA",
                    "probability": 0.776123046875
                },
                {
                    "start": 1539.87,
                    "end": 1541.15,
                    "word": ",",
                    "probability": 0.84033203125
                },
                {
                    "start": 1541.15,
                    "end": 1542.67,
                    "word": " regardless",
                    "probability": 0.900390625
                },
                {
                    "start": 1542.67,
                    "end": 1543.29,
                    "word": " of",
                    "probability": 0.90380859375
                },
                {
                    "start": 1543.29,
                    "end": 1543.67,
                    "word": " how",
                    "probability": 0.89306640625
                },
                {
                    "start": 1543.67,
                    "end": 1543.95,
                    "word": " far",
                    "probability": 0.9033203125
                },
                {
                    "start": 1543.95,
                    "end": 1544.27,
                    "word": " along",
                    "probability": 0.9111328125
                },
                {
                    "start": 1544.27,
                    "end": 1544.49,
                    "word": " the",
                    "probability": 0.814453125
                },
                {
                    "start": 1544.49,
                    "end": 1544.87,
                    "word": " training",
                    "probability": 0.8212890625
                }
            ]
        },
        {
            "id": 231,
            "seek": 156602,
            "start": 1544.87,
            "end": 1554.23,
            "text": " the agent is, both policies, the policy for the Pong agent is highly vulnerable",
            "tokens": [
                50388,
                264,
                9461,
                307,
                11,
                1293,
                7657,
                11,
                264,
                3897,
                337,
                264,
                430,
                556,
                9461,
                307,
                5405,
                10955,
                50848
            ],
            "temperature": 0.0,
            "avg_logprob": -0.15882457911290906,
            "compression_ratio": 1.5739910313901346,
            "no_speech_prob": 0.06097412109375,
            "words": [
                {
                    "start": 1544.87,
                    "end": 1545.59,
                    "word": " the",
                    "probability": 0.1259765625
                },
                {
                    "start": 1545.59,
                    "end": 1545.99,
                    "word": " agent",
                    "probability": 0.263916015625
                },
                {
                    "start": 1545.99,
                    "end": 1546.49,
                    "word": " is",
                    "probability": 0.88330078125
                },
                {
                    "start": 1546.49,
                    "end": 1547.73,
                    "word": ",",
                    "probability": 0.369140625
                },
                {
                    "start": 1547.73,
                    "end": 1548.43,
                    "word": " both",
                    "probability": 0.72119140625
                },
                {
                    "start": 1548.43,
                    "end": 1549.21,
                    "word": " policies",
                    "probability": 0.8349609375
                },
                {
                    "start": 1549.21,
                    "end": 1549.57,
                    "word": ",",
                    "probability": 0.685546875
                },
                {
                    "start": 1549.57,
                    "end": 1549.73,
                    "word": " the",
                    "probability": 0.80615234375
                },
                {
                    "start": 1549.73,
                    "end": 1550.63,
                    "word": " policy",
                    "probability": 0.80029296875
                },
                {
                    "start": 1550.63,
                    "end": 1551.95,
                    "word": " for",
                    "probability": 0.84716796875
                },
                {
                    "start": 1551.95,
                    "end": 1552.33,
                    "word": " the",
                    "probability": 0.81201171875
                },
                {
                    "start": 1552.33,
                    "end": 1552.63,
                    "word": " Pong",
                    "probability": 0.704345703125
                },
                {
                    "start": 1552.63,
                    "end": 1552.97,
                    "word": " agent",
                    "probability": 0.720703125
                },
                {
                    "start": 1552.97,
                    "end": 1553.21,
                    "word": " is",
                    "probability": 0.77490234375
                },
                {
                    "start": 1553.21,
                    "end": 1553.63,
                    "word": " highly",
                    "probability": 0.88671875
                },
                {
                    "start": 1553.63,
                    "end": 1554.23,
                    "word": " vulnerable",
                    "probability": 0.8115234375
                }
            ]
        },
        {
            "id": 232,
            "seek": 156602,
            "start": 1554.23,
            "end": 1559.63,
            "text": " to adversarial perturbations through simple techniques like FGSM and JSMA. You can see",
            "tokens": [
                50848,
                281,
                17641,
                44745,
                40468,
                763,
                807,
                2199,
                7512,
                411,
                479,
                24446,
                44,
                293,
                33063,
                9998,
                13,
                509,
                393,
                536,
                51096
            ],
            "temperature": 0.0,
            "avg_logprob": -0.15882457911290906,
            "compression_ratio": 1.5739910313901346,
            "no_speech_prob": 0.06097412109375,
            "words": [
                {
                    "start": 1554.23,
                    "end": 1554.85,
                    "word": " to",
                    "probability": 0.89111328125
                },
                {
                    "start": 1554.85,
                    "end": 1555.33,
                    "word": " adversarial",
                    "probability": 0.92236328125
                },
                {
                    "start": 1555.33,
                    "end": 1556.09,
                    "word": " perturbations",
                    "probability": 0.9091796875
                },
                {
                    "start": 1556.09,
                    "end": 1556.49,
                    "word": " through",
                    "probability": 0.6572265625
                },
                {
                    "start": 1556.49,
                    "end": 1556.91,
                    "word": " simple",
                    "probability": 0.91650390625
                },
                {
                    "start": 1556.91,
                    "end": 1557.37,
                    "word": " techniques",
                    "probability": 0.82666015625
                },
                {
                    "start": 1557.37,
                    "end": 1557.89,
                    "word": " like",
                    "probability": 0.8291015625
                },
                {
                    "start": 1557.89,
                    "end": 1558.49,
                    "word": " FGSM",
                    "probability": 0.9140625
                },
                {
                    "start": 1558.49,
                    "end": 1558.93,
                    "word": " and",
                    "probability": 0.87451171875
                },
                {
                    "start": 1558.93,
                    "end": 1559.27,
                    "word": " JSMA",
                    "probability": 0.6839599609375
                },
                {
                    "start": 1559.27,
                    "end": 1559.39,
                    "word": ".",
                    "probability": 0.8779296875
                },
                {
                    "start": 1559.39,
                    "end": 1559.39,
                    "word": " You",
                    "probability": 0.84912109375
                },
                {
                    "start": 1559.39,
                    "end": 1559.49,
                    "word": " can",
                    "probability": 0.89208984375
                },
                {
                    "start": 1559.49,
                    "end": 1559.63,
                    "word": " see",
                    "probability": 0.8125
                }
            ]
        },
        {
            "id": 233,
            "seek": 156602,
            "start": 1559.63,
            "end": 1565.87,
            "text": " that for JSMA, the success rate was 100% for all of the cases. For FGSM, it was slightly lower",
            "tokens": [
                51096,
                300,
                337,
                33063,
                9998,
                11,
                264,
                2245,
                3314,
                390,
                2319,
                4,
                337,
                439,
                295,
                264,
                3331,
                13,
                1171,
                479,
                24446,
                44,
                11,
                309,
                390,
                4748,
                3126,
                51412
            ],
            "temperature": 0.0,
            "avg_logprob": -0.15882457911290906,
            "compression_ratio": 1.5739910313901346,
            "no_speech_prob": 0.06097412109375,
            "words": [
                {
                    "start": 1559.63,
                    "end": 1559.75,
                    "word": " that",
                    "probability": 0.869140625
                },
                {
                    "start": 1559.75,
                    "end": 1559.87,
                    "word": " for",
                    "probability": 0.88525390625
                },
                {
                    "start": 1559.87,
                    "end": 1560.37,
                    "word": " JSMA",
                    "probability": 0.88525390625
                },
                {
                    "start": 1560.37,
                    "end": 1560.79,
                    "word": ",",
                    "probability": 0.59228515625
                },
                {
                    "start": 1560.79,
                    "end": 1560.91,
                    "word": " the",
                    "probability": 0.80078125
                },
                {
                    "start": 1560.91,
                    "end": 1561.15,
                    "word": " success",
                    "probability": 0.77880859375
                },
                {
                    "start": 1561.15,
                    "end": 1561.43,
                    "word": " rate",
                    "probability": 0.857421875
                },
                {
                    "start": 1561.43,
                    "end": 1561.81,
                    "word": " was",
                    "probability": 0.88623046875
                },
                {
                    "start": 1561.81,
                    "end": 1562.15,
                    "word": " 100",
                    "probability": 0.7548828125
                },
                {
                    "start": 1562.15,
                    "end": 1562.41,
                    "word": "%",
                    "probability": 0.91552734375
                },
                {
                    "start": 1562.41,
                    "end": 1562.75,
                    "word": " for",
                    "probability": 0.8984375
                },
                {
                    "start": 1562.75,
                    "end": 1562.99,
                    "word": " all",
                    "probability": 0.90283203125
                },
                {
                    "start": 1562.99,
                    "end": 1563.09,
                    "word": " of",
                    "probability": 0.89599609375
                },
                {
                    "start": 1563.09,
                    "end": 1563.19,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 1563.19,
                    "end": 1563.57,
                    "word": " cases",
                    "probability": 0.80712890625
                },
                {
                    "start": 1563.57,
                    "end": 1564.21,
                    "word": ".",
                    "probability": 0.85888671875
                },
                {
                    "start": 1564.21,
                    "end": 1564.33,
                    "word": " For",
                    "probability": 0.8935546875
                },
                {
                    "start": 1564.33,
                    "end": 1564.91,
                    "word": " FGSM",
                    "probability": 0.9752604166666666
                },
                {
                    "start": 1564.91,
                    "end": 1564.95,
                    "word": ",",
                    "probability": 0.677734375
                },
                {
                    "start": 1564.95,
                    "end": 1565.01,
                    "word": " it",
                    "probability": 0.8984375
                },
                {
                    "start": 1565.01,
                    "end": 1565.19,
                    "word": " was",
                    "probability": 0.89794921875
                },
                {
                    "start": 1565.19,
                    "end": 1565.49,
                    "word": " slightly",
                    "probability": 0.7744140625
                },
                {
                    "start": 1565.49,
                    "end": 1565.87,
                    "word": " lower",
                    "probability": 0.78076171875
                }
            ]
        },
        {
            "id": 234,
            "seek": 156602,
            "start": 1566.54,
            "end": 1572.15,
            "text": " and it was mostly because of the termination criteria and the perturbation threshold that",
            "tokens": [
                51452,
                293,
                309,
                390,
                5240,
                570,
                295,
                264,
                1433,
                2486,
                11101,
                293,
                264,
                40468,
                399,
                14678,
                300,
                51728
            ],
            "temperature": 0.0,
            "avg_logprob": -0.15882457911290906,
            "compression_ratio": 1.5739910313901346,
            "no_speech_prob": 0.06097412109375,
            "words": [
                {
                    "start": 1566.54,
                    "end": 1566.91,
                    "word": " and",
                    "probability": 0.446044921875
                },
                {
                    "start": 1566.91,
                    "end": 1567.05,
                    "word": " it",
                    "probability": 0.89892578125
                },
                {
                    "start": 1567.05,
                    "end": 1567.31,
                    "word": " was",
                    "probability": 0.89208984375
                },
                {
                    "start": 1567.31,
                    "end": 1567.67,
                    "word": " mostly",
                    "probability": 0.880859375
                },
                {
                    "start": 1567.67,
                    "end": 1567.87,
                    "word": " because",
                    "probability": 0.8330078125
                },
                {
                    "start": 1567.87,
                    "end": 1567.99,
                    "word": " of",
                    "probability": 0.90625
                },
                {
                    "start": 1567.99,
                    "end": 1568.31,
                    "word": " the",
                    "probability": 0.81982421875
                },
                {
                    "start": 1568.31,
                    "end": 1569.01,
                    "word": " termination",
                    "probability": 0.898193359375
                },
                {
                    "start": 1569.01,
                    "end": 1569.45,
                    "word": " criteria",
                    "probability": 0.77685546875
                },
                {
                    "start": 1569.45,
                    "end": 1569.97,
                    "word": " and",
                    "probability": 0.88525390625
                },
                {
                    "start": 1569.97,
                    "end": 1570.29,
                    "word": " the",
                    "probability": 0.82421875
                },
                {
                    "start": 1570.29,
                    "end": 1570.93,
                    "word": " perturbation",
                    "probability": 0.89892578125
                },
                {
                    "start": 1570.93,
                    "end": 1571.89,
                    "word": " threshold",
                    "probability": 0.65673828125
                },
                {
                    "start": 1571.89,
                    "end": 1572.15,
                    "word": " that",
                    "probability": 0.86474609375
                }
            ]
        },
        {
            "id": 235,
            "seek": 159480,
            "start": 1572.15,
            "end": 1578.75,
            "text": " I had defined. But you can see, it can be observed that the policy can be very easily",
            "tokens": [
                50364,
                286,
                632,
                7642,
                13,
                583,
                291,
                393,
                536,
                11,
                309,
                393,
                312,
                13095,
                300,
                264,
                3897,
                393,
                312,
                588,
                3612,
                50692
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23802570009900031,
            "compression_ratio": 1.6028880866425992,
            "no_speech_prob": 0.01450347900390625,
            "words": [
                {
                    "start": 1572.15,
                    "end": 1572.41,
                    "word": " I",
                    "probability": 0.75
                },
                {
                    "start": 1572.41,
                    "end": 1572.57,
                    "word": " had",
                    "probability": 0.6533203125
                },
                {
                    "start": 1572.57,
                    "end": 1572.93,
                    "word": " defined",
                    "probability": 0.7314453125
                },
                {
                    "start": 1572.93,
                    "end": 1573.15,
                    "word": ".",
                    "probability": 0.6533203125
                },
                {
                    "start": 1573.15,
                    "end": 1573.31,
                    "word": " But",
                    "probability": 0.415771484375
                },
                {
                    "start": 1573.31,
                    "end": 1573.45,
                    "word": " you",
                    "probability": 0.83154296875
                },
                {
                    "start": 1573.45,
                    "end": 1573.55,
                    "word": " can",
                    "probability": 0.89453125
                },
                {
                    "start": 1573.55,
                    "end": 1574.03,
                    "word": " see",
                    "probability": 0.81982421875
                },
                {
                    "start": 1574.03,
                    "end": 1574.27,
                    "word": ",",
                    "probability": 0.53173828125
                },
                {
                    "start": 1574.27,
                    "end": 1574.39,
                    "word": " it",
                    "probability": 0.87890625
                },
                {
                    "start": 1574.39,
                    "end": 1574.51,
                    "word": " can",
                    "probability": 0.89697265625
                },
                {
                    "start": 1574.51,
                    "end": 1574.63,
                    "word": " be",
                    "probability": 0.90625
                },
                {
                    "start": 1574.63,
                    "end": 1574.95,
                    "word": " observed",
                    "probability": 0.77685546875
                },
                {
                    "start": 1574.95,
                    "end": 1575.33,
                    "word": " that",
                    "probability": 0.7373046875
                },
                {
                    "start": 1575.33,
                    "end": 1577.07,
                    "word": " the",
                    "probability": 0.73583984375
                },
                {
                    "start": 1577.07,
                    "end": 1577.41,
                    "word": " policy",
                    "probability": 0.791015625
                },
                {
                    "start": 1577.41,
                    "end": 1577.65,
                    "word": " can",
                    "probability": 0.9013671875
                },
                {
                    "start": 1577.65,
                    "end": 1577.83,
                    "word": " be",
                    "probability": 0.9033203125
                },
                {
                    "start": 1577.83,
                    "end": 1578.23,
                    "word": " very",
                    "probability": 0.72998046875
                },
                {
                    "start": 1578.23,
                    "end": 1578.75,
                    "word": " easily",
                    "probability": 0.8505859375
                }
            ]
        },
        {
            "id": 236,
            "seek": 159480,
            "start": 1578.75,
            "end": 1583.69,
            "text": " manipulated through adversarial example attacks. So, at 100 random observations,",
            "tokens": [
                50692,
                37161,
                807,
                17641,
                44745,
                1365,
                8122,
                13,
                407,
                11,
                412,
                2319,
                4974,
                18163,
                11,
                50932
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23802570009900031,
            "compression_ratio": 1.6028880866425992,
            "no_speech_prob": 0.01450347900390625,
            "words": [
                {
                    "start": 1578.75,
                    "end": 1579.43,
                    "word": " manipulated",
                    "probability": 0.94091796875
                },
                {
                    "start": 1579.43,
                    "end": 1580.11,
                    "word": " through",
                    "probability": 0.7724609375
                },
                {
                    "start": 1580.11,
                    "end": 1580.61,
                    "word": " adversarial",
                    "probability": 0.923828125
                },
                {
                    "start": 1580.61,
                    "end": 1581.01,
                    "word": " example",
                    "probability": 0.9208984375
                },
                {
                    "start": 1581.01,
                    "end": 1581.41,
                    "word": " attacks",
                    "probability": 0.73046875
                },
                {
                    "start": 1581.41,
                    "end": 1581.61,
                    "word": ".",
                    "probability": 0.96728515625
                },
                {
                    "start": 1581.61,
                    "end": 1581.87,
                    "word": " So",
                    "probability": 0.82763671875
                },
                {
                    "start": 1581.87,
                    "end": 1582.31,
                    "word": ",",
                    "probability": 0.294677734375
                },
                {
                    "start": 1582.31,
                    "end": 1582.43,
                    "word": " at",
                    "probability": 0.4853515625
                },
                {
                    "start": 1582.43,
                    "end": 1582.67,
                    "word": " 100",
                    "probability": 0.87255859375
                },
                {
                    "start": 1582.67,
                    "end": 1582.99,
                    "word": " random",
                    "probability": 0.78515625
                },
                {
                    "start": 1582.99,
                    "end": 1583.57,
                    "word": " observations",
                    "probability": 0.6904296875
                },
                {
                    "start": 1583.57,
                    "end": 1583.69,
                    "word": ",",
                    "probability": 0.3193359375
                }
            ]
        },
        {
            "id": 237,
            "seek": 159480,
            "start": 1583.69,
            "end": 1588.81,
            "text": " perturbed with FGSM and JSMA, the results were fed, the perturbed images were fed to",
            "tokens": [
                50932,
                13269,
                374,
                2883,
                365,
                479,
                24446,
                44,
                293,
                33063,
                9998,
                11,
                264,
                3542,
                645,
                4636,
                11,
                264,
                13269,
                374,
                2883,
                5267,
                645,
                4636,
                281,
                51188
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23802570009900031,
            "compression_ratio": 1.6028880866425992,
            "no_speech_prob": 0.01450347900390625,
            "words": [
                {
                    "start": 1583.69,
                    "end": 1583.99,
                    "word": " perturbed",
                    "probability": 0.8894856770833334
                },
                {
                    "start": 1583.99,
                    "end": 1584.33,
                    "word": " with",
                    "probability": 0.80224609375
                },
                {
                    "start": 1584.33,
                    "end": 1584.91,
                    "word": " FGSM",
                    "probability": 0.5353190104166666
                },
                {
                    "start": 1584.91,
                    "end": 1584.97,
                    "word": " and",
                    "probability": 0.7822265625
                },
                {
                    "start": 1584.97,
                    "end": 1585.39,
                    "word": " JSMA",
                    "probability": 0.733154296875
                },
                {
                    "start": 1585.39,
                    "end": 1585.77,
                    "word": ",",
                    "probability": 0.8837890625
                },
                {
                    "start": 1585.77,
                    "end": 1585.87,
                    "word": " the",
                    "probability": 0.80419921875
                },
                {
                    "start": 1585.87,
                    "end": 1586.17,
                    "word": " results",
                    "probability": 0.69873046875
                },
                {
                    "start": 1586.17,
                    "end": 1586.43,
                    "word": " were",
                    "probability": 0.8857421875
                },
                {
                    "start": 1586.43,
                    "end": 1586.99,
                    "word": " fed",
                    "probability": 0.91650390625
                },
                {
                    "start": 1586.99,
                    "end": 1587.15,
                    "word": ",",
                    "probability": 0.6591796875
                },
                {
                    "start": 1587.15,
                    "end": 1587.35,
                    "word": " the",
                    "probability": 0.822265625
                },
                {
                    "start": 1587.35,
                    "end": 1587.79,
                    "word": " perturbed",
                    "probability": 0.9381510416666666
                },
                {
                    "start": 1587.79,
                    "end": 1588.11,
                    "word": " images",
                    "probability": 0.8544921875
                },
                {
                    "start": 1588.11,
                    "end": 1588.31,
                    "word": " were",
                    "probability": 0.8720703125
                },
                {
                    "start": 1588.31,
                    "end": 1588.55,
                    "word": " fed",
                    "probability": 0.87744140625
                },
                {
                    "start": 1588.55,
                    "end": 1588.81,
                    "word": " to",
                    "probability": 0.8857421875
                }
            ]
        },
        {
            "id": 238,
            "seek": 159480,
            "start": 1588.81,
            "end": 1595.79,
            "text": " the trained neural network representing the agent's policy as test input, and then the success",
            "tokens": [
                51216,
                264,
                8895,
                18161,
                3209,
                13460,
                264,
                9461,
                311,
                3897,
                382,
                1500,
                4846,
                11,
                293,
                550,
                264,
                2245,
                51544
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23802570009900031,
            "compression_ratio": 1.6028880866425992,
            "no_speech_prob": 0.01450347900390625,
            "words": [
                {
                    "start": 1588.81,
                    "end": 1589.55,
                    "word": " the",
                    "probability": 0.798828125
                },
                {
                    "start": 1589.55,
                    "end": 1590.05,
                    "word": " trained",
                    "probability": 0.83447265625
                },
                {
                    "start": 1590.05,
                    "end": 1590.71,
                    "word": " neural",
                    "probability": 0.8935546875
                },
                {
                    "start": 1590.71,
                    "end": 1590.99,
                    "word": " network",
                    "probability": 0.849609375
                },
                {
                    "start": 1590.99,
                    "end": 1591.43,
                    "word": " representing",
                    "probability": 0.51904296875
                },
                {
                    "start": 1591.43,
                    "end": 1591.79,
                    "word": " the",
                    "probability": 0.81689453125
                },
                {
                    "start": 1591.79,
                    "end": 1592.21,
                    "word": " agent's",
                    "probability": 0.876708984375
                },
                {
                    "start": 1592.21,
                    "end": 1592.63,
                    "word": " policy",
                    "probability": 0.8037109375
                },
                {
                    "start": 1592.63,
                    "end": 1592.83,
                    "word": " as",
                    "probability": 0.8779296875
                },
                {
                    "start": 1592.83,
                    "end": 1593.09,
                    "word": " test",
                    "probability": 0.73388671875
                },
                {
                    "start": 1593.09,
                    "end": 1593.33,
                    "word": " input",
                    "probability": 0.88330078125
                },
                {
                    "start": 1593.33,
                    "end": 1593.65,
                    "word": ",",
                    "probability": 0.59521484375
                },
                {
                    "start": 1593.65,
                    "end": 1593.93,
                    "word": " and",
                    "probability": 0.89111328125
                },
                {
                    "start": 1593.93,
                    "end": 1594.21,
                    "word": " then",
                    "probability": 0.71875
                },
                {
                    "start": 1594.21,
                    "end": 1595.55,
                    "word": " the",
                    "probability": 0.74755859375
                },
                {
                    "start": 1595.55,
                    "end": 1595.79,
                    "word": " success",
                    "probability": 0.75732421875
                }
            ]
        },
        {
            "id": 239,
            "seek": 159480,
            "start": 1595.79,
            "end": 1600.93,
            "text": " rate was measured. Now, is this type of attack practical? Is this really something that we should",
            "tokens": [
                51544,
                3314,
                390,
                12690,
                13,
                823,
                11,
                307,
                341,
                2010,
                295,
                2690,
                8496,
                30,
                1119,
                341,
                534,
                746,
                300,
                321,
                820,
                51796
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23802570009900031,
            "compression_ratio": 1.6028880866425992,
            "no_speech_prob": 0.01450347900390625,
            "words": [
                {
                    "start": 1595.79,
                    "end": 1596.11,
                    "word": " rate",
                    "probability": 0.87939453125
                },
                {
                    "start": 1596.11,
                    "end": 1596.41,
                    "word": " was",
                    "probability": 0.89990234375
                },
                {
                    "start": 1596.41,
                    "end": 1596.73,
                    "word": " measured",
                    "probability": 0.7587890625
                },
                {
                    "start": 1596.73,
                    "end": 1597.29,
                    "word": ".",
                    "probability": 0.998046875
                },
                {
                    "start": 1597.37,
                    "end": 1597.63,
                    "word": " Now",
                    "probability": 0.87841796875
                },
                {
                    "start": 1597.63,
                    "end": 1597.75,
                    "word": ",",
                    "probability": 0.85595703125
                },
                {
                    "start": 1597.75,
                    "end": 1597.87,
                    "word": " is",
                    "probability": 0.89990234375
                },
                {
                    "start": 1597.87,
                    "end": 1597.99,
                    "word": " this",
                    "probability": 0.87841796875
                },
                {
                    "start": 1597.99,
                    "end": 1598.15,
                    "word": " type",
                    "probability": 0.9228515625
                },
                {
                    "start": 1598.15,
                    "end": 1598.23,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 1598.23,
                    "end": 1598.43,
                    "word": " attack",
                    "probability": 0.87353515625
                },
                {
                    "start": 1598.43,
                    "end": 1598.95,
                    "word": " practical",
                    "probability": 0.8193359375
                },
                {
                    "start": 1598.95,
                    "end": 1599.13,
                    "word": "?",
                    "probability": 0.96142578125
                },
                {
                    "start": 1599.13,
                    "end": 1599.27,
                    "word": " Is",
                    "probability": 0.8486328125
                },
                {
                    "start": 1599.27,
                    "end": 1599.41,
                    "word": " this",
                    "probability": 0.865234375
                },
                {
                    "start": 1599.41,
                    "end": 1599.61,
                    "word": " really",
                    "probability": 0.765625
                },
                {
                    "start": 1599.61,
                    "end": 1599.89,
                    "word": " something",
                    "probability": 0.74169921875
                },
                {
                    "start": 1599.89,
                    "end": 1600.25,
                    "word": " that",
                    "probability": 0.87353515625
                },
                {
                    "start": 1600.25,
                    "end": 1600.79,
                    "word": " we",
                    "probability": 0.86962890625
                },
                {
                    "start": 1600.79,
                    "end": 1600.93,
                    "word": " should",
                    "probability": 0.9228515625
                }
            ]
        },
        {
            "id": 240,
            "seek": 162002,
            "start": 1600.93,
            "end": 1611.49,
            "text": " be worried about? So, a few years later, in 2018, Clark and his co-authors published the report of",
            "tokens": [
                50364,
                312,
                5804,
                466,
                30,
                407,
                11,
                257,
                1326,
                924,
                1780,
                11,
                294,
                6096,
                11,
                18572,
                293,
                702,
                598,
                12,
                40198,
                830,
                6572,
                264,
                2275,
                295,
                50904
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2356770804358853,
            "compression_ratio": 1.4393939393939394,
            "no_speech_prob": 0.0018100738525390625,
            "words": [
                {
                    "start": 1600.93,
                    "end": 1601.05,
                    "word": " be",
                    "probability": 0.35791015625
                },
                {
                    "start": 1601.05,
                    "end": 1601.35,
                    "word": " worried",
                    "probability": 0.74267578125
                },
                {
                    "start": 1601.35,
                    "end": 1601.69,
                    "word": " about",
                    "probability": 0.81201171875
                },
                {
                    "start": 1601.69,
                    "end": 1602.19,
                    "word": "?",
                    "probability": 0.041015625
                },
                {
                    "start": 1602.19,
                    "end": 1602.45,
                    "word": " So",
                    "probability": 0.69921875
                },
                {
                    "start": 1602.45,
                    "end": 1603.07,
                    "word": ",",
                    "probability": 0.367431640625
                },
                {
                    "start": 1603.17,
                    "end": 1603.29,
                    "word": " a few",
                    "probability": 0.80615234375
                },
                {
                    "start": 1603.29,
                    "end": 1603.49,
                    "word": " years",
                    "probability": 0.873046875
                },
                {
                    "start": 1603.49,
                    "end": 1603.77,
                    "word": " later",
                    "probability": 0.8740234375
                },
                {
                    "start": 1603.77,
                    "end": 1603.85,
                    "word": ",",
                    "probability": 0.410400390625
                },
                {
                    "start": 1603.85,
                    "end": 1603.91,
                    "word": " in",
                    "probability": 0.89794921875
                },
                {
                    "start": 1603.91,
                    "end": 1604.71,
                    "word": " 2018",
                    "probability": 0.7333984375
                },
                {
                    "start": 1604.71,
                    "end": 1605.37,
                    "word": ",",
                    "probability": 0.9365234375
                },
                {
                    "start": 1605.37,
                    "end": 1605.71,
                    "word": " Clark",
                    "probability": 0.6337890625
                },
                {
                    "start": 1605.71,
                    "end": 1606.37,
                    "word": " and",
                    "probability": 0.8203125
                },
                {
                    "start": 1606.37,
                    "end": 1606.63,
                    "word": " his",
                    "probability": 0.892578125
                },
                {
                    "start": 1606.63,
                    "end": 1606.87,
                    "word": " co",
                    "probability": 0.91748046875
                },
                {
                    "start": 1606.87,
                    "end": 1607.27,
                    "word": "-authors",
                    "probability": 0.7958984375
                },
                {
                    "start": 1607.27,
                    "end": 1610.23,
                    "word": " published",
                    "probability": 0.6748046875
                },
                {
                    "start": 1610.23,
                    "end": 1610.61,
                    "word": " the",
                    "probability": 0.77734375
                },
                {
                    "start": 1610.61,
                    "end": 1610.93,
                    "word": " report",
                    "probability": 0.83740234375
                },
                {
                    "start": 1610.93,
                    "end": 1611.49,
                    "word": " of",
                    "probability": 0.90185546875
                }
            ]
        },
        {
            "id": 241,
            "seek": 162002,
            "start": 1611.99,
            "end": 1619.77,
            "text": " a similar attack on an autonomous robot based on DQN policy using ultrasonic collision,",
            "tokens": [
                50904,
                257,
                2531,
                2690,
                322,
                364,
                23797,
                7881,
                2361,
                322,
                413,
                48,
                45,
                3897,
                1228,
                3725,
                81,
                39460,
                24644,
                11,
                51264
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2356770804358853,
            "compression_ratio": 1.4393939393939394,
            "no_speech_prob": 0.0018100738525390625,
            "words": [
                {
                    "start": 1611.99,
                    "end": 1612.31,
                    "word": " a similar",
                    "probability": 0.91796875
                },
                {
                    "start": 1612.31,
                    "end": 1612.85,
                    "word": " attack",
                    "probability": 0.82861328125
                },
                {
                    "start": 1612.85,
                    "end": 1613.53,
                    "word": " on",
                    "probability": 0.89404296875
                },
                {
                    "start": 1613.53,
                    "end": 1614.13,
                    "word": " an",
                    "probability": 0.8994140625
                },
                {
                    "start": 1614.13,
                    "end": 1614.55,
                    "word": " autonomous",
                    "probability": 0.845703125
                },
                {
                    "start": 1614.55,
                    "end": 1615.27,
                    "word": " robot",
                    "probability": 0.77490234375
                },
                {
                    "start": 1615.27,
                    "end": 1616.17,
                    "word": " based",
                    "probability": 0.7021484375
                },
                {
                    "start": 1616.17,
                    "end": 1616.39,
                    "word": " on",
                    "probability": 0.90576171875
                },
                {
                    "start": 1616.39,
                    "end": 1616.99,
                    "word": " DQN",
                    "probability": 0.8053385416666666
                },
                {
                    "start": 1616.99,
                    "end": 1617.49,
                    "word": " policy",
                    "probability": 0.75
                },
                {
                    "start": 1617.49,
                    "end": 1617.97,
                    "word": " using",
                    "probability": 0.60595703125
                },
                {
                    "start": 1617.97,
                    "end": 1618.49,
                    "word": " ultrasonic",
                    "probability": 0.8818359375
                },
                {
                    "start": 1618.49,
                    "end": 1618.93,
                    "word": " collision",
                    "probability": 0.63232421875
                },
                {
                    "start": 1618.93,
                    "end": 1619.77,
                    "word": ",",
                    "probability": 0.486572265625
                }
            ]
        },
        {
            "id": 242,
            "seek": 162002,
            "start": 1619.77,
            "end": 1626.15,
            "text": " ultrasonic sensory input for collision avoidance. And they had shown that they can use adversarial",
            "tokens": [
                51324,
                3725,
                81,
                39460,
                27233,
                4846,
                337,
                24644,
                5042,
                719,
                13,
                400,
                436,
                632,
                4898,
                300,
                436,
                393,
                764,
                17641,
                44745,
                51624
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2356770804358853,
            "compression_ratio": 1.4393939393939394,
            "no_speech_prob": 0.0018100738525390625,
            "words": [
                {
                    "start": 1619.77,
                    "end": 1620.75,
                    "word": " ultrasonic",
                    "probability": 0.8629557291666666
                },
                {
                    "start": 1620.75,
                    "end": 1621.55,
                    "word": " sensory",
                    "probability": 0.71337890625
                },
                {
                    "start": 1621.55,
                    "end": 1621.99,
                    "word": " input",
                    "probability": 0.65087890625
                },
                {
                    "start": 1621.99,
                    "end": 1622.23,
                    "word": " for",
                    "probability": 0.88134765625
                },
                {
                    "start": 1622.23,
                    "end": 1622.55,
                    "word": " collision",
                    "probability": 0.87646484375
                },
                {
                    "start": 1622.55,
                    "end": 1623.05,
                    "word": " avoidance",
                    "probability": 0.861083984375
                },
                {
                    "start": 1623.05,
                    "end": 1623.51,
                    "word": ".",
                    "probability": 0.499267578125
                },
                {
                    "start": 1623.51,
                    "end": 1623.67,
                    "word": " And",
                    "probability": 0.8271484375
                },
                {
                    "start": 1623.67,
                    "end": 1623.75,
                    "word": " they",
                    "probability": 0.415283203125
                },
                {
                    "start": 1623.75,
                    "end": 1623.83,
                    "word": " had",
                    "probability": 0.476806640625
                },
                {
                    "start": 1623.83,
                    "end": 1624.01,
                    "word": " shown",
                    "probability": 0.89111328125
                },
                {
                    "start": 1624.01,
                    "end": 1624.17,
                    "word": " that",
                    "probability": 0.876953125
                },
                {
                    "start": 1624.17,
                    "end": 1624.33,
                    "word": " they",
                    "probability": 0.7392578125
                },
                {
                    "start": 1624.33,
                    "end": 1624.63,
                    "word": " can",
                    "probability": 0.8935546875
                },
                {
                    "start": 1624.63,
                    "end": 1625.53,
                    "word": " use",
                    "probability": 0.8056640625
                },
                {
                    "start": 1625.53,
                    "end": 1626.15,
                    "word": " adversarial",
                    "probability": 0.93896484375
                }
            ]
        },
        {
            "id": 243,
            "seek": 164292,
            "start": 1626.15,
            "end": 1636.67,
            "text": " perturbations to manipulate the trajectory of the robot and make it follow a path defined or desired",
            "tokens": [
                50364,
                40468,
                763,
                281,
                20459,
                264,
                21512,
                295,
                264,
                7881,
                293,
                652,
                309,
                1524,
                257,
                3100,
                7642,
                420,
                14721,
                50888
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21148768319210537,
            "compression_ratio": 1.5621621621621622,
            "no_speech_prob": 0.00154876708984375,
            "words": [
                {
                    "start": 1626.15,
                    "end": 1626.85,
                    "word": " perturbations",
                    "probability": 0.737060546875
                },
                {
                    "start": 1626.85,
                    "end": 1627.09,
                    "word": " to",
                    "probability": 0.87744140625
                },
                {
                    "start": 1627.09,
                    "end": 1627.65,
                    "word": " manipulate",
                    "probability": 0.90185546875
                },
                {
                    "start": 1627.65,
                    "end": 1627.89,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 1627.89,
                    "end": 1628.33,
                    "word": " trajectory",
                    "probability": 0.759765625
                },
                {
                    "start": 1628.33,
                    "end": 1628.57,
                    "word": " of",
                    "probability": 0.89111328125
                },
                {
                    "start": 1628.57,
                    "end": 1628.71,
                    "word": " the",
                    "probability": 0.8056640625
                },
                {
                    "start": 1628.71,
                    "end": 1629.09,
                    "word": " robot",
                    "probability": 0.724609375
                },
                {
                    "start": 1629.09,
                    "end": 1629.81,
                    "word": " and",
                    "probability": 0.71630859375
                },
                {
                    "start": 1629.81,
                    "end": 1631.43,
                    "word": " make",
                    "probability": 0.8525390625
                },
                {
                    "start": 1631.43,
                    "end": 1631.69,
                    "word": " it",
                    "probability": 0.89306640625
                },
                {
                    "start": 1631.69,
                    "end": 1633.81,
                    "word": " follow",
                    "probability": 0.595703125
                },
                {
                    "start": 1634.71,
                    "end": 1635.23,
                    "word": " a path",
                    "probability": 0.87353515625
                },
                {
                    "start": 1635.23,
                    "end": 1635.99,
                    "word": " defined",
                    "probability": 0.67626953125
                },
                {
                    "start": 1635.99,
                    "end": 1636.31,
                    "word": " or",
                    "probability": 0.8623046875
                },
                {
                    "start": 1636.31,
                    "end": 1636.67,
                    "word": " desired",
                    "probability": 0.8486328125
                }
            ]
        },
        {
            "id": 244,
            "seek": 164292,
            "start": 1636.67,
            "end": 1643.03,
            "text": " by the adversary, not one that the robot itself wants to follow. There are more recent examples",
            "tokens": [
                50888,
                538,
                264,
                48222,
                11,
                406,
                472,
                300,
                264,
                7881,
                2564,
                2738,
                281,
                1524,
                13,
                821,
                366,
                544,
                5162,
                5110,
                51212
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21148768319210537,
            "compression_ratio": 1.5621621621621622,
            "no_speech_prob": 0.00154876708984375,
            "words": [
                {
                    "start": 1636.67,
                    "end": 1636.85,
                    "word": " by",
                    "probability": 0.9111328125
                },
                {
                    "start": 1636.85,
                    "end": 1636.97,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 1636.97,
                    "end": 1637.39,
                    "word": " adversary",
                    "probability": 0.796875
                },
                {
                    "start": 1637.39,
                    "end": 1637.61,
                    "word": ",",
                    "probability": 0.65185546875
                },
                {
                    "start": 1637.61,
                    "end": 1637.95,
                    "word": " not",
                    "probability": 0.884765625
                },
                {
                    "start": 1637.95,
                    "end": 1638.47,
                    "word": " one",
                    "probability": 0.81591796875
                },
                {
                    "start": 1638.47,
                    "end": 1638.73,
                    "word": " that",
                    "probability": 0.8642578125
                },
                {
                    "start": 1638.73,
                    "end": 1639.11,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 1639.11,
                    "end": 1639.43,
                    "word": " robot",
                    "probability": 0.7421875
                },
                {
                    "start": 1639.43,
                    "end": 1639.75,
                    "word": " itself",
                    "probability": 0.763671875
                },
                {
                    "start": 1639.75,
                    "end": 1639.97,
                    "word": " wants",
                    "probability": 0.72216796875
                },
                {
                    "start": 1639.97,
                    "end": 1640.09,
                    "word": " to",
                    "probability": 0.90087890625
                },
                {
                    "start": 1640.09,
                    "end": 1640.33,
                    "word": " follow",
                    "probability": 0.86181640625
                },
                {
                    "start": 1640.33,
                    "end": 1640.93,
                    "word": ".",
                    "probability": 0.966796875
                },
                {
                    "start": 1641.27,
                    "end": 1641.59,
                    "word": " There",
                    "probability": 0.73388671875
                },
                {
                    "start": 1641.59,
                    "end": 1641.99,
                    "word": " are",
                    "probability": 0.88720703125
                },
                {
                    "start": 1641.99,
                    "end": 1642.31,
                    "word": " more",
                    "probability": 0.8896484375
                },
                {
                    "start": 1642.31,
                    "end": 1642.59,
                    "word": " recent",
                    "probability": 0.79443359375
                },
                {
                    "start": 1642.59,
                    "end": 1643.03,
                    "word": " examples",
                    "probability": 0.7724609375
                }
            ]
        },
        {
            "id": 245,
            "seek": 164292,
            "start": 1643.03,
            "end": 1649.05,
            "text": " of how this sort of attack on deep RL, or in general, attacks on deep RL, can be of concern.",
            "tokens": [
                51212,
                295,
                577,
                341,
                1333,
                295,
                2690,
                322,
                2452,
                497,
                43,
                11,
                420,
                294,
                2674,
                11,
                8122,
                322,
                2452,
                497,
                43,
                11,
                393,
                312,
                295,
                3136,
                13,
                51476
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21148768319210537,
            "compression_ratio": 1.5621621621621622,
            "no_speech_prob": 0.00154876708984375,
            "words": [
                {
                    "start": 1643.03,
                    "end": 1643.41,
                    "word": " of",
                    "probability": 0.89599609375
                },
                {
                    "start": 1643.41,
                    "end": 1643.79,
                    "word": " how",
                    "probability": 0.8857421875
                },
                {
                    "start": 1643.79,
                    "end": 1644.39,
                    "word": " this",
                    "probability": 0.857421875
                },
                {
                    "start": 1644.39,
                    "end": 1644.57,
                    "word": " sort",
                    "probability": 0.814453125
                },
                {
                    "start": 1644.57,
                    "end": 1644.67,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 1644.67,
                    "end": 1644.93,
                    "word": " attack",
                    "probability": 0.83251953125
                },
                {
                    "start": 1644.93,
                    "end": 1645.11,
                    "word": " on",
                    "probability": 0.87939453125
                },
                {
                    "start": 1645.11,
                    "end": 1645.29,
                    "word": " deep",
                    "probability": 0.49560546875
                },
                {
                    "start": 1645.29,
                    "end": 1645.49,
                    "word": " RL",
                    "probability": 0.86083984375
                },
                {
                    "start": 1645.49,
                    "end": 1645.55,
                    "word": ",",
                    "probability": 0.385009765625
                },
                {
                    "start": 1645.55,
                    "end": 1645.61,
                    "word": " or",
                    "probability": 0.896484375
                },
                {
                    "start": 1645.61,
                    "end": 1645.73,
                    "word": " in",
                    "probability": 0.8505859375
                },
                {
                    "start": 1645.73,
                    "end": 1646.03,
                    "word": " general",
                    "probability": 0.80712890625
                },
                {
                    "start": 1646.03,
                    "end": 1646.33,
                    "word": ",",
                    "probability": 0.26171875
                },
                {
                    "start": 1646.33,
                    "end": 1646.55,
                    "word": " attacks",
                    "probability": 0.70751953125
                },
                {
                    "start": 1646.55,
                    "end": 1646.71,
                    "word": " on",
                    "probability": 0.90087890625
                },
                {
                    "start": 1646.71,
                    "end": 1646.87,
                    "word": " deep",
                    "probability": 0.85888671875
                },
                {
                    "start": 1646.87,
                    "end": 1647.15,
                    "word": " RL",
                    "probability": 0.97314453125
                },
                {
                    "start": 1647.15,
                    "end": 1647.49,
                    "word": ",",
                    "probability": 0.73046875
                },
                {
                    "start": 1647.49,
                    "end": 1647.69,
                    "word": " can",
                    "probability": 0.89794921875
                },
                {
                    "start": 1647.69,
                    "end": 1647.85,
                    "word": " be",
                    "probability": 0.90087890625
                },
                {
                    "start": 1647.85,
                    "end": 1647.99,
                    "word": " of",
                    "probability": 0.890625
                },
                {
                    "start": 1647.99,
                    "end": 1648.45,
                    "word": " concern",
                    "probability": 0.89892578125
                },
                {
                    "start": 1648.45,
                    "end": 1649.05,
                    "word": ".",
                    "probability": 0.93603515625
                }
            ]
        },
        {
            "id": 246,
            "seek": 167282,
            "start": 1649.53,
            "end": 1658.13,
            "text": " One of the recent works by my graduate students is on attacks on automated trading algorithms",
            "tokens": [
                50396,
                1485,
                295,
                264,
                5162,
                1985,
                538,
                452,
                8080,
                1731,
                307,
                322,
                8122,
                322,
                18473,
                9529,
                14642,
                50816
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19035456974345905,
            "compression_ratio": 1.6254416961130742,
            "no_speech_prob": 0.0001609325408935547,
            "words": [
                {
                    "start": 1649.53,
                    "end": 1650.01,
                    "word": " One",
                    "probability": 0.685546875
                },
                {
                    "start": 1650.01,
                    "end": 1650.11,
                    "word": " of",
                    "probability": 0.88525390625
                },
                {
                    "start": 1650.11,
                    "end": 1650.17,
                    "word": " the",
                    "probability": 0.8017578125
                },
                {
                    "start": 1650.17,
                    "end": 1650.41,
                    "word": " recent",
                    "probability": 0.7705078125
                },
                {
                    "start": 1650.41,
                    "end": 1650.71,
                    "word": " works",
                    "probability": 0.72265625
                },
                {
                    "start": 1650.71,
                    "end": 1651.09,
                    "word": " by",
                    "probability": 0.8896484375
                },
                {
                    "start": 1651.09,
                    "end": 1652.17,
                    "word": " my",
                    "probability": 0.90625
                },
                {
                    "start": 1652.17,
                    "end": 1652.77,
                    "word": " graduate",
                    "probability": 0.6357421875
                },
                {
                    "start": 1652.77,
                    "end": 1653.43,
                    "word": " students",
                    "probability": 0.87451171875
                },
                {
                    "start": 1653.43,
                    "end": 1654.37,
                    "word": " is",
                    "probability": 0.7666015625
                },
                {
                    "start": 1654.37,
                    "end": 1654.77,
                    "word": " on",
                    "probability": 0.84765625
                },
                {
                    "start": 1654.77,
                    "end": 1655.23,
                    "word": " attacks",
                    "probability": 0.51708984375
                },
                {
                    "start": 1655.23,
                    "end": 1656.23,
                    "word": " on",
                    "probability": 0.90185546875
                },
                {
                    "start": 1656.23,
                    "end": 1657.35,
                    "word": " automated",
                    "probability": 0.88623046875
                },
                {
                    "start": 1657.35,
                    "end": 1657.71,
                    "word": " trading",
                    "probability": 0.75439453125
                },
                {
                    "start": 1657.71,
                    "end": 1658.13,
                    "word": " algorithms",
                    "probability": 0.86474609375
                }
            ]
        },
        {
            "id": 247,
            "seek": 167282,
            "start": 1658.13,
            "end": 1662.75,
            "text": " based on reinforcement learning. The paper is coming out in a couple months, so I can't go",
            "tokens": [
                50816,
                2361,
                322,
                29280,
                2539,
                13,
                440,
                3035,
                307,
                1348,
                484,
                294,
                257,
                1916,
                2493,
                11,
                370,
                286,
                393,
                380,
                352,
                51044
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19035456974345905,
            "compression_ratio": 1.6254416961130742,
            "no_speech_prob": 0.0001609325408935547,
            "words": [
                {
                    "start": 1658.13,
                    "end": 1658.41,
                    "word": " based",
                    "probability": 0.78466796875
                },
                {
                    "start": 1658.41,
                    "end": 1658.57,
                    "word": " on",
                    "probability": 0.9013671875
                },
                {
                    "start": 1658.57,
                    "end": 1658.81,
                    "word": " reinforcement",
                    "probability": 0.7822265625
                },
                {
                    "start": 1658.81,
                    "end": 1659.41,
                    "word": " learning",
                    "probability": 0.8291015625
                },
                {
                    "start": 1659.41,
                    "end": 1659.83,
                    "word": ".",
                    "probability": 0.8984375
                },
                {
                    "start": 1659.99,
                    "end": 1660.47,
                    "word": " The",
                    "probability": 0.7978515625
                },
                {
                    "start": 1660.47,
                    "end": 1660.69,
                    "word": " paper",
                    "probability": 0.79296875
                },
                {
                    "start": 1660.69,
                    "end": 1660.85,
                    "word": " is",
                    "probability": 0.90576171875
                },
                {
                    "start": 1660.85,
                    "end": 1661.05,
                    "word": " coming",
                    "probability": 0.79296875
                },
                {
                    "start": 1661.05,
                    "end": 1661.27,
                    "word": " out",
                    "probability": 0.818359375
                },
                {
                    "start": 1661.27,
                    "end": 1661.37,
                    "word": " in",
                    "probability": 0.9072265625
                },
                {
                    "start": 1661.47,
                    "end": 1661.61,
                    "word": " a couple",
                    "probability": 0.78125
                },
                {
                    "start": 1661.61,
                    "end": 1661.99,
                    "word": " months",
                    "probability": 0.26025390625
                },
                {
                    "start": 1661.99,
                    "end": 1662.09,
                    "word": ",",
                    "probability": 0.5615234375
                },
                {
                    "start": 1662.09,
                    "end": 1662.33,
                    "word": " so",
                    "probability": 0.888671875
                },
                {
                    "start": 1662.33,
                    "end": 1662.45,
                    "word": " I",
                    "probability": 0.99072265625
                },
                {
                    "start": 1662.45,
                    "end": 1662.61,
                    "word": " can't",
                    "probability": 0.94482421875
                },
                {
                    "start": 1662.61,
                    "end": 1662.75,
                    "word": " go",
                    "probability": 0.90673828125
                }
            ]
        },
        {
            "id": 248,
            "seek": 167282,
            "start": 1662.75,
            "end": 1668.63,
            "text": " into more details, but this is one of the more severe and urgent cases for security researchers",
            "tokens": [
                51044,
                666,
                544,
                4365,
                11,
                457,
                341,
                307,
                472,
                295,
                264,
                544,
                8922,
                293,
                19022,
                3331,
                337,
                3825,
                10309,
                51348
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19035456974345905,
            "compression_ratio": 1.6254416961130742,
            "no_speech_prob": 0.0001609325408935547,
            "words": [
                {
                    "start": 1662.75,
                    "end": 1662.91,
                    "word": " into",
                    "probability": 0.73583984375
                },
                {
                    "start": 1662.91,
                    "end": 1663.11,
                    "word": " more",
                    "probability": 0.87646484375
                },
                {
                    "start": 1663.11,
                    "end": 1663.45,
                    "word": " details",
                    "probability": 0.7060546875
                },
                {
                    "start": 1663.45,
                    "end": 1663.71,
                    "word": ",",
                    "probability": 0.62353515625
                },
                {
                    "start": 1663.71,
                    "end": 1663.87,
                    "word": " but",
                    "probability": 0.89404296875
                },
                {
                    "start": 1663.87,
                    "end": 1664.31,
                    "word": " this",
                    "probability": 0.875
                },
                {
                    "start": 1664.31,
                    "end": 1664.41,
                    "word": " is",
                    "probability": 0.90478515625
                },
                {
                    "start": 1664.41,
                    "end": 1664.53,
                    "word": " one",
                    "probability": 0.81787109375
                },
                {
                    "start": 1664.53,
                    "end": 1664.61,
                    "word": " of",
                    "probability": 0.89990234375
                },
                {
                    "start": 1664.61,
                    "end": 1664.69,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 1664.69,
                    "end": 1664.85,
                    "word": " more",
                    "probability": 0.88330078125
                },
                {
                    "start": 1664.85,
                    "end": 1665.27,
                    "word": " severe",
                    "probability": 0.81787109375
                },
                {
                    "start": 1665.27,
                    "end": 1666.13,
                    "word": " and",
                    "probability": 0.88134765625
                },
                {
                    "start": 1666.13,
                    "end": 1666.63,
                    "word": " urgent",
                    "probability": 0.8017578125
                },
                {
                    "start": 1666.63,
                    "end": 1667.39,
                    "word": " cases",
                    "probability": 0.82861328125
                },
                {
                    "start": 1667.39,
                    "end": 1667.77,
                    "word": " for",
                    "probability": 0.8984375
                },
                {
                    "start": 1667.77,
                    "end": 1668.23,
                    "word": " security",
                    "probability": 0.90771484375
                },
                {
                    "start": 1668.23,
                    "end": 1668.63,
                    "word": " researchers",
                    "probability": 0.810546875
                }
            ]
        },
        {
            "id": 249,
            "seek": 167282,
            "start": 1668.63,
            "end": 1674.91,
            "text": " to consider. Deep RL is already being used by many major financial players and stock traders,",
            "tokens": [
                51348,
                281,
                1949,
                13,
                14895,
                497,
                43,
                307,
                1217,
                885,
                1143,
                538,
                867,
                2563,
                4669,
                4150,
                293,
                4127,
                26014,
                11,
                51652
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19035456974345905,
            "compression_ratio": 1.6254416961130742,
            "no_speech_prob": 0.0001609325408935547,
            "words": [
                {
                    "start": 1668.63,
                    "end": 1669.01,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 1669.01,
                    "end": 1669.55,
                    "word": " consider",
                    "probability": 0.79150390625
                },
                {
                    "start": 1669.55,
                    "end": 1670.03,
                    "word": ".",
                    "probability": 0.97705078125
                },
                {
                    "start": 1670.67,
                    "end": 1670.81,
                    "word": " Deep",
                    "probability": 0.55615234375
                },
                {
                    "start": 1670.81,
                    "end": 1671.09,
                    "word": " RL",
                    "probability": 0.708740234375
                },
                {
                    "start": 1671.09,
                    "end": 1671.31,
                    "word": " is",
                    "probability": 0.89013671875
                },
                {
                    "start": 1671.31,
                    "end": 1671.51,
                    "word": " already",
                    "probability": 0.8271484375
                },
                {
                    "start": 1671.51,
                    "end": 1671.71,
                    "word": " being",
                    "probability": 0.88427734375
                },
                {
                    "start": 1671.71,
                    "end": 1671.95,
                    "word": " used",
                    "probability": 0.8017578125
                },
                {
                    "start": 1671.95,
                    "end": 1672.19,
                    "word": " by",
                    "probability": 0.90087890625
                },
                {
                    "start": 1672.19,
                    "end": 1672.99,
                    "word": " many",
                    "probability": 0.78662109375
                },
                {
                    "start": 1672.99,
                    "end": 1673.27,
                    "word": " major",
                    "probability": 0.767578125
                },
                {
                    "start": 1673.27,
                    "end": 1673.63,
                    "word": " financial",
                    "probability": 0.79052734375
                },
                {
                    "start": 1673.63,
                    "end": 1673.93,
                    "word": " players",
                    "probability": 0.86328125
                },
                {
                    "start": 1673.93,
                    "end": 1674.09,
                    "word": " and",
                    "probability": 0.8916015625
                },
                {
                    "start": 1674.09,
                    "end": 1674.33,
                    "word": " stock",
                    "probability": 0.71044921875
                },
                {
                    "start": 1674.33,
                    "end": 1674.63,
                    "word": " traders",
                    "probability": 0.66845703125
                },
                {
                    "start": 1674.63,
                    "end": 1674.91,
                    "word": ",",
                    "probability": 0.626953125
                }
            ]
        },
        {
            "id": 250,
            "seek": 167282,
            "start": 1674.91,
            "end": 1678.95,
            "text": " and it can be easily manipulated in the real world. There are other cases, of course,",
            "tokens": [
                51652,
                293,
                309,
                393,
                312,
                3612,
                37161,
                294,
                264,
                957,
                1002,
                13,
                821,
                366,
                661,
                3331,
                11,
                295,
                1164,
                11,
                51856
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19035456974345905,
            "compression_ratio": 1.6254416961130742,
            "no_speech_prob": 0.0001609325408935547,
            "words": [
                {
                    "start": 1674.91,
                    "end": 1675.11,
                    "word": " and",
                    "probability": 0.89453125
                },
                {
                    "start": 1675.11,
                    "end": 1675.21,
                    "word": " it",
                    "probability": 0.87060546875
                },
                {
                    "start": 1675.21,
                    "end": 1675.33,
                    "word": " can",
                    "probability": 0.89794921875
                },
                {
                    "start": 1675.33,
                    "end": 1675.49,
                    "word": " be",
                    "probability": 0.89501953125
                },
                {
                    "start": 1675.49,
                    "end": 1675.79,
                    "word": " easily",
                    "probability": 0.84033203125
                },
                {
                    "start": 1675.79,
                    "end": 1676.43,
                    "word": " manipulated",
                    "probability": 0.93017578125
                },
                {
                    "start": 1676.43,
                    "end": 1677.01,
                    "word": " in",
                    "probability": 0.9013671875
                },
                {
                    "start": 1677.01,
                    "end": 1677.11,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 1677.11,
                    "end": 1677.23,
                    "word": " real",
                    "probability": 0.87255859375
                },
                {
                    "start": 1677.23,
                    "end": 1677.41,
                    "word": " world",
                    "probability": 0.880859375
                },
                {
                    "start": 1677.41,
                    "end": 1677.47,
                    "word": ".",
                    "probability": 0.98681640625
                },
                {
                    "start": 1677.47,
                    "end": 1677.55,
                    "word": " There",
                    "probability": 0.80517578125
                },
                {
                    "start": 1677.55,
                    "end": 1677.67,
                    "word": " are",
                    "probability": 0.89404296875
                },
                {
                    "start": 1677.67,
                    "end": 1677.87,
                    "word": " other",
                    "probability": 0.77587890625
                },
                {
                    "start": 1677.87,
                    "end": 1678.19,
                    "word": " cases",
                    "probability": 0.8076171875
                },
                {
                    "start": 1678.19,
                    "end": 1678.29,
                    "word": ",",
                    "probability": 0.751953125
                },
                {
                    "start": 1678.29,
                    "end": 1678.39,
                    "word": " of",
                    "probability": 0.91162109375
                },
                {
                    "start": 1678.39,
                    "end": 1678.55,
                    "word": " course",
                    "probability": 0.86181640625
                },
                {
                    "start": 1678.55,
                    "end": 1678.95,
                    "word": ",",
                    "probability": 0.8056640625
                }
            ]
        },
        {
            "id": 251,
            "seek": 170038,
            "start": 1678.95,
            "end": 1685.75,
            "text": " but this is one of the examples that demonstrates the practicality and the applicability of this",
            "tokens": [
                50412,
                457,
                341,
                307,
                472,
                295,
                264,
                5110,
                300,
                31034,
                264,
                8496,
                507,
                293,
                264,
                2580,
                2310,
                295,
                341,
                50700
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23360906263776854,
            "compression_ratio": 1.6366782006920415,
            "no_speech_prob": 2.7120113372802734e-05,
            "words": [
                {
                    "start": 1678.95,
                    "end": 1679.25,
                    "word": " but",
                    "probability": 0.24755859375
                },
                {
                    "start": 1679.25,
                    "end": 1680.31,
                    "word": " this",
                    "probability": 0.8447265625
                },
                {
                    "start": 1680.31,
                    "end": 1680.55,
                    "word": " is",
                    "probability": 0.90087890625
                },
                {
                    "start": 1680.55,
                    "end": 1680.71,
                    "word": " one",
                    "probability": 0.80615234375
                },
                {
                    "start": 1680.71,
                    "end": 1680.79,
                    "word": " of",
                    "probability": 0.88916015625
                },
                {
                    "start": 1680.79,
                    "end": 1680.89,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 1680.89,
                    "end": 1681.31,
                    "word": " examples",
                    "probability": 0.755859375
                },
                {
                    "start": 1681.31,
                    "end": 1681.61,
                    "word": " that",
                    "probability": 0.87841796875
                },
                {
                    "start": 1681.61,
                    "end": 1682.09,
                    "word": " demonstrates",
                    "probability": 0.6396484375
                },
                {
                    "start": 1682.09,
                    "end": 1682.59,
                    "word": " the",
                    "probability": 0.80517578125
                },
                {
                    "start": 1682.59,
                    "end": 1683.67,
                    "word": " practicality",
                    "probability": 0.844970703125
                },
                {
                    "start": 1683.67,
                    "end": 1684.39,
                    "word": " and",
                    "probability": 0.85107421875
                },
                {
                    "start": 1684.39,
                    "end": 1684.85,
                    "word": " the",
                    "probability": 0.79541015625
                },
                {
                    "start": 1684.85,
                    "end": 1685.43,
                    "word": " applicability",
                    "probability": 0.94189453125
                },
                {
                    "start": 1685.43,
                    "end": 1685.61,
                    "word": " of",
                    "probability": 0.8974609375
                },
                {
                    "start": 1685.61,
                    "end": 1685.75,
                    "word": " this",
                    "probability": 0.86767578125
                }
            ]
        },
        {
            "id": 252,
            "seek": 170038,
            "start": 1685.75,
            "end": 1690.37,
            "text": " attack to real world scenarios. Now, before we go further into different types of attacks,",
            "tokens": [
                50700,
                2690,
                281,
                957,
                1002,
                15077,
                13,
                823,
                11,
                949,
                321,
                352,
                3052,
                666,
                819,
                3467,
                295,
                8122,
                11,
                50928
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23360906263776854,
            "compression_ratio": 1.6366782006920415,
            "no_speech_prob": 2.7120113372802734e-05,
            "words": [
                {
                    "start": 1685.75,
                    "end": 1685.99,
                    "word": " attack",
                    "probability": 0.8310546875
                },
                {
                    "start": 1685.99,
                    "end": 1686.15,
                    "word": " to",
                    "probability": 0.8935546875
                },
                {
                    "start": 1686.15,
                    "end": 1686.33,
                    "word": " real",
                    "probability": 0.87548828125
                },
                {
                    "start": 1686.33,
                    "end": 1686.51,
                    "word": " world",
                    "probability": 0.35693359375
                },
                {
                    "start": 1686.51,
                    "end": 1687.07,
                    "word": " scenarios",
                    "probability": 0.8251953125
                },
                {
                    "start": 1687.07,
                    "end": 1687.59,
                    "word": ".",
                    "probability": 0.91015625
                },
                {
                    "start": 1687.77,
                    "end": 1688.03,
                    "word": " Now",
                    "probability": 0.7998046875
                },
                {
                    "start": 1688.03,
                    "end": 1688.13,
                    "word": ",",
                    "probability": 0.64794921875
                },
                {
                    "start": 1688.13,
                    "end": 1688.41,
                    "word": " before",
                    "probability": 0.79541015625
                },
                {
                    "start": 1688.41,
                    "end": 1688.63,
                    "word": " we",
                    "probability": 0.90625
                },
                {
                    "start": 1688.63,
                    "end": 1688.93,
                    "word": " go",
                    "probability": 0.89697265625
                },
                {
                    "start": 1688.93,
                    "end": 1689.27,
                    "word": " further",
                    "probability": 0.8642578125
                },
                {
                    "start": 1689.27,
                    "end": 1689.45,
                    "word": " into",
                    "probability": 0.744140625
                },
                {
                    "start": 1689.45,
                    "end": 1689.69,
                    "word": " different",
                    "probability": 0.68798828125
                },
                {
                    "start": 1689.69,
                    "end": 1689.93,
                    "word": " types",
                    "probability": 0.72802734375
                },
                {
                    "start": 1689.93,
                    "end": 1690.03,
                    "word": " of",
                    "probability": 0.9091796875
                },
                {
                    "start": 1690.03,
                    "end": 1690.29,
                    "word": " attacks",
                    "probability": 0.70068359375
                },
                {
                    "start": 1690.29,
                    "end": 1690.37,
                    "word": ",",
                    "probability": 0.888671875
                }
            ]
        },
        {
            "id": 253,
            "seek": 170038,
            "start": 1690.37,
            "end": 1695.49,
            "text": " let's develop a threat model for deep reinforcement learning. Again, the adversary's objectives can",
            "tokens": [
                50928,
                718,
                311,
                1499,
                257,
                4734,
                2316,
                337,
                2452,
                29280,
                2539,
                13,
                3764,
                11,
                264,
                48222,
                311,
                15961,
                393,
                51188
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23360906263776854,
            "compression_ratio": 1.6366782006920415,
            "no_speech_prob": 2.7120113372802734e-05,
            "words": [
                {
                    "start": 1690.37,
                    "end": 1690.55,
                    "word": " let's",
                    "probability": 0.9384765625
                },
                {
                    "start": 1690.55,
                    "end": 1690.93,
                    "word": " develop",
                    "probability": 0.66162109375
                },
                {
                    "start": 1691.25,
                    "end": 1691.55,
                    "word": " a threat",
                    "probability": 0.7890625
                },
                {
                    "start": 1691.55,
                    "end": 1691.87,
                    "word": " model",
                    "probability": 0.77490234375
                },
                {
                    "start": 1691.87,
                    "end": 1692.11,
                    "word": " for",
                    "probability": 0.89697265625
                },
                {
                    "start": 1692.11,
                    "end": 1692.35,
                    "word": " deep",
                    "probability": 0.7724609375
                },
                {
                    "start": 1692.35,
                    "end": 1692.77,
                    "word": " reinforcement",
                    "probability": 0.81005859375
                },
                {
                    "start": 1692.77,
                    "end": 1693.15,
                    "word": " learning",
                    "probability": 0.82373046875
                },
                {
                    "start": 1693.15,
                    "end": 1693.67,
                    "word": ".",
                    "probability": 0.97802734375
                },
                {
                    "start": 1693.67,
                    "end": 1694.07,
                    "word": " Again",
                    "probability": 0.86328125
                },
                {
                    "start": 1694.07,
                    "end": 1694.17,
                    "word": ",",
                    "probability": 0.92578125
                },
                {
                    "start": 1694.17,
                    "end": 1694.27,
                    "word": " the",
                    "probability": 0.80419921875
                },
                {
                    "start": 1694.27,
                    "end": 1694.75,
                    "word": " adversary's",
                    "probability": 0.8134765625
                },
                {
                    "start": 1694.75,
                    "end": 1695.07,
                    "word": " objectives",
                    "probability": 0.7744140625
                },
                {
                    "start": 1695.07,
                    "end": 1695.49,
                    "word": " can",
                    "probability": 0.89453125
                }
            ]
        },
        {
            "id": 254,
            "seek": 170038,
            "start": 1695.49,
            "end": 1700.89,
            "text": " follow those of the CIA triad. The adversary may wish to access the internal configurations,",
            "tokens": [
                51188,
                1524,
                729,
                295,
                264,
                25143,
                1376,
                345,
                13,
                440,
                48222,
                815,
                3172,
                281,
                2105,
                264,
                6920,
                31493,
                11,
                51452
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23360906263776854,
            "compression_ratio": 1.6366782006920415,
            "no_speech_prob": 2.7120113372802734e-05,
            "words": [
                {
                    "start": 1695.49,
                    "end": 1695.95,
                    "word": " follow",
                    "probability": 0.87939453125
                },
                {
                    "start": 1695.95,
                    "end": 1697.11,
                    "word": " those",
                    "probability": 0.8251953125
                },
                {
                    "start": 1697.11,
                    "end": 1697.21,
                    "word": " of",
                    "probability": 0.90478515625
                },
                {
                    "start": 1697.21,
                    "end": 1697.33,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 1697.33,
                    "end": 1697.61,
                    "word": " CIA",
                    "probability": 0.88427734375
                },
                {
                    "start": 1697.61,
                    "end": 1698.05,
                    "word": " triad",
                    "probability": 0.627197265625
                },
                {
                    "start": 1698.05,
                    "end": 1698.19,
                    "word": ".",
                    "probability": 0.69873046875
                },
                {
                    "start": 1698.19,
                    "end": 1698.43,
                    "word": " The",
                    "probability": 0.798828125
                },
                {
                    "start": 1698.43,
                    "end": 1698.83,
                    "word": " adversary",
                    "probability": 0.83984375
                },
                {
                    "start": 1698.83,
                    "end": 1699.03,
                    "word": " may",
                    "probability": 0.8720703125
                },
                {
                    "start": 1699.03,
                    "end": 1699.23,
                    "word": " wish",
                    "probability": 0.8017578125
                },
                {
                    "start": 1699.23,
                    "end": 1699.35,
                    "word": " to",
                    "probability": 0.90576171875
                },
                {
                    "start": 1699.35,
                    "end": 1699.75,
                    "word": " access",
                    "probability": 0.66943359375
                },
                {
                    "start": 1699.75,
                    "end": 1699.91,
                    "word": " the",
                    "probability": 0.4130859375
                },
                {
                    "start": 1699.91,
                    "end": 1700.11,
                    "word": " internal",
                    "probability": 0.70166015625
                },
                {
                    "start": 1700.11,
                    "end": 1700.73,
                    "word": " configurations",
                    "probability": 0.7841796875
                },
                {
                    "start": 1700.73,
                    "end": 1700.89,
                    "word": ",",
                    "probability": 0.2724609375
                }
            ]
        },
        {
            "id": 255,
            "seek": 170038,
            "start": 1700.89,
            "end": 1706.51,
            "text": " like model parameter, reward function, policy, and such, to still the model for intellectual",
            "tokens": [
                51452,
                411,
                2316,
                13075,
                11,
                7782,
                2445,
                11,
                3897,
                11,
                293,
                1270,
                11,
                281,
                920,
                264,
                2316,
                337,
                12576,
                51760
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23360906263776854,
            "compression_ratio": 1.6366782006920415,
            "no_speech_prob": 2.7120113372802734e-05,
            "words": [
                {
                    "start": 1700.89,
                    "end": 1700.99,
                    "word": " like",
                    "probability": 0.8671875
                },
                {
                    "start": 1700.99,
                    "end": 1701.21,
                    "word": " model",
                    "probability": 0.77294921875
                },
                {
                    "start": 1701.21,
                    "end": 1701.55,
                    "word": " parameter",
                    "probability": 0.640625
                },
                {
                    "start": 1701.55,
                    "end": 1701.69,
                    "word": ",",
                    "probability": 0.423095703125
                },
                {
                    "start": 1701.69,
                    "end": 1701.89,
                    "word": " reward",
                    "probability": 0.8876953125
                },
                {
                    "start": 1701.89,
                    "end": 1702.23,
                    "word": " function",
                    "probability": 0.87255859375
                },
                {
                    "start": 1702.23,
                    "end": 1702.41,
                    "word": ",",
                    "probability": 0.91748046875
                },
                {
                    "start": 1702.41,
                    "end": 1702.71,
                    "word": " policy",
                    "probability": 0.80224609375
                },
                {
                    "start": 1702.71,
                    "end": 1702.85,
                    "word": ",",
                    "probability": 0.466552734375
                },
                {
                    "start": 1702.85,
                    "end": 1702.89,
                    "word": " and",
                    "probability": 0.86474609375
                },
                {
                    "start": 1702.89,
                    "end": 1703.11,
                    "word": " such",
                    "probability": 0.8974609375
                },
                {
                    "start": 1703.11,
                    "end": 1703.71,
                    "word": ",",
                    "probability": 0.599609375
                },
                {
                    "start": 1703.71,
                    "end": 1703.97,
                    "word": " to",
                    "probability": 0.8828125
                },
                {
                    "start": 1703.97,
                    "end": 1704.55,
                    "word": " still",
                    "probability": 0.58935546875
                },
                {
                    "start": 1704.55,
                    "end": 1705.29,
                    "word": " the",
                    "probability": 0.79931640625
                },
                {
                    "start": 1705.29,
                    "end": 1705.53,
                    "word": " model",
                    "probability": 0.8115234375
                },
                {
                    "start": 1705.53,
                    "end": 1705.91,
                    "word": " for",
                    "probability": 0.8984375
                },
                {
                    "start": 1705.91,
                    "end": 1706.51,
                    "word": " intellectual",
                    "probability": 0.8876953125
                }
            ]
        },
        {
            "id": 256,
            "seek": 172480,
            "start": 1706.51,
            "end": 1712.27,
            "text": " property theft and such. There can be attack on integrity, which means compromising the desired",
            "tokens": [
                50380,
                4707,
                28508,
                293,
                1270,
                13,
                821,
                393,
                312,
                2690,
                322,
                16000,
                11,
                597,
                1355,
                11482,
                3436,
                264,
                14721,
                50652
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2512571969251523,
            "compression_ratio": 1.7674418604651163,
            "no_speech_prob": 0.043365478515625,
            "words": [
                {
                    "start": 1706.51,
                    "end": 1707.21,
                    "word": " property",
                    "probability": 0.1451416015625
                },
                {
                    "start": 1707.21,
                    "end": 1707.55,
                    "word": " theft",
                    "probability": 0.72021484375
                },
                {
                    "start": 1707.55,
                    "end": 1707.69,
                    "word": " and",
                    "probability": 0.724609375
                },
                {
                    "start": 1707.69,
                    "end": 1707.93,
                    "word": " such",
                    "probability": 0.87548828125
                },
                {
                    "start": 1707.93,
                    "end": 1708.45,
                    "word": ".",
                    "probability": 0.76220703125
                },
                {
                    "start": 1708.45,
                    "end": 1708.51,
                    "word": " There",
                    "probability": 0.29296875
                },
                {
                    "start": 1708.51,
                    "end": 1708.63,
                    "word": " can",
                    "probability": 0.892578125
                },
                {
                    "start": 1708.63,
                    "end": 1708.73,
                    "word": " be",
                    "probability": 0.8955078125
                },
                {
                    "start": 1708.73,
                    "end": 1709.03,
                    "word": " attack",
                    "probability": 0.2109375
                },
                {
                    "start": 1709.03,
                    "end": 1709.23,
                    "word": " on",
                    "probability": 0.89892578125
                },
                {
                    "start": 1709.23,
                    "end": 1709.75,
                    "word": " integrity",
                    "probability": 0.51220703125
                },
                {
                    "start": 1709.75,
                    "end": 1710.15,
                    "word": ",",
                    "probability": 0.498779296875
                },
                {
                    "start": 1710.15,
                    "end": 1710.83,
                    "word": " which",
                    "probability": 0.787109375
                },
                {
                    "start": 1710.83,
                    "end": 1711.23,
                    "word": " means",
                    "probability": 0.869140625
                },
                {
                    "start": 1711.23,
                    "end": 1711.79,
                    "word": " compromising",
                    "probability": 0.82763671875
                },
                {
                    "start": 1711.79,
                    "end": 1711.93,
                    "word": " the",
                    "probability": 0.77001953125
                },
                {
                    "start": 1711.93,
                    "end": 1712.27,
                    "word": " desired",
                    "probability": 0.79296875
                }
            ]
        },
        {
            "id": 257,
            "seek": 172480,
            "start": 1712.27,
            "end": 1718.51,
            "text": " learning or enactment of the policy. There can be attacks on availability, which are essentially",
            "tokens": [
                50652,
                2539,
                420,
                25909,
                518,
                295,
                264,
                3897,
                13,
                821,
                393,
                312,
                8122,
                322,
                17945,
                11,
                597,
                366,
                4476,
                50960
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2512571969251523,
            "compression_ratio": 1.7674418604651163,
            "no_speech_prob": 0.043365478515625,
            "words": [
                {
                    "start": 1712.27,
                    "end": 1712.65,
                    "word": " learning",
                    "probability": 0.853515625
                },
                {
                    "start": 1712.65,
                    "end": 1713.19,
                    "word": " or",
                    "probability": 0.85791015625
                },
                {
                    "start": 1713.19,
                    "end": 1713.81,
                    "word": " enactment",
                    "probability": 0.729248046875
                },
                {
                    "start": 1713.81,
                    "end": 1714.23,
                    "word": " of",
                    "probability": 0.90771484375
                },
                {
                    "start": 1714.23,
                    "end": 1714.65,
                    "word": " the",
                    "probability": 0.80419921875
                },
                {
                    "start": 1714.65,
                    "end": 1715.05,
                    "word": " policy",
                    "probability": 0.80078125
                },
                {
                    "start": 1715.05,
                    "end": 1715.41,
                    "word": ".",
                    "probability": 0.9228515625
                },
                {
                    "start": 1715.41,
                    "end": 1715.87,
                    "word": " There",
                    "probability": 0.7451171875
                },
                {
                    "start": 1715.87,
                    "end": 1715.99,
                    "word": " can",
                    "probability": 0.8984375
                },
                {
                    "start": 1715.99,
                    "end": 1716.11,
                    "word": " be",
                    "probability": 0.90771484375
                },
                {
                    "start": 1716.11,
                    "end": 1716.41,
                    "word": " attacks",
                    "probability": 0.73046875
                },
                {
                    "start": 1716.41,
                    "end": 1716.65,
                    "word": " on",
                    "probability": 0.90283203125
                },
                {
                    "start": 1716.65,
                    "end": 1717.21,
                    "word": " availability",
                    "probability": 0.9052734375
                },
                {
                    "start": 1717.21,
                    "end": 1717.87,
                    "word": ",",
                    "probability": 0.75341796875
                },
                {
                    "start": 1717.87,
                    "end": 1718.11,
                    "word": " which",
                    "probability": 0.91064453125
                },
                {
                    "start": 1718.11,
                    "end": 1718.21,
                    "word": " are",
                    "probability": 0.888671875
                },
                {
                    "start": 1718.21,
                    "end": 1718.51,
                    "word": " essentially",
                    "probability": 0.59619140625
                }
            ]
        },
        {
            "id": 258,
            "seek": 172480,
            "start": 1718.51,
            "end": 1725.01,
            "text": " compromises of the ability of the agent to perform training or actions when needed. Now,",
            "tokens": [
                50984,
                11482,
                3598,
                295,
                264,
                3485,
                295,
                264,
                9461,
                281,
                2042,
                3097,
                420,
                5909,
                562,
                2978,
                13,
                823,
                11,
                51280
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2512571969251523,
            "compression_ratio": 1.7674418604651163,
            "no_speech_prob": 0.043365478515625,
            "words": [
                {
                    "start": 1718.51,
                    "end": 1719.53,
                    "word": " compromises",
                    "probability": 0.822021484375
                },
                {
                    "start": 1719.53,
                    "end": 1720.19,
                    "word": " of",
                    "probability": 0.87841796875
                },
                {
                    "start": 1720.19,
                    "end": 1720.51,
                    "word": " the",
                    "probability": 0.8173828125
                },
                {
                    "start": 1720.51,
                    "end": 1721.05,
                    "word": " ability",
                    "probability": 0.865234375
                },
                {
                    "start": 1721.05,
                    "end": 1721.25,
                    "word": " of",
                    "probability": 0.89697265625
                },
                {
                    "start": 1721.25,
                    "end": 1721.33,
                    "word": " the",
                    "probability": 0.8037109375
                },
                {
                    "start": 1721.33,
                    "end": 1721.51,
                    "word": " agent",
                    "probability": 0.7802734375
                },
                {
                    "start": 1721.51,
                    "end": 1721.67,
                    "word": " to",
                    "probability": 0.8916015625
                },
                {
                    "start": 1721.67,
                    "end": 1721.89,
                    "word": " perform",
                    "probability": 0.81396484375
                },
                {
                    "start": 1721.89,
                    "end": 1722.27,
                    "word": " training",
                    "probability": 0.81787109375
                },
                {
                    "start": 1722.27,
                    "end": 1722.67,
                    "word": " or",
                    "probability": 0.89111328125
                },
                {
                    "start": 1722.67,
                    "end": 1723.29,
                    "word": " actions",
                    "probability": 0.85205078125
                },
                {
                    "start": 1723.29,
                    "end": 1723.69,
                    "word": " when",
                    "probability": 0.85400390625
                },
                {
                    "start": 1723.69,
                    "end": 1724.01,
                    "word": " needed",
                    "probability": 0.73828125
                },
                {
                    "start": 1724.01,
                    "end": 1724.61,
                    "word": ".",
                    "probability": 0.98486328125
                },
                {
                    "start": 1724.71,
                    "end": 1724.95,
                    "word": " Now",
                    "probability": 0.86376953125
                },
                {
                    "start": 1724.95,
                    "end": 1725.01,
                    "word": ",",
                    "probability": 0.277587890625
                }
            ]
        },
        {
            "id": 259,
            "seek": 172480,
            "start": 1725.01,
            "end": 1730.93,
            "text": " let's look at the attack surface of the RL. This is the general diagram. This is the general block",
            "tokens": [
                51280,
                718,
                311,
                574,
                412,
                264,
                2690,
                3753,
                295,
                264,
                497,
                43,
                13,
                639,
                307,
                264,
                2674,
                10686,
                13,
                639,
                307,
                264,
                2674,
                3461,
                51580
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2512571969251523,
            "compression_ratio": 1.7674418604651163,
            "no_speech_prob": 0.043365478515625,
            "words": [
                {
                    "start": 1725.01,
                    "end": 1725.11,
                    "word": " let's",
                    "probability": 0.9462890625
                },
                {
                    "start": 1725.11,
                    "end": 1725.41,
                    "word": " look",
                    "probability": 0.833984375
                },
                {
                    "start": 1725.41,
                    "end": 1725.51,
                    "word": " at",
                    "probability": 0.90625
                },
                {
                    "start": 1725.51,
                    "end": 1725.61,
                    "word": " the",
                    "probability": 0.8193359375
                },
                {
                    "start": 1725.61,
                    "end": 1725.85,
                    "word": " attack",
                    "probability": 0.787109375
                },
                {
                    "start": 1725.85,
                    "end": 1726.49,
                    "word": " surface",
                    "probability": 0.7900390625
                },
                {
                    "start": 1726.49,
                    "end": 1726.75,
                    "word": " of",
                    "probability": 0.89111328125
                },
                {
                    "start": 1726.75,
                    "end": 1726.85,
                    "word": " the",
                    "probability": 0.271728515625
                },
                {
                    "start": 1726.85,
                    "end": 1727.15,
                    "word": " RL",
                    "probability": 0.492034912109375
                },
                {
                    "start": 1727.15,
                    "end": 1727.29,
                    "word": ".",
                    "probability": 0.96142578125
                },
                {
                    "start": 1727.29,
                    "end": 1727.63,
                    "word": " This",
                    "probability": 0.80517578125
                },
                {
                    "start": 1727.63,
                    "end": 1727.81,
                    "word": " is",
                    "probability": 0.908203125
                },
                {
                    "start": 1727.81,
                    "end": 1728.15,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 1728.15,
                    "end": 1728.91,
                    "word": " general",
                    "probability": 0.80078125
                },
                {
                    "start": 1728.91,
                    "end": 1730.01,
                    "word": " diagram",
                    "probability": 0.341796875
                },
                {
                    "start": 1730.01,
                    "end": 1730.19,
                    "word": ".",
                    "probability": 0.6845703125
                },
                {
                    "start": 1730.19,
                    "end": 1730.29,
                    "word": " This",
                    "probability": 0.80126953125
                },
                {
                    "start": 1730.29,
                    "end": 1730.37,
                    "word": " is",
                    "probability": 0.9033203125
                },
                {
                    "start": 1730.37,
                    "end": 1730.47,
                    "word": " the",
                    "probability": 0.80029296875
                },
                {
                    "start": 1730.47,
                    "end": 1730.65,
                    "word": " general",
                    "probability": 0.76171875
                },
                {
                    "start": 1730.65,
                    "end": 1730.93,
                    "word": " block",
                    "probability": 0.7890625
                }
            ]
        },
        {
            "id": 260,
            "seek": 175480,
            "start": 1730.93,
            "end": 1738.79,
            "text": " diagram of a DRL agent or DRL system. Now, we have the agent. The agent typically has some memory",
            "tokens": [
                50364,
                10686,
                295,
                257,
                413,
                10740,
                9461,
                420,
                413,
                10740,
                1185,
                13,
                823,
                11,
                321,
                362,
                264,
                9461,
                13,
                440,
                9461,
                5850,
                575,
                512,
                4675,
                50756
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2256249962747097,
            "compression_ratio": 1.7923076923076924,
            "no_speech_prob": 0.0966796875,
            "words": [
                {
                    "start": 1730.93,
                    "end": 1731.29,
                    "word": " diagram",
                    "probability": 0.74658203125
                },
                {
                    "start": 1731.29,
                    "end": 1731.79,
                    "word": " of",
                    "probability": 0.89892578125
                },
                {
                    "start": 1732.09,
                    "end": 1732.51,
                    "word": " a DRL",
                    "probability": 0.89013671875
                },
                {
                    "start": 1732.51,
                    "end": 1733.63,
                    "word": " agent",
                    "probability": 0.859375
                },
                {
                    "start": 1733.63,
                    "end": 1733.85,
                    "word": " or",
                    "probability": 0.849609375
                },
                {
                    "start": 1733.85,
                    "end": 1734.21,
                    "word": " DRL",
                    "probability": 0.8349609375
                },
                {
                    "start": 1734.21,
                    "end": 1734.69,
                    "word": " system",
                    "probability": 0.865234375
                },
                {
                    "start": 1734.69,
                    "end": 1735.29,
                    "word": ".",
                    "probability": 0.98486328125
                },
                {
                    "start": 1735.41,
                    "end": 1735.97,
                    "word": " Now",
                    "probability": 0.89697265625
                },
                {
                    "start": 1735.97,
                    "end": 1736.05,
                    "word": ",",
                    "probability": 0.260498046875
                },
                {
                    "start": 1736.05,
                    "end": 1736.13,
                    "word": " we",
                    "probability": 0.89404296875
                },
                {
                    "start": 1736.13,
                    "end": 1736.25,
                    "word": " have",
                    "probability": 0.8837890625
                },
                {
                    "start": 1736.25,
                    "end": 1736.43,
                    "word": " the",
                    "probability": 0.72412109375
                },
                {
                    "start": 1736.43,
                    "end": 1736.65,
                    "word": " agent",
                    "probability": 0.7802734375
                },
                {
                    "start": 1736.65,
                    "end": 1736.77,
                    "word": ".",
                    "probability": 0.52783203125
                },
                {
                    "start": 1736.77,
                    "end": 1736.87,
                    "word": " The",
                    "probability": 0.81591796875
                },
                {
                    "start": 1736.87,
                    "end": 1737.19,
                    "word": " agent",
                    "probability": 0.7880859375
                },
                {
                    "start": 1737.19,
                    "end": 1737.77,
                    "word": " typically",
                    "probability": 0.82958984375
                },
                {
                    "start": 1737.77,
                    "end": 1738.07,
                    "word": " has",
                    "probability": 0.89794921875
                },
                {
                    "start": 1738.07,
                    "end": 1738.39,
                    "word": " some",
                    "probability": 0.82666015625
                },
                {
                    "start": 1738.39,
                    "end": 1738.79,
                    "word": " memory",
                    "probability": 0.80029296875
                }
            ]
        },
        {
            "id": 261,
            "seek": 175480,
            "start": 1738.79,
            "end": 1743.45,
            "text": " where it stores its experiences during training, and then those experiences help with function",
            "tokens": [
                50756,
                689,
                309,
                9512,
                1080,
                5235,
                1830,
                3097,
                11,
                293,
                550,
                729,
                5235,
                854,
                365,
                2445,
                50988
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2256249962747097,
            "compression_ratio": 1.7923076923076924,
            "no_speech_prob": 0.0966796875,
            "words": [
                {
                    "start": 1738.79,
                    "end": 1739.03,
                    "word": " where",
                    "probability": 0.87646484375
                },
                {
                    "start": 1739.03,
                    "end": 1739.15,
                    "word": " it",
                    "probability": 0.900390625
                },
                {
                    "start": 1739.15,
                    "end": 1739.39,
                    "word": " stores",
                    "probability": 0.72705078125
                },
                {
                    "start": 1739.39,
                    "end": 1739.59,
                    "word": " its",
                    "probability": 0.78857421875
                },
                {
                    "start": 1739.59,
                    "end": 1740.17,
                    "word": " experiences",
                    "probability": 0.78466796875
                },
                {
                    "start": 1740.17,
                    "end": 1740.65,
                    "word": " during",
                    "probability": 0.79931640625
                },
                {
                    "start": 1740.65,
                    "end": 1741.07,
                    "word": " training",
                    "probability": 0.7939453125
                },
                {
                    "start": 1741.07,
                    "end": 1741.25,
                    "word": ",",
                    "probability": 0.357666015625
                },
                {
                    "start": 1741.25,
                    "end": 1741.33,
                    "word": " and",
                    "probability": 0.89013671875
                },
                {
                    "start": 1741.33,
                    "end": 1741.57,
                    "word": " then",
                    "probability": 0.73291015625
                },
                {
                    "start": 1741.57,
                    "end": 1742.27,
                    "word": " those",
                    "probability": 0.80908203125
                },
                {
                    "start": 1742.27,
                    "end": 1742.73,
                    "word": " experiences",
                    "probability": 0.775390625
                },
                {
                    "start": 1742.73,
                    "end": 1743.01,
                    "word": " help",
                    "probability": 0.7880859375
                },
                {
                    "start": 1743.01,
                    "end": 1743.15,
                    "word": " with",
                    "probability": 0.546875
                },
                {
                    "start": 1743.15,
                    "end": 1743.45,
                    "word": " function",
                    "probability": 0.818359375
                }
            ]
        },
        {
            "id": 262,
            "seek": 175480,
            "start": 1743.45,
            "end": 1750.77,
            "text": " approximation, data-driven policy learning, and such. And then there is an exploration controller,",
            "tokens": [
                50988,
                28023,
                11,
                1412,
                12,
                25456,
                3897,
                2539,
                11,
                293,
                1270,
                13,
                400,
                550,
                456,
                307,
                364,
                16197,
                10561,
                11,
                51352
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2256249962747097,
            "compression_ratio": 1.7923076923076924,
            "no_speech_prob": 0.0966796875,
            "words": [
                {
                    "start": 1743.45,
                    "end": 1743.97,
                    "word": " approximation",
                    "probability": 0.76220703125
                },
                {
                    "start": 1743.97,
                    "end": 1744.19,
                    "word": ",",
                    "probability": 0.9208984375
                },
                {
                    "start": 1744.19,
                    "end": 1744.35,
                    "word": " data",
                    "probability": 0.802734375
                },
                {
                    "start": 1744.35,
                    "end": 1744.59,
                    "word": "-driven",
                    "probability": 0.822021484375
                },
                {
                    "start": 1744.59,
                    "end": 1745.01,
                    "word": " policy",
                    "probability": 0.7900390625
                },
                {
                    "start": 1745.01,
                    "end": 1745.43,
                    "word": " learning",
                    "probability": 0.8203125
                },
                {
                    "start": 1745.43,
                    "end": 1746.01,
                    "word": ",",
                    "probability": 0.6806640625
                },
                {
                    "start": 1746.01,
                    "end": 1746.23,
                    "word": " and",
                    "probability": 0.8955078125
                },
                {
                    "start": 1746.23,
                    "end": 1746.53,
                    "word": " such",
                    "probability": 0.85888671875
                },
                {
                    "start": 1746.53,
                    "end": 1747.13,
                    "word": ".",
                    "probability": 0.896484375
                },
                {
                    "start": 1747.17,
                    "end": 1747.29,
                    "word": " And",
                    "probability": 0.70751953125
                },
                {
                    "start": 1747.29,
                    "end": 1748.25,
                    "word": " then",
                    "probability": 0.71728515625
                },
                {
                    "start": 1748.25,
                    "end": 1749.43,
                    "word": " there",
                    "probability": 0.6845703125
                },
                {
                    "start": 1749.43,
                    "end": 1749.55,
                    "word": " is",
                    "probability": 0.83740234375
                },
                {
                    "start": 1749.55,
                    "end": 1749.67,
                    "word": " an",
                    "probability": 0.90869140625
                },
                {
                    "start": 1749.67,
                    "end": 1750.05,
                    "word": " exploration",
                    "probability": 0.7685546875
                },
                {
                    "start": 1750.05,
                    "end": 1750.63,
                    "word": " controller",
                    "probability": 0.66455078125
                },
                {
                    "start": 1750.63,
                    "end": 1750.77,
                    "word": ",",
                    "probability": 0.63623046875
                }
            ]
        },
        {
            "id": 263,
            "seek": 175480,
            "start": 1750.77,
            "end": 1755.21,
            "text": " which controls how the agent explores the environment during training. There's an",
            "tokens": [
                51352,
                597,
                9003,
                577,
                264,
                9461,
                45473,
                264,
                2823,
                1830,
                3097,
                13,
                821,
                311,
                364,
                51572
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2256249962747097,
            "compression_ratio": 1.7923076923076924,
            "no_speech_prob": 0.0966796875,
            "words": [
                {
                    "start": 1750.77,
                    "end": 1751.07,
                    "word": " which",
                    "probability": 0.91162109375
                },
                {
                    "start": 1751.07,
                    "end": 1751.57,
                    "word": " controls",
                    "probability": 0.59814453125
                },
                {
                    "start": 1751.57,
                    "end": 1752.15,
                    "word": " how",
                    "probability": 0.8984375
                },
                {
                    "start": 1752.15,
                    "end": 1752.37,
                    "word": " the",
                    "probability": 0.818359375
                },
                {
                    "start": 1752.37,
                    "end": 1752.55,
                    "word": " agent",
                    "probability": 0.82470703125
                },
                {
                    "start": 1752.55,
                    "end": 1753.03,
                    "word": " explores",
                    "probability": 0.73681640625
                },
                {
                    "start": 1753.03,
                    "end": 1753.21,
                    "word": " the",
                    "probability": 0.80615234375
                },
                {
                    "start": 1753.21,
                    "end": 1753.47,
                    "word": " environment",
                    "probability": 0.775390625
                },
                {
                    "start": 1753.47,
                    "end": 1753.89,
                    "word": " during",
                    "probability": 0.7744140625
                },
                {
                    "start": 1753.89,
                    "end": 1754.33,
                    "word": " training",
                    "probability": 0.798828125
                },
                {
                    "start": 1754.33,
                    "end": 1754.83,
                    "word": ".",
                    "probability": 0.9716796875
                },
                {
                    "start": 1754.83,
                    "end": 1755.13,
                    "word": " There's",
                    "probability": 0.834228515625
                },
                {
                    "start": 1755.13,
                    "end": 1755.21,
                    "word": " an",
                    "probability": 0.90380859375
                }
            ]
        },
        {
            "id": 264,
            "seek": 175480,
            "start": 1755.21,
            "end": 1760.65,
            "text": " experience selector, how to select experiences from the bank of data or observations stored.",
            "tokens": [
                51572,
                1752,
                23264,
                1672,
                11,
                577,
                281,
                3048,
                5235,
                490,
                264,
                3765,
                295,
                1412,
                420,
                18163,
                12187,
                13,
                51840
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2256249962747097,
            "compression_ratio": 1.7923076923076924,
            "no_speech_prob": 0.0966796875,
            "words": [
                {
                    "start": 1755.21,
                    "end": 1755.47,
                    "word": " experience",
                    "probability": 0.76171875
                },
                {
                    "start": 1755.47,
                    "end": 1755.97,
                    "word": " selector",
                    "probability": 0.884521484375
                },
                {
                    "start": 1755.97,
                    "end": 1756.09,
                    "word": ",",
                    "probability": 0.779296875
                },
                {
                    "start": 1756.09,
                    "end": 1756.23,
                    "word": " how",
                    "probability": 0.89404296875
                },
                {
                    "start": 1756.23,
                    "end": 1756.39,
                    "word": " to",
                    "probability": 0.9052734375
                },
                {
                    "start": 1756.39,
                    "end": 1756.67,
                    "word": " select",
                    "probability": 0.7978515625
                },
                {
                    "start": 1756.67,
                    "end": 1757.21,
                    "word": " experiences",
                    "probability": 0.7626953125
                },
                {
                    "start": 1757.21,
                    "end": 1757.63,
                    "word": " from",
                    "probability": 0.81689453125
                },
                {
                    "start": 1757.63,
                    "end": 1757.95,
                    "word": " the",
                    "probability": 0.81884765625
                },
                {
                    "start": 1757.95,
                    "end": 1758.29,
                    "word": " bank",
                    "probability": 0.83203125
                },
                {
                    "start": 1758.29,
                    "end": 1758.65,
                    "word": " of",
                    "probability": 0.908203125
                },
                {
                    "start": 1758.65,
                    "end": 1759.33,
                    "word": " data",
                    "probability": 0.8291015625
                },
                {
                    "start": 1759.33,
                    "end": 1759.57,
                    "word": " or",
                    "probability": 0.89208984375
                },
                {
                    "start": 1759.57,
                    "end": 1760.03,
                    "word": " observations",
                    "probability": 0.6005859375
                },
                {
                    "start": 1760.03,
                    "end": 1760.47,
                    "word": " stored",
                    "probability": 0.7451171875
                },
                {
                    "start": 1760.47,
                    "end": 1760.65,
                    "word": ".",
                    "probability": 0.92578125
                }
            ]
        },
        {
            "id": 265,
            "seek": 177978,
            "start": 1760.93,
            "end": 1767.99,
            "text": " In this data set, there is an actuator, which then enacts the actions of the agent inside an",
            "tokens": [
                50364,
                682,
                341,
                1412,
                992,
                11,
                456,
                307,
                364,
                34964,
                1639,
                11,
                597,
                550,
                465,
                15295,
                264,
                5909,
                295,
                264,
                9461,
                1854,
                364,
                50716
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2140436693128333,
            "compression_ratio": 1.6875,
            "no_speech_prob": 0.0006561279296875,
            "words": [
                {
                    "start": 1760.93,
                    "end": 1761.11,
                    "word": " In",
                    "probability": 0.1322021484375
                },
                {
                    "start": 1761.11,
                    "end": 1761.35,
                    "word": " this",
                    "probability": 0.86181640625
                },
                {
                    "start": 1761.35,
                    "end": 1761.63,
                    "word": " data",
                    "probability": 0.4384765625
                },
                {
                    "start": 1761.63,
                    "end": 1761.95,
                    "word": " set",
                    "probability": 0.8671875
                },
                {
                    "start": 1761.95,
                    "end": 1762.13,
                    "word": ",",
                    "probability": 0.77783203125
                },
                {
                    "start": 1762.13,
                    "end": 1762.49,
                    "word": " there",
                    "probability": 0.85107421875
                },
                {
                    "start": 1762.49,
                    "end": 1762.71,
                    "word": " is",
                    "probability": 0.57470703125
                },
                {
                    "start": 1762.71,
                    "end": 1763.05,
                    "word": " an",
                    "probability": 0.8955078125
                },
                {
                    "start": 1763.05,
                    "end": 1763.67,
                    "word": " actuator",
                    "probability": 0.880615234375
                },
                {
                    "start": 1763.67,
                    "end": 1763.85,
                    "word": ",",
                    "probability": 0.27978515625
                },
                {
                    "start": 1763.85,
                    "end": 1763.95,
                    "word": " which",
                    "probability": 0.9140625
                },
                {
                    "start": 1763.95,
                    "end": 1764.33,
                    "word": " then",
                    "probability": 0.7119140625
                },
                {
                    "start": 1764.33,
                    "end": 1765.11,
                    "word": " enacts",
                    "probability": 0.733154296875
                },
                {
                    "start": 1765.11,
                    "end": 1765.73,
                    "word": " the",
                    "probability": 0.80712890625
                },
                {
                    "start": 1765.73,
                    "end": 1766.09,
                    "word": " actions",
                    "probability": 0.8623046875
                },
                {
                    "start": 1766.09,
                    "end": 1766.31,
                    "word": " of",
                    "probability": 0.90380859375
                },
                {
                    "start": 1766.31,
                    "end": 1766.45,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 1766.45,
                    "end": 1766.85,
                    "word": " agent",
                    "probability": 0.35791015625
                },
                {
                    "start": 1766.85,
                    "end": 1767.85,
                    "word": " inside",
                    "probability": 0.71142578125
                },
                {
                    "start": 1767.85,
                    "end": 1767.99,
                    "word": " an",
                    "probability": 0.89697265625
                }
            ]
        },
        {
            "id": 266,
            "seek": 177978,
            "start": 1767.99,
            "end": 1773.97,
            "text": " environment. The environment is connected back to this agent block through an observation channel,",
            "tokens": [
                50716,
                2823,
                13,
                440,
                2823,
                307,
                4582,
                646,
                281,
                341,
                9461,
                3461,
                807,
                364,
                14816,
                2269,
                11,
                51012
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2140436693128333,
            "compression_ratio": 1.6875,
            "no_speech_prob": 0.0006561279296875,
            "words": [
                {
                    "start": 1767.99,
                    "end": 1768.59,
                    "word": " environment",
                    "probability": 0.7744140625
                },
                {
                    "start": 1768.59,
                    "end": 1768.83,
                    "word": ".",
                    "probability": 0.859375
                },
                {
                    "start": 1768.83,
                    "end": 1769.29,
                    "word": " The",
                    "probability": 0.80322265625
                },
                {
                    "start": 1769.29,
                    "end": 1769.73,
                    "word": " environment",
                    "probability": 0.73876953125
                },
                {
                    "start": 1769.73,
                    "end": 1770.33,
                    "word": " is",
                    "probability": 0.89453125
                },
                {
                    "start": 1770.33,
                    "end": 1770.75,
                    "word": " connected",
                    "probability": 0.77099609375
                },
                {
                    "start": 1770.75,
                    "end": 1771.25,
                    "word": " back",
                    "probability": 0.814453125
                },
                {
                    "start": 1771.25,
                    "end": 1771.55,
                    "word": " to",
                    "probability": 0.9033203125
                },
                {
                    "start": 1771.55,
                    "end": 1771.77,
                    "word": " this",
                    "probability": 0.86767578125
                },
                {
                    "start": 1771.77,
                    "end": 1772.07,
                    "word": " agent",
                    "probability": 0.81103515625
                },
                {
                    "start": 1772.07,
                    "end": 1772.41,
                    "word": " block",
                    "probability": 0.73828125
                },
                {
                    "start": 1772.41,
                    "end": 1772.93,
                    "word": " through",
                    "probability": 0.7666015625
                },
                {
                    "start": 1772.93,
                    "end": 1773.09,
                    "word": " an",
                    "probability": 0.89990234375
                },
                {
                    "start": 1773.09,
                    "end": 1773.53,
                    "word": " observation",
                    "probability": 0.8212890625
                },
                {
                    "start": 1773.53,
                    "end": 1773.89,
                    "word": " channel",
                    "probability": 0.734375
                },
                {
                    "start": 1773.89,
                    "end": 1773.97,
                    "word": ",",
                    "probability": 0.72802734375
                }
            ]
        },
        {
            "id": 267,
            "seek": 177978,
            "start": 1773.97,
            "end": 1779.39,
            "text": " observation of the state, and reward channel. And it's of no surprise to the more seasoned",
            "tokens": [
                51012,
                14816,
                295,
                264,
                1785,
                11,
                293,
                7782,
                2269,
                13,
                400,
                309,
                311,
                295,
                572,
                6365,
                281,
                264,
                544,
                30111,
                51288
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2140436693128333,
            "compression_ratio": 1.6875,
            "no_speech_prob": 0.0006561279296875,
            "words": [
                {
                    "start": 1773.97,
                    "end": 1774.31,
                    "word": " observation",
                    "probability": 0.80078125
                },
                {
                    "start": 1774.31,
                    "end": 1774.53,
                    "word": " of",
                    "probability": 0.900390625
                },
                {
                    "start": 1774.53,
                    "end": 1774.65,
                    "word": " the",
                    "probability": 0.7685546875
                },
                {
                    "start": 1774.65,
                    "end": 1774.97,
                    "word": " state",
                    "probability": 0.385498046875
                },
                {
                    "start": 1774.97,
                    "end": 1775.37,
                    "word": ",",
                    "probability": 0.38330078125
                },
                {
                    "start": 1775.37,
                    "end": 1775.85,
                    "word": " and",
                    "probability": 0.90087890625
                },
                {
                    "start": 1775.85,
                    "end": 1776.19,
                    "word": " reward",
                    "probability": 0.80419921875
                },
                {
                    "start": 1776.19,
                    "end": 1776.51,
                    "word": " channel",
                    "probability": 0.77001953125
                },
                {
                    "start": 1776.51,
                    "end": 1776.65,
                    "word": ".",
                    "probability": 0.94921875
                },
                {
                    "start": 1776.65,
                    "end": 1776.73,
                    "word": " And",
                    "probability": 0.63818359375
                },
                {
                    "start": 1776.73,
                    "end": 1777.15,
                    "word": " it's",
                    "probability": 0.946533203125
                },
                {
                    "start": 1777.15,
                    "end": 1777.37,
                    "word": " of",
                    "probability": 0.869140625
                },
                {
                    "start": 1777.37,
                    "end": 1777.57,
                    "word": " no",
                    "probability": 0.90673828125
                },
                {
                    "start": 1777.57,
                    "end": 1777.97,
                    "word": " surprise",
                    "probability": 0.7421875
                },
                {
                    "start": 1777.97,
                    "end": 1778.35,
                    "word": " to",
                    "probability": 0.89501953125
                },
                {
                    "start": 1778.35,
                    "end": 1778.99,
                    "word": " the",
                    "probability": 0.8046875
                },
                {
                    "start": 1778.99,
                    "end": 1779.15,
                    "word": " more",
                    "probability": 0.89013671875
                },
                {
                    "start": 1779.15,
                    "end": 1779.39,
                    "word": " seasoned",
                    "probability": 0.92236328125
                }
            ]
        },
        {
            "id": 268,
            "seek": 177978,
            "start": 1779.39,
            "end": 1785.91,
            "text": " security researcher and professional that all of these components can be a subject or target of",
            "tokens": [
                51288,
                3825,
                21751,
                293,
                4843,
                300,
                439,
                295,
                613,
                6677,
                393,
                312,
                257,
                3983,
                420,
                3779,
                295,
                51608
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2140436693128333,
            "compression_ratio": 1.6875,
            "no_speech_prob": 0.0006561279296875,
            "words": [
                {
                    "start": 1779.39,
                    "end": 1779.83,
                    "word": " security",
                    "probability": 0.9130859375
                },
                {
                    "start": 1779.83,
                    "end": 1780.27,
                    "word": " researcher",
                    "probability": 0.66845703125
                },
                {
                    "start": 1780.27,
                    "end": 1780.43,
                    "word": " and",
                    "probability": 0.8486328125
                },
                {
                    "start": 1780.43,
                    "end": 1780.77,
                    "word": " professional",
                    "probability": 0.7822265625
                },
                {
                    "start": 1780.77,
                    "end": 1781.29,
                    "word": " that",
                    "probability": 0.71435546875
                },
                {
                    "start": 1781.29,
                    "end": 1781.61,
                    "word": " all",
                    "probability": 0.8974609375
                },
                {
                    "start": 1781.61,
                    "end": 1781.73,
                    "word": " of",
                    "probability": 0.900390625
                },
                {
                    "start": 1781.73,
                    "end": 1781.91,
                    "word": " these",
                    "probability": 0.69921875
                },
                {
                    "start": 1781.91,
                    "end": 1782.29,
                    "word": " components",
                    "probability": 0.830078125
                },
                {
                    "start": 1782.29,
                    "end": 1782.67,
                    "word": " can",
                    "probability": 0.89599609375
                },
                {
                    "start": 1782.67,
                    "end": 1783.61,
                    "word": " be",
                    "probability": 0.904296875
                },
                {
                    "start": 1784.39,
                    "end": 1785.01,
                    "word": " a subject",
                    "probability": 0.90869140625
                },
                {
                    "start": 1785.01,
                    "end": 1785.29,
                    "word": " or",
                    "probability": 0.88427734375
                },
                {
                    "start": 1785.29,
                    "end": 1785.69,
                    "word": " target",
                    "probability": 0.83056640625
                },
                {
                    "start": 1785.69,
                    "end": 1785.91,
                    "word": " of",
                    "probability": 0.8974609375
                }
            ]
        },
        {
            "id": 269,
            "seek": 180616,
            "start": 1785.91,
            "end": 1794.59,
            "text": " attack. As we go forward, we'll cover some examples of attacks that can occur on each of",
            "tokens": [
                50404,
                2690,
                13,
                1018,
                321,
                352,
                2128,
                11,
                321,
                603,
                2060,
                512,
                5110,
                295,
                8122,
                300,
                393,
                5160,
                322,
                1184,
                295,
                50792
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21161684442473494,
            "compression_ratio": 1.711111111111111,
            "no_speech_prob": 0.1519775390625,
            "words": [
                {
                    "start": 1785.91,
                    "end": 1786.47,
                    "word": " attack",
                    "probability": 0.00180816650390625
                },
                {
                    "start": 1786.47,
                    "end": 1786.87,
                    "word": ".",
                    "probability": 0.63232421875
                },
                {
                    "start": 1786.97,
                    "end": 1787.53,
                    "word": " As",
                    "probability": 0.66650390625
                },
                {
                    "start": 1787.53,
                    "end": 1787.77,
                    "word": " we",
                    "probability": 0.9013671875
                },
                {
                    "start": 1787.77,
                    "end": 1788.61,
                    "word": " go",
                    "probability": 0.84716796875
                },
                {
                    "start": 1788.61,
                    "end": 1790.17,
                    "word": " forward",
                    "probability": 0.81982421875
                },
                {
                    "start": 1790.17,
                    "end": 1790.35,
                    "word": ",",
                    "probability": 0.802734375
                },
                {
                    "start": 1790.35,
                    "end": 1790.77,
                    "word": " we'll",
                    "probability": 0.813232421875
                },
                {
                    "start": 1790.77,
                    "end": 1791.01,
                    "word": " cover",
                    "probability": 0.767578125
                },
                {
                    "start": 1791.01,
                    "end": 1791.77,
                    "word": " some",
                    "probability": 0.802734375
                },
                {
                    "start": 1791.77,
                    "end": 1792.09,
                    "word": " examples",
                    "probability": 0.7705078125
                },
                {
                    "start": 1792.09,
                    "end": 1792.21,
                    "word": " of",
                    "probability": 0.90234375
                },
                {
                    "start": 1792.21,
                    "end": 1792.55,
                    "word": " attacks",
                    "probability": 0.7333984375
                },
                {
                    "start": 1792.55,
                    "end": 1792.77,
                    "word": " that",
                    "probability": 0.869140625
                },
                {
                    "start": 1792.77,
                    "end": 1792.87,
                    "word": " can",
                    "probability": 0.87890625
                },
                {
                    "start": 1792.87,
                    "end": 1793.11,
                    "word": " occur",
                    "probability": 0.845703125
                },
                {
                    "start": 1793.11,
                    "end": 1793.47,
                    "word": " on",
                    "probability": 0.90478515625
                },
                {
                    "start": 1793.47,
                    "end": 1794.49,
                    "word": " each",
                    "probability": 0.82421875
                },
                {
                    "start": 1794.49,
                    "end": 1794.59,
                    "word": " of",
                    "probability": 0.90576171875
                }
            ]
        },
        {
            "id": 270,
            "seek": 180616,
            "start": 1794.59,
            "end": 1800.19,
            "text": " these components. We've already seen an attack on the observation channel. There are attacks on the",
            "tokens": [
                50792,
                613,
                6677,
                13,
                492,
                600,
                1217,
                1612,
                364,
                2690,
                322,
                264,
                14816,
                2269,
                13,
                821,
                366,
                8122,
                322,
                264,
                51072
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21161684442473494,
            "compression_ratio": 1.711111111111111,
            "no_speech_prob": 0.1519775390625,
            "words": [
                {
                    "start": 1794.59,
                    "end": 1794.85,
                    "word": " these",
                    "probability": 0.69580078125
                },
                {
                    "start": 1794.85,
                    "end": 1795.33,
                    "word": " components",
                    "probability": 0.822265625
                },
                {
                    "start": 1795.33,
                    "end": 1795.33,
                    "word": ".",
                    "probability": 0.90380859375
                },
                {
                    "start": 1795.35,
                    "end": 1795.91,
                    "word": " We've",
                    "probability": 0.918701171875
                },
                {
                    "start": 1795.91,
                    "end": 1796.05,
                    "word": " already",
                    "probability": 0.8076171875
                },
                {
                    "start": 1796.05,
                    "end": 1796.23,
                    "word": " seen",
                    "probability": 0.8876953125
                },
                {
                    "start": 1796.23,
                    "end": 1796.37,
                    "word": " an",
                    "probability": 0.89501953125
                },
                {
                    "start": 1796.37,
                    "end": 1796.59,
                    "word": " attack",
                    "probability": 0.85009765625
                },
                {
                    "start": 1796.59,
                    "end": 1796.91,
                    "word": " on",
                    "probability": 0.90771484375
                },
                {
                    "start": 1796.91,
                    "end": 1797.09,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 1797.09,
                    "end": 1797.51,
                    "word": " observation",
                    "probability": 0.69287109375
                },
                {
                    "start": 1797.51,
                    "end": 1798.01,
                    "word": " channel",
                    "probability": 0.77587890625
                },
                {
                    "start": 1798.01,
                    "end": 1798.57,
                    "word": ".",
                    "probability": 0.423583984375
                },
                {
                    "start": 1799.07,
                    "end": 1799.63,
                    "word": " There",
                    "probability": 0.63427734375
                },
                {
                    "start": 1799.63,
                    "end": 1799.77,
                    "word": " are",
                    "probability": 0.873046875
                },
                {
                    "start": 1799.77,
                    "end": 1799.99,
                    "word": " attacks",
                    "probability": 0.697265625
                },
                {
                    "start": 1799.99,
                    "end": 1800.09,
                    "word": " on",
                    "probability": 0.8017578125
                },
                {
                    "start": 1800.09,
                    "end": 1800.19,
                    "word": " the",
                    "probability": 0.740234375
                }
            ]
        },
        {
            "id": 271,
            "seek": 180616,
            "start": 1800.19,
            "end": 1807.93,
            "text": " reward channel, which we'll hopefully touch on. And there are attacks on the agents during training",
            "tokens": [
                51072,
                7782,
                2269,
                11,
                597,
                321,
                603,
                4696,
                2557,
                322,
                13,
                400,
                456,
                366,
                8122,
                322,
                264,
                12554,
                1830,
                3097,
                51464
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21161684442473494,
            "compression_ratio": 1.711111111111111,
            "no_speech_prob": 0.1519775390625,
            "words": [
                {
                    "start": 1800.19,
                    "end": 1800.51,
                    "word": " reward",
                    "probability": 0.861328125
                },
                {
                    "start": 1800.51,
                    "end": 1800.95,
                    "word": " channel",
                    "probability": 0.7744140625
                },
                {
                    "start": 1800.95,
                    "end": 1801.33,
                    "word": ",",
                    "probability": 0.86376953125
                },
                {
                    "start": 1801.33,
                    "end": 1801.77,
                    "word": " which",
                    "probability": 0.9052734375
                },
                {
                    "start": 1801.77,
                    "end": 1802.23,
                    "word": " we'll",
                    "probability": 0.841064453125
                },
                {
                    "start": 1802.23,
                    "end": 1802.87,
                    "word": " hopefully",
                    "probability": 0.7041015625
                },
                {
                    "start": 1802.87,
                    "end": 1803.73,
                    "word": " touch",
                    "probability": 0.69873046875
                },
                {
                    "start": 1803.73,
                    "end": 1804.45,
                    "word": " on",
                    "probability": 0.90625
                },
                {
                    "start": 1804.45,
                    "end": 1804.57,
                    "word": ".",
                    "probability": 0.6240234375
                },
                {
                    "start": 1804.57,
                    "end": 1804.99,
                    "word": " And",
                    "probability": 0.82958984375
                },
                {
                    "start": 1804.99,
                    "end": 1805.79,
                    "word": " there",
                    "probability": 0.8349609375
                },
                {
                    "start": 1805.79,
                    "end": 1805.89,
                    "word": " are",
                    "probability": 0.888671875
                },
                {
                    "start": 1805.89,
                    "end": 1806.07,
                    "word": " attacks",
                    "probability": 0.7099609375
                },
                {
                    "start": 1806.07,
                    "end": 1806.19,
                    "word": " on",
                    "probability": 0.8837890625
                },
                {
                    "start": 1806.19,
                    "end": 1806.29,
                    "word": " the",
                    "probability": 0.79931640625
                },
                {
                    "start": 1806.29,
                    "end": 1806.83,
                    "word": " agents",
                    "probability": 0.59375
                },
                {
                    "start": 1806.83,
                    "end": 1807.53,
                    "word": " during",
                    "probability": 0.521484375
                },
                {
                    "start": 1807.53,
                    "end": 1807.93,
                    "word": " training",
                    "probability": 0.79736328125
                }
            ]
        },
        {
            "id": 272,
            "seek": 180616,
            "start": 1807.93,
            "end": 1812.29,
            "text": " and the actuator. Let me give you a quick example of the actuator attack. Let's assume we have a",
            "tokens": [
                51464,
                293,
                264,
                34964,
                1639,
                13,
                961,
                385,
                976,
                291,
                257,
                1702,
                1365,
                295,
                264,
                34964,
                1639,
                2690,
                13,
                961,
                311,
                6552,
                321,
                362,
                257,
                51684
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21161684442473494,
            "compression_ratio": 1.711111111111111,
            "no_speech_prob": 0.1519775390625,
            "words": [
                {
                    "start": 1807.93,
                    "end": 1808.15,
                    "word": " and",
                    "probability": 0.7529296875
                },
                {
                    "start": 1808.15,
                    "end": 1808.29,
                    "word": " the",
                    "probability": 0.78955078125
                },
                {
                    "start": 1808.29,
                    "end": 1808.71,
                    "word": " actuator",
                    "probability": 0.8291015625
                },
                {
                    "start": 1808.71,
                    "end": 1808.87,
                    "word": ".",
                    "probability": 0.83935546875
                },
                {
                    "start": 1808.87,
                    "end": 1808.95,
                    "word": " Let",
                    "probability": 0.8828125
                },
                {
                    "start": 1808.95,
                    "end": 1809.05,
                    "word": " me",
                    "probability": 0.8984375
                },
                {
                    "start": 1809.05,
                    "end": 1809.17,
                    "word": " give",
                    "probability": 0.82763671875
                },
                {
                    "start": 1809.17,
                    "end": 1809.23,
                    "word": " you",
                    "probability": 0.8935546875
                },
                {
                    "start": 1809.33,
                    "end": 1809.47,
                    "word": " a quick",
                    "probability": 0.92041015625
                },
                {
                    "start": 1809.47,
                    "end": 1809.81,
                    "word": " example",
                    "probability": 0.9228515625
                },
                {
                    "start": 1809.81,
                    "end": 1809.97,
                    "word": " of",
                    "probability": 0.76416015625
                },
                {
                    "start": 1809.97,
                    "end": 1810.09,
                    "word": " the",
                    "probability": 0.802734375
                },
                {
                    "start": 1810.09,
                    "end": 1810.45,
                    "word": " actuator",
                    "probability": 0.868408203125
                },
                {
                    "start": 1810.45,
                    "end": 1810.83,
                    "word": " attack",
                    "probability": 0.853515625
                },
                {
                    "start": 1810.83,
                    "end": 1811.21,
                    "word": ".",
                    "probability": 0.9619140625
                },
                {
                    "start": 1811.21,
                    "end": 1811.39,
                    "word": " Let's",
                    "probability": 0.943115234375
                },
                {
                    "start": 1811.39,
                    "end": 1811.57,
                    "word": " assume",
                    "probability": 0.828125
                },
                {
                    "start": 1811.57,
                    "end": 1811.71,
                    "word": " we",
                    "probability": 0.787109375
                },
                {
                    "start": 1811.71,
                    "end": 1812.01,
                    "word": " have",
                    "probability": 0.88037109375
                },
                {
                    "start": 1812.01,
                    "end": 1812.29,
                    "word": " a",
                    "probability": 0.93359375
                }
            ]
        },
        {
            "id": 273,
            "seek": 183426,
            "start": 1812.29,
            "end": 1817.97,
            "text": " robot, an actual robot, learning to navigate in an environment while avoiding obstacles.",
            "tokens": [
                50368,
                7881,
                11,
                364,
                3539,
                7881,
                11,
                2539,
                281,
                12350,
                294,
                364,
                2823,
                1339,
                20220,
                17735,
                13,
                50620
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23125000140246224,
            "compression_ratio": 1.8316326530612246,
            "no_speech_prob": 0.01450347900390625,
            "words": [
                {
                    "start": 1812.29,
                    "end": 1812.75,
                    "word": " robot",
                    "probability": 0.192138671875
                },
                {
                    "start": 1812.75,
                    "end": 1812.85,
                    "word": ",",
                    "probability": 0.242431640625
                },
                {
                    "start": 1812.85,
                    "end": 1812.93,
                    "word": " an",
                    "probability": 0.65771484375
                },
                {
                    "start": 1812.93,
                    "end": 1813.29,
                    "word": " actual",
                    "probability": 0.814453125
                },
                {
                    "start": 1813.29,
                    "end": 1813.83,
                    "word": " robot",
                    "probability": 0.7685546875
                },
                {
                    "start": 1813.83,
                    "end": 1814.31,
                    "word": ",",
                    "probability": 0.533203125
                },
                {
                    "start": 1814.31,
                    "end": 1814.97,
                    "word": " learning",
                    "probability": 0.81640625
                },
                {
                    "start": 1814.97,
                    "end": 1815.19,
                    "word": " to",
                    "probability": 0.89990234375
                },
                {
                    "start": 1815.19,
                    "end": 1815.51,
                    "word": " navigate",
                    "probability": 0.8193359375
                },
                {
                    "start": 1815.51,
                    "end": 1815.75,
                    "word": " in",
                    "probability": 0.77587890625
                },
                {
                    "start": 1815.75,
                    "end": 1815.87,
                    "word": " an",
                    "probability": 0.8740234375
                },
                {
                    "start": 1815.87,
                    "end": 1816.19,
                    "word": " environment",
                    "probability": 0.759765625
                },
                {
                    "start": 1816.19,
                    "end": 1816.55,
                    "word": " while",
                    "probability": 0.87939453125
                },
                {
                    "start": 1816.55,
                    "end": 1816.87,
                    "word": " avoiding",
                    "probability": 0.93798828125
                },
                {
                    "start": 1816.87,
                    "end": 1817.45,
                    "word": " obstacles",
                    "probability": 0.765625
                },
                {
                    "start": 1817.45,
                    "end": 1817.97,
                    "word": ".",
                    "probability": 0.904296875
                }
            ]
        },
        {
            "id": 274,
            "seek": 183426,
            "start": 1818.13,
            "end": 1828.59,
            "text": " If the robot commands or decides to, let's say, move the left wheel forward,",
            "tokens": [
                50668,
                759,
                264,
                7881,
                16901,
                420,
                14898,
                281,
                11,
                718,
                311,
                584,
                11,
                1286,
                264,
                1411,
                5589,
                2128,
                11,
                51168
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23125000140246224,
            "compression_ratio": 1.8316326530612246,
            "no_speech_prob": 0.01450347900390625,
            "words": [
                {
                    "start": 1818.13,
                    "end": 1818.55,
                    "word": " If",
                    "probability": 0.859375
                },
                {
                    "start": 1818.55,
                    "end": 1818.99,
                    "word": " the",
                    "probability": 0.79931640625
                },
                {
                    "start": 1818.99,
                    "end": 1819.39,
                    "word": " robot",
                    "probability": 0.7880859375
                },
                {
                    "start": 1819.39,
                    "end": 1823.11,
                    "word": " commands",
                    "probability": 0.8193359375
                },
                {
                    "start": 1823.11,
                    "end": 1823.53,
                    "word": " or",
                    "probability": 0.73876953125
                },
                {
                    "start": 1823.53,
                    "end": 1825.05,
                    "word": " decides",
                    "probability": 0.88623046875
                },
                {
                    "start": 1825.05,
                    "end": 1826.09,
                    "word": " to",
                    "probability": 0.8818359375
                },
                {
                    "start": 1826.09,
                    "end": 1826.39,
                    "word": ",",
                    "probability": 0.384521484375
                },
                {
                    "start": 1826.39,
                    "end": 1826.65,
                    "word": " let's",
                    "probability": 0.937744140625
                },
                {
                    "start": 1826.65,
                    "end": 1826.79,
                    "word": " say",
                    "probability": 0.89501953125
                },
                {
                    "start": 1826.79,
                    "end": 1826.83,
                    "word": ",",
                    "probability": 0.861328125
                },
                {
                    "start": 1826.83,
                    "end": 1827.01,
                    "word": " move",
                    "probability": 0.8671875
                },
                {
                    "start": 1827.01,
                    "end": 1827.21,
                    "word": " the",
                    "probability": 0.8076171875
                },
                {
                    "start": 1827.21,
                    "end": 1827.47,
                    "word": " left",
                    "probability": 0.84033203125
                },
                {
                    "start": 1827.47,
                    "end": 1827.73,
                    "word": " wheel",
                    "probability": 0.8388671875
                },
                {
                    "start": 1827.73,
                    "end": 1828.33,
                    "word": " forward",
                    "probability": 0.8486328125
                },
                {
                    "start": 1828.33,
                    "end": 1828.59,
                    "word": ",",
                    "probability": 0.634765625
                }
            ]
        },
        {
            "id": 275,
            "seek": 183426,
            "start": 1829.25,
            "end": 1833.25,
            "text": " but there is some sort of obstacle in front of the left wheel, and the left wheel doesn't actually",
            "tokens": [
                51224,
                457,
                456,
                307,
                512,
                1333,
                295,
                23112,
                294,
                1868,
                295,
                264,
                1411,
                5589,
                11,
                293,
                264,
                1411,
                5589,
                1177,
                380,
                767,
                51416
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23125000140246224,
            "compression_ratio": 1.8316326530612246,
            "no_speech_prob": 0.01450347900390625,
            "words": [
                {
                    "start": 1829.25,
                    "end": 1829.77,
                    "word": " but",
                    "probability": 0.88623046875
                },
                {
                    "start": 1829.77,
                    "end": 1829.97,
                    "word": " there",
                    "probability": 0.85888671875
                },
                {
                    "start": 1829.97,
                    "end": 1830.15,
                    "word": " is",
                    "probability": 0.88427734375
                },
                {
                    "start": 1830.15,
                    "end": 1830.33,
                    "word": " some",
                    "probability": 0.8125
                },
                {
                    "start": 1830.33,
                    "end": 1830.43,
                    "word": " sort",
                    "probability": 0.673828125
                },
                {
                    "start": 1830.43,
                    "end": 1830.57,
                    "word": " of",
                    "probability": 0.90576171875
                },
                {
                    "start": 1830.57,
                    "end": 1830.91,
                    "word": " obstacle",
                    "probability": 0.82080078125
                },
                {
                    "start": 1830.91,
                    "end": 1831.13,
                    "word": " in",
                    "probability": 0.90234375
                },
                {
                    "start": 1831.13,
                    "end": 1831.27,
                    "word": " front",
                    "probability": 0.92041015625
                },
                {
                    "start": 1831.27,
                    "end": 1831.39,
                    "word": " of",
                    "probability": 0.90625
                },
                {
                    "start": 1831.39,
                    "end": 1831.53,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 1831.53,
                    "end": 1831.71,
                    "word": " left",
                    "probability": 0.85595703125
                },
                {
                    "start": 1831.71,
                    "end": 1832.07,
                    "word": " wheel",
                    "probability": 0.83447265625
                },
                {
                    "start": 1832.07,
                    "end": 1832.25,
                    "word": ",",
                    "probability": 0.390625
                },
                {
                    "start": 1832.25,
                    "end": 1832.45,
                    "word": " and",
                    "probability": 0.88330078125
                },
                {
                    "start": 1832.45,
                    "end": 1832.57,
                    "word": " the",
                    "probability": 0.80517578125
                },
                {
                    "start": 1832.57,
                    "end": 1832.71,
                    "word": " left",
                    "probability": 0.83544921875
                },
                {
                    "start": 1832.71,
                    "end": 1832.79,
                    "word": " wheel",
                    "probability": 0.6484375
                },
                {
                    "start": 1832.79,
                    "end": 1833.01,
                    "word": " doesn't",
                    "probability": 0.9306640625
                },
                {
                    "start": 1833.01,
                    "end": 1833.25,
                    "word": " actually",
                    "probability": 0.78271484375
                }
            ]
        },
        {
            "id": 276,
            "seek": 183426,
            "start": 1833.25,
            "end": 1840.39,
            "text": " move, then the actuator, the resulting data is going to be, the resulting observation is going",
            "tokens": [
                51416,
                1286,
                11,
                550,
                264,
                34964,
                1639,
                11,
                264,
                16505,
                1412,
                307,
                516,
                281,
                312,
                11,
                264,
                16505,
                14816,
                307,
                516,
                51768
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23125000140246224,
            "compression_ratio": 1.8316326530612246,
            "no_speech_prob": 0.01450347900390625,
            "words": [
                {
                    "start": 1833.25,
                    "end": 1833.69,
                    "word": " move",
                    "probability": 0.88818359375
                },
                {
                    "start": 1833.69,
                    "end": 1834.19,
                    "word": ",",
                    "probability": 0.9287109375
                },
                {
                    "start": 1834.19,
                    "end": 1834.55,
                    "word": " then",
                    "probability": 0.7060546875
                },
                {
                    "start": 1834.55,
                    "end": 1835.19,
                    "word": " the",
                    "probability": 0.8017578125
                },
                {
                    "start": 1835.19,
                    "end": 1835.85,
                    "word": " actuator",
                    "probability": 0.704833984375
                },
                {
                    "start": 1835.85,
                    "end": 1836.57,
                    "word": ",",
                    "probability": 0.5048828125
                },
                {
                    "start": 1836.57,
                    "end": 1837.57,
                    "word": " the",
                    "probability": 0.77001953125
                },
                {
                    "start": 1837.57,
                    "end": 1838.03,
                    "word": " resulting",
                    "probability": 0.81982421875
                },
                {
                    "start": 1838.03,
                    "end": 1838.45,
                    "word": " data",
                    "probability": 0.4140625
                },
                {
                    "start": 1838.45,
                    "end": 1838.63,
                    "word": " is",
                    "probability": 0.81787109375
                },
                {
                    "start": 1838.63,
                    "end": 1838.85,
                    "word": " going",
                    "probability": 0.83837890625
                },
                {
                    "start": 1838.85,
                    "end": 1839.03,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 1839.03,
                    "end": 1839.31,
                    "word": " be",
                    "probability": 0.88671875
                },
                {
                    "start": 1839.31,
                    "end": 1839.37,
                    "word": ",",
                    "probability": 0.52783203125
                },
                {
                    "start": 1839.37,
                    "end": 1839.57,
                    "word": " the",
                    "probability": 0.7978515625
                },
                {
                    "start": 1839.57,
                    "end": 1839.81,
                    "word": " resulting",
                    "probability": 0.80517578125
                },
                {
                    "start": 1839.81,
                    "end": 1840.13,
                    "word": " observation",
                    "probability": 0.7900390625
                },
                {
                    "start": 1840.13,
                    "end": 1840.29,
                    "word": " is",
                    "probability": 0.8486328125
                },
                {
                    "start": 1840.29,
                    "end": 1840.39,
                    "word": " going",
                    "probability": 0.85205078125
                }
            ]
        },
        {
            "id": 277,
            "seek": 186370,
            "start": 1840.39,
            "end": 1845.91,
            "text": " to be skewed, because the agent is going to assume that the actuation has happened and then",
            "tokens": [
                50364,
                281,
                312,
                8756,
                26896,
                11,
                570,
                264,
                9461,
                307,
                516,
                281,
                6552,
                300,
                264,
                605,
                16073,
                575,
                2011,
                293,
                550,
                50636
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21686921627433212,
            "compression_ratio": 1.755639097744361,
            "no_speech_prob": 0.00075531005859375,
            "words": [
                {
                    "start": 1840.39,
                    "end": 1840.45,
                    "word": " to",
                    "probability": 0.21337890625
                },
                {
                    "start": 1840.45,
                    "end": 1840.55,
                    "word": " be",
                    "probability": 0.84130859375
                },
                {
                    "start": 1840.55,
                    "end": 1841.01,
                    "word": " skewed",
                    "probability": 0.748779296875
                },
                {
                    "start": 1841.01,
                    "end": 1841.43,
                    "word": ",",
                    "probability": 0.35595703125
                },
                {
                    "start": 1841.43,
                    "end": 1841.67,
                    "word": " because",
                    "probability": 0.8330078125
                },
                {
                    "start": 1841.67,
                    "end": 1841.83,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 1841.83,
                    "end": 1842.03,
                    "word": " agent",
                    "probability": 0.806640625
                },
                {
                    "start": 1842.03,
                    "end": 1842.17,
                    "word": " is",
                    "probability": 0.90185546875
                },
                {
                    "start": 1842.17,
                    "end": 1842.29,
                    "word": " going",
                    "probability": 0.85400390625
                },
                {
                    "start": 1842.29,
                    "end": 1842.57,
                    "word": " to",
                    "probability": 0.91162109375
                },
                {
                    "start": 1842.57,
                    "end": 1843.29,
                    "word": " assume",
                    "probability": 0.82421875
                },
                {
                    "start": 1843.29,
                    "end": 1843.99,
                    "word": " that",
                    "probability": 0.8662109375
                },
                {
                    "start": 1843.99,
                    "end": 1844.23,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 1844.23,
                    "end": 1844.53,
                    "word": " actuation",
                    "probability": 0.947021484375
                },
                {
                    "start": 1844.53,
                    "end": 1844.71,
                    "word": " has",
                    "probability": 0.904296875
                },
                {
                    "start": 1844.71,
                    "end": 1845.05,
                    "word": " happened",
                    "probability": 0.71142578125
                },
                {
                    "start": 1845.05,
                    "end": 1845.79,
                    "word": " and",
                    "probability": 0.32177734375
                },
                {
                    "start": 1845.79,
                    "end": 1845.91,
                    "word": " then",
                    "probability": 0.728515625
                }
            ]
        },
        {
            "id": 278,
            "seek": 186370,
            "start": 1845.91,
            "end": 1852.87,
            "text": " look at the changes in the observation and use that to retrain its policy, to optimize its policy",
            "tokens": [
                50636,
                574,
                412,
                264,
                2962,
                294,
                264,
                14816,
                293,
                764,
                300,
                281,
                1533,
                7146,
                1080,
                3897,
                11,
                281,
                19719,
                1080,
                3897,
                50988
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21686921627433212,
            "compression_ratio": 1.755639097744361,
            "no_speech_prob": 0.00075531005859375,
            "words": [
                {
                    "start": 1845.91,
                    "end": 1846.09,
                    "word": " look",
                    "probability": 0.9140625
                },
                {
                    "start": 1846.09,
                    "end": 1846.23,
                    "word": " at",
                    "probability": 0.9091796875
                },
                {
                    "start": 1846.23,
                    "end": 1846.41,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 1846.41,
                    "end": 1846.73,
                    "word": " changes",
                    "probability": 0.794921875
                },
                {
                    "start": 1846.73,
                    "end": 1846.85,
                    "word": " in",
                    "probability": 0.89794921875
                },
                {
                    "start": 1846.85,
                    "end": 1846.95,
                    "word": " the",
                    "probability": 0.80419921875
                },
                {
                    "start": 1846.95,
                    "end": 1847.41,
                    "word": " observation",
                    "probability": 0.8291015625
                },
                {
                    "start": 1847.41,
                    "end": 1847.97,
                    "word": " and",
                    "probability": 0.80029296875
                },
                {
                    "start": 1847.97,
                    "end": 1848.29,
                    "word": " use",
                    "probability": 0.80859375
                },
                {
                    "start": 1848.29,
                    "end": 1848.63,
                    "word": " that",
                    "probability": 0.87890625
                },
                {
                    "start": 1848.63,
                    "end": 1849.39,
                    "word": " to",
                    "probability": 0.89794921875
                },
                {
                    "start": 1849.39,
                    "end": 1850.19,
                    "word": " retrain",
                    "probability": 0.792236328125
                },
                {
                    "start": 1850.19,
                    "end": 1850.35,
                    "word": " its",
                    "probability": 0.5546875
                },
                {
                    "start": 1850.35,
                    "end": 1850.75,
                    "word": " policy",
                    "probability": 0.8134765625
                },
                {
                    "start": 1850.75,
                    "end": 1850.95,
                    "word": ",",
                    "probability": 0.6044921875
                },
                {
                    "start": 1850.95,
                    "end": 1851.15,
                    "word": " to",
                    "probability": 0.8994140625
                },
                {
                    "start": 1851.15,
                    "end": 1852.23,
                    "word": " optimize",
                    "probability": 0.6552734375
                },
                {
                    "start": 1852.23,
                    "end": 1852.49,
                    "word": " its",
                    "probability": 0.80029296875
                },
                {
                    "start": 1852.49,
                    "end": 1852.87,
                    "word": " policy",
                    "probability": 0.81689453125
                }
            ]
        },
        {
            "id": 279,
            "seek": 186370,
            "start": 1852.87,
            "end": 1857.35,
            "text": " based on faulty data. What are the adversarial capabilities? Well, we first look at different",
            "tokens": [
                50988,
                2361,
                322,
                2050,
                5773,
                1412,
                13,
                708,
                366,
                264,
                17641,
                44745,
                10862,
                30,
                1042,
                11,
                321,
                700,
                574,
                412,
                819,
                51212
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21686921627433212,
            "compression_ratio": 1.755639097744361,
            "no_speech_prob": 0.00075531005859375,
            "words": [
                {
                    "start": 1852.87,
                    "end": 1853.15,
                    "word": " based",
                    "probability": 0.75732421875
                },
                {
                    "start": 1853.15,
                    "end": 1853.35,
                    "word": " on",
                    "probability": 0.90673828125
                },
                {
                    "start": 1853.35,
                    "end": 1853.73,
                    "word": " faulty",
                    "probability": 0.859619140625
                },
                {
                    "start": 1853.73,
                    "end": 1854.11,
                    "word": " data",
                    "probability": 0.80029296875
                },
                {
                    "start": 1854.11,
                    "end": 1854.67,
                    "word": ".",
                    "probability": 0.99072265625
                },
                {
                    "start": 1854.81,
                    "end": 1855.05,
                    "word": " What",
                    "probability": 0.80810546875
                },
                {
                    "start": 1855.05,
                    "end": 1855.15,
                    "word": " are",
                    "probability": 0.87841796875
                },
                {
                    "start": 1855.15,
                    "end": 1855.23,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 1855.23,
                    "end": 1855.59,
                    "word": " adversarial",
                    "probability": 0.927978515625
                },
                {
                    "start": 1855.59,
                    "end": 1856.13,
                    "word": " capabilities",
                    "probability": 0.79541015625
                },
                {
                    "start": 1856.13,
                    "end": 1856.41,
                    "word": "?",
                    "probability": 0.98828125
                },
                {
                    "start": 1856.41,
                    "end": 1856.53,
                    "word": " Well",
                    "probability": 0.67919921875
                },
                {
                    "start": 1856.53,
                    "end": 1856.55,
                    "word": ",",
                    "probability": 0.95166015625
                },
                {
                    "start": 1856.55,
                    "end": 1856.65,
                    "word": " we",
                    "probability": 0.86865234375
                },
                {
                    "start": 1856.65,
                    "end": 1856.81,
                    "word": " first",
                    "probability": 0.78125
                },
                {
                    "start": 1856.81,
                    "end": 1856.93,
                    "word": " look",
                    "probability": 0.904296875
                },
                {
                    "start": 1856.93,
                    "end": 1857.05,
                    "word": " at",
                    "probability": 0.912109375
                },
                {
                    "start": 1857.05,
                    "end": 1857.35,
                    "word": " different",
                    "probability": 0.49462890625
                }
            ]
        },
        {
            "id": 280,
            "seek": 186370,
            "start": 1857.35,
            "end": 1864.53,
            "text": " attack modes. The attacker can perform a passive attack where it's only observing the target,",
            "tokens": [
                51212,
                2690,
                14068,
                13,
                440,
                35871,
                393,
                2042,
                257,
                14975,
                2690,
                689,
                309,
                311,
                787,
                22107,
                264,
                3779,
                11,
                51572
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21686921627433212,
            "compression_ratio": 1.755639097744361,
            "no_speech_prob": 0.00075531005859375,
            "words": [
                {
                    "start": 1857.35,
                    "end": 1857.71,
                    "word": " attack",
                    "probability": 0.81640625
                },
                {
                    "start": 1857.71,
                    "end": 1857.95,
                    "word": " modes",
                    "probability": 0.734375
                },
                {
                    "start": 1857.95,
                    "end": 1858.05,
                    "word": ".",
                    "probability": 0.8984375
                },
                {
                    "start": 1858.05,
                    "end": 1858.13,
                    "word": " The",
                    "probability": 0.81396484375
                },
                {
                    "start": 1858.13,
                    "end": 1858.43,
                    "word": " attacker",
                    "probability": 0.78759765625
                },
                {
                    "start": 1858.43,
                    "end": 1859.19,
                    "word": " can",
                    "probability": 0.8974609375
                },
                {
                    "start": 1859.19,
                    "end": 1859.47,
                    "word": " perform",
                    "probability": 0.7890625
                },
                {
                    "start": 1859.63,
                    "end": 1859.91,
                    "word": " a passive",
                    "probability": 0.7763671875
                },
                {
                    "start": 1859.91,
                    "end": 1860.25,
                    "word": " attack",
                    "probability": 0.84228515625
                },
                {
                    "start": 1860.25,
                    "end": 1860.45,
                    "word": " where",
                    "probability": 0.64208984375
                },
                {
                    "start": 1860.45,
                    "end": 1860.73,
                    "word": " it's",
                    "probability": 0.91259765625
                },
                {
                    "start": 1860.73,
                    "end": 1860.99,
                    "word": " only",
                    "probability": 0.85546875
                },
                {
                    "start": 1860.99,
                    "end": 1861.59,
                    "word": " observing",
                    "probability": 0.71240234375
                },
                {
                    "start": 1861.59,
                    "end": 1862.31,
                    "word": " the",
                    "probability": 0.806640625
                },
                {
                    "start": 1862.31,
                    "end": 1864.27,
                    "word": " target",
                    "probability": 0.70458984375
                },
                {
                    "start": 1864.27,
                    "end": 1864.53,
                    "word": ",",
                    "probability": 0.485595703125
                }
            ]
        },
        {
            "id": 281,
            "seek": 186370,
            "start": 1864.53,
            "end": 1869.83,
            "text": " it's not changing anything, or it can perform an active manipulation. In passive attacks,",
            "tokens": [
                51572,
                309,
                311,
                406,
                4473,
                1340,
                11,
                420,
                309,
                393,
                2042,
                364,
                4967,
                26475,
                13,
                682,
                14975,
                8122,
                11,
                51824
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21686921627433212,
            "compression_ratio": 1.755639097744361,
            "no_speech_prob": 0.00075531005859375,
            "words": [
                {
                    "start": 1864.53,
                    "end": 1864.83,
                    "word": " it's",
                    "probability": 0.948486328125
                },
                {
                    "start": 1864.83,
                    "end": 1865.03,
                    "word": " not",
                    "probability": 0.8896484375
                },
                {
                    "start": 1865.03,
                    "end": 1865.45,
                    "word": " changing",
                    "probability": 0.8046875
                },
                {
                    "start": 1865.45,
                    "end": 1865.81,
                    "word": " anything",
                    "probability": 0.86181640625
                },
                {
                    "start": 1865.81,
                    "end": 1866.11,
                    "word": ",",
                    "probability": 0.6669921875
                },
                {
                    "start": 1866.11,
                    "end": 1866.35,
                    "word": " or",
                    "probability": 0.90625
                },
                {
                    "start": 1866.35,
                    "end": 1866.45,
                    "word": " it",
                    "probability": 0.89208984375
                },
                {
                    "start": 1866.45,
                    "end": 1866.55,
                    "word": " can",
                    "probability": 0.89453125
                },
                {
                    "start": 1866.55,
                    "end": 1866.77,
                    "word": " perform",
                    "probability": 0.79833984375
                },
                {
                    "start": 1866.77,
                    "end": 1866.95,
                    "word": " an",
                    "probability": 0.90966796875
                },
                {
                    "start": 1866.95,
                    "end": 1867.31,
                    "word": " active",
                    "probability": 0.841796875
                },
                {
                    "start": 1867.31,
                    "end": 1867.93,
                    "word": " manipulation",
                    "probability": 0.92236328125
                },
                {
                    "start": 1867.93,
                    "end": 1868.49,
                    "word": ".",
                    "probability": 0.93994140625
                },
                {
                    "start": 1868.53,
                    "end": 1868.75,
                    "word": " In",
                    "probability": 0.90234375
                },
                {
                    "start": 1868.75,
                    "end": 1869.11,
                    "word": " passive",
                    "probability": 0.7685546875
                },
                {
                    "start": 1869.11,
                    "end": 1869.55,
                    "word": " attacks",
                    "probability": 0.681640625
                },
                {
                    "start": 1869.55,
                    "end": 1869.83,
                    "word": ",",
                    "probability": 0.97265625
                }
            ]
        },
        {
            "id": 282,
            "seek": 189020,
            "start": 1869.83,
            "end": 1876.77,
            "text": " the attacker can perform inverse reinforcement learning to learn about the reward function",
            "tokens": [
                50376,
                264,
                35871,
                393,
                2042,
                17340,
                29280,
                2539,
                281,
                1466,
                466,
                264,
                7782,
                2445,
                50704
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22544642160465192,
            "compression_ratio": 1.829268292682927,
            "no_speech_prob": 2.753734588623047e-05,
            "words": [
                {
                    "start": 1869.83,
                    "end": 1870.45,
                    "word": " the",
                    "probability": 0.09857177734375
                },
                {
                    "start": 1870.45,
                    "end": 1871.27,
                    "word": " attacker",
                    "probability": 0.75634765625
                },
                {
                    "start": 1871.27,
                    "end": 1871.75,
                    "word": " can",
                    "probability": 0.88330078125
                },
                {
                    "start": 1871.75,
                    "end": 1872.09,
                    "word": " perform",
                    "probability": 0.751953125
                },
                {
                    "start": 1872.09,
                    "end": 1872.43,
                    "word": " inverse",
                    "probability": 0.56787109375
                },
                {
                    "start": 1872.43,
                    "end": 1872.67,
                    "word": " reinforcement",
                    "probability": 0.80419921875
                },
                {
                    "start": 1872.67,
                    "end": 1873.17,
                    "word": " learning",
                    "probability": 0.84326171875
                },
                {
                    "start": 1873.17,
                    "end": 1873.49,
                    "word": " to",
                    "probability": 0.869140625
                },
                {
                    "start": 1873.49,
                    "end": 1874.41,
                    "word": " learn",
                    "probability": 0.89697265625
                },
                {
                    "start": 1874.41,
                    "end": 1874.93,
                    "word": " about",
                    "probability": 0.8046875
                },
                {
                    "start": 1874.93,
                    "end": 1875.87,
                    "word": " the",
                    "probability": 0.7802734375
                },
                {
                    "start": 1875.87,
                    "end": 1876.31,
                    "word": " reward",
                    "probability": 0.8896484375
                },
                {
                    "start": 1876.31,
                    "end": 1876.77,
                    "word": " function",
                    "probability": 0.8671875
                }
            ]
        },
        {
            "id": 283,
            "seek": 189020,
            "start": 1877.43,
            "end": 1883.25,
            "text": " of the agent, or later on we'll see it can perform imitation learning to steal the policy.",
            "tokens": [
                50744,
                295,
                264,
                9461,
                11,
                420,
                1780,
                322,
                321,
                603,
                536,
                309,
                393,
                2042,
                47624,
                2539,
                281,
                11009,
                264,
                3897,
                13,
                50996
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22544642160465192,
            "compression_ratio": 1.829268292682927,
            "no_speech_prob": 2.753734588623047e-05,
            "words": [
                {
                    "start": 1877.43,
                    "end": 1877.87,
                    "word": " of",
                    "probability": 0.86279296875
                },
                {
                    "start": 1877.87,
                    "end": 1878.07,
                    "word": " the",
                    "probability": 0.7998046875
                },
                {
                    "start": 1878.07,
                    "end": 1878.43,
                    "word": " agent",
                    "probability": 0.7392578125
                },
                {
                    "start": 1878.43,
                    "end": 1878.83,
                    "word": ",",
                    "probability": 0.34619140625
                },
                {
                    "start": 1878.83,
                    "end": 1879.19,
                    "word": " or",
                    "probability": 0.8916015625
                },
                {
                    "start": 1879.19,
                    "end": 1879.89,
                    "word": " later",
                    "probability": 0.5751953125
                },
                {
                    "start": 1879.89,
                    "end": 1880.03,
                    "word": " on",
                    "probability": 0.83056640625
                },
                {
                    "start": 1880.03,
                    "end": 1880.17,
                    "word": " we'll",
                    "probability": 0.53662109375
                },
                {
                    "start": 1880.17,
                    "end": 1880.35,
                    "word": " see",
                    "probability": 0.80029296875
                },
                {
                    "start": 1880.35,
                    "end": 1880.45,
                    "word": " it",
                    "probability": 0.71044921875
                },
                {
                    "start": 1880.45,
                    "end": 1880.55,
                    "word": " can",
                    "probability": 0.8837890625
                },
                {
                    "start": 1880.55,
                    "end": 1880.79,
                    "word": " perform",
                    "probability": 0.78955078125
                },
                {
                    "start": 1880.79,
                    "end": 1881.13,
                    "word": " imitation",
                    "probability": 0.919921875
                },
                {
                    "start": 1881.13,
                    "end": 1881.49,
                    "word": " learning",
                    "probability": 0.85498046875
                },
                {
                    "start": 1881.49,
                    "end": 1881.69,
                    "word": " to",
                    "probability": 0.89697265625
                },
                {
                    "start": 1881.69,
                    "end": 1881.93,
                    "word": " steal",
                    "probability": 0.83056640625
                },
                {
                    "start": 1881.93,
                    "end": 1882.11,
                    "word": " the",
                    "probability": 0.8056640625
                },
                {
                    "start": 1882.11,
                    "end": 1882.53,
                    "word": " policy",
                    "probability": 0.8134765625
                },
                {
                    "start": 1882.53,
                    "end": 1883.25,
                    "word": ".",
                    "probability": 0.931640625
                }
            ]
        },
        {
            "id": 284,
            "seek": 189020,
            "start": 1883.83,
            "end": 1890.73,
            "text": " Active measures include attacks on the actuation, observation, or the reward channel. And attacks on",
            "tokens": [
                51060,
                26635,
                8000,
                4090,
                8122,
                322,
                264,
                605,
                16073,
                11,
                14816,
                11,
                420,
                264,
                7782,
                2269,
                13,
                400,
                8122,
                322,
                51408
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22544642160465192,
            "compression_ratio": 1.829268292682927,
            "no_speech_prob": 2.753734588623047e-05,
            "words": [
                {
                    "start": 1883.83,
                    "end": 1884.19,
                    "word": " Active",
                    "probability": 0.86474609375
                },
                {
                    "start": 1884.19,
                    "end": 1884.47,
                    "word": " measures",
                    "probability": 0.76708984375
                },
                {
                    "start": 1884.47,
                    "end": 1884.99,
                    "word": " include",
                    "probability": 0.7783203125
                },
                {
                    "start": 1884.99,
                    "end": 1885.73,
                    "word": " attacks",
                    "probability": 0.68994140625
                },
                {
                    "start": 1885.73,
                    "end": 1885.99,
                    "word": " on",
                    "probability": 0.9033203125
                },
                {
                    "start": 1885.99,
                    "end": 1886.09,
                    "word": " the",
                    "probability": 0.798828125
                },
                {
                    "start": 1886.09,
                    "end": 1886.57,
                    "word": " actuation",
                    "probability": 0.923583984375
                },
                {
                    "start": 1886.57,
                    "end": 1887.15,
                    "word": ",",
                    "probability": 0.83154296875
                },
                {
                    "start": 1887.15,
                    "end": 1887.65,
                    "word": " observation",
                    "probability": 0.79833984375
                },
                {
                    "start": 1887.65,
                    "end": 1887.89,
                    "word": ",",
                    "probability": 0.625
                },
                {
                    "start": 1887.89,
                    "end": 1887.99,
                    "word": " or",
                    "probability": 0.86376953125
                },
                {
                    "start": 1887.99,
                    "end": 1888.09,
                    "word": " the",
                    "probability": 0.75732421875
                },
                {
                    "start": 1888.09,
                    "end": 1888.39,
                    "word": " reward",
                    "probability": 0.8359375
                },
                {
                    "start": 1888.39,
                    "end": 1888.79,
                    "word": " channel",
                    "probability": 0.74658203125
                },
                {
                    "start": 1888.79,
                    "end": 1889.41,
                    "word": ".",
                    "probability": 0.662109375
                },
                {
                    "start": 1889.41,
                    "end": 1889.79,
                    "word": " And",
                    "probability": 0.55859375
                },
                {
                    "start": 1889.79,
                    "end": 1890.43,
                    "word": " attacks",
                    "probability": 0.69091796875
                },
                {
                    "start": 1890.43,
                    "end": 1890.73,
                    "word": " on",
                    "probability": 0.90087890625
                }
            ]
        },
        {
            "id": 285,
            "seek": 189020,
            "start": 1890.73,
            "end": 1896.33,
            "text": " observation can be targeting their representation model, how the agent sees the environment,",
            "tokens": [
                51408,
                14816,
                393,
                312,
                17918,
                641,
                10290,
                2316,
                11,
                577,
                264,
                9461,
                8194,
                264,
                2823,
                11,
                51684
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22544642160465192,
            "compression_ratio": 1.829268292682927,
            "no_speech_prob": 2.753734588623047e-05,
            "words": [
                {
                    "start": 1890.73,
                    "end": 1891.35,
                    "word": " observation",
                    "probability": 0.80078125
                },
                {
                    "start": 1891.35,
                    "end": 1892.15,
                    "word": " can",
                    "probability": 0.8837890625
                },
                {
                    "start": 1892.15,
                    "end": 1892.33,
                    "word": " be",
                    "probability": 0.9111328125
                },
                {
                    "start": 1892.33,
                    "end": 1892.65,
                    "word": " targeting",
                    "probability": 0.822265625
                },
                {
                    "start": 1892.65,
                    "end": 1892.89,
                    "word": " their",
                    "probability": 0.455078125
                },
                {
                    "start": 1892.89,
                    "end": 1893.35,
                    "word": " representation",
                    "probability": 0.765625
                },
                {
                    "start": 1893.35,
                    "end": 1893.91,
                    "word": " model",
                    "probability": 0.794921875
                },
                {
                    "start": 1893.91,
                    "end": 1894.05,
                    "word": ",",
                    "probability": 0.74609375
                },
                {
                    "start": 1894.05,
                    "end": 1894.19,
                    "word": " how",
                    "probability": 0.89453125
                },
                {
                    "start": 1894.19,
                    "end": 1894.33,
                    "word": " the",
                    "probability": 0.81494140625
                },
                {
                    "start": 1894.33,
                    "end": 1894.51,
                    "word": " agent",
                    "probability": 0.619140625
                },
                {
                    "start": 1894.51,
                    "end": 1894.99,
                    "word": " sees",
                    "probability": 0.87548828125
                },
                {
                    "start": 1894.99,
                    "end": 1895.65,
                    "word": " the",
                    "probability": 0.8037109375
                },
                {
                    "start": 1895.65,
                    "end": 1896.03,
                    "word": " environment",
                    "probability": 0.7529296875
                },
                {
                    "start": 1896.03,
                    "end": 1896.33,
                    "word": ",",
                    "probability": 0.60302734375
                }
            ]
        },
        {
            "id": 286,
            "seek": 191960,
            "start": 1896.33,
            "end": 1901.15,
            "text": " or perturbing the transition dynamics, how the agent sees the changes in the environment.",
            "tokens": [
                50364,
                420,
                13269,
                374,
                4324,
                264,
                6034,
                15679,
                11,
                577,
                264,
                9461,
                8194,
                264,
                2962,
                294,
                264,
                2823,
                13,
                50584
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22150734733132754,
            "compression_ratio": 1.5560344827586208,
            "no_speech_prob": 0.0010166168212890625,
            "words": [
                {
                    "start": 1896.33,
                    "end": 1896.53,
                    "word": " or",
                    "probability": 0.591796875
                },
                {
                    "start": 1896.53,
                    "end": 1896.99,
                    "word": " perturbing",
                    "probability": 0.84423828125
                },
                {
                    "start": 1896.99,
                    "end": 1897.13,
                    "word": " the",
                    "probability": 0.79931640625
                },
                {
                    "start": 1897.13,
                    "end": 1897.49,
                    "word": " transition",
                    "probability": 0.92333984375
                },
                {
                    "start": 1897.49,
                    "end": 1898.03,
                    "word": " dynamics",
                    "probability": 0.7548828125
                },
                {
                    "start": 1898.03,
                    "end": 1898.21,
                    "word": ",",
                    "probability": 0.57470703125
                },
                {
                    "start": 1898.21,
                    "end": 1898.35,
                    "word": " how",
                    "probability": 0.8720703125
                },
                {
                    "start": 1898.35,
                    "end": 1898.47,
                    "word": " the",
                    "probability": 0.7763671875
                },
                {
                    "start": 1898.47,
                    "end": 1898.67,
                    "word": " agent",
                    "probability": 0.1182861328125
                },
                {
                    "start": 1898.67,
                    "end": 1899.01,
                    "word": " sees",
                    "probability": 0.86669921875
                },
                {
                    "start": 1899.01,
                    "end": 1899.25,
                    "word": " the",
                    "probability": 0.79296875
                },
                {
                    "start": 1899.25,
                    "end": 1899.73,
                    "word": " changes",
                    "probability": 0.796875
                },
                {
                    "start": 1899.73,
                    "end": 1900.07,
                    "word": " in",
                    "probability": 0.9033203125
                },
                {
                    "start": 1900.07,
                    "end": 1900.19,
                    "word": " the",
                    "probability": 0.80712890625
                },
                {
                    "start": 1900.19,
                    "end": 1900.55,
                    "word": " environment",
                    "probability": 0.70849609375
                },
                {
                    "start": 1900.55,
                    "end": 1901.15,
                    "word": ".",
                    "probability": 0.92431640625
                }
            ]
        },
        {
            "id": 287,
            "seek": 191960,
            "start": 1902.05,
            "end": 1906.29,
            "text": " Okay, going back to our initial proof of concept, remember our original goal was",
            "tokens": [
                50648,
                1033,
                11,
                516,
                646,
                281,
                527,
                5883,
                8177,
                295,
                3410,
                11,
                1604,
                527,
                3380,
                3387,
                390,
                50860
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22150734733132754,
            "compression_ratio": 1.5560344827586208,
            "no_speech_prob": 0.0010166168212890625,
            "words": [
                {
                    "start": 1902.05,
                    "end": 1902.31,
                    "word": " Okay",
                    "probability": 0.405029296875
                },
                {
                    "start": 1902.31,
                    "end": 1902.43,
                    "word": ",",
                    "probability": 0.78857421875
                },
                {
                    "start": 1902.43,
                    "end": 1902.71,
                    "word": " going",
                    "probability": 0.84033203125
                },
                {
                    "start": 1902.71,
                    "end": 1902.95,
                    "word": " back",
                    "probability": 0.81494140625
                },
                {
                    "start": 1902.95,
                    "end": 1903.17,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 1903.17,
                    "end": 1903.43,
                    "word": " our",
                    "probability": 0.82421875
                },
                {
                    "start": 1903.43,
                    "end": 1903.87,
                    "word": " initial",
                    "probability": 0.84423828125
                },
                {
                    "start": 1903.87,
                    "end": 1904.11,
                    "word": " proof",
                    "probability": 0.845703125
                },
                {
                    "start": 1904.11,
                    "end": 1904.25,
                    "word": " of",
                    "probability": 0.84716796875
                },
                {
                    "start": 1904.25,
                    "end": 1904.63,
                    "word": " concept",
                    "probability": 0.7939453125
                },
                {
                    "start": 1904.63,
                    "end": 1904.99,
                    "word": ",",
                    "probability": 0.638671875
                },
                {
                    "start": 1904.99,
                    "end": 1905.25,
                    "word": " remember",
                    "probability": 0.7861328125
                },
                {
                    "start": 1905.25,
                    "end": 1905.61,
                    "word": " our",
                    "probability": 0.544921875
                },
                {
                    "start": 1905.61,
                    "end": 1905.87,
                    "word": " original",
                    "probability": 0.8037109375
                },
                {
                    "start": 1905.87,
                    "end": 1906.13,
                    "word": " goal",
                    "probability": 0.9091796875
                },
                {
                    "start": 1906.13,
                    "end": 1906.29,
                    "word": " was",
                    "probability": 0.9013671875
                }
            ]
        },
        {
            "id": 288,
            "seek": 191960,
            "start": 1906.29,
            "end": 1919.59,
            "text": " to perform a black box attack. So, to achieve this objective, we introduce an approach based",
            "tokens": [
                50860,
                281,
                2042,
                257,
                2211,
                2424,
                2690,
                13,
                407,
                11,
                281,
                4584,
                341,
                10024,
                11,
                321,
                5366,
                364,
                3109,
                2361,
                51528
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22150734733132754,
            "compression_ratio": 1.5560344827586208,
            "no_speech_prob": 0.0010166168212890625,
            "words": [
                {
                    "start": 1906.29,
                    "end": 1906.39,
                    "word": " to",
                    "probability": 0.90185546875
                },
                {
                    "start": 1906.39,
                    "end": 1906.63,
                    "word": " perform",
                    "probability": 0.80029296875
                },
                {
                    "start": 1906.77,
                    "end": 1906.99,
                    "word": " a black",
                    "probability": 0.70947265625
                },
                {
                    "start": 1906.99,
                    "end": 1907.25,
                    "word": " box",
                    "probability": 0.64794921875
                },
                {
                    "start": 1907.25,
                    "end": 1907.65,
                    "word": " attack",
                    "probability": 0.83349609375
                },
                {
                    "start": 1907.65,
                    "end": 1908.25,
                    "word": ".",
                    "probability": 0.94482421875
                },
                {
                    "start": 1910.65,
                    "end": 1910.83,
                    "word": " So",
                    "probability": 0.78515625
                },
                {
                    "start": 1910.83,
                    "end": 1911.59,
                    "word": ",",
                    "probability": 0.40673828125
                },
                {
                    "start": 1911.59,
                    "end": 1912.03,
                    "word": " to",
                    "probability": 0.90576171875
                },
                {
                    "start": 1912.03,
                    "end": 1915.09,
                    "word": " achieve",
                    "probability": 0.72705078125
                },
                {
                    "start": 1915.09,
                    "end": 1915.39,
                    "word": " this",
                    "probability": 0.88134765625
                },
                {
                    "start": 1915.39,
                    "end": 1915.93,
                    "word": " objective",
                    "probability": 0.8291015625
                },
                {
                    "start": 1915.93,
                    "end": 1916.91,
                    "word": ",",
                    "probability": 0.892578125
                },
                {
                    "start": 1916.91,
                    "end": 1917.89,
                    "word": " we",
                    "probability": 0.89599609375
                },
                {
                    "start": 1917.89,
                    "end": 1918.49,
                    "word": " introduce",
                    "probability": 0.56298828125
                },
                {
                    "start": 1918.49,
                    "end": 1918.91,
                    "word": " an",
                    "probability": 0.89892578125
                },
                {
                    "start": 1918.91,
                    "end": 1919.23,
                    "word": " approach",
                    "probability": 0.7861328125
                },
                {
                    "start": 1919.23,
                    "end": 1919.59,
                    "word": " based",
                    "probability": 0.802734375
                }
            ]
        },
        {
            "id": 289,
            "seek": 191960,
            "start": 1919.59,
            "end": 1925.73,
            "text": " on the transferability of adversarial examples. So, we create a second DQN, the adversary creates",
            "tokens": [
                51528,
                322,
                264,
                5003,
                2310,
                295,
                17641,
                44745,
                5110,
                13,
                407,
                11,
                321,
                1884,
                257,
                1150,
                413,
                48,
                45,
                11,
                264,
                48222,
                7829,
                51832
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22150734733132754,
            "compression_ratio": 1.5560344827586208,
            "no_speech_prob": 0.0010166168212890625,
            "words": [
                {
                    "start": 1919.59,
                    "end": 1919.83,
                    "word": " on",
                    "probability": 0.90673828125
                },
                {
                    "start": 1919.83,
                    "end": 1919.95,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 1919.95,
                    "end": 1920.55,
                    "word": " transferability",
                    "probability": 0.9423828125
                },
                {
                    "start": 1920.55,
                    "end": 1920.71,
                    "word": " of",
                    "probability": 0.88720703125
                },
                {
                    "start": 1920.71,
                    "end": 1921.11,
                    "word": " adversarial",
                    "probability": 0.912353515625
                },
                {
                    "start": 1921.11,
                    "end": 1921.63,
                    "word": " examples",
                    "probability": 0.775390625
                },
                {
                    "start": 1921.63,
                    "end": 1921.97,
                    "word": ".",
                    "probability": 0.89697265625
                },
                {
                    "start": 1921.97,
                    "end": 1922.15,
                    "word": " So",
                    "probability": 0.892578125
                },
                {
                    "start": 1922.15,
                    "end": 1922.71,
                    "word": ",",
                    "probability": 0.6953125
                },
                {
                    "start": 1922.71,
                    "end": 1922.97,
                    "word": " we",
                    "probability": 0.8896484375
                },
                {
                    "start": 1922.97,
                    "end": 1923.33,
                    "word": " create",
                    "probability": 0.865234375
                },
                {
                    "start": 1923.51,
                    "end": 1923.91,
                    "word": " a second",
                    "probability": 0.81640625
                },
                {
                    "start": 1923.91,
                    "end": 1924.61,
                    "word": " DQN",
                    "probability": 0.9459635416666666
                },
                {
                    "start": 1924.61,
                    "end": 1924.93,
                    "word": ",",
                    "probability": 0.7939453125
                },
                {
                    "start": 1924.93,
                    "end": 1925.11,
                    "word": " the",
                    "probability": 0.80908203125
                },
                {
                    "start": 1925.11,
                    "end": 1925.41,
                    "word": " adversary",
                    "probability": 0.81201171875
                },
                {
                    "start": 1925.41,
                    "end": 1925.73,
                    "word": " creates",
                    "probability": 0.75830078125
                }
            ]
        },
        {
            "id": 290,
            "seek": 194960,
            "start": 1925.83,
            "end": 1929.63,
            "text": " a second DQN with similar architecture, but different initial parameters. When I say similar,",
            "tokens": [
                50364,
                257,
                1150,
                413,
                48,
                45,
                365,
                2531,
                9482,
                11,
                457,
                819,
                5883,
                9834,
                13,
                1133,
                286,
                584,
                2531,
                11,
                50556
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2190504867870074,
            "compression_ratio": 1.7282608695652173,
            "no_speech_prob": 0.0008692741394042969,
            "words": [
                {
                    "start": 1925.83,
                    "end": 1926.03,
                    "word": " a second",
                    "probability": 0.75927734375
                },
                {
                    "start": 1926.03,
                    "end": 1926.37,
                    "word": " DQN",
                    "probability": 0.3015543619791667
                },
                {
                    "start": 1926.37,
                    "end": 1926.51,
                    "word": " with",
                    "probability": 0.7255859375
                },
                {
                    "start": 1926.51,
                    "end": 1926.71,
                    "word": " similar",
                    "probability": 0.8662109375
                },
                {
                    "start": 1926.71,
                    "end": 1927.21,
                    "word": " architecture",
                    "probability": 0.8193359375
                },
                {
                    "start": 1927.21,
                    "end": 1927.47,
                    "word": ",",
                    "probability": 0.50244140625
                },
                {
                    "start": 1927.47,
                    "end": 1927.65,
                    "word": " but",
                    "probability": 0.8857421875
                },
                {
                    "start": 1927.65,
                    "end": 1928.15,
                    "word": " different",
                    "probability": 0.68994140625
                },
                {
                    "start": 1928.15,
                    "end": 1928.47,
                    "word": " initial",
                    "probability": 0.830078125
                },
                {
                    "start": 1928.47,
                    "end": 1928.93,
                    "word": " parameters",
                    "probability": 0.88818359375
                },
                {
                    "start": 1928.93,
                    "end": 1929.05,
                    "word": ".",
                    "probability": 0.72998046875
                },
                {
                    "start": 1929.05,
                    "end": 1929.07,
                    "word": " When",
                    "probability": 0.595703125
                },
                {
                    "start": 1929.07,
                    "end": 1929.15,
                    "word": " I",
                    "probability": 0.990234375
                },
                {
                    "start": 1929.15,
                    "end": 1929.25,
                    "word": " say",
                    "probability": 0.8916015625
                },
                {
                    "start": 1929.25,
                    "end": 1929.55,
                    "word": " similar",
                    "probability": 0.84765625
                },
                {
                    "start": 1929.55,
                    "end": 1929.63,
                    "word": ",",
                    "probability": 0.8876953125
                }
            ]
        },
        {
            "id": 291,
            "seek": 194960,
            "start": 1929.63,
            "end": 1932.81,
            "text": " it doesn't necessarily have to be a match, it just needs to be a convolutional neural network",
            "tokens": [
                50556,
                309,
                1177,
                380,
                4725,
                362,
                281,
                312,
                257,
                2995,
                11,
                309,
                445,
                2203,
                281,
                312,
                257,
                45216,
                304,
                18161,
                3209,
                50740
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2190504867870074,
            "compression_ratio": 1.7282608695652173,
            "no_speech_prob": 0.0008692741394042969,
            "words": [
                {
                    "start": 1929.63,
                    "end": 1929.71,
                    "word": " it",
                    "probability": 0.9013671875
                },
                {
                    "start": 1929.71,
                    "end": 1929.87,
                    "word": " doesn't",
                    "probability": 0.93359375
                },
                {
                    "start": 1929.87,
                    "end": 1930.19,
                    "word": " necessarily",
                    "probability": 0.76416015625
                },
                {
                    "start": 1930.19,
                    "end": 1930.35,
                    "word": " have",
                    "probability": 0.88427734375
                },
                {
                    "start": 1930.35,
                    "end": 1930.41,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 1930.41,
                    "end": 1930.51,
                    "word": " be",
                    "probability": 0.90283203125
                },
                {
                    "start": 1930.61,
                    "end": 1930.77,
                    "word": " a match",
                    "probability": 0.6337890625
                },
                {
                    "start": 1930.77,
                    "end": 1930.81,
                    "word": ",",
                    "probability": 0.63330078125
                },
                {
                    "start": 1930.81,
                    "end": 1930.91,
                    "word": " it",
                    "probability": 0.86669921875
                },
                {
                    "start": 1930.91,
                    "end": 1931.07,
                    "word": " just",
                    "probability": 0.873046875
                },
                {
                    "start": 1931.07,
                    "end": 1931.29,
                    "word": " needs",
                    "probability": 0.7490234375
                },
                {
                    "start": 1931.29,
                    "end": 1931.41,
                    "word": " to",
                    "probability": 0.9091796875
                },
                {
                    "start": 1931.41,
                    "end": 1931.61,
                    "word": " be",
                    "probability": 0.90380859375
                },
                {
                    "start": 1931.77,
                    "end": 1932.25,
                    "word": " a convolutional",
                    "probability": 0.894775390625
                },
                {
                    "start": 1932.25,
                    "end": 1932.43,
                    "word": " neural",
                    "probability": 0.92578125
                },
                {
                    "start": 1932.43,
                    "end": 1932.81,
                    "word": " network",
                    "probability": 0.82080078125
                }
            ]
        },
        {
            "id": 292,
            "seek": 194960,
            "start": 1932.81,
            "end": 1938.15,
            "text": " with some functional approximation techniques, but it doesn't need to have the same parameters",
            "tokens": [
                50740,
                365,
                512,
                11745,
                28023,
                7512,
                11,
                457,
                309,
                1177,
                380,
                643,
                281,
                362,
                264,
                912,
                9834,
                50988
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2190504867870074,
            "compression_ratio": 1.7282608695652173,
            "no_speech_prob": 0.0008692741394042969,
            "words": [
                {
                    "start": 1932.81,
                    "end": 1933.61,
                    "word": " with",
                    "probability": 0.736328125
                },
                {
                    "start": 1933.61,
                    "end": 1934.23,
                    "word": " some",
                    "probability": 0.81494140625
                },
                {
                    "start": 1934.23,
                    "end": 1935.05,
                    "word": " functional",
                    "probability": 0.345947265625
                },
                {
                    "start": 1935.05,
                    "end": 1935.47,
                    "word": " approximation",
                    "probability": 0.76806640625
                },
                {
                    "start": 1935.47,
                    "end": 1936.07,
                    "word": " techniques",
                    "probability": 0.83447265625
                },
                {
                    "start": 1936.07,
                    "end": 1936.07,
                    "word": ",",
                    "probability": 0.7109375
                },
                {
                    "start": 1936.07,
                    "end": 1936.49,
                    "word": " but",
                    "probability": 0.8974609375
                },
                {
                    "start": 1936.49,
                    "end": 1936.63,
                    "word": " it",
                    "probability": 0.9052734375
                },
                {
                    "start": 1936.63,
                    "end": 1936.93,
                    "word": " doesn't",
                    "probability": 0.93505859375
                },
                {
                    "start": 1936.93,
                    "end": 1937.31,
                    "word": " need",
                    "probability": 0.81640625
                },
                {
                    "start": 1937.31,
                    "end": 1937.41,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 1937.41,
                    "end": 1937.53,
                    "word": " have",
                    "probability": 0.88623046875
                },
                {
                    "start": 1937.53,
                    "end": 1937.65,
                    "word": " the",
                    "probability": 0.80859375
                },
                {
                    "start": 1937.65,
                    "end": 1937.81,
                    "word": " same",
                    "probability": 0.8125
                },
                {
                    "start": 1937.81,
                    "end": 1938.15,
                    "word": " parameters",
                    "probability": 0.89306640625
                }
            ]
        },
        {
            "id": 293,
            "seek": 194960,
            "start": 1938.15,
            "end": 1947.47,
            "text": " or the exact same architecture. And it trains that agent, that model, on the same environment.",
            "tokens": [
                50988,
                420,
                264,
                1900,
                912,
                9482,
                13,
                400,
                309,
                16329,
                300,
                9461,
                11,
                300,
                2316,
                11,
                322,
                264,
                912,
                2823,
                13,
                51444
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2190504867870074,
            "compression_ratio": 1.7282608695652173,
            "no_speech_prob": 0.0008692741394042969,
            "words": [
                {
                    "start": 1938.15,
                    "end": 1938.39,
                    "word": " or",
                    "probability": 0.8681640625
                },
                {
                    "start": 1938.39,
                    "end": 1938.57,
                    "word": " the",
                    "probability": 0.8076171875
                },
                {
                    "start": 1938.57,
                    "end": 1938.75,
                    "word": " exact",
                    "probability": 0.87646484375
                },
                {
                    "start": 1938.75,
                    "end": 1938.97,
                    "word": " same",
                    "probability": 0.80712890625
                },
                {
                    "start": 1938.97,
                    "end": 1939.35,
                    "word": " architecture",
                    "probability": 0.84765625
                },
                {
                    "start": 1939.35,
                    "end": 1939.83,
                    "word": ".",
                    "probability": 0.90625
                },
                {
                    "start": 1940.65,
                    "end": 1941.13,
                    "word": " And",
                    "probability": 0.7998046875
                },
                {
                    "start": 1941.13,
                    "end": 1941.79,
                    "word": " it",
                    "probability": 0.8486328125
                },
                {
                    "start": 1941.79,
                    "end": 1942.21,
                    "word": " trains",
                    "probability": 0.75732421875
                },
                {
                    "start": 1942.21,
                    "end": 1942.57,
                    "word": " that",
                    "probability": 0.88037109375
                },
                {
                    "start": 1942.57,
                    "end": 1943.35,
                    "word": " agent",
                    "probability": 0.8408203125
                },
                {
                    "start": 1943.35,
                    "end": 1943.61,
                    "word": ",",
                    "probability": 0.76904296875
                },
                {
                    "start": 1943.61,
                    "end": 1943.77,
                    "word": " that",
                    "probability": 0.8798828125
                },
                {
                    "start": 1943.77,
                    "end": 1946.11,
                    "word": " model",
                    "probability": 0.7890625
                },
                {
                    "start": 1946.11,
                    "end": 1946.45,
                    "word": ",",
                    "probability": 0.275146484375
                },
                {
                    "start": 1946.45,
                    "end": 1946.65,
                    "word": " on",
                    "probability": 0.9052734375
                },
                {
                    "start": 1946.65,
                    "end": 1946.83,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 1946.83,
                    "end": 1947.01,
                    "word": " same",
                    "probability": 0.8125
                },
                {
                    "start": 1947.01,
                    "end": 1947.37,
                    "word": " environment",
                    "probability": 0.7685546875
                },
                {
                    "start": 1947.37,
                    "end": 1947.47,
                    "word": ".",
                    "probability": 0.6513671875
                }
            ]
        },
        {
            "id": 294,
            "seek": 194960,
            "start": 1947.47,
            "end": 1954.11,
            "text": " The assumption is the adversary has access to that environment, and then uses the knowledge of that",
            "tokens": [
                51444,
                440,
                15302,
                307,
                264,
                48222,
                575,
                2105,
                281,
                300,
                2823,
                11,
                293,
                550,
                4960,
                264,
                3601,
                295,
                300,
                51776
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2190504867870074,
            "compression_ratio": 1.7282608695652173,
            "no_speech_prob": 0.0008692741394042969,
            "words": [
                {
                    "start": 1947.47,
                    "end": 1947.55,
                    "word": " The",
                    "probability": 0.77734375
                },
                {
                    "start": 1947.55,
                    "end": 1947.73,
                    "word": " assumption",
                    "probability": 0.931640625
                },
                {
                    "start": 1947.73,
                    "end": 1948.07,
                    "word": " is",
                    "probability": 0.908203125
                },
                {
                    "start": 1948.07,
                    "end": 1948.81,
                    "word": " the",
                    "probability": 0.7431640625
                },
                {
                    "start": 1948.81,
                    "end": 1949.43,
                    "word": " adversary",
                    "probability": 0.8203125
                },
                {
                    "start": 1949.43,
                    "end": 1949.65,
                    "word": " has",
                    "probability": 0.904296875
                },
                {
                    "start": 1949.65,
                    "end": 1949.95,
                    "word": " access",
                    "probability": 0.791015625
                },
                {
                    "start": 1949.95,
                    "end": 1950.09,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 1950.09,
                    "end": 1950.25,
                    "word": " that",
                    "probability": 0.8798828125
                },
                {
                    "start": 1950.25,
                    "end": 1950.67,
                    "word": " environment",
                    "probability": 0.76806640625
                },
                {
                    "start": 1950.67,
                    "end": 1951.29,
                    "word": ",",
                    "probability": 0.393798828125
                },
                {
                    "start": 1951.29,
                    "end": 1951.89,
                    "word": " and",
                    "probability": 0.8955078125
                },
                {
                    "start": 1951.89,
                    "end": 1952.17,
                    "word": " then",
                    "probability": 0.72998046875
                },
                {
                    "start": 1952.17,
                    "end": 1952.73,
                    "word": " uses",
                    "probability": 0.81201171875
                },
                {
                    "start": 1952.73,
                    "end": 1953.01,
                    "word": " the",
                    "probability": 0.8173828125
                },
                {
                    "start": 1953.01,
                    "end": 1953.37,
                    "word": " knowledge",
                    "probability": 0.8330078125
                },
                {
                    "start": 1953.37,
                    "end": 1953.71,
                    "word": " of",
                    "probability": 0.90185546875
                },
                {
                    "start": 1953.71,
                    "end": 1954.11,
                    "word": " that",
                    "probability": 0.8828125
                }
            ]
        },
        {
            "id": 295,
            "seek": 197624,
            "start": 1955.75,
            "end": 1961.79,
            "text": " architecture and the trained replica policy. It crafts adversarial examples the same way it did",
            "tokens": [
                50364,
                9482,
                293,
                264,
                8895,
                35456,
                3897,
                13,
                467,
                27831,
                17641,
                44745,
                5110,
                264,
                912,
                636,
                309,
                630,
                50664
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22436797350979923,
            "compression_ratio": 1.6297872340425532,
            "no_speech_prob": 0.00045108795166015625,
            "words": [
                {
                    "start": 1955.75,
                    "end": 1956.31,
                    "word": " architecture",
                    "probability": 0.369140625
                },
                {
                    "start": 1956.31,
                    "end": 1956.49,
                    "word": " and",
                    "probability": 0.765625
                },
                {
                    "start": 1956.49,
                    "end": 1956.65,
                    "word": " the",
                    "probability": 0.42041015625
                },
                {
                    "start": 1956.65,
                    "end": 1957.01,
                    "word": " trained",
                    "probability": 0.4931640625
                },
                {
                    "start": 1957.01,
                    "end": 1957.79,
                    "word": " replica",
                    "probability": 0.82421875
                },
                {
                    "start": 1957.79,
                    "end": 1958.35,
                    "word": " policy",
                    "probability": 0.771484375
                },
                {
                    "start": 1958.35,
                    "end": 1958.91,
                    "word": ".",
                    "probability": 0.58642578125
                },
                {
                    "start": 1959.21,
                    "end": 1959.53,
                    "word": " It",
                    "probability": 0.89306640625
                },
                {
                    "start": 1959.53,
                    "end": 1960.07,
                    "word": " crafts",
                    "probability": 0.6953125
                },
                {
                    "start": 1960.07,
                    "end": 1960.55,
                    "word": " adversarial",
                    "probability": 0.934814453125
                },
                {
                    "start": 1960.55,
                    "end": 1960.93,
                    "word": " examples",
                    "probability": 0.79541015625
                },
                {
                    "start": 1960.93,
                    "end": 1961.11,
                    "word": " the",
                    "probability": 0.78515625
                },
                {
                    "start": 1961.11,
                    "end": 1961.25,
                    "word": " same",
                    "probability": 0.81396484375
                },
                {
                    "start": 1961.25,
                    "end": 1961.45,
                    "word": " way",
                    "probability": 0.88134765625
                },
                {
                    "start": 1961.45,
                    "end": 1961.59,
                    "word": " it",
                    "probability": 0.8173828125
                },
                {
                    "start": 1961.59,
                    "end": 1961.79,
                    "word": " did",
                    "probability": 0.89501953125
                }
            ]
        },
        {
            "id": 296,
            "seek": 197624,
            "start": 1961.79,
            "end": 1966.25,
            "text": " for the white box case, and we know that many of those adversarial examples can transfer",
            "tokens": [
                50664,
                337,
                264,
                2418,
                2424,
                1389,
                11,
                293,
                321,
                458,
                300,
                867,
                295,
                729,
                17641,
                44745,
                5110,
                393,
                5003,
                50912
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22436797350979923,
            "compression_ratio": 1.6297872340425532,
            "no_speech_prob": 0.00045108795166015625,
            "words": [
                {
                    "start": 1961.79,
                    "end": 1962.05,
                    "word": " for",
                    "probability": 0.88525390625
                },
                {
                    "start": 1962.05,
                    "end": 1962.23,
                    "word": " the",
                    "probability": 0.8173828125
                },
                {
                    "start": 1962.23,
                    "end": 1962.45,
                    "word": " white",
                    "probability": 0.7578125
                },
                {
                    "start": 1962.45,
                    "end": 1962.67,
                    "word": " box",
                    "probability": 0.5791015625
                },
                {
                    "start": 1962.67,
                    "end": 1962.99,
                    "word": " case",
                    "probability": 0.8291015625
                },
                {
                    "start": 1962.99,
                    "end": 1963.19,
                    "word": ",",
                    "probability": 0.348388671875
                },
                {
                    "start": 1963.19,
                    "end": 1963.31,
                    "word": " and",
                    "probability": 0.89404296875
                },
                {
                    "start": 1963.31,
                    "end": 1963.43,
                    "word": " we",
                    "probability": 0.9013671875
                },
                {
                    "start": 1963.43,
                    "end": 1963.61,
                    "word": " know",
                    "probability": 0.81640625
                },
                {
                    "start": 1963.61,
                    "end": 1963.87,
                    "word": " that",
                    "probability": 0.8740234375
                },
                {
                    "start": 1963.87,
                    "end": 1964.27,
                    "word": " many",
                    "probability": 0.81005859375
                },
                {
                    "start": 1964.27,
                    "end": 1964.39,
                    "word": " of",
                    "probability": 0.896484375
                },
                {
                    "start": 1964.39,
                    "end": 1964.53,
                    "word": " those",
                    "probability": 0.8330078125
                },
                {
                    "start": 1964.53,
                    "end": 1964.87,
                    "word": " adversarial",
                    "probability": 0.9296875
                },
                {
                    "start": 1964.87,
                    "end": 1965.25,
                    "word": " examples",
                    "probability": 0.79345703125
                },
                {
                    "start": 1965.25,
                    "end": 1965.73,
                    "word": " can",
                    "probability": 0.8916015625
                },
                {
                    "start": 1965.73,
                    "end": 1966.25,
                    "word": " transfer",
                    "probability": 0.83935546875
                }
            ]
        },
        {
            "id": 297,
            "seek": 197624,
            "start": 1966.25,
            "end": 1975.93,
            "text": " to similar models, train on similar data, which also you can see applies to DQN policies as well.",
            "tokens": [
                50912,
                281,
                2531,
                5245,
                11,
                3847,
                322,
                2531,
                1412,
                11,
                597,
                611,
                291,
                393,
                536,
                13165,
                281,
                413,
                48,
                45,
                7657,
                382,
                731,
                13,
                51364
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22436797350979923,
            "compression_ratio": 1.6297872340425532,
            "no_speech_prob": 0.00045108795166015625,
            "words": [
                {
                    "start": 1966.25,
                    "end": 1966.97,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 1966.97,
                    "end": 1967.49,
                    "word": " similar",
                    "probability": 0.87158203125
                },
                {
                    "start": 1967.49,
                    "end": 1968.85,
                    "word": " models",
                    "probability": 0.78564453125
                },
                {
                    "start": 1968.85,
                    "end": 1969.39,
                    "word": ",",
                    "probability": 0.369384765625
                },
                {
                    "start": 1969.39,
                    "end": 1969.57,
                    "word": " train",
                    "probability": 0.482177734375
                },
                {
                    "start": 1969.57,
                    "end": 1969.83,
                    "word": " on",
                    "probability": 0.88037109375
                },
                {
                    "start": 1969.83,
                    "end": 1970.23,
                    "word": " similar",
                    "probability": 0.91650390625
                },
                {
                    "start": 1970.23,
                    "end": 1970.77,
                    "word": " data",
                    "probability": 0.80712890625
                },
                {
                    "start": 1970.77,
                    "end": 1971.37,
                    "word": ",",
                    "probability": 0.84716796875
                },
                {
                    "start": 1971.37,
                    "end": 1971.97,
                    "word": " which",
                    "probability": 0.9111328125
                },
                {
                    "start": 1971.97,
                    "end": 1972.25,
                    "word": " also",
                    "probability": 0.7939453125
                },
                {
                    "start": 1972.25,
                    "end": 1972.55,
                    "word": " you",
                    "probability": 0.63525390625
                },
                {
                    "start": 1972.55,
                    "end": 1972.65,
                    "word": " can",
                    "probability": 0.896484375
                },
                {
                    "start": 1972.65,
                    "end": 1972.81,
                    "word": " see",
                    "probability": 0.818359375
                },
                {
                    "start": 1972.81,
                    "end": 1973.25,
                    "word": " applies",
                    "probability": 0.7626953125
                },
                {
                    "start": 1973.25,
                    "end": 1974.11,
                    "word": " to",
                    "probability": 0.90478515625
                },
                {
                    "start": 1974.11,
                    "end": 1975.05,
                    "word": " DQN",
                    "probability": 0.70263671875
                },
                {
                    "start": 1975.05,
                    "end": 1975.43,
                    "word": " policies",
                    "probability": 0.8427734375
                },
                {
                    "start": 1975.43,
                    "end": 1975.67,
                    "word": " as",
                    "probability": 0.873046875
                },
                {
                    "start": 1975.67,
                    "end": 1975.81,
                    "word": " well",
                    "probability": 0.880859375
                },
                {
                    "start": 1975.81,
                    "end": 1975.93,
                    "word": ".",
                    "probability": 0.763671875
                }
            ]
        },
        {
            "id": 298,
            "seek": 197624,
            "start": 1975.93,
            "end": 1982.37,
            "text": " And this is how we implemented a black box attack against deep RL policies. Now, what about training",
            "tokens": [
                51364,
                400,
                341,
                307,
                577,
                321,
                12270,
                257,
                2211,
                2424,
                2690,
                1970,
                2452,
                497,
                43,
                7657,
                13,
                823,
                11,
                437,
                466,
                3097,
                51692
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22436797350979923,
            "compression_ratio": 1.6297872340425532,
            "no_speech_prob": 0.00045108795166015625,
            "words": [
                {
                    "start": 1975.93,
                    "end": 1976.01,
                    "word": " And",
                    "probability": 0.75732421875
                },
                {
                    "start": 1976.01,
                    "end": 1976.13,
                    "word": " this",
                    "probability": 0.87353515625
                },
                {
                    "start": 1976.13,
                    "end": 1976.25,
                    "word": " is",
                    "probability": 0.908203125
                },
                {
                    "start": 1976.25,
                    "end": 1976.41,
                    "word": " how",
                    "probability": 0.89599609375
                },
                {
                    "start": 1976.41,
                    "end": 1976.59,
                    "word": " we",
                    "probability": 0.90673828125
                },
                {
                    "start": 1976.59,
                    "end": 1977.19,
                    "word": " implemented",
                    "probability": 0.708984375
                },
                {
                    "start": 1977.53,
                    "end": 1977.73,
                    "word": " a black",
                    "probability": 0.8232421875
                },
                {
                    "start": 1977.73,
                    "end": 1977.93,
                    "word": " box",
                    "probability": 0.716796875
                },
                {
                    "start": 1977.93,
                    "end": 1978.21,
                    "word": " attack",
                    "probability": 0.857421875
                },
                {
                    "start": 1978.21,
                    "end": 1978.71,
                    "word": " against",
                    "probability": 0.865234375
                },
                {
                    "start": 1978.71,
                    "end": 1979.67,
                    "word": " deep",
                    "probability": 0.44287109375
                },
                {
                    "start": 1979.67,
                    "end": 1979.87,
                    "word": " RL",
                    "probability": 0.892333984375
                },
                {
                    "start": 1979.87,
                    "end": 1980.31,
                    "word": " policies",
                    "probability": 0.82470703125
                },
                {
                    "start": 1980.31,
                    "end": 1980.87,
                    "word": ".",
                    "probability": 0.99365234375
                },
                {
                    "start": 1981.43,
                    "end": 1981.79,
                    "word": " Now",
                    "probability": 0.8779296875
                },
                {
                    "start": 1981.79,
                    "end": 1981.83,
                    "word": ",",
                    "probability": 0.52197265625
                },
                {
                    "start": 1981.83,
                    "end": 1981.93,
                    "word": " what",
                    "probability": 0.873046875
                },
                {
                    "start": 1981.93,
                    "end": 1982.09,
                    "word": " about",
                    "probability": 0.818359375
                },
                {
                    "start": 1982.09,
                    "end": 1982.37,
                    "word": " training",
                    "probability": 0.78173828125
                }
            ]
        },
        {
            "id": 299,
            "seek": 200142,
            "start": 1982.37,
            "end": 1990.69,
            "text": " time attacks? In the same paper we introduced the policy induction attack, where the attack",
            "tokens": [
                50364,
                565,
                8122,
                30,
                682,
                264,
                912,
                3035,
                321,
                7268,
                264,
                3897,
                33371,
                2690,
                11,
                689,
                264,
                2690,
                50800
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22668851455373148,
            "compression_ratio": 1.5808383233532934,
            "no_speech_prob": 0.0005192756652832031,
            "words": [
                {
                    "start": 1982.37,
                    "end": 1982.61,
                    "word": " time",
                    "probability": 0.189208984375
                },
                {
                    "start": 1982.61,
                    "end": 1983.01,
                    "word": " attacks",
                    "probability": 0.70556640625
                },
                {
                    "start": 1983.01,
                    "end": 1983.39,
                    "word": "?",
                    "probability": 0.04583740234375
                },
                {
                    "start": 1983.39,
                    "end": 1983.57,
                    "word": " In",
                    "probability": 0.87255859375
                },
                {
                    "start": 1983.57,
                    "end": 1983.69,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 1983.69,
                    "end": 1983.87,
                    "word": " same",
                    "probability": 0.81005859375
                },
                {
                    "start": 1983.87,
                    "end": 1984.11,
                    "word": " paper",
                    "probability": 0.82373046875
                },
                {
                    "start": 1984.11,
                    "end": 1984.27,
                    "word": " we",
                    "probability": 0.53271484375
                },
                {
                    "start": 1984.27,
                    "end": 1984.71,
                    "word": " introduced",
                    "probability": 0.52734375
                },
                {
                    "start": 1984.71,
                    "end": 1984.89,
                    "word": " the",
                    "probability": 0.642578125
                },
                {
                    "start": 1984.89,
                    "end": 1985.17,
                    "word": " policy",
                    "probability": 0.73583984375
                },
                {
                    "start": 1985.17,
                    "end": 1985.53,
                    "word": " induction",
                    "probability": 0.7392578125
                },
                {
                    "start": 1985.53,
                    "end": 1985.93,
                    "word": " attack",
                    "probability": 0.830078125
                },
                {
                    "start": 1985.93,
                    "end": 1986.11,
                    "word": ",",
                    "probability": 0.35107421875
                },
                {
                    "start": 1986.11,
                    "end": 1986.25,
                    "word": " where",
                    "probability": 0.8173828125
                },
                {
                    "start": 1986.25,
                    "end": 1987.39,
                    "word": " the",
                    "probability": 0.56640625
                },
                {
                    "start": 1987.39,
                    "end": 1990.69,
                    "word": " attack",
                    "probability": 0.65869140625
                }
            ]
        },
        {
            "id": 300,
            "seek": 200142,
            "start": 1990.69,
            "end": 2000.23,
            "text": " is of the adversarial example type against a DQN agent during training. Now, what are",
            "tokens": [
                50800,
                307,
                295,
                264,
                17641,
                44745,
                1365,
                2010,
                1970,
                257,
                413,
                48,
                45,
                9461,
                1830,
                3097,
                13,
                823,
                11,
                437,
                366,
                51252
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22668851455373148,
            "compression_ratio": 1.5808383233532934,
            "no_speech_prob": 0.0005192756652832031,
            "words": [
                {
                    "start": 1990.69,
                    "end": 1991.53,
                    "word": " is",
                    "probability": 0.8662109375
                },
                {
                    "start": 1991.53,
                    "end": 1992.77,
                    "word": " of",
                    "probability": 0.8408203125
                },
                {
                    "start": 1992.77,
                    "end": 1992.89,
                    "word": " the",
                    "probability": 0.802734375
                },
                {
                    "start": 1992.89,
                    "end": 1993.31,
                    "word": " adversarial",
                    "probability": 0.910400390625
                },
                {
                    "start": 1993.31,
                    "end": 1993.83,
                    "word": " example",
                    "probability": 0.9189453125
                },
                {
                    "start": 1993.83,
                    "end": 1994.35,
                    "word": " type",
                    "probability": 0.9248046875
                },
                {
                    "start": 1994.35,
                    "end": 1995.17,
                    "word": " against",
                    "probability": 0.7509765625
                },
                {
                    "start": 1995.61,
                    "end": 1996.17,
                    "word": " a DQN",
                    "probability": 0.9342447916666666
                },
                {
                    "start": 1996.17,
                    "end": 1996.65,
                    "word": " agent",
                    "probability": 0.8271484375
                },
                {
                    "start": 1996.65,
                    "end": 1997.21,
                    "word": " during",
                    "probability": 0.7421875
                },
                {
                    "start": 1997.21,
                    "end": 1997.65,
                    "word": " training",
                    "probability": 0.7900390625
                },
                {
                    "start": 1997.65,
                    "end": 1998.41,
                    "word": ".",
                    "probability": 0.98046875
                },
                {
                    "start": 1998.85,
                    "end": 1999.01,
                    "word": " Now",
                    "probability": 0.72900390625
                },
                {
                    "start": 1999.01,
                    "end": 1999.23,
                    "word": ",",
                    "probability": 0.42919921875
                },
                {
                    "start": 1999.23,
                    "end": 2000.13,
                    "word": " what",
                    "probability": 0.77587890625
                },
                {
                    "start": 2000.13,
                    "end": 2000.23,
                    "word": " are",
                    "probability": 0.89453125
                }
            ]
        },
        {
            "id": 301,
            "seek": 200142,
            "start": 2000.23,
            "end": 2007.55,
            "text": " the different steps in this attack? First, the adversary derives an adversarial policy",
            "tokens": [
                51252,
                264,
                819,
                4439,
                294,
                341,
                2690,
                30,
                2386,
                11,
                264,
                48222,
                1163,
                1539,
                364,
                17641,
                44745,
                3897,
                51644
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22668851455373148,
            "compression_ratio": 1.5808383233532934,
            "no_speech_prob": 0.0005192756652832031,
            "words": [
                {
                    "start": 2000.23,
                    "end": 2000.33,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 2000.33,
                    "end": 2000.53,
                    "word": " different",
                    "probability": 0.67333984375
                },
                {
                    "start": 2000.53,
                    "end": 2000.85,
                    "word": " steps",
                    "probability": 0.703125
                },
                {
                    "start": 2000.85,
                    "end": 2001.21,
                    "word": " in",
                    "probability": 0.873046875
                },
                {
                    "start": 2001.21,
                    "end": 2001.49,
                    "word": " this",
                    "probability": 0.88232421875
                },
                {
                    "start": 2001.49,
                    "end": 2001.87,
                    "word": " attack",
                    "probability": 0.85693359375
                },
                {
                    "start": 2001.87,
                    "end": 2002.11,
                    "word": "?",
                    "probability": 0.9697265625
                },
                {
                    "start": 2002.45,
                    "end": 2003.21,
                    "word": " First",
                    "probability": 0.88330078125
                },
                {
                    "start": 2003.21,
                    "end": 2004.13,
                    "word": ",",
                    "probability": 0.92431640625
                },
                {
                    "start": 2004.13,
                    "end": 2004.51,
                    "word": " the",
                    "probability": 0.81689453125
                },
                {
                    "start": 2004.51,
                    "end": 2005.37,
                    "word": " adversary",
                    "probability": 0.81591796875
                },
                {
                    "start": 2005.37,
                    "end": 2006.01,
                    "word": " derives",
                    "probability": 0.892578125
                },
                {
                    "start": 2006.01,
                    "end": 2006.23,
                    "word": " an",
                    "probability": 0.90673828125
                },
                {
                    "start": 2006.23,
                    "end": 2006.91,
                    "word": " adversarial",
                    "probability": 0.916748046875
                },
                {
                    "start": 2006.91,
                    "end": 2007.55,
                    "word": " policy",
                    "probability": 0.7998046875
                }
            ]
        },
        {
            "id": 302,
            "seek": 202950,
            "start": 2007.55,
            "end": 2014.39,
            "text": " from the adversarial goal by training on the same environment where the target is going to",
            "tokens": [
                50384,
                490,
                264,
                17641,
                44745,
                3387,
                538,
                3097,
                322,
                264,
                912,
                2823,
                689,
                264,
                3779,
                307,
                516,
                281,
                50700
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23401988975026392,
            "compression_ratio": 1.984375,
            "no_speech_prob": 0.00555419921875,
            "words": [
                {
                    "start": 2007.55,
                    "end": 2008.21,
                    "word": " from",
                    "probability": 0.2371826171875
                },
                {
                    "start": 2008.21,
                    "end": 2009.05,
                    "word": " the",
                    "probability": 0.315185546875
                },
                {
                    "start": 2009.05,
                    "end": 2009.49,
                    "word": " adversarial",
                    "probability": 0.722900390625
                },
                {
                    "start": 2009.49,
                    "end": 2009.79,
                    "word": " goal",
                    "probability": 0.88623046875
                },
                {
                    "start": 2009.79,
                    "end": 2010.05,
                    "word": " by",
                    "probability": 0.791015625
                },
                {
                    "start": 2010.05,
                    "end": 2010.43,
                    "word": " training",
                    "probability": 0.8134765625
                },
                {
                    "start": 2010.43,
                    "end": 2010.97,
                    "word": " on",
                    "probability": 0.8291015625
                },
                {
                    "start": 2010.97,
                    "end": 2011.25,
                    "word": " the",
                    "probability": 0.806640625
                },
                {
                    "start": 2011.25,
                    "end": 2011.61,
                    "word": " same",
                    "probability": 0.80126953125
                },
                {
                    "start": 2011.61,
                    "end": 2012.29,
                    "word": " environment",
                    "probability": 0.7353515625
                },
                {
                    "start": 2012.29,
                    "end": 2012.97,
                    "word": " where",
                    "probability": 0.82373046875
                },
                {
                    "start": 2012.97,
                    "end": 2013.13,
                    "word": " the",
                    "probability": 0.80859375
                },
                {
                    "start": 2013.13,
                    "end": 2013.41,
                    "word": " target",
                    "probability": 0.80029296875
                },
                {
                    "start": 2013.41,
                    "end": 2014.03,
                    "word": " is",
                    "probability": 0.876953125
                },
                {
                    "start": 2014.03,
                    "end": 2014.27,
                    "word": " going",
                    "probability": 0.8486328125
                },
                {
                    "start": 2014.27,
                    "end": 2014.39,
                    "word": " to",
                    "probability": 0.90625
                }
            ]
        },
        {
            "id": 303,
            "seek": 202950,
            "start": 2014.39,
            "end": 2022.29,
            "text": " perform or be trained in. So, if the adversary wants to minimize the reward gained in the game",
            "tokens": [
                50700,
                2042,
                420,
                312,
                8895,
                294,
                13,
                407,
                11,
                498,
                264,
                48222,
                2738,
                281,
                17522,
                264,
                7782,
                12634,
                294,
                264,
                1216,
                51104
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23401988975026392,
            "compression_ratio": 1.984375,
            "no_speech_prob": 0.00555419921875,
            "words": [
                {
                    "start": 2014.39,
                    "end": 2014.79,
                    "word": " perform",
                    "probability": 0.7353515625
                },
                {
                    "start": 2014.79,
                    "end": 2015.09,
                    "word": " or",
                    "probability": 0.69970703125
                },
                {
                    "start": 2015.09,
                    "end": 2015.27,
                    "word": " be",
                    "probability": 0.896484375
                },
                {
                    "start": 2015.27,
                    "end": 2015.59,
                    "word": " trained",
                    "probability": 0.82861328125
                },
                {
                    "start": 2015.59,
                    "end": 2015.85,
                    "word": " in",
                    "probability": 0.77587890625
                },
                {
                    "start": 2015.85,
                    "end": 2016.31,
                    "word": ".",
                    "probability": 0.90625
                },
                {
                    "start": 2016.31,
                    "end": 2016.63,
                    "word": " So",
                    "probability": 0.75048828125
                },
                {
                    "start": 2016.63,
                    "end": 2016.89,
                    "word": ",",
                    "probability": 0.2054443359375
                },
                {
                    "start": 2016.89,
                    "end": 2016.91,
                    "word": " if",
                    "probability": 0.90478515625
                },
                {
                    "start": 2016.91,
                    "end": 2017.59,
                    "word": " the",
                    "probability": 0.8076171875
                },
                {
                    "start": 2017.59,
                    "end": 2017.95,
                    "word": " adversary",
                    "probability": 0.732421875
                },
                {
                    "start": 2017.95,
                    "end": 2018.23,
                    "word": " wants",
                    "probability": 0.75634765625
                },
                {
                    "start": 2018.23,
                    "end": 2018.47,
                    "word": " to",
                    "probability": 0.90576171875
                },
                {
                    "start": 2018.47,
                    "end": 2018.79,
                    "word": " minimize",
                    "probability": 0.7275390625
                },
                {
                    "start": 2018.79,
                    "end": 2019.37,
                    "word": " the",
                    "probability": 0.81005859375
                },
                {
                    "start": 2019.37,
                    "end": 2020.19,
                    "word": " reward",
                    "probability": 0.81982421875
                },
                {
                    "start": 2020.19,
                    "end": 2020.61,
                    "word": " gained",
                    "probability": 0.6884765625
                },
                {
                    "start": 2020.61,
                    "end": 2020.85,
                    "word": " in",
                    "probability": 0.744140625
                },
                {
                    "start": 2020.85,
                    "end": 2021.15,
                    "word": " the",
                    "probability": 0.313232421875
                },
                {
                    "start": 2021.15,
                    "end": 2022.29,
                    "word": " game",
                    "probability": 0.7666015625
                }
            ]
        },
        {
            "id": 304,
            "seek": 202950,
            "start": 2022.29,
            "end": 2030.25,
            "text": " by the agent, then the goal, the optimization goal, is going to be the exact opposite of that of the",
            "tokens": [
                51104,
                538,
                264,
                9461,
                11,
                550,
                264,
                3387,
                11,
                264,
                19618,
                3387,
                11,
                307,
                516,
                281,
                312,
                264,
                1900,
                6182,
                295,
                300,
                295,
                264,
                51492
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23401988975026392,
            "compression_ratio": 1.984375,
            "no_speech_prob": 0.00555419921875,
            "words": [
                {
                    "start": 2022.29,
                    "end": 2022.69,
                    "word": " by",
                    "probability": 0.896484375
                },
                {
                    "start": 2022.69,
                    "end": 2023.15,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 2023.15,
                    "end": 2023.47,
                    "word": " agent",
                    "probability": 0.736328125
                },
                {
                    "start": 2023.47,
                    "end": 2024.13,
                    "word": ",",
                    "probability": 0.83251953125
                },
                {
                    "start": 2024.13,
                    "end": 2024.25,
                    "word": " then",
                    "probability": 0.69677734375
                },
                {
                    "start": 2024.25,
                    "end": 2025.07,
                    "word": " the",
                    "probability": 0.80029296875
                },
                {
                    "start": 2025.07,
                    "end": 2025.87,
                    "word": " goal",
                    "probability": 0.60302734375
                },
                {
                    "start": 2025.87,
                    "end": 2026.09,
                    "word": ",",
                    "probability": 0.85400390625
                },
                {
                    "start": 2026.09,
                    "end": 2026.19,
                    "word": " the",
                    "probability": 0.8046875
                },
                {
                    "start": 2026.19,
                    "end": 2026.55,
                    "word": " optimization",
                    "probability": 0.7421875
                },
                {
                    "start": 2026.55,
                    "end": 2027.13,
                    "word": " goal",
                    "probability": 0.90966796875
                },
                {
                    "start": 2027.13,
                    "end": 2027.35,
                    "word": ",",
                    "probability": 0.6962890625
                },
                {
                    "start": 2027.35,
                    "end": 2027.41,
                    "word": " is",
                    "probability": 0.8984375
                },
                {
                    "start": 2027.41,
                    "end": 2027.71,
                    "word": " going",
                    "probability": 0.85009765625
                },
                {
                    "start": 2027.71,
                    "end": 2027.95,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 2027.95,
                    "end": 2028.09,
                    "word": " be",
                    "probability": 0.9111328125
                },
                {
                    "start": 2028.09,
                    "end": 2028.27,
                    "word": " the",
                    "probability": 0.8076171875
                },
                {
                    "start": 2028.27,
                    "end": 2028.53,
                    "word": " exact",
                    "probability": 0.86669921875
                },
                {
                    "start": 2028.53,
                    "end": 2028.93,
                    "word": " opposite",
                    "probability": 0.86083984375
                },
                {
                    "start": 2028.93,
                    "end": 2029.43,
                    "word": " of",
                    "probability": 0.9033203125
                },
                {
                    "start": 2029.43,
                    "end": 2029.73,
                    "word": " that",
                    "probability": 0.830078125
                },
                {
                    "start": 2029.73,
                    "end": 2030.03,
                    "word": " of",
                    "probability": 0.90185546875
                },
                {
                    "start": 2030.03,
                    "end": 2030.25,
                    "word": " the",
                    "probability": 0.806640625
                }
            ]
        },
        {
            "id": 305,
            "seek": 202950,
            "start": 2030.25,
            "end": 2035.63,
            "text": " target policy. The target wants to maximize the reward and the adversary wants to minimize it.",
            "tokens": [
                51492,
                3779,
                3897,
                13,
                440,
                3779,
                2738,
                281,
                19874,
                264,
                7782,
                293,
                264,
                48222,
                2738,
                281,
                17522,
                309,
                13,
                51760
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23401988975026392,
            "compression_ratio": 1.984375,
            "no_speech_prob": 0.00555419921875,
            "words": [
                {
                    "start": 2030.25,
                    "end": 2030.55,
                    "word": " target",
                    "probability": 0.81640625
                },
                {
                    "start": 2030.55,
                    "end": 2031.69,
                    "word": " policy",
                    "probability": 0.47021484375
                },
                {
                    "start": 2031.69,
                    "end": 2032.05,
                    "word": ".",
                    "probability": 0.94580078125
                },
                {
                    "start": 2032.05,
                    "end": 2032.49,
                    "word": " The",
                    "probability": 0.81640625
                },
                {
                    "start": 2032.49,
                    "end": 2032.69,
                    "word": " target",
                    "probability": 0.86865234375
                },
                {
                    "start": 2032.69,
                    "end": 2032.91,
                    "word": " wants",
                    "probability": 0.75390625
                },
                {
                    "start": 2032.91,
                    "end": 2033.05,
                    "word": " to",
                    "probability": 0.88916015625
                },
                {
                    "start": 2033.05,
                    "end": 2033.41,
                    "word": " maximize",
                    "probability": 0.78271484375
                },
                {
                    "start": 2033.41,
                    "end": 2034.05,
                    "word": " the",
                    "probability": 0.7587890625
                },
                {
                    "start": 2034.05,
                    "end": 2034.29,
                    "word": " reward",
                    "probability": 0.841796875
                },
                {
                    "start": 2034.29,
                    "end": 2034.49,
                    "word": " and",
                    "probability": 0.6943359375
                },
                {
                    "start": 2034.49,
                    "end": 2034.63,
                    "word": " the",
                    "probability": 0.80517578125
                },
                {
                    "start": 2034.63,
                    "end": 2034.91,
                    "word": " adversary",
                    "probability": 0.77490234375
                },
                {
                    "start": 2034.91,
                    "end": 2035.15,
                    "word": " wants",
                    "probability": 0.71044921875
                },
                {
                    "start": 2035.15,
                    "end": 2035.25,
                    "word": " to",
                    "probability": 0.90283203125
                },
                {
                    "start": 2035.25,
                    "end": 2035.41,
                    "word": " minimize",
                    "probability": 0.828125
                },
                {
                    "start": 2035.41,
                    "end": 2035.61,
                    "word": " it",
                    "probability": 0.88134765625
                },
                {
                    "start": 2035.61,
                    "end": 2035.63,
                    "word": ".",
                    "probability": 0.9892578125
                }
            ]
        },
        {
            "id": 306,
            "seek": 205846,
            "start": 2035.63,
            "end": 2038.73,
            "text": " Of course, there are different ways of formulating this adversarial goal.",
            "tokens": [
                50364,
                2720,
                1164,
                11,
                456,
                366,
                819,
                2098,
                295,
                1254,
                12162,
                341,
                17641,
                44745,
                3387,
                13,
                50480
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26777344942092896,
            "compression_ratio": 1.5903083700440528,
            "no_speech_prob": 0.00044417381286621094,
            "words": [
                {
                    "start": 2035.63,
                    "end": 2035.69,
                    "word": " Of",
                    "probability": 0.0017347335815429688
                },
                {
                    "start": 2035.69,
                    "end": 2035.69,
                    "word": " course",
                    "probability": 0.8583984375
                },
                {
                    "start": 2035.69,
                    "end": 2035.69,
                    "word": ",",
                    "probability": 0.68603515625
                },
                {
                    "start": 2035.69,
                    "end": 2035.69,
                    "word": " there",
                    "probability": 0.603515625
                },
                {
                    "start": 2035.69,
                    "end": 2035.79,
                    "word": " are",
                    "probability": 0.8779296875
                },
                {
                    "start": 2035.79,
                    "end": 2036.21,
                    "word": " different",
                    "probability": 0.640625
                },
                {
                    "start": 2036.21,
                    "end": 2036.39,
                    "word": " ways",
                    "probability": 0.83544921875
                },
                {
                    "start": 2036.39,
                    "end": 2036.51,
                    "word": " of",
                    "probability": 0.86865234375
                },
                {
                    "start": 2036.51,
                    "end": 2036.91,
                    "word": " formulating",
                    "probability": 0.93115234375
                },
                {
                    "start": 2036.91,
                    "end": 2037.15,
                    "word": " this",
                    "probability": 0.82177734375
                },
                {
                    "start": 2037.15,
                    "end": 2037.63,
                    "word": " adversarial",
                    "probability": 0.921630859375
                },
                {
                    "start": 2037.63,
                    "end": 2038.01,
                    "word": " goal",
                    "probability": 0.89013671875
                },
                {
                    "start": 2038.01,
                    "end": 2038.73,
                    "word": ".",
                    "probability": 0.93701171875
                }
            ]
        },
        {
            "id": 307,
            "seek": 205846,
            "start": 2039.79,
            "end": 2048.09,
            "text": " Then, the adversary creates a replica of target's DQN and initializes it randomly. And then",
            "tokens": [
                50584,
                1396,
                11,
                264,
                48222,
                7829,
                257,
                35456,
                295,
                3779,
                311,
                413,
                48,
                45,
                293,
                5883,
                5660,
                309,
                16979,
                13,
                400,
                550,
                50980
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26777344942092896,
            "compression_ratio": 1.5903083700440528,
            "no_speech_prob": 0.00044417381286621094,
            "words": [
                {
                    "start": 2039.79,
                    "end": 2040.37,
                    "word": " Then",
                    "probability": 0.68505859375
                },
                {
                    "start": 2040.37,
                    "end": 2040.99,
                    "word": ",",
                    "probability": 0.24072265625
                },
                {
                    "start": 2040.99,
                    "end": 2042.15,
                    "word": " the",
                    "probability": 0.79833984375
                },
                {
                    "start": 2042.15,
                    "end": 2043.53,
                    "word": " adversary",
                    "probability": 0.822265625
                },
                {
                    "start": 2043.53,
                    "end": 2043.89,
                    "word": " creates",
                    "probability": 0.76953125
                },
                {
                    "start": 2044.05,
                    "end": 2044.33,
                    "word": " a replica",
                    "probability": 0.7802734375
                },
                {
                    "start": 2044.33,
                    "end": 2044.59,
                    "word": " of",
                    "probability": 0.859375
                },
                {
                    "start": 2044.59,
                    "end": 2044.97,
                    "word": " target's",
                    "probability": 0.638671875
                },
                {
                    "start": 2044.97,
                    "end": 2045.53,
                    "word": " DQN",
                    "probability": 0.7496744791666666
                },
                {
                    "start": 2045.53,
                    "end": 2045.83,
                    "word": " and",
                    "probability": 0.7509765625
                },
                {
                    "start": 2045.83,
                    "end": 2046.43,
                    "word": " initializes",
                    "probability": 0.877197265625
                },
                {
                    "start": 2046.43,
                    "end": 2046.61,
                    "word": " it",
                    "probability": 0.89794921875
                },
                {
                    "start": 2046.61,
                    "end": 2047.21,
                    "word": " randomly",
                    "probability": 0.6201171875
                },
                {
                    "start": 2047.21,
                    "end": 2047.73,
                    "word": ".",
                    "probability": 0.51708984375
                },
                {
                    "start": 2047.73,
                    "end": 2047.83,
                    "word": " And",
                    "probability": 0.51123046875
                },
                {
                    "start": 2047.83,
                    "end": 2048.09,
                    "word": " then",
                    "probability": 0.7109375
                }
            ]
        },
        {
            "id": 308,
            "seek": 205846,
            "start": 2050.81,
            "end": 2057.51,
            "text": " comes the exploitation phase, where the attacker observes the current state and transitions in the",
            "tokens": [
                51144,
                1487,
                264,
                33122,
                5574,
                11,
                689,
                264,
                35871,
                3181,
                9054,
                264,
                2190,
                1785,
                293,
                23767,
                294,
                264,
                51452
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26777344942092896,
            "compression_ratio": 1.5903083700440528,
            "no_speech_prob": 0.00044417381286621094,
            "words": [
                {
                    "start": 2050.81,
                    "end": 2051.53,
                    "word": " comes",
                    "probability": 0.5185546875
                },
                {
                    "start": 2051.53,
                    "end": 2051.65,
                    "word": " the",
                    "probability": 0.78369140625
                },
                {
                    "start": 2051.65,
                    "end": 2052.05,
                    "word": " exploitation",
                    "probability": 0.837890625
                },
                {
                    "start": 2052.05,
                    "end": 2052.49,
                    "word": " phase",
                    "probability": 0.9033203125
                },
                {
                    "start": 2052.49,
                    "end": 2052.71,
                    "word": ",",
                    "probability": 0.611328125
                },
                {
                    "start": 2052.71,
                    "end": 2052.99,
                    "word": " where",
                    "probability": 0.900390625
                },
                {
                    "start": 2052.99,
                    "end": 2054.21,
                    "word": " the",
                    "probability": 0.8095703125
                },
                {
                    "start": 2054.21,
                    "end": 2054.87,
                    "word": " attacker",
                    "probability": 0.7685546875
                },
                {
                    "start": 2054.87,
                    "end": 2055.95,
                    "word": " observes",
                    "probability": 0.8349609375
                },
                {
                    "start": 2055.95,
                    "end": 2056.15,
                    "word": " the",
                    "probability": 0.81201171875
                },
                {
                    "start": 2056.15,
                    "end": 2056.35,
                    "word": " current",
                    "probability": 0.7275390625
                },
                {
                    "start": 2056.35,
                    "end": 2056.73,
                    "word": " state",
                    "probability": 0.88330078125
                },
                {
                    "start": 2056.73,
                    "end": 2056.91,
                    "word": " and",
                    "probability": 0.884765625
                },
                {
                    "start": 2056.91,
                    "end": 2057.23,
                    "word": " transitions",
                    "probability": 0.83935546875
                },
                {
                    "start": 2057.23,
                    "end": 2057.41,
                    "word": " in",
                    "probability": 0.8779296875
                },
                {
                    "start": 2057.41,
                    "end": 2057.51,
                    "word": " the",
                    "probability": 0.80615234375
                }
            ]
        },
        {
            "id": 309,
            "seek": 205846,
            "start": 2057.51,
            "end": 2064.59,
            "text": " environment, then estimates best action according to the adversarial policy derived in step one.",
            "tokens": [
                51452,
                2823,
                11,
                550,
                20561,
                1151,
                3069,
                4650,
                281,
                264,
                17641,
                44745,
                3897,
                18949,
                294,
                1823,
                472,
                13,
                51776
            ],
            "temperature": 0.0,
            "avg_logprob": -0.26777344942092896,
            "compression_ratio": 1.5903083700440528,
            "no_speech_prob": 0.00044417381286621094,
            "words": [
                {
                    "start": 2057.51,
                    "end": 2057.89,
                    "word": " environment",
                    "probability": 0.75732421875
                },
                {
                    "start": 2057.89,
                    "end": 2058.69,
                    "word": ",",
                    "probability": 0.75244140625
                },
                {
                    "start": 2058.69,
                    "end": 2059.49,
                    "word": " then",
                    "probability": 0.7255859375
                },
                {
                    "start": 2059.49,
                    "end": 2060.25,
                    "word": " estimates",
                    "probability": 0.7841796875
                },
                {
                    "start": 2060.25,
                    "end": 2060.57,
                    "word": " best",
                    "probability": 0.794921875
                },
                {
                    "start": 2060.57,
                    "end": 2060.91,
                    "word": " action",
                    "probability": 0.80029296875
                },
                {
                    "start": 2060.91,
                    "end": 2061.17,
                    "word": " according",
                    "probability": 0.8076171875
                },
                {
                    "start": 2061.17,
                    "end": 2061.33,
                    "word": " to",
                    "probability": 0.90283203125
                },
                {
                    "start": 2061.33,
                    "end": 2061.43,
                    "word": " the",
                    "probability": 0.814453125
                },
                {
                    "start": 2061.43,
                    "end": 2061.89,
                    "word": " adversarial",
                    "probability": 0.919921875
                },
                {
                    "start": 2061.89,
                    "end": 2062.47,
                    "word": " policy",
                    "probability": 0.8095703125
                },
                {
                    "start": 2062.47,
                    "end": 2063.15,
                    "word": " derived",
                    "probability": 0.66748046875
                },
                {
                    "start": 2063.15,
                    "end": 2063.35,
                    "word": " in",
                    "probability": 0.90380859375
                },
                {
                    "start": 2063.35,
                    "end": 2063.57,
                    "word": " step",
                    "probability": 0.70947265625
                },
                {
                    "start": 2063.57,
                    "end": 2063.91,
                    "word": " one",
                    "probability": 0.658203125
                },
                {
                    "start": 2063.91,
                    "end": 2064.59,
                    "word": ".",
                    "probability": 0.615234375
                }
            ]
        },
        {
            "id": 310,
            "seek": 208328,
            "start": 2064.59,
            "end": 2072.21,
            "text": " Then, the attacker crafts perturbations to induce adversarial action based on the replica of",
            "tokens": [
                50384,
                1396,
                11,
                264,
                35871,
                27831,
                40468,
                763,
                281,
                41263,
                17641,
                44745,
                3069,
                2361,
                322,
                264,
                35456,
                295,
                50740
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20003434393432115,
            "compression_ratio": 1.7272727272727273,
            "no_speech_prob": 0.00013554096221923828,
            "words": [
                {
                    "start": 2064.59,
                    "end": 2065.25,
                    "word": " Then",
                    "probability": 0.408203125
                },
                {
                    "start": 2065.25,
                    "end": 2065.51,
                    "word": ",",
                    "probability": 0.09625244140625
                },
                {
                    "start": 2065.51,
                    "end": 2065.69,
                    "word": " the",
                    "probability": 0.8056640625
                },
                {
                    "start": 2065.69,
                    "end": 2065.99,
                    "word": " attacker",
                    "probability": 0.767578125
                },
                {
                    "start": 2065.99,
                    "end": 2066.35,
                    "word": " crafts",
                    "probability": 0.7958984375
                },
                {
                    "start": 2066.35,
                    "end": 2067.15,
                    "word": " perturbations",
                    "probability": 0.884521484375
                },
                {
                    "start": 2067.15,
                    "end": 2067.69,
                    "word": " to",
                    "probability": 0.87060546875
                },
                {
                    "start": 2067.69,
                    "end": 2067.97,
                    "word": " induce",
                    "probability": 0.77294921875
                },
                {
                    "start": 2067.97,
                    "end": 2068.47,
                    "word": " adversarial",
                    "probability": 0.937255859375
                },
                {
                    "start": 2068.47,
                    "end": 2068.97,
                    "word": " action",
                    "probability": 0.80615234375
                },
                {
                    "start": 2068.97,
                    "end": 2069.69,
                    "word": " based",
                    "probability": 0.63037109375
                },
                {
                    "start": 2069.69,
                    "end": 2070.13,
                    "word": " on",
                    "probability": 0.90869140625
                },
                {
                    "start": 2070.13,
                    "end": 2070.75,
                    "word": " the",
                    "probability": 0.7841796875
                },
                {
                    "start": 2070.75,
                    "end": 2071.49,
                    "word": " replica",
                    "probability": 0.8134765625
                },
                {
                    "start": 2071.49,
                    "end": 2072.21,
                    "word": " of",
                    "probability": 0.8935546875
                }
            ]
        },
        {
            "id": 311,
            "seek": 208328,
            "start": 2072.21,
            "end": 2077.95,
            "text": " target's DQN. This is exactly the same as our black box test time attack. The attacker applies",
            "tokens": [
                50740,
                3779,
                311,
                413,
                48,
                45,
                13,
                639,
                307,
                2293,
                264,
                912,
                382,
                527,
                2211,
                2424,
                1500,
                565,
                2690,
                13,
                440,
                35871,
                13165,
                51028
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20003434393432115,
            "compression_ratio": 1.7272727272727273,
            "no_speech_prob": 0.00013554096221923828,
            "words": [
                {
                    "start": 2072.21,
                    "end": 2072.61,
                    "word": " target's",
                    "probability": 0.6968994140625
                },
                {
                    "start": 2072.61,
                    "end": 2073.17,
                    "word": " DQN",
                    "probability": 0.82958984375
                },
                {
                    "start": 2073.17,
                    "end": 2073.43,
                    "word": ".",
                    "probability": 0.70556640625
                },
                {
                    "start": 2073.43,
                    "end": 2073.57,
                    "word": " This",
                    "probability": 0.7490234375
                },
                {
                    "start": 2073.57,
                    "end": 2073.77,
                    "word": " is",
                    "probability": 0.904296875
                },
                {
                    "start": 2073.77,
                    "end": 2074.29,
                    "word": " exactly",
                    "probability": 0.76318359375
                },
                {
                    "start": 2074.29,
                    "end": 2074.51,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 2074.51,
                    "end": 2074.77,
                    "word": " same",
                    "probability": 0.81494140625
                },
                {
                    "start": 2074.77,
                    "end": 2075.09,
                    "word": " as",
                    "probability": 0.904296875
                },
                {
                    "start": 2075.09,
                    "end": 2075.33,
                    "word": " our",
                    "probability": 0.78515625
                },
                {
                    "start": 2075.33,
                    "end": 2075.63,
                    "word": " black",
                    "probability": 0.6845703125
                },
                {
                    "start": 2075.63,
                    "end": 2075.83,
                    "word": " box",
                    "probability": 0.51513671875
                },
                {
                    "start": 2075.83,
                    "end": 2076.09,
                    "word": " test",
                    "probability": 0.7724609375
                },
                {
                    "start": 2076.09,
                    "end": 2076.27,
                    "word": " time",
                    "probability": 0.5673828125
                },
                {
                    "start": 2076.27,
                    "end": 2076.65,
                    "word": " attack",
                    "probability": 0.822265625
                },
                {
                    "start": 2076.65,
                    "end": 2077.15,
                    "word": ".",
                    "probability": 0.95751953125
                },
                {
                    "start": 2077.15,
                    "end": 2077.27,
                    "word": " The",
                    "probability": 0.806640625
                },
                {
                    "start": 2077.27,
                    "end": 2077.53,
                    "word": " attacker",
                    "probability": 0.7451171875
                },
                {
                    "start": 2077.53,
                    "end": 2077.95,
                    "word": " applies",
                    "probability": 0.7275390625
                }
            ]
        },
        {
            "id": 312,
            "seek": 208328,
            "start": 2077.95,
            "end": 2082.69,
            "text": " the perturbation as a man in the middle in the observation channel. The perturbed input is",
            "tokens": [
                51028,
                264,
                40468,
                399,
                382,
                257,
                587,
                294,
                264,
                2808,
                294,
                264,
                14816,
                2269,
                13,
                440,
                13269,
                374,
                2883,
                4846,
                307,
                51268
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20003434393432115,
            "compression_ratio": 1.7272727272727273,
            "no_speech_prob": 0.00013554096221923828,
            "words": [
                {
                    "start": 2077.95,
                    "end": 2078.09,
                    "word": " the",
                    "probability": 0.73828125
                },
                {
                    "start": 2078.09,
                    "end": 2078.65,
                    "word": " perturbation",
                    "probability": 0.902587890625
                },
                {
                    "start": 2078.65,
                    "end": 2079.05,
                    "word": " as",
                    "probability": 0.88720703125
                },
                {
                    "start": 2079.21,
                    "end": 2079.43,
                    "word": " a man",
                    "probability": 0.8759765625
                },
                {
                    "start": 2079.43,
                    "end": 2079.57,
                    "word": " in",
                    "probability": 0.88623046875
                },
                {
                    "start": 2079.57,
                    "end": 2079.73,
                    "word": " the",
                    "probability": 0.814453125
                },
                {
                    "start": 2079.73,
                    "end": 2079.91,
                    "word": " middle",
                    "probability": 0.9267578125
                },
                {
                    "start": 2079.91,
                    "end": 2080.05,
                    "word": " in",
                    "probability": 0.61279296875
                },
                {
                    "start": 2080.05,
                    "end": 2080.15,
                    "word": " the",
                    "probability": 0.80810546875
                },
                {
                    "start": 2080.15,
                    "end": 2080.49,
                    "word": " observation",
                    "probability": 0.83984375
                },
                {
                    "start": 2080.49,
                    "end": 2080.99,
                    "word": " channel",
                    "probability": 0.7861328125
                },
                {
                    "start": 2080.99,
                    "end": 2081.55,
                    "word": ".",
                    "probability": 0.771484375
                },
                {
                    "start": 2081.65,
                    "end": 2081.91,
                    "word": " The",
                    "probability": 0.80859375
                },
                {
                    "start": 2081.91,
                    "end": 2082.31,
                    "word": " perturbed",
                    "probability": 0.9041341145833334
                },
                {
                    "start": 2082.31,
                    "end": 2082.51,
                    "word": " input",
                    "probability": 0.87109375
                },
                {
                    "start": 2082.51,
                    "end": 2082.69,
                    "word": " is",
                    "probability": 0.90185546875
                }
            ]
        },
        {
            "id": 313,
            "seek": 208328,
            "start": 2082.69,
            "end": 2089.41,
            "text": " revealed to the target and the attacker waits for target's actions. And this is a loop. This loop can",
            "tokens": [
                51268,
                9599,
                281,
                264,
                3779,
                293,
                264,
                35871,
                40597,
                337,
                3779,
                311,
                5909,
                13,
                400,
                341,
                307,
                257,
                6367,
                13,
                639,
                6367,
                393,
                51604
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20003434393432115,
            "compression_ratio": 1.7272727272727273,
            "no_speech_prob": 0.00013554096221923828,
            "words": [
                {
                    "start": 2082.69,
                    "end": 2083.09,
                    "word": " revealed",
                    "probability": 0.87890625
                },
                {
                    "start": 2083.09,
                    "end": 2083.25,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 2083.25,
                    "end": 2083.39,
                    "word": " the",
                    "probability": 0.8076171875
                },
                {
                    "start": 2083.39,
                    "end": 2083.75,
                    "word": " target",
                    "probability": 0.833984375
                },
                {
                    "start": 2083.75,
                    "end": 2084.41,
                    "word": " and",
                    "probability": 0.390380859375
                },
                {
                    "start": 2084.41,
                    "end": 2084.53,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 2084.53,
                    "end": 2084.79,
                    "word": " attacker",
                    "probability": 0.7412109375
                },
                {
                    "start": 2084.79,
                    "end": 2085.23,
                    "word": " waits",
                    "probability": 0.80908203125
                },
                {
                    "start": 2085.23,
                    "end": 2085.49,
                    "word": " for",
                    "probability": 0.9033203125
                },
                {
                    "start": 2085.49,
                    "end": 2086.09,
                    "word": " target's",
                    "probability": 0.8271484375
                },
                {
                    "start": 2086.09,
                    "end": 2086.61,
                    "word": " actions",
                    "probability": 0.85009765625
                },
                {
                    "start": 2086.61,
                    "end": 2087.17,
                    "word": ".",
                    "probability": 0.89306640625
                },
                {
                    "start": 2087.53,
                    "end": 2087.63,
                    "word": " And",
                    "probability": 0.6533203125
                },
                {
                    "start": 2087.63,
                    "end": 2087.81,
                    "word": " this",
                    "probability": 0.8525390625
                },
                {
                    "start": 2087.81,
                    "end": 2087.89,
                    "word": " is",
                    "probability": 0.90380859375
                },
                {
                    "start": 2088.01,
                    "end": 2088.29,
                    "word": " a loop",
                    "probability": 0.89306640625
                },
                {
                    "start": 2088.29,
                    "end": 2088.49,
                    "word": ".",
                    "probability": 0.89990234375
                },
                {
                    "start": 2088.49,
                    "end": 2088.83,
                    "word": " This",
                    "probability": 0.794921875
                },
                {
                    "start": 2088.83,
                    "end": 2089.17,
                    "word": " loop",
                    "probability": 0.89111328125
                },
                {
                    "start": 2089.17,
                    "end": 2089.41,
                    "word": " can",
                    "probability": 0.90234375
                }
            ]
        },
        {
            "id": 314,
            "seek": 210554,
            "start": 2089.41,
            "end": 2096.87,
            "text": " go on until the training process of the target either converges to a suboptimal policy or up to",
            "tokens": [
                50364,
                352,
                322,
                1826,
                264,
                3097,
                1399,
                295,
                264,
                3779,
                2139,
                9652,
                2880,
                281,
                257,
                1422,
                5747,
                10650,
                3897,
                420,
                493,
                281,
                50748
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25147805141436086,
            "compression_ratio": 1.5621621621621622,
            "no_speech_prob": 0.0185394287109375,
            "words": [
                {
                    "start": 2089.41,
                    "end": 2089.63,
                    "word": " go",
                    "probability": 0.16552734375
                },
                {
                    "start": 2089.63,
                    "end": 2089.93,
                    "word": " on",
                    "probability": 0.8935546875
                },
                {
                    "start": 2089.93,
                    "end": 2090.39,
                    "word": " until",
                    "probability": 0.791015625
                },
                {
                    "start": 2090.39,
                    "end": 2091.05,
                    "word": " the",
                    "probability": 0.7861328125
                },
                {
                    "start": 2091.05,
                    "end": 2092.57,
                    "word": " training",
                    "probability": 0.60986328125
                },
                {
                    "start": 2092.57,
                    "end": 2093.25,
                    "word": " process",
                    "probability": 0.7685546875
                },
                {
                    "start": 2093.25,
                    "end": 2093.37,
                    "word": " of",
                    "probability": 0.87646484375
                },
                {
                    "start": 2093.37,
                    "end": 2093.47,
                    "word": " the",
                    "probability": 0.7958984375
                },
                {
                    "start": 2093.47,
                    "end": 2093.63,
                    "word": " target",
                    "probability": 0.85498046875
                },
                {
                    "start": 2093.63,
                    "end": 2093.91,
                    "word": " either",
                    "probability": 0.72314453125
                },
                {
                    "start": 2093.91,
                    "end": 2094.85,
                    "word": " converges",
                    "probability": 0.95361328125
                },
                {
                    "start": 2094.85,
                    "end": 2095.05,
                    "word": " to",
                    "probability": 0.8955078125
                },
                {
                    "start": 2095.21,
                    "end": 2095.71,
                    "word": " a suboptimal",
                    "probability": 0.7642415364583334
                },
                {
                    "start": 2095.71,
                    "end": 2096.11,
                    "word": " policy",
                    "probability": 0.76953125
                },
                {
                    "start": 2096.11,
                    "end": 2096.43,
                    "word": " or",
                    "probability": 0.75048828125
                },
                {
                    "start": 2096.43,
                    "end": 2096.65,
                    "word": " up",
                    "probability": 0.90234375
                },
                {
                    "start": 2096.65,
                    "end": 2096.87,
                    "word": " to",
                    "probability": 0.89599609375
                }
            ]
        },
        {
            "id": 315,
            "seek": 210554,
            "start": 2097.31,
            "end": 2104.09,
            "text": " a certain number of iterations. This is a very rough plot. This is not smooth yet, but you can",
            "tokens": [
                50748,
                257,
                1629,
                1230,
                295,
                36540,
                13,
                639,
                307,
                257,
                588,
                5903,
                7542,
                13,
                639,
                307,
                406,
                5508,
                1939,
                11,
                457,
                291,
                393,
                51096
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25147805141436086,
            "compression_ratio": 1.5621621621621622,
            "no_speech_prob": 0.0185394287109375,
            "words": [
                {
                    "start": 2097.31,
                    "end": 2097.49,
                    "word": " a certain",
                    "probability": 0.8134765625
                },
                {
                    "start": 2097.49,
                    "end": 2097.75,
                    "word": " number",
                    "probability": 0.875
                },
                {
                    "start": 2097.75,
                    "end": 2097.87,
                    "word": " of",
                    "probability": 0.90478515625
                },
                {
                    "start": 2097.87,
                    "end": 2098.27,
                    "word": " iterations",
                    "probability": 0.72998046875
                },
                {
                    "start": 2098.27,
                    "end": 2098.79,
                    "word": ".",
                    "probability": 0.91650390625
                },
                {
                    "start": 2099.47,
                    "end": 2099.99,
                    "word": " This",
                    "probability": 0.78857421875
                },
                {
                    "start": 2099.99,
                    "end": 2100.39,
                    "word": " is",
                    "probability": 0.89990234375
                },
                {
                    "start": 2100.61,
                    "end": 2100.79,
                    "word": " a very",
                    "probability": 0.71435546875
                },
                {
                    "start": 2100.79,
                    "end": 2100.99,
                    "word": " rough",
                    "probability": 0.7353515625
                },
                {
                    "start": 2100.99,
                    "end": 2101.27,
                    "word": " plot",
                    "probability": 0.884765625
                },
                {
                    "start": 2101.27,
                    "end": 2101.37,
                    "word": ".",
                    "probability": 0.32763671875
                },
                {
                    "start": 2101.37,
                    "end": 2101.51,
                    "word": " This",
                    "probability": 0.77392578125
                },
                {
                    "start": 2101.51,
                    "end": 2101.63,
                    "word": " is",
                    "probability": 0.8876953125
                },
                {
                    "start": 2101.63,
                    "end": 2101.83,
                    "word": " not",
                    "probability": 0.89990234375
                },
                {
                    "start": 2101.83,
                    "end": 2102.47,
                    "word": " smooth",
                    "probability": 0.76220703125
                },
                {
                    "start": 2102.47,
                    "end": 2103.33,
                    "word": " yet",
                    "probability": 0.46630859375
                },
                {
                    "start": 2103.33,
                    "end": 2103.69,
                    "word": ",",
                    "probability": 0.63232421875
                },
                {
                    "start": 2103.69,
                    "end": 2103.71,
                    "word": " but",
                    "probability": 0.90087890625
                },
                {
                    "start": 2103.71,
                    "end": 2103.95,
                    "word": " you",
                    "probability": 0.88623046875
                },
                {
                    "start": 2103.95,
                    "end": 2104.09,
                    "word": " can",
                    "probability": 0.89990234375
                }
            ]
        },
        {
            "id": 316,
            "seek": 210554,
            "start": 2104.09,
            "end": 2111.67,
            "text": " see that the unperturbed agent moves towards... this is in the game of Pong still... moves towards",
            "tokens": [
                51096,
                536,
                300,
                264,
                517,
                15346,
                374,
                2883,
                9461,
                6067,
                3030,
                485,
                341,
                307,
                294,
                264,
                1216,
                295,
                430,
                556,
                920,
                485,
                6067,
                3030,
                51476
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25147805141436086,
            "compression_ratio": 1.5621621621621622,
            "no_speech_prob": 0.0185394287109375,
            "words": [
                {
                    "start": 2104.09,
                    "end": 2104.33,
                    "word": " see",
                    "probability": 0.82080078125
                },
                {
                    "start": 2104.33,
                    "end": 2104.65,
                    "word": " that",
                    "probability": 0.876953125
                },
                {
                    "start": 2104.65,
                    "end": 2105.15,
                    "word": " the",
                    "probability": 0.78759765625
                },
                {
                    "start": 2105.15,
                    "end": 2106.17,
                    "word": " unperturbed",
                    "probability": 0.8841552734375
                },
                {
                    "start": 2106.17,
                    "end": 2106.79,
                    "word": " agent",
                    "probability": 0.837890625
                },
                {
                    "start": 2106.79,
                    "end": 2107.77,
                    "word": " moves",
                    "probability": 0.6953125
                },
                {
                    "start": 2107.77,
                    "end": 2108.49,
                    "word": " towards...",
                    "probability": 0.333709716796875
                },
                {
                    "start": 2108.49,
                    "end": 2109.19,
                    "word": " this",
                    "probability": 0.384521484375
                },
                {
                    "start": 2109.19,
                    "end": 2109.31,
                    "word": " is",
                    "probability": 0.8984375
                },
                {
                    "start": 2109.31,
                    "end": 2109.45,
                    "word": " in",
                    "probability": 0.88037109375
                },
                {
                    "start": 2109.45,
                    "end": 2109.55,
                    "word": " the",
                    "probability": 0.806640625
                },
                {
                    "start": 2109.55,
                    "end": 2109.67,
                    "word": " game",
                    "probability": 0.6376953125
                },
                {
                    "start": 2109.67,
                    "end": 2109.81,
                    "word": " of",
                    "probability": 0.88916015625
                },
                {
                    "start": 2109.81,
                    "end": 2110.03,
                    "word": " Pong",
                    "probability": 0.745849609375
                },
                {
                    "start": 2110.03,
                    "end": 2110.83,
                    "word": " still...",
                    "probability": 0.6640625
                },
                {
                    "start": 2110.83,
                    "end": 2110.97,
                    "word": " moves",
                    "probability": 0.60302734375
                },
                {
                    "start": 2110.97,
                    "end": 2111.67,
                    "word": " towards",
                    "probability": 0.493408203125
                }
            ]
        },
        {
            "id": 317,
            "seek": 213532,
            "start": 2111.67,
            "end": 2121.73,
            "text": " convergence to an optimal total sum of rewards, while the attack agent moves towards a convergence",
            "tokens": [
                50364,
                32181,
                281,
                364,
                16252,
                3217,
                2408,
                295,
                17203,
                11,
                1339,
                264,
                2690,
                9461,
                6067,
                3030,
                257,
                32181,
                50872
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22247869436713782,
            "compression_ratio": 1.5975103734439835,
            "no_speech_prob": 0.010986328125,
            "words": [
                {
                    "start": 2111.67,
                    "end": 2112.11,
                    "word": " convergence",
                    "probability": 0.246337890625
                },
                {
                    "start": 2112.11,
                    "end": 2112.47,
                    "word": " to",
                    "probability": 0.865234375
                },
                {
                    "start": 2112.47,
                    "end": 2113.05,
                    "word": " an",
                    "probability": 0.85498046875
                },
                {
                    "start": 2113.05,
                    "end": 2113.69,
                    "word": " optimal",
                    "probability": 0.72802734375
                },
                {
                    "start": 2113.69,
                    "end": 2114.47,
                    "word": " total",
                    "probability": 0.7548828125
                },
                {
                    "start": 2114.47,
                    "end": 2115.03,
                    "word": " sum",
                    "probability": 0.888671875
                },
                {
                    "start": 2115.03,
                    "end": 2115.15,
                    "word": " of",
                    "probability": 0.89306640625
                },
                {
                    "start": 2115.15,
                    "end": 2115.65,
                    "word": " rewards",
                    "probability": 0.66357421875
                },
                {
                    "start": 2115.65,
                    "end": 2116.21,
                    "word": ",",
                    "probability": 0.5859375
                },
                {
                    "start": 2116.21,
                    "end": 2116.67,
                    "word": " while",
                    "probability": 0.892578125
                },
                {
                    "start": 2116.67,
                    "end": 2117.09,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 2117.09,
                    "end": 2117.51,
                    "word": " attack",
                    "probability": 0.7138671875
                },
                {
                    "start": 2117.51,
                    "end": 2118.15,
                    "word": " agent",
                    "probability": 0.7666015625
                },
                {
                    "start": 2118.15,
                    "end": 2118.97,
                    "word": " moves",
                    "probability": 0.8388671875
                },
                {
                    "start": 2118.97,
                    "end": 2119.63,
                    "word": " towards",
                    "probability": 0.60888671875
                },
                {
                    "start": 2120.73,
                    "end": 2121.73,
                    "word": " a convergence",
                    "probability": 0.8291015625
                }
            ]
        },
        {
            "id": 318,
            "seek": 213532,
            "start": 2121.73,
            "end": 2129.45,
            "text": " to the minimum possible return of zero. It's getting closer and closer to zero, which indicates that",
            "tokens": [
                50872,
                281,
                264,
                7285,
                1944,
                2736,
                295,
                4018,
                13,
                467,
                311,
                1242,
                4966,
                293,
                4966,
                281,
                4018,
                11,
                597,
                16203,
                300,
                51264
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22247869436713782,
            "compression_ratio": 1.5975103734439835,
            "no_speech_prob": 0.010986328125,
            "words": [
                {
                    "start": 2121.73,
                    "end": 2122.13,
                    "word": " to",
                    "probability": 0.89599609375
                },
                {
                    "start": 2122.13,
                    "end": 2122.61,
                    "word": " the",
                    "probability": 0.7978515625
                },
                {
                    "start": 2122.61,
                    "end": 2122.97,
                    "word": " minimum",
                    "probability": 0.9033203125
                },
                {
                    "start": 2122.97,
                    "end": 2123.61,
                    "word": " possible",
                    "probability": 0.8486328125
                },
                {
                    "start": 2123.61,
                    "end": 2124.23,
                    "word": " return",
                    "probability": 0.79833984375
                },
                {
                    "start": 2124.23,
                    "end": 2124.65,
                    "word": " of",
                    "probability": 0.8896484375
                },
                {
                    "start": 2124.65,
                    "end": 2125.19,
                    "word": " zero",
                    "probability": 0.61474609375
                },
                {
                    "start": 2125.19,
                    "end": 2125.31,
                    "word": ".",
                    "probability": 0.77197265625
                },
                {
                    "start": 2125.31,
                    "end": 2125.35,
                    "word": " It's",
                    "probability": 0.933837890625
                },
                {
                    "start": 2125.35,
                    "end": 2125.53,
                    "word": " getting",
                    "probability": 0.89892578125
                },
                {
                    "start": 2125.53,
                    "end": 2125.85,
                    "word": " closer",
                    "probability": 0.78759765625
                },
                {
                    "start": 2125.85,
                    "end": 2126.01,
                    "word": " and",
                    "probability": 0.89404296875
                },
                {
                    "start": 2126.01,
                    "end": 2126.29,
                    "word": " closer",
                    "probability": 0.78271484375
                },
                {
                    "start": 2126.29,
                    "end": 2127.07,
                    "word": " to",
                    "probability": 0.8984375
                },
                {
                    "start": 2127.07,
                    "end": 2127.51,
                    "word": " zero",
                    "probability": 0.814453125
                },
                {
                    "start": 2127.51,
                    "end": 2127.91,
                    "word": ",",
                    "probability": 0.7763671875
                },
                {
                    "start": 2127.91,
                    "end": 2128.45,
                    "word": " which",
                    "probability": 0.91357421875
                },
                {
                    "start": 2128.45,
                    "end": 2129.07,
                    "word": " indicates",
                    "probability": 0.89599609375
                },
                {
                    "start": 2129.07,
                    "end": 2129.45,
                    "word": " that",
                    "probability": 0.87109375
                }
            ]
        },
        {
            "id": 319,
            "seek": 213532,
            "start": 2129.45,
            "end": 2135.33,
            "text": " the training process of DQN and DeepRL in general can also be targeted through adversarial attacks.",
            "tokens": [
                51264,
                264,
                3097,
                1399,
                295,
                413,
                48,
                45,
                293,
                14895,
                49,
                43,
                294,
                2674,
                393,
                611,
                312,
                15045,
                807,
                17641,
                44745,
                8122,
                13,
                51516
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22247869436713782,
            "compression_ratio": 1.5975103734439835,
            "no_speech_prob": 0.010986328125,
            "words": [
                {
                    "start": 2129.45,
                    "end": 2130.05,
                    "word": " the",
                    "probability": 0.81494140625
                },
                {
                    "start": 2130.05,
                    "end": 2130.45,
                    "word": " training",
                    "probability": 0.79833984375
                },
                {
                    "start": 2130.45,
                    "end": 2130.89,
                    "word": " process",
                    "probability": 0.8583984375
                },
                {
                    "start": 2130.89,
                    "end": 2131.21,
                    "word": " of",
                    "probability": 0.90576171875
                },
                {
                    "start": 2131.21,
                    "end": 2131.59,
                    "word": " DQN",
                    "probability": 0.8972981770833334
                },
                {
                    "start": 2131.59,
                    "end": 2131.71,
                    "word": " and",
                    "probability": 0.83203125
                },
                {
                    "start": 2131.71,
                    "end": 2131.99,
                    "word": " DeepRL",
                    "probability": 0.3742268880208333
                },
                {
                    "start": 2131.99,
                    "end": 2132.13,
                    "word": " in",
                    "probability": 0.837890625
                },
                {
                    "start": 2132.13,
                    "end": 2132.35,
                    "word": " general",
                    "probability": 0.7861328125
                },
                {
                    "start": 2132.35,
                    "end": 2132.57,
                    "word": " can",
                    "probability": 0.8671875
                },
                {
                    "start": 2132.57,
                    "end": 2132.87,
                    "word": " also",
                    "probability": 0.818359375
                },
                {
                    "start": 2132.87,
                    "end": 2133.05,
                    "word": " be",
                    "probability": 0.912109375
                },
                {
                    "start": 2133.05,
                    "end": 2133.41,
                    "word": " targeted",
                    "probability": 0.81396484375
                },
                {
                    "start": 2133.41,
                    "end": 2133.79,
                    "word": " through",
                    "probability": 0.74365234375
                },
                {
                    "start": 2133.79,
                    "end": 2134.27,
                    "word": " adversarial",
                    "probability": 0.93212890625
                },
                {
                    "start": 2134.27,
                    "end": 2134.73,
                    "word": " attacks",
                    "probability": 0.7333984375
                },
                {
                    "start": 2134.73,
                    "end": 2135.33,
                    "word": ".",
                    "probability": 0.9921875
                }
            ]
        },
        {
            "id": 320,
            "seek": 213532,
            "start": 2135.33,
            "end": 2141.45,
            "text": " Now, I'm going to introduce another type of training time attack. Again, this aims to",
            "tokens": [
                51548,
                823,
                11,
                286,
                478,
                516,
                281,
                5366,
                1071,
                2010,
                295,
                3097,
                565,
                2690,
                13,
                3764,
                11,
                341,
                24683,
                281,
                51852
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22247869436713782,
            "compression_ratio": 1.5975103734439835,
            "no_speech_prob": 0.010986328125,
            "words": [
                {
                    "start": 2135.33,
                    "end": 2135.63,
                    "word": " Now",
                    "probability": 0.85791015625
                },
                {
                    "start": 2135.63,
                    "end": 2135.67,
                    "word": ",",
                    "probability": 0.3193359375
                },
                {
                    "start": 2135.67,
                    "end": 2135.75,
                    "word": " I'm",
                    "probability": 0.99267578125
                },
                {
                    "start": 2135.75,
                    "end": 2135.87,
                    "word": " going",
                    "probability": 0.857421875
                },
                {
                    "start": 2135.87,
                    "end": 2136.01,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 2136.01,
                    "end": 2136.33,
                    "word": " introduce",
                    "probability": 0.71044921875
                },
                {
                    "start": 2136.33,
                    "end": 2136.69,
                    "word": " another",
                    "probability": 0.8193359375
                },
                {
                    "start": 2136.69,
                    "end": 2137.05,
                    "word": " type",
                    "probability": 0.93359375
                },
                {
                    "start": 2137.05,
                    "end": 2137.75,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 2137.75,
                    "end": 2138.33,
                    "word": " training",
                    "probability": 0.7490234375
                },
                {
                    "start": 2138.33,
                    "end": 2138.59,
                    "word": " time",
                    "probability": 0.70849609375
                },
                {
                    "start": 2138.59,
                    "end": 2139.03,
                    "word": " attack",
                    "probability": 0.85986328125
                },
                {
                    "start": 2139.03,
                    "end": 2139.33,
                    "word": ".",
                    "probability": 0.9423828125
                },
                {
                    "start": 2139.33,
                    "end": 2139.63,
                    "word": " Again",
                    "probability": 0.88720703125
                },
                {
                    "start": 2139.63,
                    "end": 2139.79,
                    "word": ",",
                    "probability": 0.958984375
                },
                {
                    "start": 2139.79,
                    "end": 2140.83,
                    "word": " this",
                    "probability": 0.87939453125
                },
                {
                    "start": 2140.83,
                    "end": 2141.09,
                    "word": " aims",
                    "probability": 0.71630859375
                },
                {
                    "start": 2141.09,
                    "end": 2141.45,
                    "word": " to",
                    "probability": 0.90576171875
                }
            ]
        },
        {
            "id": 321,
            "seek": 216444,
            "start": 2141.45,
            "end": 2148.35,
            "text": " induce some form of misbehavior. We call this misbehavior addiction, and this is a follow-up to",
            "tokens": [
                50408,
                41263,
                512,
                1254,
                295,
                3346,
                29437,
                38387,
                13,
                492,
                818,
                341,
                3346,
                29437,
                38387,
                16835,
                11,
                293,
                341,
                307,
                257,
                1524,
                12,
                1010,
                281,
                50704
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24484535590889528,
            "compression_ratio": 1.548,
            "no_speech_prob": 0.0005884170532226562,
            "words": [
                {
                    "start": 2141.45,
                    "end": 2142.07,
                    "word": " induce",
                    "probability": 0.1597900390625
                },
                {
                    "start": 2142.07,
                    "end": 2142.67,
                    "word": " some",
                    "probability": 0.80810546875
                },
                {
                    "start": 2142.67,
                    "end": 2142.95,
                    "word": " form",
                    "probability": 0.810546875
                },
                {
                    "start": 2142.95,
                    "end": 2143.13,
                    "word": " of",
                    "probability": 0.90625
                },
                {
                    "start": 2143.13,
                    "end": 2143.91,
                    "word": " misbehavior",
                    "probability": 0.7911783854166666
                },
                {
                    "start": 2143.91,
                    "end": 2144.55,
                    "word": ".",
                    "probability": 0.84033203125
                },
                {
                    "start": 2144.75,
                    "end": 2145.35,
                    "word": " We",
                    "probability": 0.849609375
                },
                {
                    "start": 2145.35,
                    "end": 2145.53,
                    "word": " call",
                    "probability": 0.7880859375
                },
                {
                    "start": 2145.53,
                    "end": 2145.71,
                    "word": " this",
                    "probability": 0.87158203125
                },
                {
                    "start": 2145.71,
                    "end": 2146.13,
                    "word": " misbehavior",
                    "probability": 0.8816731770833334
                },
                {
                    "start": 2146.13,
                    "end": 2146.59,
                    "word": " addiction",
                    "probability": 0.55078125
                },
                {
                    "start": 2146.59,
                    "end": 2147.21,
                    "word": ",",
                    "probability": 0.2230224609375
                },
                {
                    "start": 2147.21,
                    "end": 2147.29,
                    "word": " and",
                    "probability": 0.77001953125
                },
                {
                    "start": 2147.29,
                    "end": 2147.45,
                    "word": " this",
                    "probability": 0.88720703125
                },
                {
                    "start": 2147.45,
                    "end": 2147.71,
                    "word": " is",
                    "probability": 0.90234375
                },
                {
                    "start": 2147.91,
                    "end": 2148.09,
                    "word": " a follow",
                    "probability": 0.869140625
                },
                {
                    "start": 2148.09,
                    "end": 2148.23,
                    "word": "-up",
                    "probability": 0.62152099609375
                },
                {
                    "start": 2148.23,
                    "end": 2148.35,
                    "word": " to",
                    "probability": 0.89892578125
                }
            ]
        },
        {
            "id": 322,
            "seek": 216444,
            "start": 2148.45,
            "end": 2154.07,
            "text": " a work I did with my colleague Roman Niempolski on psychopathological modeling of AI safety problems.",
            "tokens": [
                50704,
                257,
                589,
                286,
                630,
                365,
                452,
                13532,
                8566,
                426,
                4907,
                12892,
                18020,
                322,
                47577,
                4383,
                15983,
                295,
                7318,
                4514,
                2740,
                13,
                50960
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24484535590889528,
            "compression_ratio": 1.548,
            "no_speech_prob": 0.0005884170532226562,
            "words": [
                {
                    "start": 2148.45,
                    "end": 2148.61,
                    "word": " a work",
                    "probability": 0.79736328125
                },
                {
                    "start": 2148.61,
                    "end": 2148.71,
                    "word": " I",
                    "probability": 0.98828125
                },
                {
                    "start": 2148.71,
                    "end": 2148.85,
                    "word": " did",
                    "probability": 0.896484375
                },
                {
                    "start": 2148.85,
                    "end": 2149.01,
                    "word": " with",
                    "probability": 0.81640625
                },
                {
                    "start": 2149.01,
                    "end": 2149.23,
                    "word": " my",
                    "probability": 0.90185546875
                },
                {
                    "start": 2149.23,
                    "end": 2149.47,
                    "word": " colleague",
                    "probability": 0.74853515625
                },
                {
                    "start": 2149.47,
                    "end": 2149.69,
                    "word": " Roman",
                    "probability": 0.830078125
                },
                {
                    "start": 2149.69,
                    "end": 2150.31,
                    "word": " Niempolski",
                    "probability": 0.5503997802734375
                },
                {
                    "start": 2150.31,
                    "end": 2150.79,
                    "word": " on",
                    "probability": 0.8486328125
                },
                {
                    "start": 2150.79,
                    "end": 2151.95,
                    "word": " psychopathological",
                    "probability": 0.91357421875
                },
                {
                    "start": 2151.95,
                    "end": 2152.27,
                    "word": " modeling",
                    "probability": 0.6337890625
                },
                {
                    "start": 2152.27,
                    "end": 2152.43,
                    "word": " of",
                    "probability": 0.8974609375
                },
                {
                    "start": 2152.43,
                    "end": 2152.63,
                    "word": " AI",
                    "probability": 0.888671875
                },
                {
                    "start": 2152.63,
                    "end": 2152.95,
                    "word": " safety",
                    "probability": 0.90869140625
                },
                {
                    "start": 2152.95,
                    "end": 2153.43,
                    "word": " problems",
                    "probability": 0.69873046875
                },
                {
                    "start": 2153.43,
                    "end": 2154.07,
                    "word": ".",
                    "probability": 0.98583984375
                }
            ]
        },
        {
            "id": 323,
            "seek": 216444,
            "start": 2154.27,
            "end": 2161.51,
            "text": " This is a proof of concept. We consider the game of Snake. Many of you probably remember Snake from",
            "tokens": [
                51020,
                639,
                307,
                257,
                8177,
                295,
                3410,
                13,
                492,
                1949,
                264,
                1216,
                295,
                33885,
                13,
                5126,
                295,
                291,
                1391,
                1604,
                33885,
                490,
                51360
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24484535590889528,
            "compression_ratio": 1.548,
            "no_speech_prob": 0.0005884170532226562,
            "words": [
                {
                    "start": 2154.27,
                    "end": 2154.91,
                    "word": " This",
                    "probability": 0.8154296875
                },
                {
                    "start": 2154.91,
                    "end": 2155.01,
                    "word": " is",
                    "probability": 0.9033203125
                },
                {
                    "start": 2155.11,
                    "end": 2155.27,
                    "word": " a proof",
                    "probability": 0.85009765625
                },
                {
                    "start": 2155.27,
                    "end": 2155.39,
                    "word": " of",
                    "probability": 0.8251953125
                },
                {
                    "start": 2155.39,
                    "end": 2155.79,
                    "word": " concept",
                    "probability": 0.7880859375
                },
                {
                    "start": 2155.79,
                    "end": 2156.07,
                    "word": ".",
                    "probability": 0.58203125
                },
                {
                    "start": 2156.07,
                    "end": 2156.27,
                    "word": " We",
                    "probability": 0.8818359375
                },
                {
                    "start": 2156.27,
                    "end": 2157.29,
                    "word": " consider",
                    "probability": 0.5556640625
                },
                {
                    "start": 2157.29,
                    "end": 2157.67,
                    "word": " the",
                    "probability": 0.80224609375
                },
                {
                    "start": 2157.67,
                    "end": 2158.43,
                    "word": " game",
                    "probability": 0.64892578125
                },
                {
                    "start": 2158.43,
                    "end": 2158.63,
                    "word": " of",
                    "probability": 0.888671875
                },
                {
                    "start": 2158.63,
                    "end": 2159.11,
                    "word": " Snake",
                    "probability": 0.33203125
                },
                {
                    "start": 2159.11,
                    "end": 2159.51,
                    "word": ".",
                    "probability": 0.55419921875
                },
                {
                    "start": 2159.51,
                    "end": 2159.89,
                    "word": " Many",
                    "probability": 0.8544921875
                },
                {
                    "start": 2159.89,
                    "end": 2160.03,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 2160.03,
                    "end": 2160.13,
                    "word": " you",
                    "probability": 0.89599609375
                },
                {
                    "start": 2160.13,
                    "end": 2160.45,
                    "word": " probably",
                    "probability": 0.88916015625
                },
                {
                    "start": 2160.45,
                    "end": 2160.89,
                    "word": " remember",
                    "probability": 0.74609375
                },
                {
                    "start": 2160.89,
                    "end": 2161.17,
                    "word": " Snake",
                    "probability": 0.7333984375
                },
                {
                    "start": 2161.17,
                    "end": 2161.51,
                    "word": " from",
                    "probability": 0.81787109375
                }
            ]
        },
        {
            "id": 324,
            "seek": 216444,
            "start": 2162.33,
            "end": 2170.57,
            "text": " older Nokia phones. The DQN agent is a snake and is learning to play in this environment.",
            "tokens": [
                51408,
                4906,
                43980,
                10216,
                13,
                440,
                413,
                48,
                45,
                9461,
                307,
                257,
                12650,
                293,
                307,
                2539,
                281,
                862,
                294,
                341,
                2823,
                13,
                51812
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24484535590889528,
            "compression_ratio": 1.548,
            "no_speech_prob": 0.0005884170532226562,
            "words": [
                {
                    "start": 2162.33,
                    "end": 2162.73,
                    "word": " older",
                    "probability": 0.80322265625
                },
                {
                    "start": 2162.73,
                    "end": 2163.05,
                    "word": " Nokia",
                    "probability": 0.80859375
                },
                {
                    "start": 2163.05,
                    "end": 2163.47,
                    "word": " phones",
                    "probability": 0.78955078125
                },
                {
                    "start": 2163.47,
                    "end": 2164.11,
                    "word": ".",
                    "probability": 0.97509765625
                },
                {
                    "start": 2164.31,
                    "end": 2164.55,
                    "word": " The",
                    "probability": 0.8232421875
                },
                {
                    "start": 2164.55,
                    "end": 2165.13,
                    "word": " DQN",
                    "probability": 0.9396158854166666
                },
                {
                    "start": 2165.13,
                    "end": 2165.53,
                    "word": " agent",
                    "probability": 0.78173828125
                },
                {
                    "start": 2165.53,
                    "end": 2166.43,
                    "word": " is",
                    "probability": 0.89990234375
                },
                {
                    "start": 2166.61,
                    "end": 2166.87,
                    "word": " a snake",
                    "probability": 0.70947265625
                },
                {
                    "start": 2166.87,
                    "end": 2167.19,
                    "word": " and",
                    "probability": 0.673828125
                },
                {
                    "start": 2167.19,
                    "end": 2167.33,
                    "word": " is",
                    "probability": 0.7626953125
                },
                {
                    "start": 2167.33,
                    "end": 2167.95,
                    "word": " learning",
                    "probability": 0.7587890625
                },
                {
                    "start": 2167.95,
                    "end": 2168.17,
                    "word": " to",
                    "probability": 0.90625
                },
                {
                    "start": 2168.17,
                    "end": 2168.55,
                    "word": " play",
                    "probability": 0.88134765625
                },
                {
                    "start": 2168.55,
                    "end": 2169.49,
                    "word": " in",
                    "probability": 0.8994140625
                },
                {
                    "start": 2169.49,
                    "end": 2169.67,
                    "word": " this",
                    "probability": 0.8798828125
                },
                {
                    "start": 2169.67,
                    "end": 2170.09,
                    "word": " environment",
                    "probability": 0.75537109375
                },
                {
                    "start": 2170.09,
                    "end": 2170.57,
                    "word": ".",
                    "probability": 0.98974609375
                }
            ]
        },
        {
            "id": 325,
            "seek": 219044,
            "start": 2170.57,
            "end": 2176.45,
            "text": " The attacker adds a drug seed with more instantaneous reward than the typical seed,",
            "tokens": [
                50364,
                440,
                35871,
                10860,
                257,
                4110,
                8871,
                365,
                544,
                45596,
                7782,
                813,
                264,
                7476,
                8871,
                11,
                50648
            ],
            "temperature": 0.0,
            "avg_logprob": -0.28239888347247066,
            "compression_ratio": 1.6428571428571428,
            "no_speech_prob": 0.002216339111328125,
            "words": [
                {
                    "start": 2170.57,
                    "end": 2170.81,
                    "word": " The",
                    "probability": 0.0419921875
                },
                {
                    "start": 2170.81,
                    "end": 2171.19,
                    "word": " attacker",
                    "probability": 0.7724609375
                },
                {
                    "start": 2171.19,
                    "end": 2172.17,
                    "word": " adds",
                    "probability": 0.6669921875
                },
                {
                    "start": 2172.41,
                    "end": 2172.67,
                    "word": " a drug",
                    "probability": 0.11962890625
                },
                {
                    "start": 2172.67,
                    "end": 2173.03,
                    "word": " seed",
                    "probability": 0.8974609375
                },
                {
                    "start": 2173.03,
                    "end": 2173.99,
                    "word": " with",
                    "probability": 0.76025390625
                },
                {
                    "start": 2173.99,
                    "end": 2174.49,
                    "word": " more",
                    "probability": 0.771484375
                },
                {
                    "start": 2174.49,
                    "end": 2175.05,
                    "word": " instantaneous",
                    "probability": 0.8349609375
                },
                {
                    "start": 2175.05,
                    "end": 2175.35,
                    "word": " reward",
                    "probability": 0.7939453125
                },
                {
                    "start": 2175.35,
                    "end": 2175.51,
                    "word": " than",
                    "probability": 0.83837890625
                },
                {
                    "start": 2175.51,
                    "end": 2175.65,
                    "word": " the",
                    "probability": 0.744140625
                },
                {
                    "start": 2175.65,
                    "end": 2175.99,
                    "word": " typical",
                    "probability": 0.76513671875
                },
                {
                    "start": 2175.99,
                    "end": 2176.33,
                    "word": " seed",
                    "probability": 0.78564453125
                },
                {
                    "start": 2176.33,
                    "end": 2176.45,
                    "word": ",",
                    "probability": 0.4677734375
                }
            ]
        },
        {
            "id": 326,
            "seek": 219044,
            "start": 2176.45,
            "end": 2182.65,
            "text": " but it also results in more increase in the length of the tail. You can see that this can end up...",
            "tokens": [
                50680,
                457,
                309,
                611,
                3542,
                294,
                544,
                3488,
                294,
                264,
                4641,
                295,
                264,
                6838,
                13,
                509,
                393,
                536,
                300,
                341,
                393,
                917,
                493,
                485,
                50960
            ],
            "temperature": 0.0,
            "avg_logprob": -0.28239888347247066,
            "compression_ratio": 1.6428571428571428,
            "no_speech_prob": 0.002216339111328125,
            "words": [
                {
                    "start": 2176.45,
                    "end": 2177.19,
                    "word": " but",
                    "probability": 0.8427734375
                },
                {
                    "start": 2177.19,
                    "end": 2177.33,
                    "word": " it",
                    "probability": 0.7255859375
                },
                {
                    "start": 2177.33,
                    "end": 2177.69,
                    "word": " also",
                    "probability": 0.73291015625
                },
                {
                    "start": 2177.69,
                    "end": 2178.11,
                    "word": " results",
                    "probability": 0.73046875
                },
                {
                    "start": 2178.11,
                    "end": 2178.29,
                    "word": " in",
                    "probability": 0.89453125
                },
                {
                    "start": 2178.29,
                    "end": 2178.51,
                    "word": " more",
                    "probability": 0.73193359375
                },
                {
                    "start": 2178.51,
                    "end": 2178.87,
                    "word": " increase",
                    "probability": 0.69482421875
                },
                {
                    "start": 2178.87,
                    "end": 2178.97,
                    "word": " in",
                    "probability": 0.85595703125
                },
                {
                    "start": 2178.97,
                    "end": 2179.05,
                    "word": " the",
                    "probability": 0.6748046875
                },
                {
                    "start": 2179.05,
                    "end": 2179.25,
                    "word": " length",
                    "probability": 0.72119140625
                },
                {
                    "start": 2179.25,
                    "end": 2179.37,
                    "word": " of",
                    "probability": 0.8994140625
                },
                {
                    "start": 2179.37,
                    "end": 2179.47,
                    "word": " the",
                    "probability": 0.806640625
                },
                {
                    "start": 2179.47,
                    "end": 2179.69,
                    "word": " tail",
                    "probability": 0.78173828125
                },
                {
                    "start": 2179.69,
                    "end": 2179.91,
                    "word": ".",
                    "probability": 0.9013671875
                },
                {
                    "start": 2179.91,
                    "end": 2180.07,
                    "word": " You",
                    "probability": 0.40087890625
                },
                {
                    "start": 2180.07,
                    "end": 2180.23,
                    "word": " can",
                    "probability": 0.87841796875
                },
                {
                    "start": 2180.23,
                    "end": 2180.47,
                    "word": " see",
                    "probability": 0.81298828125
                },
                {
                    "start": 2180.47,
                    "end": 2180.79,
                    "word": " that",
                    "probability": 0.7958984375
                },
                {
                    "start": 2180.79,
                    "end": 2181.61,
                    "word": " this",
                    "probability": 0.7900390625
                },
                {
                    "start": 2181.61,
                    "end": 2181.97,
                    "word": " can",
                    "probability": 0.896484375
                },
                {
                    "start": 2181.97,
                    "end": 2182.35,
                    "word": " end",
                    "probability": 0.814453125
                },
                {
                    "start": 2182.35,
                    "end": 2182.65,
                    "word": " up...",
                    "probability": 0.713623046875
                }
            ]
        },
        {
            "id": 327,
            "seek": 219044,
            "start": 2184.73,
            "end": 2196.57,
            "text": " well, if the increase in the tail length is more than a certain amount, then a longer-tailed",
            "tokens": [
                51064,
                731,
                11,
                498,
                264,
                3488,
                294,
                264,
                6838,
                4641,
                307,
                544,
                813,
                257,
                1629,
                2372,
                11,
                550,
                257,
                2854,
                12,
                14430,
                292,
                51660
            ],
            "temperature": 0.0,
            "avg_logprob": -0.28239888347247066,
            "compression_ratio": 1.6428571428571428,
            "no_speech_prob": 0.002216339111328125,
            "words": [
                {
                    "start": 2184.73,
                    "end": 2185.33,
                    "word": " well",
                    "probability": 0.10882568359375
                },
                {
                    "start": 2185.33,
                    "end": 2185.93,
                    "word": ",",
                    "probability": 0.435546875
                },
                {
                    "start": 2185.93,
                    "end": 2186.27,
                    "word": " if",
                    "probability": 0.8779296875
                },
                {
                    "start": 2186.27,
                    "end": 2187.95,
                    "word": " the",
                    "probability": 0.75732421875
                },
                {
                    "start": 2187.95,
                    "end": 2189.49,
                    "word": " increase",
                    "probability": 0.771484375
                },
                {
                    "start": 2189.49,
                    "end": 2189.65,
                    "word": " in",
                    "probability": 0.859375
                },
                {
                    "start": 2189.65,
                    "end": 2189.83,
                    "word": " the",
                    "probability": 0.79345703125
                },
                {
                    "start": 2189.83,
                    "end": 2190.41,
                    "word": " tail",
                    "probability": 0.7548828125
                },
                {
                    "start": 2190.41,
                    "end": 2190.71,
                    "word": " length",
                    "probability": 0.7568359375
                },
                {
                    "start": 2190.71,
                    "end": 2191.05,
                    "word": " is",
                    "probability": 0.90283203125
                },
                {
                    "start": 2191.05,
                    "end": 2191.33,
                    "word": " more",
                    "probability": 0.8583984375
                },
                {
                    "start": 2191.33,
                    "end": 2191.55,
                    "word": " than",
                    "probability": 0.8720703125
                },
                {
                    "start": 2191.73,
                    "end": 2192.03,
                    "word": " a certain",
                    "probability": 0.81640625
                },
                {
                    "start": 2192.03,
                    "end": 2193.21,
                    "word": " amount",
                    "probability": 0.837890625
                },
                {
                    "start": 2193.21,
                    "end": 2194.11,
                    "word": ",",
                    "probability": 0.8623046875
                },
                {
                    "start": 2194.11,
                    "end": 2194.35,
                    "word": " then",
                    "probability": 0.6201171875
                },
                {
                    "start": 2195.67,
                    "end": 2195.89,
                    "word": " a longer",
                    "probability": 0.82666015625
                },
                {
                    "start": 2195.89,
                    "end": 2196.57,
                    "word": "-tailed",
                    "probability": 0.8056640625
                }
            ]
        },
        {
            "id": 328,
            "seek": 221792,
            "start": 2196.57,
            "end": 2205.17,
            "text": " snake is bound to eat its own tail sooner rather than later. So, we show that we actually derive",
            "tokens": [
                50364,
                12650,
                307,
                5472,
                281,
                1862,
                1080,
                1065,
                6838,
                15324,
                2831,
                813,
                1780,
                13,
                407,
                11,
                321,
                855,
                300,
                321,
                767,
                28446,
                50796
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23933823529411766,
            "compression_ratio": 1.6782608695652175,
            "no_speech_prob": 0.00701141357421875,
            "words": [
                {
                    "start": 2196.57,
                    "end": 2196.87,
                    "word": " snake",
                    "probability": 0.0469970703125
                },
                {
                    "start": 2196.87,
                    "end": 2197.03,
                    "word": " is",
                    "probability": 0.90380859375
                },
                {
                    "start": 2197.03,
                    "end": 2197.41,
                    "word": " bound",
                    "probability": 0.8359375
                },
                {
                    "start": 2197.41,
                    "end": 2198.39,
                    "word": " to",
                    "probability": 0.90283203125
                },
                {
                    "start": 2198.39,
                    "end": 2198.83,
                    "word": " eat",
                    "probability": 0.87939453125
                },
                {
                    "start": 2198.83,
                    "end": 2198.99,
                    "word": " its",
                    "probability": 0.7939453125
                },
                {
                    "start": 2198.99,
                    "end": 2199.21,
                    "word": " own",
                    "probability": 0.77294921875
                },
                {
                    "start": 2199.21,
                    "end": 2199.43,
                    "word": " tail",
                    "probability": 0.72509765625
                },
                {
                    "start": 2199.43,
                    "end": 2199.91,
                    "word": " sooner",
                    "probability": 0.8349609375
                },
                {
                    "start": 2199.91,
                    "end": 2200.71,
                    "word": " rather",
                    "probability": 0.69287109375
                },
                {
                    "start": 2200.71,
                    "end": 2200.91,
                    "word": " than",
                    "probability": 0.8818359375
                },
                {
                    "start": 2200.91,
                    "end": 2201.19,
                    "word": " later",
                    "probability": 0.8994140625
                },
                {
                    "start": 2201.19,
                    "end": 2201.79,
                    "word": ".",
                    "probability": 0.943359375
                },
                {
                    "start": 2202.09,
                    "end": 2202.47,
                    "word": " So",
                    "probability": 0.83447265625
                },
                {
                    "start": 2202.47,
                    "end": 2202.67,
                    "word": ",",
                    "probability": 0.1910400390625
                },
                {
                    "start": 2202.67,
                    "end": 2202.93,
                    "word": " we",
                    "probability": 0.90673828125
                },
                {
                    "start": 2202.93,
                    "end": 2203.25,
                    "word": " show",
                    "probability": 0.818359375
                },
                {
                    "start": 2203.25,
                    "end": 2203.59,
                    "word": " that",
                    "probability": 0.884765625
                },
                {
                    "start": 2203.59,
                    "end": 2204.45,
                    "word": " we",
                    "probability": 0.70361328125
                },
                {
                    "start": 2204.45,
                    "end": 2204.85,
                    "word": " actually",
                    "probability": 0.7666015625
                },
                {
                    "start": 2204.85,
                    "end": 2205.17,
                    "word": " derive",
                    "probability": 0.60791015625
                }
            ]
        },
        {
            "id": 329,
            "seek": 221792,
            "start": 2205.17,
            "end": 2213.09,
            "text": " some theoretical closed-form solutions for what the additional reward and increase in the tail",
            "tokens": [
                50796,
                512,
                20864,
                5395,
                12,
                837,
                6547,
                337,
                437,
                264,
                4497,
                7782,
                293,
                3488,
                294,
                264,
                6838,
                51188
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23933823529411766,
            "compression_ratio": 1.6782608695652175,
            "no_speech_prob": 0.00701141357421875,
            "words": [
                {
                    "start": 2205.17,
                    "end": 2205.63,
                    "word": " some",
                    "probability": 0.7763671875
                },
                {
                    "start": 2205.63,
                    "end": 2206.11,
                    "word": " theoretical",
                    "probability": 0.72265625
                },
                {
                    "start": 2206.11,
                    "end": 2206.41,
                    "word": " closed",
                    "probability": 0.71630859375
                },
                {
                    "start": 2206.41,
                    "end": 2206.53,
                    "word": "-form",
                    "probability": 0.5906982421875
                },
                {
                    "start": 2206.53,
                    "end": 2206.85,
                    "word": " solutions",
                    "probability": 0.80419921875
                },
                {
                    "start": 2206.85,
                    "end": 2207.09,
                    "word": " for",
                    "probability": 0.89501953125
                },
                {
                    "start": 2207.09,
                    "end": 2207.51,
                    "word": " what",
                    "probability": 0.876953125
                },
                {
                    "start": 2207.51,
                    "end": 2207.81,
                    "word": " the",
                    "probability": 0.81689453125
                },
                {
                    "start": 2207.81,
                    "end": 2209.17,
                    "word": " additional",
                    "probability": 0.861328125
                },
                {
                    "start": 2209.17,
                    "end": 2209.97,
                    "word": " reward",
                    "probability": 0.86279296875
                },
                {
                    "start": 2209.97,
                    "end": 2210.47,
                    "word": " and",
                    "probability": 0.88818359375
                },
                {
                    "start": 2210.47,
                    "end": 2210.95,
                    "word": " increase",
                    "probability": 0.7919921875
                },
                {
                    "start": 2210.95,
                    "end": 2211.41,
                    "word": " in",
                    "probability": 0.90185546875
                },
                {
                    "start": 2211.41,
                    "end": 2212.19,
                    "word": " the",
                    "probability": 0.8056640625
                },
                {
                    "start": 2212.19,
                    "end": 2213.09,
                    "word": " tail",
                    "probability": 0.74853515625
                }
            ]
        },
        {
            "id": 330,
            "seek": 221792,
            "start": 2213.09,
            "end": 2218.61,
            "text": " length should be for addiction to emerge, meaning that the agent learns a more myopic policy instead",
            "tokens": [
                51188,
                4641,
                820,
                312,
                337,
                16835,
                281,
                21511,
                11,
                3620,
                300,
                264,
                9461,
                27152,
                257,
                544,
                452,
                40216,
                3897,
                2602,
                51464
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23933823529411766,
            "compression_ratio": 1.6782608695652175,
            "no_speech_prob": 0.00701141357421875,
            "words": [
                {
                    "start": 2213.09,
                    "end": 2213.29,
                    "word": " length",
                    "probability": 0.73095703125
                },
                {
                    "start": 2213.29,
                    "end": 2213.53,
                    "word": " should",
                    "probability": 0.93115234375
                },
                {
                    "start": 2213.53,
                    "end": 2213.87,
                    "word": " be",
                    "probability": 0.9091796875
                },
                {
                    "start": 2213.87,
                    "end": 2214.17,
                    "word": " for",
                    "probability": 0.876953125
                },
                {
                    "start": 2214.17,
                    "end": 2214.75,
                    "word": " addiction",
                    "probability": 0.818359375
                },
                {
                    "start": 2214.75,
                    "end": 2215.03,
                    "word": " to",
                    "probability": 0.90576171875
                },
                {
                    "start": 2215.03,
                    "end": 2215.45,
                    "word": " emerge",
                    "probability": 0.79931640625
                },
                {
                    "start": 2215.45,
                    "end": 2215.73,
                    "word": ",",
                    "probability": 0.77880859375
                },
                {
                    "start": 2215.73,
                    "end": 2216.03,
                    "word": " meaning",
                    "probability": 0.8154296875
                },
                {
                    "start": 2216.03,
                    "end": 2216.25,
                    "word": " that",
                    "probability": 0.8740234375
                },
                {
                    "start": 2216.25,
                    "end": 2216.41,
                    "word": " the",
                    "probability": 0.80810546875
                },
                {
                    "start": 2216.41,
                    "end": 2216.61,
                    "word": " agent",
                    "probability": 0.802734375
                },
                {
                    "start": 2216.61,
                    "end": 2216.85,
                    "word": " learns",
                    "probability": 0.84033203125
                },
                {
                    "start": 2216.97,
                    "end": 2217.11,
                    "word": " a more",
                    "probability": 0.87890625
                },
                {
                    "start": 2217.11,
                    "end": 2217.51,
                    "word": " myopic",
                    "probability": 0.97119140625
                },
                {
                    "start": 2217.51,
                    "end": 2218.39,
                    "word": " policy",
                    "probability": 0.806640625
                },
                {
                    "start": 2218.39,
                    "end": 2218.61,
                    "word": " instead",
                    "probability": 0.806640625
                }
            ]
        },
        {
            "id": 331,
            "seek": 221792,
            "start": 2218.61,
            "end": 2224.05,
            "text": " of the optimal policy, and you can see that it's actually possible to make the agent addicted",
            "tokens": [
                51464,
                295,
                264,
                16252,
                3897,
                11,
                293,
                291,
                393,
                536,
                300,
                309,
                311,
                767,
                1944,
                281,
                652,
                264,
                9461,
                24629,
                51744
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23933823529411766,
            "compression_ratio": 1.6782608695652175,
            "no_speech_prob": 0.00701141357421875,
            "words": [
                {
                    "start": 2218.61,
                    "end": 2218.71,
                    "word": " of",
                    "probability": 0.90380859375
                },
                {
                    "start": 2218.71,
                    "end": 2218.85,
                    "word": " the",
                    "probability": 0.80859375
                },
                {
                    "start": 2218.85,
                    "end": 2219.15,
                    "word": " optimal",
                    "probability": 0.7138671875
                },
                {
                    "start": 2219.15,
                    "end": 2219.63,
                    "word": " policy",
                    "probability": 0.8203125
                },
                {
                    "start": 2219.63,
                    "end": 2219.97,
                    "word": ",",
                    "probability": 0.350341796875
                },
                {
                    "start": 2219.97,
                    "end": 2220.37,
                    "word": " and",
                    "probability": 0.89599609375
                },
                {
                    "start": 2220.37,
                    "end": 2220.51,
                    "word": " you",
                    "probability": 0.89599609375
                },
                {
                    "start": 2220.51,
                    "end": 2220.65,
                    "word": " can",
                    "probability": 0.89794921875
                },
                {
                    "start": 2220.65,
                    "end": 2220.83,
                    "word": " see",
                    "probability": 0.81982421875
                },
                {
                    "start": 2220.83,
                    "end": 2221.01,
                    "word": " that",
                    "probability": 0.880859375
                },
                {
                    "start": 2221.01,
                    "end": 2221.23,
                    "word": " it's",
                    "probability": 0.947265625
                },
                {
                    "start": 2221.23,
                    "end": 2221.51,
                    "word": " actually",
                    "probability": 0.78466796875
                },
                {
                    "start": 2221.51,
                    "end": 2221.97,
                    "word": " possible",
                    "probability": 0.86279296875
                },
                {
                    "start": 2221.97,
                    "end": 2222.39,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 2222.39,
                    "end": 2223.13,
                    "word": " make",
                    "probability": 0.8779296875
                },
                {
                    "start": 2223.13,
                    "end": 2223.27,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 2223.27,
                    "end": 2223.53,
                    "word": " agent",
                    "probability": 0.81298828125
                },
                {
                    "start": 2223.53,
                    "end": 2224.05,
                    "word": " addicted",
                    "probability": 0.77099609375
                }
            ]
        },
        {
            "id": 332,
            "seek": 224568,
            "start": 2225.43,
            "end": 2231.55,
            "text": " to the drug seed, as we call it, and this results in learning a sub-optimal policy.",
            "tokens": [
                50392,
                281,
                264,
                4110,
                8871,
                11,
                382,
                321,
                818,
                309,
                11,
                293,
                341,
                3542,
                294,
                2539,
                257,
                1422,
                12,
                5747,
                10650,
                3897,
                13,
                50716
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2522321356521858,
            "compression_ratio": 1.5462962962962963,
            "no_speech_prob": 0.00017404556274414062,
            "words": [
                {
                    "start": 2225.43,
                    "end": 2225.91,
                    "word": " to",
                    "probability": 0.55126953125
                },
                {
                    "start": 2225.91,
                    "end": 2226.39,
                    "word": " the",
                    "probability": 0.70703125
                },
                {
                    "start": 2226.39,
                    "end": 2226.83,
                    "word": " drug",
                    "probability": 0.44970703125
                },
                {
                    "start": 2226.83,
                    "end": 2227.01,
                    "word": " seed",
                    "probability": 0.7998046875
                },
                {
                    "start": 2227.01,
                    "end": 2227.11,
                    "word": ",",
                    "probability": 0.59326171875
                },
                {
                    "start": 2227.11,
                    "end": 2227.19,
                    "word": " as",
                    "probability": 0.900390625
                },
                {
                    "start": 2227.19,
                    "end": 2227.31,
                    "word": " we",
                    "probability": 0.8984375
                },
                {
                    "start": 2227.31,
                    "end": 2227.51,
                    "word": " call",
                    "probability": 0.80712890625
                },
                {
                    "start": 2227.51,
                    "end": 2227.73,
                    "word": " it",
                    "probability": 0.90625
                },
                {
                    "start": 2227.73,
                    "end": 2228.09,
                    "word": ",",
                    "probability": 0.380126953125
                },
                {
                    "start": 2228.09,
                    "end": 2228.59,
                    "word": " and",
                    "probability": 0.89111328125
                },
                {
                    "start": 2228.59,
                    "end": 2229.13,
                    "word": " this",
                    "probability": 0.8857421875
                },
                {
                    "start": 2229.13,
                    "end": 2229.41,
                    "word": " results",
                    "probability": 0.66845703125
                },
                {
                    "start": 2229.41,
                    "end": 2229.53,
                    "word": " in",
                    "probability": 0.90576171875
                },
                {
                    "start": 2229.53,
                    "end": 2229.77,
                    "word": " learning",
                    "probability": 0.8408203125
                },
                {
                    "start": 2229.99,
                    "end": 2230.21,
                    "word": " a sub",
                    "probability": 0.8271484375
                },
                {
                    "start": 2230.21,
                    "end": 2230.57,
                    "word": "-optimal",
                    "probability": 0.7113444010416666
                },
                {
                    "start": 2230.57,
                    "end": 2231.07,
                    "word": " policy",
                    "probability": 0.81103515625
                },
                {
                    "start": 2231.07,
                    "end": 2231.55,
                    "word": ".",
                    "probability": 0.9521484375
                }
            ]
        },
        {
            "id": 333,
            "seek": 224568,
            "start": 2232.39,
            "end": 2239.51,
            "text": " And, of course, due to time limitations, I'm going to introduce only one more type of attack,",
            "tokens": [
                50772,
                400,
                11,
                295,
                1164,
                11,
                3462,
                281,
                565,
                15705,
                11,
                286,
                478,
                516,
                281,
                5366,
                787,
                472,
                544,
                2010,
                295,
                2690,
                11,
                51128
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2522321356521858,
            "compression_ratio": 1.5462962962962963,
            "no_speech_prob": 0.00017404556274414062,
            "words": [
                {
                    "start": 2232.39,
                    "end": 2232.75,
                    "word": " And",
                    "probability": 0.853515625
                },
                {
                    "start": 2232.75,
                    "end": 2232.93,
                    "word": ",",
                    "probability": 0.44970703125
                },
                {
                    "start": 2232.93,
                    "end": 2233.93,
                    "word": " of",
                    "probability": 0.89599609375
                },
                {
                    "start": 2233.93,
                    "end": 2234.15,
                    "word": " course",
                    "probability": 0.8564453125
                },
                {
                    "start": 2234.15,
                    "end": 2234.55,
                    "word": ",",
                    "probability": 0.94482421875
                },
                {
                    "start": 2234.55,
                    "end": 2235.93,
                    "word": " due",
                    "probability": 0.90576171875
                },
                {
                    "start": 2235.93,
                    "end": 2236.29,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 2236.29,
                    "end": 2236.65,
                    "word": " time",
                    "probability": 0.796875
                },
                {
                    "start": 2236.65,
                    "end": 2237.05,
                    "word": " limitations",
                    "probability": 0.6357421875
                },
                {
                    "start": 2237.05,
                    "end": 2237.37,
                    "word": ",",
                    "probability": 0.8505859375
                },
                {
                    "start": 2237.37,
                    "end": 2237.53,
                    "word": " I'm",
                    "probability": 0.990966796875
                },
                {
                    "start": 2237.53,
                    "end": 2237.67,
                    "word": " going",
                    "probability": 0.857421875
                },
                {
                    "start": 2237.67,
                    "end": 2237.89,
                    "word": " to",
                    "probability": 0.90185546875
                },
                {
                    "start": 2237.89,
                    "end": 2238.17,
                    "word": " introduce",
                    "probability": 0.66796875
                },
                {
                    "start": 2238.17,
                    "end": 2238.31,
                    "word": " only",
                    "probability": 0.7763671875
                },
                {
                    "start": 2238.31,
                    "end": 2238.57,
                    "word": " one",
                    "probability": 0.81689453125
                },
                {
                    "start": 2238.57,
                    "end": 2238.77,
                    "word": " more",
                    "probability": 0.87744140625
                },
                {
                    "start": 2238.77,
                    "end": 2238.97,
                    "word": " type",
                    "probability": 0.9111328125
                },
                {
                    "start": 2238.97,
                    "end": 2239.07,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 2239.07,
                    "end": 2239.37,
                    "word": " attack",
                    "probability": 0.8701171875
                },
                {
                    "start": 2239.37,
                    "end": 2239.51,
                    "word": ",",
                    "probability": 0.51611328125
                }
            ]
        },
        {
            "id": 334,
            "seek": 224568,
            "start": 2239.51,
            "end": 2246.33,
            "text": " and it's that of targeting the confidentiality of a deep RL policy.",
            "tokens": [
                51128,
                293,
                309,
                311,
                300,
                295,
                17918,
                264,
                27054,
                507,
                295,
                257,
                2452,
                497,
                43,
                3897,
                13,
                51456
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2522321356521858,
            "compression_ratio": 1.5462962962962963,
            "no_speech_prob": 0.00017404556274414062,
            "words": [
                {
                    "start": 2239.51,
                    "end": 2239.61,
                    "word": " and",
                    "probability": 0.89306640625
                },
                {
                    "start": 2239.61,
                    "end": 2239.85,
                    "word": " it's",
                    "probability": 0.943603515625
                },
                {
                    "start": 2239.85,
                    "end": 2240.03,
                    "word": " that",
                    "probability": 0.87548828125
                },
                {
                    "start": 2240.03,
                    "end": 2240.41,
                    "word": " of",
                    "probability": 0.89453125
                },
                {
                    "start": 2240.41,
                    "end": 2241.19,
                    "word": " targeting",
                    "probability": 0.78271484375
                },
                {
                    "start": 2241.19,
                    "end": 2241.37,
                    "word": " the",
                    "probability": 0.81201171875
                },
                {
                    "start": 2241.37,
                    "end": 2242.43,
                    "word": " confidentiality",
                    "probability": 0.906982421875
                },
                {
                    "start": 2242.43,
                    "end": 2243.59,
                    "word": " of",
                    "probability": 0.9013671875
                },
                {
                    "start": 2245.11,
                    "end": 2245.25,
                    "word": " a deep",
                    "probability": 0.54296875
                },
                {
                    "start": 2245.25,
                    "end": 2245.43,
                    "word": " RL",
                    "probability": 0.548126220703125
                },
                {
                    "start": 2245.43,
                    "end": 2245.85,
                    "word": " policy",
                    "probability": 0.75537109375
                },
                {
                    "start": 2245.85,
                    "end": 2246.33,
                    "word": ".",
                    "probability": 0.98876953125
                }
            ]
        },
        {
            "id": 335,
            "seek": 224568,
            "start": 2247.23,
            "end": 2251.81,
            "text": " The problem here is, or the question is, is it possible to extract a deep RL policy from",
            "tokens": [
                51536,
                440,
                1154,
                510,
                307,
                11,
                420,
                264,
                1168,
                307,
                11,
                307,
                309,
                1944,
                281,
                8947,
                257,
                2452,
                497,
                43,
                3897,
                490,
                51752
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2522321356521858,
            "compression_ratio": 1.5462962962962963,
            "no_speech_prob": 0.00017404556274414062,
            "words": [
                {
                    "start": 2247.23,
                    "end": 2247.71,
                    "word": " The",
                    "probability": 0.81982421875
                },
                {
                    "start": 2247.71,
                    "end": 2247.99,
                    "word": " problem",
                    "probability": 0.677734375
                },
                {
                    "start": 2247.99,
                    "end": 2248.33,
                    "word": " here",
                    "probability": 0.79296875
                },
                {
                    "start": 2248.33,
                    "end": 2249.01,
                    "word": " is",
                    "probability": 0.837890625
                },
                {
                    "start": 2249.01,
                    "end": 2249.15,
                    "word": ",",
                    "probability": 0.63671875
                },
                {
                    "start": 2249.15,
                    "end": 2249.25,
                    "word": " or",
                    "probability": 0.89306640625
                },
                {
                    "start": 2249.25,
                    "end": 2249.35,
                    "word": " the",
                    "probability": 0.80859375
                },
                {
                    "start": 2249.35,
                    "end": 2249.55,
                    "word": " question",
                    "probability": 0.79443359375
                },
                {
                    "start": 2249.55,
                    "end": 2249.83,
                    "word": " is",
                    "probability": 0.88671875
                },
                {
                    "start": 2249.83,
                    "end": 2249.87,
                    "word": ",",
                    "probability": 0.95849609375
                },
                {
                    "start": 2249.87,
                    "end": 2249.99,
                    "word": " is",
                    "probability": 0.85888671875
                },
                {
                    "start": 2249.99,
                    "end": 2250.11,
                    "word": " it",
                    "probability": 0.8994140625
                },
                {
                    "start": 2250.11,
                    "end": 2250.39,
                    "word": " possible",
                    "probability": 0.828125
                },
                {
                    "start": 2250.39,
                    "end": 2250.55,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 2250.55,
                    "end": 2250.83,
                    "word": " extract",
                    "probability": 0.861328125
                },
                {
                    "start": 2250.95,
                    "end": 2251.09,
                    "word": " a deep",
                    "probability": 0.91455078125
                },
                {
                    "start": 2251.09,
                    "end": 2251.25,
                    "word": " RL",
                    "probability": 0.969970703125
                },
                {
                    "start": 2251.25,
                    "end": 2251.59,
                    "word": " policy",
                    "probability": 0.81298828125
                },
                {
                    "start": 2251.59,
                    "end": 2251.81,
                    "word": " from",
                    "probability": 0.81396484375
                }
            ]
        },
        {
            "id": 336,
            "seek": 227140,
            "start": 2251.81,
            "end": 2258.83,
            "text": " observations of its actions? Why does this matter? Well, the security challenge posed by",
            "tokens": [
                50364,
                18163,
                295,
                1080,
                5909,
                30,
                1545,
                775,
                341,
                1871,
                30,
                1042,
                11,
                264,
                3825,
                3430,
                31399,
                538,
                50708
            ],
            "temperature": 0.0,
            "avg_logprob": -0.16831596708960003,
            "compression_ratio": 1.5403225806451613,
            "no_speech_prob": 0.00010389089584350586,
            "words": [
                {
                    "start": 2251.81,
                    "end": 2252.43,
                    "word": " observations",
                    "probability": 0.332275390625
                },
                {
                    "start": 2252.43,
                    "end": 2252.77,
                    "word": " of",
                    "probability": 0.8994140625
                },
                {
                    "start": 2252.77,
                    "end": 2252.91,
                    "word": " its",
                    "probability": 0.81982421875
                },
                {
                    "start": 2252.91,
                    "end": 2253.45,
                    "word": " actions",
                    "probability": 0.86572265625
                },
                {
                    "start": 2253.45,
                    "end": 2254.01,
                    "word": "?",
                    "probability": 0.03338623046875
                },
                {
                    "start": 2254.73,
                    "end": 2255.29,
                    "word": " Why",
                    "probability": 0.755859375
                },
                {
                    "start": 2255.29,
                    "end": 2255.43,
                    "word": " does",
                    "probability": 0.92578125
                },
                {
                    "start": 2255.43,
                    "end": 2255.57,
                    "word": " this",
                    "probability": 0.8798828125
                },
                {
                    "start": 2255.57,
                    "end": 2255.79,
                    "word": " matter",
                    "probability": 0.81787109375
                },
                {
                    "start": 2255.79,
                    "end": 2256.07,
                    "word": "?",
                    "probability": 0.9873046875
                },
                {
                    "start": 2256.07,
                    "end": 2256.25,
                    "word": " Well",
                    "probability": 0.72607421875
                },
                {
                    "start": 2256.25,
                    "end": 2256.35,
                    "word": ",",
                    "probability": 0.921875
                },
                {
                    "start": 2256.35,
                    "end": 2256.49,
                    "word": " the",
                    "probability": 0.78662109375
                },
                {
                    "start": 2256.49,
                    "end": 2256.77,
                    "word": " security",
                    "probability": 0.92578125
                },
                {
                    "start": 2256.77,
                    "end": 2257.33,
                    "word": " challenge",
                    "probability": 0.763671875
                },
                {
                    "start": 2257.33,
                    "end": 2258.09,
                    "word": " posed",
                    "probability": 0.81494140625
                },
                {
                    "start": 2258.09,
                    "end": 2258.83,
                    "word": " by",
                    "probability": 0.91552734375
                }
            ]
        },
        {
            "id": 337,
            "seek": 227140,
            "start": 2259.57,
            "end": 2265.51,
            "text": " this sort of action is, of course, model theft. A company like Google or, let's say, Uber or Waymo",
            "tokens": [
                50752,
                341,
                1333,
                295,
                3069,
                307,
                11,
                295,
                1164,
                11,
                2316,
                28508,
                13,
                316,
                2237,
                411,
                3329,
                420,
                11,
                718,
                311,
                584,
                11,
                21839,
                420,
                9558,
                3280,
                51048
            ],
            "temperature": 0.0,
            "avg_logprob": -0.16831596708960003,
            "compression_ratio": 1.5403225806451613,
            "no_speech_prob": 0.00010389089584350586,
            "words": [
                {
                    "start": 2259.57,
                    "end": 2259.85,
                    "word": " this",
                    "probability": 0.873046875
                },
                {
                    "start": 2259.85,
                    "end": 2260.01,
                    "word": " sort",
                    "probability": 0.82080078125
                },
                {
                    "start": 2260.01,
                    "end": 2260.13,
                    "word": " of",
                    "probability": 0.8984375
                },
                {
                    "start": 2260.13,
                    "end": 2260.47,
                    "word": " action",
                    "probability": 0.7978515625
                },
                {
                    "start": 2260.47,
                    "end": 2260.93,
                    "word": " is",
                    "probability": 0.87548828125
                },
                {
                    "start": 2260.93,
                    "end": 2261.03,
                    "word": ",",
                    "probability": 0.5146484375
                },
                {
                    "start": 2261.03,
                    "end": 2261.13,
                    "word": " of",
                    "probability": 0.8974609375
                },
                {
                    "start": 2261.13,
                    "end": 2261.21,
                    "word": " course",
                    "probability": 0.8642578125
                },
                {
                    "start": 2261.21,
                    "end": 2261.27,
                    "word": ",",
                    "probability": 0.9384765625
                },
                {
                    "start": 2261.27,
                    "end": 2261.51,
                    "word": " model",
                    "probability": 0.7998046875
                },
                {
                    "start": 2261.51,
                    "end": 2261.89,
                    "word": " theft",
                    "probability": 0.7998046875
                },
                {
                    "start": 2261.89,
                    "end": 2262.17,
                    "word": ".",
                    "probability": 0.53515625
                },
                {
                    "start": 2262.17,
                    "end": 2262.37,
                    "word": " A",
                    "probability": 0.98193359375
                },
                {
                    "start": 2262.37,
                    "end": 2262.85,
                    "word": " company",
                    "probability": 0.796875
                },
                {
                    "start": 2262.85,
                    "end": 2263.09,
                    "word": " like",
                    "probability": 0.8681640625
                },
                {
                    "start": 2263.09,
                    "end": 2263.39,
                    "word": " Google",
                    "probability": 0.88134765625
                },
                {
                    "start": 2263.39,
                    "end": 2263.97,
                    "word": " or",
                    "probability": 0.43017578125
                },
                {
                    "start": 2263.97,
                    "end": 2264.17,
                    "word": ",",
                    "probability": 0.31494140625
                },
                {
                    "start": 2264.17,
                    "end": 2264.25,
                    "word": " let's",
                    "probability": 0.9404296875
                },
                {
                    "start": 2264.25,
                    "end": 2264.47,
                    "word": " say",
                    "probability": 0.8916015625
                },
                {
                    "start": 2264.47,
                    "end": 2264.69,
                    "word": ",",
                    "probability": 0.91796875
                },
                {
                    "start": 2264.69,
                    "end": 2264.97,
                    "word": " Uber",
                    "probability": 0.8837890625
                },
                {
                    "start": 2264.97,
                    "end": 2265.23,
                    "word": " or",
                    "probability": 0.83740234375
                },
                {
                    "start": 2265.23,
                    "end": 2265.51,
                    "word": " Waymo",
                    "probability": 0.96875
                }
            ]
        },
        {
            "id": 338,
            "seek": 227140,
            "start": 2265.51,
            "end": 2271.87,
            "text": " may have spent billions or millions of dollars on coming up with a very accurate deep RL policy for",
            "tokens": [
                51048,
                815,
                362,
                4418,
                17375,
                420,
                6803,
                295,
                3808,
                322,
                1348,
                493,
                365,
                257,
                588,
                8559,
                2452,
                497,
                43,
                3897,
                337,
                51364
            ],
            "temperature": 0.0,
            "avg_logprob": -0.16831596708960003,
            "compression_ratio": 1.5403225806451613,
            "no_speech_prob": 0.00010389089584350586,
            "words": [
                {
                    "start": 2265.51,
                    "end": 2265.69,
                    "word": " may",
                    "probability": 0.759765625
                },
                {
                    "start": 2265.69,
                    "end": 2265.87,
                    "word": " have",
                    "probability": 0.88623046875
                },
                {
                    "start": 2265.87,
                    "end": 2266.11,
                    "word": " spent",
                    "probability": 0.8125
                },
                {
                    "start": 2266.11,
                    "end": 2266.67,
                    "word": " billions",
                    "probability": 0.85205078125
                },
                {
                    "start": 2266.67,
                    "end": 2266.83,
                    "word": " or",
                    "probability": 0.841796875
                },
                {
                    "start": 2266.83,
                    "end": 2267.03,
                    "word": " millions",
                    "probability": 0.84033203125
                },
                {
                    "start": 2267.03,
                    "end": 2267.19,
                    "word": " of",
                    "probability": 0.8984375
                },
                {
                    "start": 2267.19,
                    "end": 2267.39,
                    "word": " dollars",
                    "probability": 0.9150390625
                },
                {
                    "start": 2267.39,
                    "end": 2267.65,
                    "word": " on",
                    "probability": 0.86181640625
                },
                {
                    "start": 2267.65,
                    "end": 2268.25,
                    "word": " coming",
                    "probability": 0.79052734375
                },
                {
                    "start": 2268.25,
                    "end": 2268.51,
                    "word": " up",
                    "probability": 0.8955078125
                },
                {
                    "start": 2268.51,
                    "end": 2268.99,
                    "word": " with",
                    "probability": 0.8173828125
                },
                {
                    "start": 2269.49,
                    "end": 2269.93,
                    "word": " a very",
                    "probability": 0.74755859375
                },
                {
                    "start": 2269.93,
                    "end": 2270.45,
                    "word": " accurate",
                    "probability": 0.69970703125
                },
                {
                    "start": 2270.45,
                    "end": 2270.75,
                    "word": " deep",
                    "probability": 0.771484375
                },
                {
                    "start": 2270.75,
                    "end": 2271.05,
                    "word": " RL",
                    "probability": 0.7587890625
                },
                {
                    "start": 2271.05,
                    "end": 2271.53,
                    "word": " policy",
                    "probability": 0.7646484375
                },
                {
                    "start": 2271.53,
                    "end": 2271.87,
                    "word": " for",
                    "probability": 0.90185546875
                }
            ]
        },
        {
            "id": 339,
            "seek": 227140,
            "start": 2271.87,
            "end": 2277.53,
            "text": " autonomous navigation, and if it can be stolen by an adversary, then the intellectual property",
            "tokens": [
                51364,
                23797,
                17346,
                11,
                293,
                498,
                309,
                393,
                312,
                15900,
                538,
                364,
                48222,
                11,
                550,
                264,
                12576,
                4707,
                51652
            ],
            "temperature": 0.0,
            "avg_logprob": -0.16831596708960003,
            "compression_ratio": 1.5403225806451613,
            "no_speech_prob": 0.00010389089584350586,
            "words": [
                {
                    "start": 2271.87,
                    "end": 2272.21,
                    "word": " autonomous",
                    "probability": 0.8544921875
                },
                {
                    "start": 2272.21,
                    "end": 2272.67,
                    "word": " navigation",
                    "probability": 0.73291015625
                },
                {
                    "start": 2272.67,
                    "end": 2272.91,
                    "word": ",",
                    "probability": 0.2939453125
                },
                {
                    "start": 2272.91,
                    "end": 2273.23,
                    "word": " and",
                    "probability": 0.89501953125
                },
                {
                    "start": 2273.23,
                    "end": 2273.33,
                    "word": " if",
                    "probability": 0.90087890625
                },
                {
                    "start": 2273.33,
                    "end": 2273.41,
                    "word": " it",
                    "probability": 0.90625
                },
                {
                    "start": 2273.41,
                    "end": 2273.57,
                    "word": " can",
                    "probability": 0.90234375
                },
                {
                    "start": 2273.57,
                    "end": 2273.69,
                    "word": " be",
                    "probability": 0.86474609375
                },
                {
                    "start": 2273.69,
                    "end": 2274.01,
                    "word": " stolen",
                    "probability": 0.75146484375
                },
                {
                    "start": 2274.01,
                    "end": 2275.01,
                    "word": " by",
                    "probability": 0.90283203125
                },
                {
                    "start": 2275.01,
                    "end": 2275.83,
                    "word": " an",
                    "probability": 0.90478515625
                },
                {
                    "start": 2275.83,
                    "end": 2276.31,
                    "word": " adversary",
                    "probability": 0.79931640625
                },
                {
                    "start": 2276.31,
                    "end": 2276.59,
                    "word": ",",
                    "probability": 0.84228515625
                },
                {
                    "start": 2276.59,
                    "end": 2276.71,
                    "word": " then",
                    "probability": 0.69091796875
                },
                {
                    "start": 2276.71,
                    "end": 2276.83,
                    "word": " the",
                    "probability": 0.424072265625
                },
                {
                    "start": 2276.83,
                    "end": 2277.17,
                    "word": " intellectual",
                    "probability": 0.88525390625
                },
                {
                    "start": 2277.17,
                    "end": 2277.53,
                    "word": " property",
                    "probability": 0.8173828125
                }
            ]
        },
        {
            "id": 340,
            "seek": 229880,
            "start": 2277.53,
            "end": 2284.33,
            "text": " becomes worthless. And also, a stolen policy, an extracted policy, can be leveraged in integrity",
            "tokens": [
                50364,
                3643,
                34857,
                13,
                400,
                611,
                11,
                257,
                15900,
                3897,
                11,
                364,
                34086,
                3897,
                11,
                393,
                312,
                12451,
                2980,
                294,
                16000,
                50704
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20279605702350015,
            "compression_ratio": 1.5875,
            "no_speech_prob": 0.00037407875061035156,
            "words": [
                {
                    "start": 2277.53,
                    "end": 2277.77,
                    "word": " becomes",
                    "probability": 0.740234375
                },
                {
                    "start": 2277.77,
                    "end": 2278.25,
                    "word": " worthless",
                    "probability": 0.85107421875
                },
                {
                    "start": 2278.25,
                    "end": 2278.79,
                    "word": ".",
                    "probability": 0.80859375
                },
                {
                    "start": 2279.27,
                    "end": 2279.35,
                    "word": " And",
                    "probability": 0.74169921875
                },
                {
                    "start": 2279.35,
                    "end": 2279.79,
                    "word": " also",
                    "probability": 0.8046875
                },
                {
                    "start": 2279.79,
                    "end": 2280.41,
                    "word": ",",
                    "probability": 0.55859375
                },
                {
                    "start": 2280.63,
                    "end": 2281.03,
                    "word": " a stolen",
                    "probability": 0.7353515625
                },
                {
                    "start": 2281.03,
                    "end": 2281.73,
                    "word": " policy",
                    "probability": 0.75341796875
                },
                {
                    "start": 2281.73,
                    "end": 2281.85,
                    "word": ",",
                    "probability": 0.4404296875
                },
                {
                    "start": 2281.85,
                    "end": 2281.91,
                    "word": " an",
                    "probability": 0.87548828125
                },
                {
                    "start": 2281.91,
                    "end": 2282.25,
                    "word": " extracted",
                    "probability": 0.572265625
                },
                {
                    "start": 2282.25,
                    "end": 2282.65,
                    "word": " policy",
                    "probability": 0.818359375
                },
                {
                    "start": 2282.65,
                    "end": 2282.75,
                    "word": ",",
                    "probability": 0.364013671875
                },
                {
                    "start": 2282.75,
                    "end": 2282.83,
                    "word": " can",
                    "probability": 0.89111328125
                },
                {
                    "start": 2282.83,
                    "end": 2282.99,
                    "word": " be",
                    "probability": 0.89404296875
                },
                {
                    "start": 2282.99,
                    "end": 2283.45,
                    "word": " leveraged",
                    "probability": 0.904541015625
                },
                {
                    "start": 2283.45,
                    "end": 2283.81,
                    "word": " in",
                    "probability": 0.86181640625
                },
                {
                    "start": 2283.81,
                    "end": 2284.33,
                    "word": " integrity",
                    "probability": 0.7197265625
                }
            ]
        },
        {
            "id": 341,
            "seek": 229880,
            "start": 2284.33,
            "end": 2294.23,
            "text": " attacks in the same way that we mounted black box attacks on DQN policies. So, let's see.",
            "tokens": [
                50704,
                8122,
                294,
                264,
                912,
                636,
                300,
                321,
                19138,
                2211,
                2424,
                8122,
                322,
                413,
                48,
                45,
                7657,
                13,
                407,
                11,
                718,
                311,
                536,
                13,
                51196
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20279605702350015,
            "compression_ratio": 1.5875,
            "no_speech_prob": 0.00037407875061035156,
            "words": [
                {
                    "start": 2284.33,
                    "end": 2284.71,
                    "word": " attacks",
                    "probability": 0.71533203125
                },
                {
                    "start": 2284.71,
                    "end": 2284.79,
                    "word": " in",
                    "probability": 0.7626953125
                },
                {
                    "start": 2284.79,
                    "end": 2284.93,
                    "word": " the",
                    "probability": 0.81103515625
                },
                {
                    "start": 2284.93,
                    "end": 2285.09,
                    "word": " same",
                    "probability": 0.8095703125
                },
                {
                    "start": 2285.09,
                    "end": 2285.35,
                    "word": " way",
                    "probability": 0.896484375
                },
                {
                    "start": 2285.35,
                    "end": 2286.37,
                    "word": " that",
                    "probability": 0.86572265625
                },
                {
                    "start": 2286.37,
                    "end": 2287.77,
                    "word": " we",
                    "probability": 0.84814453125
                },
                {
                    "start": 2287.77,
                    "end": 2288.11,
                    "word": " mounted",
                    "probability": 0.728515625
                },
                {
                    "start": 2288.11,
                    "end": 2288.47,
                    "word": " black",
                    "probability": 0.80126953125
                },
                {
                    "start": 2288.47,
                    "end": 2288.73,
                    "word": " box",
                    "probability": 0.5888671875
                },
                {
                    "start": 2288.73,
                    "end": 2289.15,
                    "word": " attacks",
                    "probability": 0.73046875
                },
                {
                    "start": 2289.15,
                    "end": 2290.59,
                    "word": " on",
                    "probability": 0.88427734375
                },
                {
                    "start": 2290.59,
                    "end": 2291.23,
                    "word": " DQN",
                    "probability": 0.7672526041666666
                },
                {
                    "start": 2291.23,
                    "end": 2291.77,
                    "word": " policies",
                    "probability": 0.82177734375
                },
                {
                    "start": 2291.77,
                    "end": 2292.31,
                    "word": ".",
                    "probability": 0.99267578125
                },
                {
                    "start": 2292.75,
                    "end": 2292.99,
                    "word": " So",
                    "probability": 0.8740234375
                },
                {
                    "start": 2292.99,
                    "end": 2293.25,
                    "word": ",",
                    "probability": 0.4619140625
                },
                {
                    "start": 2293.25,
                    "end": 2293.63,
                    "word": " let's",
                    "probability": 0.938232421875
                },
                {
                    "start": 2293.63,
                    "end": 2293.81,
                    "word": " see",
                    "probability": 0.8115234375
                },
                {
                    "start": 2293.81,
                    "end": 2294.23,
                    "word": ".",
                    "probability": 0.7158203125
                }
            ]
        },
        {
            "id": 342,
            "seek": 229880,
            "start": 2294.23,
            "end": 2299.61,
            "text": " As it happens, a branch of reinforcement learning, or in general, one solution to the sequential",
            "tokens": [
                51196,
                1018,
                309,
                2314,
                11,
                257,
                9819,
                295,
                29280,
                2539,
                11,
                420,
                294,
                2674,
                11,
                472,
                3827,
                281,
                264,
                42881,
                51468
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20279605702350015,
            "compression_ratio": 1.5875,
            "no_speech_prob": 0.00037407875061035156,
            "words": [
                {
                    "start": 2294.23,
                    "end": 2294.49,
                    "word": " As",
                    "probability": 0.90625
                },
                {
                    "start": 2294.49,
                    "end": 2294.61,
                    "word": " it",
                    "probability": 0.91064453125
                },
                {
                    "start": 2294.61,
                    "end": 2294.99,
                    "word": " happens",
                    "probability": 0.90869140625
                },
                {
                    "start": 2294.99,
                    "end": 2295.39,
                    "word": ",",
                    "probability": 0.9560546875
                },
                {
                    "start": 2295.67,
                    "end": 2296.51,
                    "word": " a branch",
                    "probability": 0.8056640625
                },
                {
                    "start": 2296.51,
                    "end": 2296.83,
                    "word": " of",
                    "probability": 0.89697265625
                },
                {
                    "start": 2296.83,
                    "end": 2297.27,
                    "word": " reinforcement",
                    "probability": 0.8193359375
                },
                {
                    "start": 2297.27,
                    "end": 2297.53,
                    "word": " learning",
                    "probability": 0.83642578125
                },
                {
                    "start": 2297.53,
                    "end": 2297.67,
                    "word": ",",
                    "probability": 0.59912109375
                },
                {
                    "start": 2297.67,
                    "end": 2297.81,
                    "word": " or",
                    "probability": 0.896484375
                },
                {
                    "start": 2297.81,
                    "end": 2298.01,
                    "word": " in",
                    "probability": 0.7763671875
                },
                {
                    "start": 2298.01,
                    "end": 2298.27,
                    "word": " general",
                    "probability": 0.79833984375
                },
                {
                    "start": 2298.27,
                    "end": 2298.37,
                    "word": ",",
                    "probability": 0.58740234375
                },
                {
                    "start": 2298.37,
                    "end": 2298.53,
                    "word": " one",
                    "probability": 0.81689453125
                },
                {
                    "start": 2298.53,
                    "end": 2298.87,
                    "word": " solution",
                    "probability": 0.8271484375
                },
                {
                    "start": 2298.87,
                    "end": 2299.11,
                    "word": " to",
                    "probability": 0.908203125
                },
                {
                    "start": 2299.11,
                    "end": 2299.29,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 2299.29,
                    "end": 2299.61,
                    "word": " sequential",
                    "probability": 0.90478515625
                }
            ]
        },
        {
            "id": 343,
            "seek": 229880,
            "start": 2299.61,
            "end": 2304.93,
            "text": " decision-making problem, is not in the RL domain, but in the supervised learning domain, and it's",
            "tokens": [
                51468,
                3537,
                12,
                12402,
                1154,
                11,
                307,
                406,
                294,
                264,
                497,
                43,
                9274,
                11,
                457,
                294,
                264,
                46533,
                2539,
                9274,
                11,
                293,
                309,
                311,
                51728
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20279605702350015,
            "compression_ratio": 1.5875,
            "no_speech_prob": 0.00037407875061035156,
            "words": [
                {
                    "start": 2299.61,
                    "end": 2299.99,
                    "word": " decision",
                    "probability": 0.85205078125
                },
                {
                    "start": 2299.99,
                    "end": 2300.21,
                    "word": "-making",
                    "probability": 0.695068359375
                },
                {
                    "start": 2300.21,
                    "end": 2300.71,
                    "word": " problem",
                    "probability": 0.77490234375
                },
                {
                    "start": 2300.71,
                    "end": 2301.37,
                    "word": ",",
                    "probability": 0.78955078125
                },
                {
                    "start": 2301.37,
                    "end": 2301.85,
                    "word": " is",
                    "probability": 0.90234375
                },
                {
                    "start": 2301.85,
                    "end": 2302.61,
                    "word": " not",
                    "probability": 0.892578125
                },
                {
                    "start": 2302.61,
                    "end": 2302.77,
                    "word": " in",
                    "probability": 0.88916015625
                },
                {
                    "start": 2302.77,
                    "end": 2302.87,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 2302.87,
                    "end": 2303.09,
                    "word": " RL",
                    "probability": 0.937744140625
                },
                {
                    "start": 2303.09,
                    "end": 2303.39,
                    "word": " domain",
                    "probability": 0.85302734375
                },
                {
                    "start": 2303.39,
                    "end": 2303.49,
                    "word": ",",
                    "probability": 0.63134765625
                },
                {
                    "start": 2303.49,
                    "end": 2303.65,
                    "word": " but",
                    "probability": 0.89599609375
                },
                {
                    "start": 2303.65,
                    "end": 2303.75,
                    "word": " in",
                    "probability": 0.8564453125
                },
                {
                    "start": 2303.75,
                    "end": 2303.85,
                    "word": " the",
                    "probability": 0.78271484375
                },
                {
                    "start": 2303.85,
                    "end": 2304.13,
                    "word": " supervised",
                    "probability": 0.80517578125
                },
                {
                    "start": 2304.13,
                    "end": 2304.43,
                    "word": " learning",
                    "probability": 0.82958984375
                },
                {
                    "start": 2304.43,
                    "end": 2304.69,
                    "word": " domain",
                    "probability": 0.8759765625
                },
                {
                    "start": 2304.69,
                    "end": 2304.75,
                    "word": ",",
                    "probability": 0.6708984375
                },
                {
                    "start": 2304.75,
                    "end": 2304.83,
                    "word": " and",
                    "probability": 0.89208984375
                },
                {
                    "start": 2304.83,
                    "end": 2304.93,
                    "word": " it's",
                    "probability": 0.92236328125
                }
            ]
        },
        {
            "id": 344,
            "seek": 232126,
            "start": 2304.93,
            "end": 2309.67,
            "text": " called imitation learning. Imitation learning is the supervised learning of policies from observed",
            "tokens": [
                50364,
                1219,
                47624,
                2539,
                13,
                286,
                3508,
                399,
                2539,
                307,
                264,
                46533,
                2539,
                295,
                7657,
                490,
                13095,
                50604
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21083332896232604,
            "compression_ratio": 1.5767195767195767,
            "no_speech_prob": 0.00609588623046875,
            "words": [
                {
                    "start": 2304.93,
                    "end": 2305.07,
                    "word": " called",
                    "probability": 0.544921875
                },
                {
                    "start": 2305.07,
                    "end": 2305.53,
                    "word": " imitation",
                    "probability": 0.7373046875
                },
                {
                    "start": 2305.53,
                    "end": 2305.89,
                    "word": " learning",
                    "probability": 0.82763671875
                },
                {
                    "start": 2305.89,
                    "end": 2306.07,
                    "word": ".",
                    "probability": 0.8623046875
                },
                {
                    "start": 2306.07,
                    "end": 2306.47,
                    "word": " Imitation",
                    "probability": 0.9348958333333334
                },
                {
                    "start": 2306.47,
                    "end": 2306.71,
                    "word": " learning",
                    "probability": 0.76416015625
                },
                {
                    "start": 2306.71,
                    "end": 2306.89,
                    "word": " is",
                    "probability": 0.90478515625
                },
                {
                    "start": 2306.89,
                    "end": 2307.03,
                    "word": " the",
                    "probability": 0.78857421875
                },
                {
                    "start": 2307.03,
                    "end": 2307.35,
                    "word": " supervised",
                    "probability": 0.78076171875
                },
                {
                    "start": 2307.35,
                    "end": 2307.61,
                    "word": " learning",
                    "probability": 0.84228515625
                },
                {
                    "start": 2307.61,
                    "end": 2307.79,
                    "word": " of",
                    "probability": 0.8388671875
                },
                {
                    "start": 2307.79,
                    "end": 2308.23,
                    "word": " policies",
                    "probability": 0.82568359375
                },
                {
                    "start": 2308.23,
                    "end": 2308.99,
                    "word": " from",
                    "probability": 0.77587890625
                },
                {
                    "start": 2308.99,
                    "end": 2309.67,
                    "word": " observed",
                    "probability": 0.7626953125
                }
            ]
        },
        {
            "id": 345,
            "seek": 232126,
            "start": 2309.67,
            "end": 2316.05,
            "text": " behavior of an expert, and by behavior, I mean state action behavior. What is the policy of an expert?",
            "tokens": [
                50604,
                5223,
                295,
                364,
                5844,
                11,
                293,
                538,
                5223,
                11,
                286,
                914,
                1785,
                3069,
                5223,
                13,
                708,
                307,
                264,
                3897,
                295,
                364,
                5844,
                30,
                50908
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21083332896232604,
            "compression_ratio": 1.5767195767195767,
            "no_speech_prob": 0.00609588623046875,
            "words": [
                {
                    "start": 2309.67,
                    "end": 2310.29,
                    "word": " behavior",
                    "probability": 0.74609375
                },
                {
                    "start": 2310.29,
                    "end": 2310.57,
                    "word": " of",
                    "probability": 0.896484375
                },
                {
                    "start": 2310.57,
                    "end": 2310.69,
                    "word": " an",
                    "probability": 0.90869140625
                },
                {
                    "start": 2310.69,
                    "end": 2310.91,
                    "word": " expert",
                    "probability": 0.79638671875
                },
                {
                    "start": 2310.91,
                    "end": 2311.03,
                    "word": ",",
                    "probability": 0.173828125
                },
                {
                    "start": 2311.03,
                    "end": 2311.09,
                    "word": " and",
                    "probability": 0.87060546875
                },
                {
                    "start": 2311.09,
                    "end": 2311.23,
                    "word": " by",
                    "probability": 0.90673828125
                },
                {
                    "start": 2311.23,
                    "end": 2311.55,
                    "word": " behavior",
                    "probability": 0.8505859375
                },
                {
                    "start": 2311.55,
                    "end": 2311.71,
                    "word": ",",
                    "probability": 0.3818359375
                },
                {
                    "start": 2311.71,
                    "end": 2311.71,
                    "word": " I",
                    "probability": 0.99169921875
                },
                {
                    "start": 2311.71,
                    "end": 2311.83,
                    "word": " mean",
                    "probability": 0.92822265625
                },
                {
                    "start": 2311.83,
                    "end": 2312.05,
                    "word": " state",
                    "probability": 0.80908203125
                },
                {
                    "start": 2312.05,
                    "end": 2312.61,
                    "word": " action",
                    "probability": 0.70947265625
                },
                {
                    "start": 2312.61,
                    "end": 2314.07,
                    "word": " behavior",
                    "probability": 0.77490234375
                },
                {
                    "start": 2314.07,
                    "end": 2314.23,
                    "word": ".",
                    "probability": 0.76171875
                },
                {
                    "start": 2314.23,
                    "end": 2314.33,
                    "word": " What",
                    "probability": 0.79296875
                },
                {
                    "start": 2314.33,
                    "end": 2314.43,
                    "word": " is",
                    "probability": 0.8955078125
                },
                {
                    "start": 2314.43,
                    "end": 2314.57,
                    "word": " the",
                    "probability": 0.7060546875
                },
                {
                    "start": 2314.57,
                    "end": 2314.97,
                    "word": " policy",
                    "probability": 0.82177734375
                },
                {
                    "start": 2314.97,
                    "end": 2315.15,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 2315.15,
                    "end": 2315.41,
                    "word": " an",
                    "probability": 0.904296875
                },
                {
                    "start": 2315.41,
                    "end": 2315.77,
                    "word": " expert",
                    "probability": 0.78515625
                },
                {
                    "start": 2315.77,
                    "end": 2316.05,
                    "word": "?",
                    "probability": 0.904296875
                }
            ]
        },
        {
            "id": 346,
            "seek": 232126,
            "start": 2317.49,
            "end": 2327.39,
            "text": " Based on this concept, in 2018, Hester et al. proposed DQFD, or DQ Learning from Demonstrations,",
            "tokens": [
                50996,
                18785,
                322,
                341,
                3410,
                11,
                294,
                6096,
                11,
                389,
                3011,
                1030,
                419,
                13,
                10348,
                413,
                48,
                37,
                35,
                11,
                420,
                413,
                48,
                15205,
                490,
                4686,
                4068,
                12154,
                11,
                51464
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21083332896232604,
            "compression_ratio": 1.5767195767195767,
            "no_speech_prob": 0.00609588623046875,
            "words": [
                {
                    "start": 2317.49,
                    "end": 2317.87,
                    "word": " Based",
                    "probability": 0.86181640625
                },
                {
                    "start": 2317.87,
                    "end": 2318.33,
                    "word": " on",
                    "probability": 0.90673828125
                },
                {
                    "start": 2318.33,
                    "end": 2319.13,
                    "word": " this",
                    "probability": 0.87841796875
                },
                {
                    "start": 2319.13,
                    "end": 2319.79,
                    "word": " concept",
                    "probability": 0.8037109375
                },
                {
                    "start": 2319.79,
                    "end": 2320.41,
                    "word": ",",
                    "probability": 0.775390625
                },
                {
                    "start": 2320.41,
                    "end": 2321.01,
                    "word": " in",
                    "probability": 0.89697265625
                },
                {
                    "start": 2321.01,
                    "end": 2321.71,
                    "word": " 2018",
                    "probability": 0.7490234375
                },
                {
                    "start": 2321.71,
                    "end": 2321.99,
                    "word": ",",
                    "probability": 0.72314453125
                },
                {
                    "start": 2321.99,
                    "end": 2322.21,
                    "word": " Hester",
                    "probability": 0.779052734375
                },
                {
                    "start": 2322.21,
                    "end": 2322.37,
                    "word": " et",
                    "probability": 0.76953125
                },
                {
                    "start": 2322.37,
                    "end": 2322.53,
                    "word": " al",
                    "probability": 0.88427734375
                },
                {
                    "start": 2322.53,
                    "end": 2323.09,
                    "word": ".",
                    "probability": 0.353759765625
                },
                {
                    "start": 2323.41,
                    "end": 2323.97,
                    "word": " proposed",
                    "probability": 0.75830078125
                },
                {
                    "start": 2323.97,
                    "end": 2325.15,
                    "word": " DQFD",
                    "probability": 0.85888671875
                },
                {
                    "start": 2325.15,
                    "end": 2325.25,
                    "word": ",",
                    "probability": 0.487548828125
                },
                {
                    "start": 2325.25,
                    "end": 2325.35,
                    "word": " or",
                    "probability": 0.869140625
                },
                {
                    "start": 2325.35,
                    "end": 2325.71,
                    "word": " DQ",
                    "probability": 0.87744140625
                },
                {
                    "start": 2325.71,
                    "end": 2325.97,
                    "word": " Learning",
                    "probability": 0.32080078125
                },
                {
                    "start": 2325.97,
                    "end": 2326.15,
                    "word": " from",
                    "probability": 0.497314453125
                },
                {
                    "start": 2326.15,
                    "end": 2326.81,
                    "word": " Demonstrations",
                    "probability": 0.884765625
                },
                {
                    "start": 2326.81,
                    "end": 2327.39,
                    "word": ",",
                    "probability": 0.91357421875
                }
            ]
        },
        {
            "id": 347,
            "seek": 234564,
            "start": 2327.39,
            "end": 2338.49,
            "text": " which is DQN, where the initial training is done based on observed data using deep learning. So,",
            "tokens": [
                50368,
                597,
                307,
                413,
                48,
                45,
                11,
                689,
                264,
                5883,
                3097,
                307,
                1096,
                2361,
                322,
                13095,
                1412,
                1228,
                2452,
                2539,
                13,
                407,
                11,
                50912
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19701087388439456,
            "compression_ratio": 1.6184971098265897,
            "no_speech_prob": 0.0005273818969726562,
            "words": [
                {
                    "start": 2327.39,
                    "end": 2327.77,
                    "word": " which",
                    "probability": 0.52880859375
                },
                {
                    "start": 2327.77,
                    "end": 2328.19,
                    "word": " is",
                    "probability": 0.90478515625
                },
                {
                    "start": 2328.19,
                    "end": 2329.17,
                    "word": " DQN",
                    "probability": 0.9220377604166666
                },
                {
                    "start": 2329.17,
                    "end": 2330.29,
                    "word": ",",
                    "probability": 0.55810546875
                },
                {
                    "start": 2330.29,
                    "end": 2331.39,
                    "word": " where",
                    "probability": 0.85986328125
                },
                {
                    "start": 2331.39,
                    "end": 2332.13,
                    "word": " the",
                    "probability": 0.73486328125
                },
                {
                    "start": 2332.13,
                    "end": 2332.39,
                    "word": " initial",
                    "probability": 0.8515625
                },
                {
                    "start": 2332.39,
                    "end": 2332.87,
                    "word": " training",
                    "probability": 0.806640625
                },
                {
                    "start": 2332.87,
                    "end": 2334.23,
                    "word": " is",
                    "probability": 0.89404296875
                },
                {
                    "start": 2334.23,
                    "end": 2334.95,
                    "word": " done",
                    "probability": 0.79296875
                },
                {
                    "start": 2334.95,
                    "end": 2335.71,
                    "word": " based",
                    "probability": 0.79638671875
                },
                {
                    "start": 2335.71,
                    "end": 2335.89,
                    "word": " on",
                    "probability": 0.90625
                },
                {
                    "start": 2335.89,
                    "end": 2336.41,
                    "word": " observed",
                    "probability": 0.748046875
                },
                {
                    "start": 2336.41,
                    "end": 2336.85,
                    "word": " data",
                    "probability": 0.80810546875
                },
                {
                    "start": 2336.85,
                    "end": 2337.21,
                    "word": " using",
                    "probability": 0.75
                },
                {
                    "start": 2337.21,
                    "end": 2337.47,
                    "word": " deep",
                    "probability": 0.85205078125
                },
                {
                    "start": 2337.47,
                    "end": 2337.75,
                    "word": " learning",
                    "probability": 0.82958984375
                },
                {
                    "start": 2337.75,
                    "end": 2337.95,
                    "word": ".",
                    "probability": 0.724609375
                },
                {
                    "start": 2337.95,
                    "end": 2338.09,
                    "word": " So",
                    "probability": 0.8173828125
                },
                {
                    "start": 2338.09,
                    "end": 2338.49,
                    "word": ",",
                    "probability": 0.2052001953125
                }
            ]
        },
        {
            "id": 348,
            "seek": 234564,
            "start": 2338.49,
            "end": 2346.15,
            "text": " they have data from human players playing a certain game, or human performance doing a",
            "tokens": [
                50912,
                436,
                362,
                1412,
                490,
                1952,
                4150,
                2433,
                257,
                1629,
                1216,
                11,
                420,
                1952,
                3389,
                884,
                257,
                51284
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19701087388439456,
            "compression_ratio": 1.6184971098265897,
            "no_speech_prob": 0.0005273818969726562,
            "words": [
                {
                    "start": 2338.49,
                    "end": 2338.59,
                    "word": " they",
                    "probability": 0.73779296875
                },
                {
                    "start": 2338.59,
                    "end": 2338.87,
                    "word": " have",
                    "probability": 0.8759765625
                },
                {
                    "start": 2338.87,
                    "end": 2339.19,
                    "word": " data",
                    "probability": 0.78173828125
                },
                {
                    "start": 2339.19,
                    "end": 2339.59,
                    "word": " from",
                    "probability": 0.81884765625
                },
                {
                    "start": 2339.59,
                    "end": 2339.99,
                    "word": " human",
                    "probability": 0.7900390625
                },
                {
                    "start": 2339.99,
                    "end": 2340.29,
                    "word": " players",
                    "probability": 0.84130859375
                },
                {
                    "start": 2340.29,
                    "end": 2340.63,
                    "word": " playing",
                    "probability": 0.77734375
                },
                {
                    "start": 2340.95,
                    "end": 2341.15,
                    "word": " a certain",
                    "probability": 0.798828125
                },
                {
                    "start": 2341.15,
                    "end": 2341.47,
                    "word": " game",
                    "probability": 0.81884765625
                },
                {
                    "start": 2341.47,
                    "end": 2341.53,
                    "word": ",",
                    "probability": 0.2646484375
                },
                {
                    "start": 2341.53,
                    "end": 2341.63,
                    "word": " or",
                    "probability": 0.90380859375
                },
                {
                    "start": 2341.63,
                    "end": 2342.03,
                    "word": " human",
                    "probability": 0.74609375
                },
                {
                    "start": 2342.03,
                    "end": 2343.97,
                    "word": " performance",
                    "probability": 0.61181640625
                },
                {
                    "start": 2343.97,
                    "end": 2345.11,
                    "word": " doing",
                    "probability": 0.85498046875
                },
                {
                    "start": 2345.71,
                    "end": 2346.15,
                    "word": " a certain",
                    "probability": 0.79833984375
                }
            ]
        },
        {
            "id": 349,
            "seek": 234564,
            "start": 2346.15,
            "end": 2351.77,
            "text": " certain task that they want the agent to learn. The initial step of training for a DQFD agent is",
            "tokens": [
                51284,
                1629,
                5633,
                300,
                436,
                528,
                264,
                9461,
                281,
                1466,
                13,
                440,
                5883,
                1823,
                295,
                3097,
                337,
                257,
                413,
                48,
                37,
                35,
                9461,
                307,
                51576
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19701087388439456,
            "compression_ratio": 1.6184971098265897,
            "no_speech_prob": 0.0005273818969726562,
            "words": [
                {
                    "start": 2346.15,
                    "end": 2346.53,
                    "word": " task",
                    "probability": 0.8408203125
                },
                {
                    "start": 2346.53,
                    "end": 2346.69,
                    "word": " that",
                    "probability": 0.84033203125
                },
                {
                    "start": 2346.69,
                    "end": 2346.81,
                    "word": " they",
                    "probability": 0.72998046875
                },
                {
                    "start": 2346.81,
                    "end": 2346.93,
                    "word": " want",
                    "probability": 0.810546875
                },
                {
                    "start": 2346.93,
                    "end": 2347.03,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 2347.03,
                    "end": 2347.21,
                    "word": " agent",
                    "probability": 0.80322265625
                },
                {
                    "start": 2347.21,
                    "end": 2347.37,
                    "word": " to",
                    "probability": 0.90234375
                },
                {
                    "start": 2347.37,
                    "end": 2347.61,
                    "word": " learn",
                    "probability": 0.92041015625
                },
                {
                    "start": 2347.61,
                    "end": 2348.03,
                    "word": ".",
                    "probability": 0.94775390625
                },
                {
                    "start": 2348.03,
                    "end": 2348.49,
                    "word": " The",
                    "probability": 0.82177734375
                },
                {
                    "start": 2348.49,
                    "end": 2349.01,
                    "word": " initial",
                    "probability": 0.8701171875
                },
                {
                    "start": 2349.01,
                    "end": 2349.29,
                    "word": " step",
                    "probability": 0.81982421875
                },
                {
                    "start": 2349.29,
                    "end": 2349.41,
                    "word": " of",
                    "probability": 0.865234375
                },
                {
                    "start": 2349.41,
                    "end": 2349.67,
                    "word": " training",
                    "probability": 0.80029296875
                },
                {
                    "start": 2349.67,
                    "end": 2350.03,
                    "word": " for",
                    "probability": 0.8916015625
                },
                {
                    "start": 2350.71,
                    "end": 2351.37,
                    "word": " a DQFD",
                    "probability": 0.922119140625
                },
                {
                    "start": 2351.37,
                    "end": 2351.61,
                    "word": " agent",
                    "probability": 0.8056640625
                },
                {
                    "start": 2351.61,
                    "end": 2351.77,
                    "word": " is",
                    "probability": 0.73291015625
                }
            ]
        },
        {
            "id": 350,
            "seek": 237216,
            "start": 2351.77,
            "end": 2359.81,
            "text": " supervised learning on observed data, and then it starts building on top of it through reinforcement",
            "tokens": [
                50364,
                46533,
                2539,
                322,
                13095,
                1412,
                11,
                293,
                550,
                309,
                3719,
                2390,
                322,
                1192,
                295,
                309,
                807,
                29280,
                50764
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22665895061728394,
            "compression_ratio": 1.5708333333333333,
            "no_speech_prob": 0.0010480880737304688,
            "words": [
                {
                    "start": 2351.77,
                    "end": 2352.05,
                    "word": " supervised",
                    "probability": 0.372314453125
                },
                {
                    "start": 2352.05,
                    "end": 2352.59,
                    "word": " learning",
                    "probability": 0.84619140625
                },
                {
                    "start": 2352.59,
                    "end": 2353.49,
                    "word": " on",
                    "probability": 0.8681640625
                },
                {
                    "start": 2353.49,
                    "end": 2354.45,
                    "word": " observed",
                    "probability": 0.70654296875
                },
                {
                    "start": 2354.45,
                    "end": 2354.85,
                    "word": " data",
                    "probability": 0.79541015625
                },
                {
                    "start": 2354.85,
                    "end": 2355.09,
                    "word": ",",
                    "probability": 0.313232421875
                },
                {
                    "start": 2355.09,
                    "end": 2355.41,
                    "word": " and",
                    "probability": 0.892578125
                },
                {
                    "start": 2355.41,
                    "end": 2356.01,
                    "word": " then",
                    "probability": 0.72705078125
                },
                {
                    "start": 2356.01,
                    "end": 2357.17,
                    "word": " it",
                    "probability": 0.87353515625
                },
                {
                    "start": 2357.17,
                    "end": 2357.57,
                    "word": " starts",
                    "probability": 0.78857421875
                },
                {
                    "start": 2357.57,
                    "end": 2358.05,
                    "word": " building",
                    "probability": 0.78564453125
                },
                {
                    "start": 2358.05,
                    "end": 2358.23,
                    "word": " on",
                    "probability": 0.9013671875
                },
                {
                    "start": 2358.23,
                    "end": 2358.35,
                    "word": " top",
                    "probability": 0.8203125
                },
                {
                    "start": 2358.35,
                    "end": 2358.47,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 2358.47,
                    "end": 2358.63,
                    "word": " it",
                    "probability": 0.9091796875
                },
                {
                    "start": 2358.63,
                    "end": 2359.11,
                    "word": " through",
                    "probability": 0.759765625
                },
                {
                    "start": 2359.11,
                    "end": 2359.81,
                    "word": " reinforcement",
                    "probability": 0.80224609375
                }
            ]
        },
        {
            "id": 351,
            "seek": 237216,
            "start": 2359.81,
            "end": 2364.73,
            "text": " learning approaches, and it was shown that it can result in faster convergence, better sample",
            "tokens": [
                50764,
                2539,
                11587,
                11,
                293,
                309,
                390,
                4898,
                300,
                309,
                393,
                1874,
                294,
                4663,
                32181,
                11,
                1101,
                6889,
                51012
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22665895061728394,
            "compression_ratio": 1.5708333333333333,
            "no_speech_prob": 0.0010480880737304688,
            "words": [
                {
                    "start": 2359.81,
                    "end": 2360.11,
                    "word": " learning",
                    "probability": 0.8388671875
                },
                {
                    "start": 2360.11,
                    "end": 2360.49,
                    "word": " approaches",
                    "probability": 0.69189453125
                },
                {
                    "start": 2360.49,
                    "end": 2360.73,
                    "word": ",",
                    "probability": 0.1650390625
                },
                {
                    "start": 2360.73,
                    "end": 2360.83,
                    "word": " and",
                    "probability": 0.89306640625
                },
                {
                    "start": 2360.83,
                    "end": 2360.91,
                    "word": " it",
                    "probability": 0.890625
                },
                {
                    "start": 2360.91,
                    "end": 2361.03,
                    "word": " was",
                    "probability": 0.89599609375
                },
                {
                    "start": 2361.03,
                    "end": 2361.27,
                    "word": " shown",
                    "probability": 0.88232421875
                },
                {
                    "start": 2361.27,
                    "end": 2361.43,
                    "word": " that",
                    "probability": 0.8818359375
                },
                {
                    "start": 2361.43,
                    "end": 2361.55,
                    "word": " it",
                    "probability": 0.8828125
                },
                {
                    "start": 2361.55,
                    "end": 2361.75,
                    "word": " can",
                    "probability": 0.90771484375
                },
                {
                    "start": 2361.75,
                    "end": 2362.17,
                    "word": " result",
                    "probability": 0.8291015625
                },
                {
                    "start": 2362.17,
                    "end": 2362.85,
                    "word": " in",
                    "probability": 0.9033203125
                },
                {
                    "start": 2362.85,
                    "end": 2363.81,
                    "word": " faster",
                    "probability": 0.55810546875
                },
                {
                    "start": 2363.81,
                    "end": 2364.23,
                    "word": " convergence",
                    "probability": 0.75537109375
                },
                {
                    "start": 2364.23,
                    "end": 2364.35,
                    "word": ",",
                    "probability": 0.96240234375
                },
                {
                    "start": 2364.35,
                    "end": 2364.43,
                    "word": " better",
                    "probability": 0.81689453125
                },
                {
                    "start": 2364.43,
                    "end": 2364.73,
                    "word": " sample",
                    "probability": 0.74072265625
                }
            ]
        },
        {
            "id": 352,
            "seek": 237216,
            "start": 2364.73,
            "end": 2372.33,
            "text": " complexity, and sometimes more interesting and robust policies. As security researchers,",
            "tokens": [
                51012,
                14024,
                11,
                293,
                2171,
                544,
                1880,
                293,
                13956,
                7657,
                13,
                1018,
                3825,
                10309,
                11,
                51388
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22665895061728394,
            "compression_ratio": 1.5708333333333333,
            "no_speech_prob": 0.0010480880737304688,
            "words": [
                {
                    "start": 2364.73,
                    "end": 2365.21,
                    "word": " complexity",
                    "probability": 0.6962890625
                },
                {
                    "start": 2365.21,
                    "end": 2365.31,
                    "word": ",",
                    "probability": 0.767578125
                },
                {
                    "start": 2365.31,
                    "end": 2365.45,
                    "word": " and",
                    "probability": 0.8994140625
                },
                {
                    "start": 2365.45,
                    "end": 2365.79,
                    "word": " sometimes",
                    "probability": 0.9072265625
                },
                {
                    "start": 2365.79,
                    "end": 2366.19,
                    "word": " more",
                    "probability": 0.8515625
                },
                {
                    "start": 2366.19,
                    "end": 2366.75,
                    "word": " interesting",
                    "probability": 0.634765625
                },
                {
                    "start": 2366.75,
                    "end": 2367.17,
                    "word": " and",
                    "probability": 0.8359375
                },
                {
                    "start": 2367.17,
                    "end": 2367.41,
                    "word": " robust",
                    "probability": 0.74658203125
                },
                {
                    "start": 2367.41,
                    "end": 2368.03,
                    "word": " policies",
                    "probability": 0.853515625
                },
                {
                    "start": 2368.03,
                    "end": 2368.63,
                    "word": ".",
                    "probability": 0.9912109375
                },
                {
                    "start": 2369.39,
                    "end": 2369.93,
                    "word": " As",
                    "probability": 0.92578125
                },
                {
                    "start": 2369.93,
                    "end": 2371.51,
                    "word": " security",
                    "probability": 0.73583984375
                },
                {
                    "start": 2371.51,
                    "end": 2371.97,
                    "word": " researchers",
                    "probability": 0.79541015625
                },
                {
                    "start": 2371.97,
                    "end": 2372.33,
                    "word": ",",
                    "probability": 0.9208984375
                }
            ]
        },
        {
            "id": 353,
            "seek": 237216,
            "start": 2372.33,
            "end": 2378.29,
            "text": " you can probably see where this is going. This wonderful algorithm, DQFD, can also be used to",
            "tokens": [
                51388,
                291,
                393,
                1391,
                536,
                689,
                341,
                307,
                516,
                13,
                639,
                3715,
                9284,
                11,
                413,
                48,
                37,
                35,
                11,
                393,
                611,
                312,
                1143,
                281,
                51692
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22665895061728394,
            "compression_ratio": 1.5708333333333333,
            "no_speech_prob": 0.0010480880737304688,
            "words": [
                {
                    "start": 2372.33,
                    "end": 2372.47,
                    "word": " you",
                    "probability": 0.8974609375
                },
                {
                    "start": 2372.47,
                    "end": 2372.61,
                    "word": " can",
                    "probability": 0.88671875
                },
                {
                    "start": 2372.61,
                    "end": 2372.83,
                    "word": " probably",
                    "probability": 0.90966796875
                },
                {
                    "start": 2372.83,
                    "end": 2373.01,
                    "word": " see",
                    "probability": 0.82666015625
                },
                {
                    "start": 2373.01,
                    "end": 2373.13,
                    "word": " where",
                    "probability": 0.91015625
                },
                {
                    "start": 2373.13,
                    "end": 2373.29,
                    "word": " this",
                    "probability": 0.8818359375
                },
                {
                    "start": 2373.29,
                    "end": 2373.41,
                    "word": " is",
                    "probability": 0.9072265625
                },
                {
                    "start": 2373.41,
                    "end": 2373.69,
                    "word": " going",
                    "probability": 0.857421875
                },
                {
                    "start": 2373.69,
                    "end": 2374.25,
                    "word": ".",
                    "probability": 0.83984375
                },
                {
                    "start": 2374.25,
                    "end": 2374.41,
                    "word": " This",
                    "probability": 0.783203125
                },
                {
                    "start": 2374.41,
                    "end": 2374.93,
                    "word": " wonderful",
                    "probability": 0.7568359375
                },
                {
                    "start": 2374.93,
                    "end": 2375.67,
                    "word": " algorithm",
                    "probability": 0.77783203125
                },
                {
                    "start": 2375.67,
                    "end": 2375.79,
                    "word": ",",
                    "probability": 0.66455078125
                },
                {
                    "start": 2375.79,
                    "end": 2376.39,
                    "word": " DQFD",
                    "probability": 0.80242919921875
                },
                {
                    "start": 2376.39,
                    "end": 2376.53,
                    "word": ",",
                    "probability": 0.89697265625
                },
                {
                    "start": 2376.53,
                    "end": 2376.79,
                    "word": " can",
                    "probability": 0.896484375
                },
                {
                    "start": 2376.79,
                    "end": 2377.09,
                    "word": " also",
                    "probability": 0.8310546875
                },
                {
                    "start": 2377.09,
                    "end": 2377.29,
                    "word": " be",
                    "probability": 0.904296875
                },
                {
                    "start": 2377.29,
                    "end": 2377.73,
                    "word": " used",
                    "probability": 0.810546875
                },
                {
                    "start": 2377.73,
                    "end": 2378.29,
                    "word": " to",
                    "probability": 0.904296875
                }
            ]
        },
        {
            "id": 354,
            "seek": 240088,
            "start": 2378.29,
            "end": 2385.39,
            "text": " replicate policies. Instead of applying it on observed data collected from human performance,",
            "tokens": [
                50368,
                25356,
                7657,
                13,
                7156,
                295,
                9275,
                309,
                322,
                13095,
                1412,
                11087,
                490,
                1952,
                3389,
                11,
                50712
            ],
            "temperature": 0.0,
            "avg_logprob": -0.28273810126951765,
            "compression_ratio": 1.663716814159292,
            "no_speech_prob": 0.0012254714965820312,
            "words": [
                {
                    "start": 2378.29,
                    "end": 2378.93,
                    "word": " replicate",
                    "probability": 0.1495361328125
                },
                {
                    "start": 2378.93,
                    "end": 2381.15,
                    "word": " policies",
                    "probability": 0.794921875
                },
                {
                    "start": 2381.15,
                    "end": 2381.83,
                    "word": ".",
                    "probability": 0.2568359375
                },
                {
                    "start": 2381.83,
                    "end": 2381.93,
                    "word": " Instead",
                    "probability": 0.87646484375
                },
                {
                    "start": 2381.93,
                    "end": 2382.27,
                    "word": " of",
                    "probability": 0.89892578125
                },
                {
                    "start": 2382.27,
                    "end": 2382.73,
                    "word": " applying",
                    "probability": 0.72412109375
                },
                {
                    "start": 2382.73,
                    "end": 2382.93,
                    "word": " it",
                    "probability": 0.8740234375
                },
                {
                    "start": 2382.93,
                    "end": 2383.17,
                    "word": " on",
                    "probability": 0.8916015625
                },
                {
                    "start": 2383.17,
                    "end": 2383.63,
                    "word": " observed",
                    "probability": 0.70263671875
                },
                {
                    "start": 2383.63,
                    "end": 2384.05,
                    "word": " data",
                    "probability": 0.8076171875
                },
                {
                    "start": 2384.05,
                    "end": 2384.35,
                    "word": " collected",
                    "probability": 0.73291015625
                },
                {
                    "start": 2384.35,
                    "end": 2384.55,
                    "word": " from",
                    "probability": 0.8046875
                },
                {
                    "start": 2384.55,
                    "end": 2384.81,
                    "word": " human",
                    "probability": 0.77294921875
                },
                {
                    "start": 2384.81,
                    "end": 2385.27,
                    "word": " performance",
                    "probability": 0.68798828125
                },
                {
                    "start": 2385.27,
                    "end": 2385.39,
                    "word": ",",
                    "probability": 0.75634765625
                }
            ]
        },
        {
            "id": 355,
            "seek": 240088,
            "start": 2385.39,
            "end": 2391.81,
            "text": " it can be applied on observed data from a target policy. So, here's a proof of concept attack the",
            "tokens": [
                50712,
                309,
                393,
                312,
                6456,
                322,
                13095,
                1412,
                490,
                257,
                3779,
                3897,
                13,
                407,
                11,
                510,
                311,
                257,
                8177,
                295,
                3410,
                2690,
                264,
                51040
            ],
            "temperature": 0.0,
            "avg_logprob": -0.28273810126951765,
            "compression_ratio": 1.663716814159292,
            "no_speech_prob": 0.0012254714965820312,
            "words": [
                {
                    "start": 2385.39,
                    "end": 2385.43,
                    "word": " it",
                    "probability": 0.822265625
                },
                {
                    "start": 2385.43,
                    "end": 2385.57,
                    "word": " can",
                    "probability": 0.89794921875
                },
                {
                    "start": 2385.57,
                    "end": 2385.69,
                    "word": " be",
                    "probability": 0.8955078125
                },
                {
                    "start": 2385.69,
                    "end": 2386.09,
                    "word": " applied",
                    "probability": 0.7099609375
                },
                {
                    "start": 2386.09,
                    "end": 2386.95,
                    "word": " on",
                    "probability": 0.88720703125
                },
                {
                    "start": 2386.95,
                    "end": 2387.29,
                    "word": " observed",
                    "probability": 0.74560546875
                },
                {
                    "start": 2387.29,
                    "end": 2387.67,
                    "word": " data",
                    "probability": 0.80419921875
                },
                {
                    "start": 2387.67,
                    "end": 2388.15,
                    "word": " from",
                    "probability": 0.81103515625
                },
                {
                    "start": 2388.53,
                    "end": 2388.79,
                    "word": " a target",
                    "probability": 0.83642578125
                },
                {
                    "start": 2388.79,
                    "end": 2389.33,
                    "word": " policy",
                    "probability": 0.7939453125
                },
                {
                    "start": 2389.33,
                    "end": 2390.01,
                    "word": ".",
                    "probability": 0.98681640625
                },
                {
                    "start": 2390.11,
                    "end": 2390.21,
                    "word": " So",
                    "probability": 0.69384765625
                },
                {
                    "start": 2390.21,
                    "end": 2390.29,
                    "word": ",",
                    "probability": 0.1448974609375
                },
                {
                    "start": 2390.29,
                    "end": 2390.45,
                    "word": " here's",
                    "probability": 0.8779296875
                },
                {
                    "start": 2390.55,
                    "end": 2390.73,
                    "word": " a proof",
                    "probability": 0.8486328125
                },
                {
                    "start": 2390.73,
                    "end": 2390.89,
                    "word": " of",
                    "probability": 0.67138671875
                },
                {
                    "start": 2390.89,
                    "end": 2391.25,
                    "word": " concept",
                    "probability": 0.73779296875
                },
                {
                    "start": 2391.25,
                    "end": 2391.61,
                    "word": " attack",
                    "probability": 0.33203125
                },
                {
                    "start": 2391.61,
                    "end": 2391.81,
                    "word": " the",
                    "probability": 0.52197265625
                }
            ]
        },
        {
            "id": 356,
            "seek": 240088,
            "start": 2391.81,
            "end": 2399.43,
            "text": " procedure. The attacker observes and records n interactions of the SARSA type, state action,",
            "tokens": [
                51040,
                10747,
                13,
                440,
                35871,
                3181,
                9054,
                293,
                7724,
                297,
                13280,
                295,
                264,
                18748,
                8886,
                2010,
                11,
                1785,
                3069,
                11,
                51416
            ],
            "temperature": 0.0,
            "avg_logprob": -0.28273810126951765,
            "compression_ratio": 1.663716814159292,
            "no_speech_prob": 0.0012254714965820312,
            "words": [
                {
                    "start": 2391.81,
                    "end": 2392.27,
                    "word": " procedure",
                    "probability": 0.418212890625
                },
                {
                    "start": 2392.27,
                    "end": 2392.95,
                    "word": ".",
                    "probability": 0.309326171875
                },
                {
                    "start": 2393.01,
                    "end": 2393.17,
                    "word": " The",
                    "probability": 0.79833984375
                },
                {
                    "start": 2393.17,
                    "end": 2394.57,
                    "word": " attacker",
                    "probability": 0.78515625
                },
                {
                    "start": 2394.57,
                    "end": 2395.15,
                    "word": " observes",
                    "probability": 0.888427734375
                },
                {
                    "start": 2395.15,
                    "end": 2395.37,
                    "word": " and",
                    "probability": 0.8056640625
                },
                {
                    "start": 2395.37,
                    "end": 2395.73,
                    "word": " records",
                    "probability": 0.8544921875
                },
                {
                    "start": 2395.73,
                    "end": 2396.21,
                    "word": " n",
                    "probability": 0.29443359375
                },
                {
                    "start": 2396.21,
                    "end": 2396.77,
                    "word": " interactions",
                    "probability": 0.7841796875
                },
                {
                    "start": 2396.77,
                    "end": 2397.65,
                    "word": " of",
                    "probability": 0.8974609375
                },
                {
                    "start": 2397.65,
                    "end": 2397.97,
                    "word": " the",
                    "probability": 0.79736328125
                },
                {
                    "start": 2397.97,
                    "end": 2398.29,
                    "word": " SARSA",
                    "probability": 0.6392822265625
                },
                {
                    "start": 2398.29,
                    "end": 2398.63,
                    "word": " type",
                    "probability": 0.8310546875
                },
                {
                    "start": 2398.63,
                    "end": 2398.75,
                    "word": ",",
                    "probability": 0.408447265625
                },
                {
                    "start": 2398.75,
                    "end": 2399.01,
                    "word": " state",
                    "probability": 0.6796875
                },
                {
                    "start": 2399.01,
                    "end": 2399.35,
                    "word": " action",
                    "probability": 0.60693359375
                },
                {
                    "start": 2399.35,
                    "end": 2399.43,
                    "word": ",",
                    "probability": 0.84521484375
                }
            ]
        },
        {
            "id": 357,
            "seek": 240088,
            "start": 2399.43,
            "end": 2407.01,
            "text": " the next state, and the reward based on this transition of the target agent in a particular",
            "tokens": [
                51416,
                264,
                958,
                1785,
                11,
                293,
                264,
                7782,
                2361,
                322,
                341,
                6034,
                295,
                264,
                3779,
                9461,
                294,
                257,
                1729,
                51804
            ],
            "temperature": 0.0,
            "avg_logprob": -0.28273810126951765,
            "compression_ratio": 1.663716814159292,
            "no_speech_prob": 0.0012254714965820312,
            "words": [
                {
                    "start": 2399.43,
                    "end": 2399.51,
                    "word": " the",
                    "probability": 0.7939453125
                },
                {
                    "start": 2399.51,
                    "end": 2399.75,
                    "word": " next",
                    "probability": 0.85302734375
                },
                {
                    "start": 2399.75,
                    "end": 2400.33,
                    "word": " state",
                    "probability": 0.875
                },
                {
                    "start": 2400.33,
                    "end": 2400.49,
                    "word": ",",
                    "probability": 0.5390625
                },
                {
                    "start": 2400.49,
                    "end": 2400.77,
                    "word": " and",
                    "probability": 0.9013671875
                },
                {
                    "start": 2400.77,
                    "end": 2401.17,
                    "word": " the",
                    "probability": 0.73193359375
                },
                {
                    "start": 2401.17,
                    "end": 2401.69,
                    "word": " reward",
                    "probability": 0.8603515625
                },
                {
                    "start": 2401.69,
                    "end": 2402.15,
                    "word": " based",
                    "probability": 0.455078125
                },
                {
                    "start": 2402.15,
                    "end": 2402.23,
                    "word": " on",
                    "probability": 0.90380859375
                },
                {
                    "start": 2402.23,
                    "end": 2402.35,
                    "word": " this",
                    "probability": 0.84619140625
                },
                {
                    "start": 2402.35,
                    "end": 2402.73,
                    "word": " transition",
                    "probability": 0.9130859375
                },
                {
                    "start": 2402.73,
                    "end": 2402.89,
                    "word": " of",
                    "probability": 0.88916015625
                },
                {
                    "start": 2402.89,
                    "end": 2402.99,
                    "word": " the",
                    "probability": 0.80517578125
                },
                {
                    "start": 2402.99,
                    "end": 2403.21,
                    "word": " target",
                    "probability": 0.8642578125
                },
                {
                    "start": 2403.21,
                    "end": 2403.67,
                    "word": " agent",
                    "probability": 0.751953125
                },
                {
                    "start": 2403.67,
                    "end": 2406.13,
                    "word": " in",
                    "probability": 0.74462890625
                },
                {
                    "start": 2406.71,
                    "end": 2407.01,
                    "word": " a particular",
                    "probability": 0.79052734375
                }
            ]
        },
        {
            "id": 358,
            "seek": 242924,
            "start": 2407.01,
            "end": 2413.17,
            "text": " environment, and then the attacker applies DQFD to learn an imitation of the target policy and",
            "tokens": [
                50364,
                2823,
                11,
                293,
                550,
                264,
                35871,
                13165,
                413,
                48,
                37,
                35,
                281,
                1466,
                364,
                47624,
                295,
                264,
                3779,
                3897,
                293,
                50664
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19438844150112522,
            "compression_ratio": 1.5819672131147542,
            "no_speech_prob": 0.0004801750183105469,
            "words": [
                {
                    "start": 2407.01,
                    "end": 2407.41,
                    "word": " environment",
                    "probability": 0.1256103515625
                },
                {
                    "start": 2407.41,
                    "end": 2407.87,
                    "word": ",",
                    "probability": 0.223876953125
                },
                {
                    "start": 2407.87,
                    "end": 2408.01,
                    "word": " and",
                    "probability": 0.86279296875
                },
                {
                    "start": 2408.01,
                    "end": 2408.27,
                    "word": " then",
                    "probability": 0.7353515625
                },
                {
                    "start": 2408.27,
                    "end": 2408.55,
                    "word": " the",
                    "probability": 0.7919921875
                },
                {
                    "start": 2408.55,
                    "end": 2408.75,
                    "word": " attacker",
                    "probability": 0.73974609375
                },
                {
                    "start": 2408.75,
                    "end": 2409.15,
                    "word": " applies",
                    "probability": 0.71728515625
                },
                {
                    "start": 2409.15,
                    "end": 2409.75,
                    "word": " DQFD",
                    "probability": 0.767578125
                },
                {
                    "start": 2409.75,
                    "end": 2409.93,
                    "word": " to",
                    "probability": 0.89599609375
                },
                {
                    "start": 2409.93,
                    "end": 2410.19,
                    "word": " learn",
                    "probability": 0.91162109375
                },
                {
                    "start": 2410.19,
                    "end": 2410.33,
                    "word": " an",
                    "probability": 0.8857421875
                },
                {
                    "start": 2410.33,
                    "end": 2410.77,
                    "word": " imitation",
                    "probability": 0.80419921875
                },
                {
                    "start": 2410.77,
                    "end": 2411.41,
                    "word": " of",
                    "probability": 0.90234375
                },
                {
                    "start": 2411.41,
                    "end": 2412.11,
                    "word": " the",
                    "probability": 0.79638671875
                },
                {
                    "start": 2412.11,
                    "end": 2412.39,
                    "word": " target",
                    "probability": 0.83203125
                },
                {
                    "start": 2412.39,
                    "end": 2412.83,
                    "word": " policy",
                    "probability": 0.77099609375
                },
                {
                    "start": 2412.83,
                    "end": 2413.17,
                    "word": " and",
                    "probability": 0.89599609375
                }
            ]
        },
        {
            "id": 359,
            "seek": 242924,
            "start": 2413.17,
            "end": 2422.65,
            "text": " Q function. Now, at this point, the attacker may either just go away and sell the extracted policy,",
            "tokens": [
                50664,
                1249,
                2445,
                13,
                823,
                11,
                412,
                341,
                935,
                11,
                264,
                35871,
                815,
                2139,
                445,
                352,
                1314,
                293,
                3607,
                264,
                34086,
                3897,
                11,
                51132
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19438844150112522,
            "compression_ratio": 1.5819672131147542,
            "no_speech_prob": 0.0004801750183105469,
            "words": [
                {
                    "start": 2413.17,
                    "end": 2413.35,
                    "word": " Q",
                    "probability": 0.50390625
                },
                {
                    "start": 2413.35,
                    "end": 2413.75,
                    "word": " function",
                    "probability": 0.7509765625
                },
                {
                    "start": 2413.75,
                    "end": 2414.27,
                    "word": ".",
                    "probability": 0.935546875
                },
                {
                    "start": 2414.31,
                    "end": 2414.51,
                    "word": " Now",
                    "probability": 0.86865234375
                },
                {
                    "start": 2414.51,
                    "end": 2414.61,
                    "word": ",",
                    "probability": 0.4677734375
                },
                {
                    "start": 2414.61,
                    "end": 2414.71,
                    "word": " at",
                    "probability": 0.900390625
                },
                {
                    "start": 2414.71,
                    "end": 2414.89,
                    "word": " this",
                    "probability": 0.88134765625
                },
                {
                    "start": 2414.89,
                    "end": 2415.05,
                    "word": " point",
                    "probability": 0.86083984375
                },
                {
                    "start": 2415.05,
                    "end": 2415.09,
                    "word": ",",
                    "probability": 0.71484375
                },
                {
                    "start": 2415.09,
                    "end": 2415.19,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 2415.19,
                    "end": 2415.45,
                    "word": " attacker",
                    "probability": 0.72607421875
                },
                {
                    "start": 2415.45,
                    "end": 2415.75,
                    "word": " may",
                    "probability": 0.896484375
                },
                {
                    "start": 2415.75,
                    "end": 2416.57,
                    "word": " either",
                    "probability": 0.34423828125
                },
                {
                    "start": 2416.57,
                    "end": 2418.65,
                    "word": " just",
                    "probability": 0.818359375
                },
                {
                    "start": 2418.65,
                    "end": 2419.49,
                    "word": " go",
                    "probability": 0.90234375
                },
                {
                    "start": 2419.49,
                    "end": 2419.81,
                    "word": " away",
                    "probability": 0.80908203125
                },
                {
                    "start": 2419.81,
                    "end": 2420.21,
                    "word": " and",
                    "probability": 0.88671875
                },
                {
                    "start": 2420.21,
                    "end": 2420.89,
                    "word": " sell",
                    "probability": 0.7568359375
                },
                {
                    "start": 2420.89,
                    "end": 2421.43,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 2421.43,
                    "end": 2421.99,
                    "word": " extracted",
                    "probability": 0.70849609375
                },
                {
                    "start": 2421.99,
                    "end": 2422.43,
                    "word": " policy",
                    "probability": 0.80078125
                },
                {
                    "start": 2422.43,
                    "end": 2422.65,
                    "word": ",",
                    "probability": 0.53564453125
                }
            ]
        },
        {
            "id": 360,
            "seek": 242924,
            "start": 2422.65,
            "end": 2428.45,
            "text": " or it may decide to target it using different adversarial perturbation attacks, some of which",
            "tokens": [
                51132,
                420,
                309,
                815,
                4536,
                281,
                3779,
                309,
                1228,
                819,
                17641,
                44745,
                40468,
                399,
                8122,
                11,
                512,
                295,
                597,
                51428
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19438844150112522,
            "compression_ratio": 1.5819672131147542,
            "no_speech_prob": 0.0004801750183105469,
            "words": [
                {
                    "start": 2422.65,
                    "end": 2422.71,
                    "word": " or",
                    "probability": 0.8984375
                },
                {
                    "start": 2422.71,
                    "end": 2422.83,
                    "word": " it",
                    "probability": 0.8505859375
                },
                {
                    "start": 2422.83,
                    "end": 2422.91,
                    "word": " may",
                    "probability": 0.89697265625
                },
                {
                    "start": 2422.91,
                    "end": 2423.17,
                    "word": " decide",
                    "probability": 0.8779296875
                },
                {
                    "start": 2423.17,
                    "end": 2423.31,
                    "word": " to",
                    "probability": 0.8984375
                },
                {
                    "start": 2423.31,
                    "end": 2423.55,
                    "word": " target",
                    "probability": 0.841796875
                },
                {
                    "start": 2423.55,
                    "end": 2423.73,
                    "word": " it",
                    "probability": 0.9033203125
                },
                {
                    "start": 2423.73,
                    "end": 2423.99,
                    "word": " using",
                    "probability": 0.8076171875
                },
                {
                    "start": 2423.99,
                    "end": 2424.77,
                    "word": " different",
                    "probability": 0.70458984375
                },
                {
                    "start": 2424.77,
                    "end": 2426.11,
                    "word": " adversarial",
                    "probability": 0.93310546875
                },
                {
                    "start": 2426.11,
                    "end": 2426.91,
                    "word": " perturbation",
                    "probability": 0.89208984375
                },
                {
                    "start": 2426.91,
                    "end": 2427.63,
                    "word": " attacks",
                    "probability": 0.732421875
                },
                {
                    "start": 2427.63,
                    "end": 2427.87,
                    "word": ",",
                    "probability": 0.87158203125
                },
                {
                    "start": 2427.87,
                    "end": 2428.25,
                    "word": " some",
                    "probability": 0.78515625
                },
                {
                    "start": 2428.25,
                    "end": 2428.35,
                    "word": " of",
                    "probability": 0.89453125
                },
                {
                    "start": 2428.35,
                    "end": 2428.45,
                    "word": " which",
                    "probability": 0.90869140625
                }
            ]
        },
        {
            "id": 361,
            "seek": 242924,
            "start": 2428.45,
            "end": 2435.37,
            "text": " we've covered so far in this talk. So, as a proof of concept, we consider a slightly less complex",
            "tokens": [
                51428,
                321,
                600,
                5343,
                370,
                1400,
                294,
                341,
                751,
                13,
                407,
                11,
                382,
                257,
                8177,
                295,
                3410,
                11,
                321,
                1949,
                257,
                4748,
                1570,
                3997,
                51784
            ],
            "temperature": 0.0,
            "avg_logprob": -0.19438844150112522,
            "compression_ratio": 1.5819672131147542,
            "no_speech_prob": 0.0004801750183105469,
            "words": [
                {
                    "start": 2428.45,
                    "end": 2428.61,
                    "word": " we've",
                    "probability": 0.904541015625
                },
                {
                    "start": 2428.61,
                    "end": 2428.83,
                    "word": " covered",
                    "probability": 0.716796875
                },
                {
                    "start": 2428.83,
                    "end": 2429.01,
                    "word": " so",
                    "probability": 0.89892578125
                },
                {
                    "start": 2429.01,
                    "end": 2429.13,
                    "word": " far",
                    "probability": 0.89111328125
                },
                {
                    "start": 2429.13,
                    "end": 2429.25,
                    "word": " in",
                    "probability": 0.90283203125
                },
                {
                    "start": 2429.25,
                    "end": 2429.39,
                    "word": " this",
                    "probability": 0.87841796875
                },
                {
                    "start": 2429.39,
                    "end": 2429.59,
                    "word": " talk",
                    "probability": 0.82373046875
                },
                {
                    "start": 2429.59,
                    "end": 2430.11,
                    "word": ".",
                    "probability": 0.99609375
                },
                {
                    "start": 2430.85,
                    "end": 2431.37,
                    "word": " So",
                    "probability": 0.86474609375
                },
                {
                    "start": 2431.37,
                    "end": 2431.63,
                    "word": ",",
                    "probability": 0.58984375
                },
                {
                    "start": 2431.63,
                    "end": 2431.85,
                    "word": " as",
                    "probability": 0.90087890625
                },
                {
                    "start": 2431.93,
                    "end": 2432.07,
                    "word": " a proof",
                    "probability": 0.82568359375
                },
                {
                    "start": 2432.07,
                    "end": 2432.19,
                    "word": " of",
                    "probability": 0.8505859375
                },
                {
                    "start": 2432.19,
                    "end": 2432.51,
                    "word": " concept",
                    "probability": 0.8046875
                },
                {
                    "start": 2432.51,
                    "end": 2432.59,
                    "word": ",",
                    "probability": 0.90966796875
                },
                {
                    "start": 2432.59,
                    "end": 2432.71,
                    "word": " we",
                    "probability": 0.9033203125
                },
                {
                    "start": 2432.71,
                    "end": 2433.17,
                    "word": " consider",
                    "probability": 0.78271484375
                },
                {
                    "start": 2433.87,
                    "end": 2434.27,
                    "word": " a slightly",
                    "probability": 0.77978515625
                },
                {
                    "start": 2434.27,
                    "end": 2434.77,
                    "word": " less",
                    "probability": 0.88330078125
                },
                {
                    "start": 2434.77,
                    "end": 2435.37,
                    "word": " complex",
                    "probability": 0.72216796875
                }
            ]
        },
        {
            "id": 362,
            "seek": 245786,
            "start": 2435.37,
            "end": 2441.73,
            "text": " environment data cart poll, where the objective is to stabilize this poll on this cart by moving",
            "tokens": [
                50368,
                2823,
                1412,
                5467,
                6418,
                11,
                689,
                264,
                10024,
                307,
                281,
                31870,
                341,
                6418,
                322,
                341,
                5467,
                538,
                2684,
                50680
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2996002848065177,
            "compression_ratio": 1.5791666666666666,
            "no_speech_prob": 0.0074615478515625,
            "words": [
                {
                    "start": 2435.37,
                    "end": 2435.97,
                    "word": " environment",
                    "probability": 0.517578125
                },
                {
                    "start": 2435.97,
                    "end": 2436.25,
                    "word": " data",
                    "probability": 0.082763671875
                },
                {
                    "start": 2436.25,
                    "end": 2436.67,
                    "word": " cart",
                    "probability": 0.4267578125
                },
                {
                    "start": 2436.67,
                    "end": 2436.89,
                    "word": " poll",
                    "probability": 0.356201171875
                },
                {
                    "start": 2436.89,
                    "end": 2436.97,
                    "word": ",",
                    "probability": 0.2103271484375
                },
                {
                    "start": 2436.97,
                    "end": 2437.07,
                    "word": " where",
                    "probability": 0.8984375
                },
                {
                    "start": 2437.07,
                    "end": 2437.37,
                    "word": " the",
                    "probability": 0.81494140625
                },
                {
                    "start": 2437.37,
                    "end": 2438.05,
                    "word": " objective",
                    "probability": 0.7998046875
                },
                {
                    "start": 2438.05,
                    "end": 2438.33,
                    "word": " is",
                    "probability": 0.90576171875
                },
                {
                    "start": 2438.33,
                    "end": 2438.57,
                    "word": " to",
                    "probability": 0.91064453125
                },
                {
                    "start": 2438.57,
                    "end": 2439.21,
                    "word": " stabilize",
                    "probability": 0.59521484375
                },
                {
                    "start": 2439.21,
                    "end": 2440.09,
                    "word": " this",
                    "probability": 0.85791015625
                },
                {
                    "start": 2440.09,
                    "end": 2440.39,
                    "word": " poll",
                    "probability": 0.60400390625
                },
                {
                    "start": 2440.39,
                    "end": 2440.59,
                    "word": " on",
                    "probability": 0.8056640625
                },
                {
                    "start": 2440.59,
                    "end": 2440.73,
                    "word": " this",
                    "probability": 0.86669921875
                },
                {
                    "start": 2440.73,
                    "end": 2441.05,
                    "word": " cart",
                    "probability": 0.8271484375
                },
                {
                    "start": 2441.05,
                    "end": 2441.33,
                    "word": " by",
                    "probability": 0.8798828125
                },
                {
                    "start": 2441.33,
                    "end": 2441.73,
                    "word": " moving",
                    "probability": 0.853515625
                }
            ]
        },
        {
            "id": 363,
            "seek": 245786,
            "start": 2441.73,
            "end": 2448.97,
            "text": " the cart to right and left. The reason for choosing a simple environment is merely economical,",
            "tokens": [
                50680,
                264,
                5467,
                281,
                558,
                293,
                1411,
                13,
                440,
                1778,
                337,
                10875,
                257,
                2199,
                2823,
                307,
                17003,
                42473,
                11,
                51028
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2996002848065177,
            "compression_ratio": 1.5791666666666666,
            "no_speech_prob": 0.0074615478515625,
            "words": [
                {
                    "start": 2441.73,
                    "end": 2441.99,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 2441.99,
                    "end": 2442.25,
                    "word": " cart",
                    "probability": 0.5712890625
                },
                {
                    "start": 2442.25,
                    "end": 2442.49,
                    "word": " to",
                    "probability": 0.88427734375
                },
                {
                    "start": 2442.49,
                    "end": 2442.89,
                    "word": " right",
                    "probability": 0.79052734375
                },
                {
                    "start": 2442.89,
                    "end": 2443.03,
                    "word": " and",
                    "probability": 0.6474609375
                },
                {
                    "start": 2443.03,
                    "end": 2443.31,
                    "word": " left",
                    "probability": 0.85986328125
                },
                {
                    "start": 2443.31,
                    "end": 2443.53,
                    "word": ".",
                    "probability": 0.85888671875
                },
                {
                    "start": 2443.53,
                    "end": 2443.91,
                    "word": " The",
                    "probability": 0.8017578125
                },
                {
                    "start": 2443.91,
                    "end": 2444.21,
                    "word": " reason",
                    "probability": 0.916015625
                },
                {
                    "start": 2444.21,
                    "end": 2444.37,
                    "word": " for",
                    "probability": 0.8876953125
                },
                {
                    "start": 2444.37,
                    "end": 2444.53,
                    "word": " choosing",
                    "probability": 0.84619140625
                },
                {
                    "start": 2444.69,
                    "end": 2444.89,
                    "word": " a simple",
                    "probability": 0.91796875
                },
                {
                    "start": 2444.89,
                    "end": 2445.33,
                    "word": " environment",
                    "probability": 0.728515625
                },
                {
                    "start": 2445.33,
                    "end": 2446.03,
                    "word": " is",
                    "probability": 0.9013671875
                },
                {
                    "start": 2446.03,
                    "end": 2446.57,
                    "word": " merely",
                    "probability": 0.51953125
                },
                {
                    "start": 2446.57,
                    "end": 2448.59,
                    "word": " economical",
                    "probability": 0.701171875
                },
                {
                    "start": 2448.59,
                    "end": 2448.97,
                    "word": ",",
                    "probability": 0.4638671875
                }
            ]
        },
        {
            "id": 364,
            "seek": 245786,
            "start": 2449.99,
            "end": 2457.31,
            "text": " because we didn't want the experiments to take days or weeks. We start with a simple case,",
            "tokens": [
                51104,
                570,
                321,
                994,
                380,
                528,
                264,
                12050,
                281,
                747,
                1708,
                420,
                3259,
                13,
                492,
                722,
                365,
                257,
                2199,
                1389,
                11,
                51452
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2996002848065177,
            "compression_ratio": 1.5791666666666666,
            "no_speech_prob": 0.0074615478515625,
            "words": [
                {
                    "start": 2449.99,
                    "end": 2450.59,
                    "word": " because",
                    "probability": 0.82421875
                },
                {
                    "start": 2450.59,
                    "end": 2451.17,
                    "word": " we",
                    "probability": 0.90234375
                },
                {
                    "start": 2451.17,
                    "end": 2451.35,
                    "word": " didn't",
                    "probability": 0.91357421875
                },
                {
                    "start": 2451.35,
                    "end": 2451.51,
                    "word": " want",
                    "probability": 0.82373046875
                },
                {
                    "start": 2451.51,
                    "end": 2451.63,
                    "word": " the",
                    "probability": 0.79833984375
                },
                {
                    "start": 2451.63,
                    "end": 2451.95,
                    "word": " experiments",
                    "probability": 0.80224609375
                },
                {
                    "start": 2451.95,
                    "end": 2452.27,
                    "word": " to",
                    "probability": 0.90234375
                },
                {
                    "start": 2452.27,
                    "end": 2453.13,
                    "word": " take",
                    "probability": 0.82373046875
                },
                {
                    "start": 2453.13,
                    "end": 2454.15,
                    "word": " days",
                    "probability": 0.81005859375
                },
                {
                    "start": 2454.15,
                    "end": 2454.93,
                    "word": " or",
                    "probability": 0.8447265625
                },
                {
                    "start": 2454.93,
                    "end": 2455.35,
                    "word": " weeks",
                    "probability": 0.73828125
                },
                {
                    "start": 2455.35,
                    "end": 2455.51,
                    "word": ".",
                    "probability": 0.642578125
                },
                {
                    "start": 2455.51,
                    "end": 2455.77,
                    "word": " We",
                    "probability": 0.89453125
                },
                {
                    "start": 2455.77,
                    "end": 2456.07,
                    "word": " start",
                    "probability": 0.07275390625
                },
                {
                    "start": 2456.07,
                    "end": 2456.19,
                    "word": " with",
                    "probability": 0.8203125
                },
                {
                    "start": 2456.33,
                    "end": 2456.55,
                    "word": " a simple",
                    "probability": 0.9052734375
                },
                {
                    "start": 2456.55,
                    "end": 2456.93,
                    "word": " case",
                    "probability": 0.81689453125
                },
                {
                    "start": 2456.93,
                    "end": 2457.31,
                    "word": ",",
                    "probability": 0.32080078125
                }
            ]
        },
        {
            "id": 365,
            "seek": 245786,
            "start": 2457.31,
            "end": 2463.99,
            "text": " and we consider different types of policies. DQN with prioritized replay, an enhanced version of",
            "tokens": [
                51452,
                293,
                321,
                1949,
                819,
                3467,
                295,
                7657,
                13,
                413,
                48,
                45,
                365,
                14846,
                1602,
                23836,
                11,
                364,
                21191,
                3037,
                295,
                51788
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2996002848065177,
            "compression_ratio": 1.5791666666666666,
            "no_speech_prob": 0.0074615478515625,
            "words": [
                {
                    "start": 2457.31,
                    "end": 2457.43,
                    "word": " and",
                    "probability": 0.89208984375
                },
                {
                    "start": 2457.43,
                    "end": 2457.53,
                    "word": " we",
                    "probability": 0.90087890625
                },
                {
                    "start": 2457.53,
                    "end": 2457.81,
                    "word": " consider",
                    "probability": 0.6083984375
                },
                {
                    "start": 2457.81,
                    "end": 2458.15,
                    "word": " different",
                    "probability": 0.69140625
                },
                {
                    "start": 2458.15,
                    "end": 2458.59,
                    "word": " types",
                    "probability": 0.751953125
                },
                {
                    "start": 2458.59,
                    "end": 2459.23,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 2459.23,
                    "end": 2459.89,
                    "word": " policies",
                    "probability": 0.8330078125
                },
                {
                    "start": 2459.89,
                    "end": 2460.13,
                    "word": ".",
                    "probability": 0.7373046875
                },
                {
                    "start": 2460.13,
                    "end": 2460.53,
                    "word": " DQN",
                    "probability": 0.8653971354166666
                },
                {
                    "start": 2460.53,
                    "end": 2460.95,
                    "word": " with",
                    "probability": 0.75830078125
                },
                {
                    "start": 2460.95,
                    "end": 2462.01,
                    "word": " prioritized",
                    "probability": 0.7451171875
                },
                {
                    "start": 2462.01,
                    "end": 2462.25,
                    "word": " replay",
                    "probability": 0.84814453125
                },
                {
                    "start": 2462.25,
                    "end": 2462.45,
                    "word": ",",
                    "probability": 0.79736328125
                },
                {
                    "start": 2462.45,
                    "end": 2462.69,
                    "word": " an",
                    "probability": 0.46337890625
                },
                {
                    "start": 2462.69,
                    "end": 2463.39,
                    "word": " enhanced",
                    "probability": 0.80810546875
                },
                {
                    "start": 2463.39,
                    "end": 2463.69,
                    "word": " version",
                    "probability": 0.71044921875
                },
                {
                    "start": 2463.69,
                    "end": 2463.99,
                    "word": " of",
                    "probability": 0.90869140625
                }
            ]
        },
        {
            "id": 366,
            "seek": 248346,
            "start": 2463.99,
            "end": 2470.91,
            "text": " the classical DQN proximal, policy optimization, and asynchronous actor-critic. And, of course,",
            "tokens": [
                50364,
                264,
                13735,
                413,
                48,
                45,
                21932,
                304,
                11,
                3897,
                19618,
                11,
                293,
                49174,
                8747,
                12,
                32255,
                299,
                13,
                400,
                11,
                295,
                1164,
                11,
                50704
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24365234033515057,
            "compression_ratio": 1.5311203319502074,
            "no_speech_prob": 0.0011157989501953125,
            "words": [
                {
                    "start": 2463.99,
                    "end": 2464.09,
                    "word": " the",
                    "probability": 0.26220703125
                },
                {
                    "start": 2464.09,
                    "end": 2464.41,
                    "word": " classical",
                    "probability": 0.8671875
                },
                {
                    "start": 2464.41,
                    "end": 2464.77,
                    "word": " DQN",
                    "probability": 0.8720703125
                },
                {
                    "start": 2464.77,
                    "end": 2465.35,
                    "word": " proximal",
                    "probability": 0.811279296875
                },
                {
                    "start": 2465.35,
                    "end": 2465.79,
                    "word": ",",
                    "probability": 0.455322265625
                },
                {
                    "start": 2465.79,
                    "end": 2466.13,
                    "word": " policy",
                    "probability": 0.7998046875
                },
                {
                    "start": 2466.13,
                    "end": 2466.69,
                    "word": " optimization",
                    "probability": 0.78662109375
                },
                {
                    "start": 2466.69,
                    "end": 2467.15,
                    "word": ",",
                    "probability": 0.517578125
                },
                {
                    "start": 2467.15,
                    "end": 2467.41,
                    "word": " and",
                    "probability": 0.8740234375
                },
                {
                    "start": 2467.41,
                    "end": 2467.83,
                    "word": " asynchronous",
                    "probability": 0.83349609375
                },
                {
                    "start": 2467.83,
                    "end": 2468.35,
                    "word": " actor",
                    "probability": 0.48876953125
                },
                {
                    "start": 2468.35,
                    "end": 2468.85,
                    "word": "-critic",
                    "probability": 0.7676595052083334
                },
                {
                    "start": 2468.85,
                    "end": 2469.41,
                    "word": ".",
                    "probability": 0.88818359375
                },
                {
                    "start": 2469.83,
                    "end": 2470.05,
                    "word": " And",
                    "probability": 0.78369140625
                },
                {
                    "start": 2470.05,
                    "end": 2470.31,
                    "word": ",",
                    "probability": 0.284912109375
                },
                {
                    "start": 2470.31,
                    "end": 2470.57,
                    "word": " of",
                    "probability": 0.9072265625
                },
                {
                    "start": 2470.57,
                    "end": 2470.77,
                    "word": " course",
                    "probability": 0.86181640625
                },
                {
                    "start": 2470.77,
                    "end": 2470.91,
                    "word": ",",
                    "probability": 0.953125
                }
            ]
        },
        {
            "id": 367,
            "seek": 248346,
            "start": 2470.91,
            "end": 2478.73,
            "text": " we also train an adversarial RL agent, a DQN agent, whose objective is to incur maximum loss",
            "tokens": [
                50704,
                321,
                611,
                3847,
                364,
                17641,
                44745,
                497,
                43,
                9461,
                11,
                257,
                413,
                48,
                45,
                9461,
                11,
                6104,
                10024,
                307,
                281,
                35774,
                6674,
                4470,
                51100
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24365234033515057,
            "compression_ratio": 1.5311203319502074,
            "no_speech_prob": 0.0011157989501953125,
            "words": [
                {
                    "start": 2470.91,
                    "end": 2471.07,
                    "word": " we",
                    "probability": 0.904296875
                },
                {
                    "start": 2471.07,
                    "end": 2471.43,
                    "word": " also",
                    "probability": 0.8154296875
                },
                {
                    "start": 2471.43,
                    "end": 2471.71,
                    "word": " train",
                    "probability": 0.6484375
                },
                {
                    "start": 2471.71,
                    "end": 2471.89,
                    "word": " an",
                    "probability": 0.87744140625
                },
                {
                    "start": 2471.89,
                    "end": 2472.69,
                    "word": " adversarial",
                    "probability": 0.928466796875
                },
                {
                    "start": 2472.69,
                    "end": 2473.03,
                    "word": " RL",
                    "probability": 0.86572265625
                },
                {
                    "start": 2473.03,
                    "end": 2473.29,
                    "word": " agent",
                    "probability": 0.80859375
                },
                {
                    "start": 2473.29,
                    "end": 2473.39,
                    "word": ",",
                    "probability": 0.921875
                },
                {
                    "start": 2473.47,
                    "end": 2473.89,
                    "word": " a DQN",
                    "probability": 0.9788411458333334
                },
                {
                    "start": 2473.89,
                    "end": 2474.17,
                    "word": " agent",
                    "probability": 0.7998046875
                },
                {
                    "start": 2474.17,
                    "end": 2474.19,
                    "word": ",",
                    "probability": 0.74853515625
                },
                {
                    "start": 2474.19,
                    "end": 2474.37,
                    "word": " whose",
                    "probability": 0.8271484375
                },
                {
                    "start": 2474.37,
                    "end": 2474.79,
                    "word": " objective",
                    "probability": 0.84521484375
                },
                {
                    "start": 2474.79,
                    "end": 2475.77,
                    "word": " is",
                    "probability": 0.89697265625
                },
                {
                    "start": 2475.77,
                    "end": 2476.25,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 2476.25,
                    "end": 2477.25,
                    "word": " incur",
                    "probability": 0.8935546875
                },
                {
                    "start": 2477.25,
                    "end": 2478.01,
                    "word": " maximum",
                    "probability": 0.81298828125
                },
                {
                    "start": 2478.01,
                    "end": 2478.73,
                    "word": " loss",
                    "probability": 0.828125
                }
            ]
        },
        {
            "id": 368,
            "seek": 248346,
            "start": 2478.73,
            "end": 2484.33,
            "text": " of reward, or in other terms, in more technical terms, maximize the regret of its target.",
            "tokens": [
                51100,
                295,
                7782,
                11,
                420,
                294,
                661,
                2115,
                11,
                294,
                544,
                6191,
                2115,
                11,
                19874,
                264,
                10879,
                295,
                1080,
                3779,
                13,
                51368
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24365234033515057,
            "compression_ratio": 1.5311203319502074,
            "no_speech_prob": 0.0011157989501953125,
            "words": [
                {
                    "start": 2478.73,
                    "end": 2478.91,
                    "word": " of",
                    "probability": 0.8828125
                },
                {
                    "start": 2478.91,
                    "end": 2479.33,
                    "word": " reward",
                    "probability": 0.86083984375
                },
                {
                    "start": 2479.33,
                    "end": 2479.61,
                    "word": ",",
                    "probability": 0.5703125
                },
                {
                    "start": 2479.61,
                    "end": 2479.83,
                    "word": " or",
                    "probability": 0.89599609375
                },
                {
                    "start": 2479.83,
                    "end": 2480.21,
                    "word": " in",
                    "probability": 0.62548828125
                },
                {
                    "start": 2480.21,
                    "end": 2480.35,
                    "word": " other",
                    "probability": 0.63427734375
                },
                {
                    "start": 2480.35,
                    "end": 2480.71,
                    "word": " terms",
                    "probability": 0.771484375
                },
                {
                    "start": 2480.71,
                    "end": 2480.87,
                    "word": ",",
                    "probability": 0.8681640625
                },
                {
                    "start": 2480.87,
                    "end": 2480.97,
                    "word": " in",
                    "probability": 0.8955078125
                },
                {
                    "start": 2480.97,
                    "end": 2481.15,
                    "word": " more",
                    "probability": 0.87841796875
                },
                {
                    "start": 2481.15,
                    "end": 2481.43,
                    "word": " technical",
                    "probability": 0.85546875
                },
                {
                    "start": 2481.43,
                    "end": 2481.77,
                    "word": " terms",
                    "probability": 0.77392578125
                },
                {
                    "start": 2481.77,
                    "end": 2482.19,
                    "word": ",",
                    "probability": 0.98388671875
                },
                {
                    "start": 2482.19,
                    "end": 2482.55,
                    "word": " maximize",
                    "probability": 0.74072265625
                },
                {
                    "start": 2482.55,
                    "end": 2482.79,
                    "word": " the",
                    "probability": 0.7978515625
                },
                {
                    "start": 2482.79,
                    "end": 2483.15,
                    "word": " regret",
                    "probability": 0.662109375
                },
                {
                    "start": 2483.15,
                    "end": 2483.55,
                    "word": " of",
                    "probability": 0.8525390625
                },
                {
                    "start": 2483.55,
                    "end": 2483.71,
                    "word": " its",
                    "probability": 0.806640625
                },
                {
                    "start": 2483.71,
                    "end": 2484.05,
                    "word": " target",
                    "probability": 0.82470703125
                },
                {
                    "start": 2484.05,
                    "end": 2484.33,
                    "word": ".",
                    "probability": 0.99609375
                }
            ]
        },
        {
            "id": 369,
            "seek": 248346,
            "start": 2485.39,
            "end": 2489.59,
            "text": " And here are the results. First, with regards to replication progress, based on only 5,000",
            "tokens": [
                51436,
                400,
                510,
                366,
                264,
                3542,
                13,
                2386,
                11,
                365,
                14258,
                281,
                39911,
                4205,
                11,
                2361,
                322,
                787,
                1025,
                11,
                1360,
                51640
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24365234033515057,
            "compression_ratio": 1.5311203319502074,
            "no_speech_prob": 0.0011157989501953125,
            "words": [
                {
                    "start": 2485.39,
                    "end": 2485.71,
                    "word": " And",
                    "probability": 0.87353515625
                },
                {
                    "start": 2485.71,
                    "end": 2485.89,
                    "word": " here",
                    "probability": 0.8095703125
                },
                {
                    "start": 2485.89,
                    "end": 2485.99,
                    "word": " are",
                    "probability": 0.89208984375
                },
                {
                    "start": 2485.99,
                    "end": 2486.13,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 2486.13,
                    "end": 2486.57,
                    "word": " results",
                    "probability": 0.6845703125
                },
                {
                    "start": 2486.57,
                    "end": 2486.73,
                    "word": ".",
                    "probability": 0.67724609375
                },
                {
                    "start": 2486.73,
                    "end": 2487.01,
                    "word": " First",
                    "probability": 0.8955078125
                },
                {
                    "start": 2487.01,
                    "end": 2487.07,
                    "word": ",",
                    "probability": 0.75634765625
                },
                {
                    "start": 2487.07,
                    "end": 2487.13,
                    "word": " with",
                    "probability": 0.80078125
                },
                {
                    "start": 2487.13,
                    "end": 2487.35,
                    "word": " regards",
                    "probability": 0.8173828125
                },
                {
                    "start": 2487.35,
                    "end": 2487.51,
                    "word": " to",
                    "probability": 0.90771484375
                },
                {
                    "start": 2487.51,
                    "end": 2487.87,
                    "word": " replication",
                    "probability": 0.69970703125
                },
                {
                    "start": 2487.87,
                    "end": 2488.47,
                    "word": " progress",
                    "probability": 0.7001953125
                },
                {
                    "start": 2488.47,
                    "end": 2488.53,
                    "word": ",",
                    "probability": 0.452880859375
                },
                {
                    "start": 2488.53,
                    "end": 2488.75,
                    "word": " based",
                    "probability": 0.83447265625
                },
                {
                    "start": 2488.75,
                    "end": 2488.87,
                    "word": " on",
                    "probability": 0.90576171875
                },
                {
                    "start": 2488.87,
                    "end": 2489.05,
                    "word": " only",
                    "probability": 0.84765625
                },
                {
                    "start": 2489.05,
                    "end": 2489.33,
                    "word": " 5",
                    "probability": 0.7734375
                },
                {
                    "start": 2489.33,
                    "end": 2489.59,
                    "word": ",000",
                    "probability": 0.99609375
                }
            ]
        },
        {
            "id": 370,
            "seek": 250830,
            "start": 2489.59,
            "end": 2496.65,
            "text": " demonstrations, 5,000 state action, next state reward observations, we see that all three",
            "tokens": [
                50364,
                34714,
                11,
                1025,
                11,
                1360,
                1785,
                3069,
                11,
                958,
                1785,
                7782,
                18163,
                11,
                321,
                536,
                300,
                439,
                1045,
                50708
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2243303627307926,
            "compression_ratio": 1.5411764705882354,
            "no_speech_prob": 0.0012845993041992188,
            "words": [
                {
                    "start": 2489.59,
                    "end": 2490.21,
                    "word": " demonstrations",
                    "probability": 0.1805419921875
                },
                {
                    "start": 2490.21,
                    "end": 2490.31,
                    "word": ",",
                    "probability": 0.619140625
                },
                {
                    "start": 2490.31,
                    "end": 2490.51,
                    "word": " 5",
                    "probability": 0.37744140625
                },
                {
                    "start": 2490.51,
                    "end": 2490.75,
                    "word": ",000",
                    "probability": 0.931640625
                },
                {
                    "start": 2490.75,
                    "end": 2491.03,
                    "word": " state",
                    "probability": 0.79638671875
                },
                {
                    "start": 2491.03,
                    "end": 2491.41,
                    "word": " action",
                    "probability": 0.77392578125
                },
                {
                    "start": 2491.41,
                    "end": 2491.55,
                    "word": ",",
                    "probability": 0.7099609375
                },
                {
                    "start": 2491.55,
                    "end": 2491.77,
                    "word": " next",
                    "probability": 0.8046875
                },
                {
                    "start": 2491.77,
                    "end": 2492.01,
                    "word": " state",
                    "probability": 0.75439453125
                },
                {
                    "start": 2492.01,
                    "end": 2492.43,
                    "word": " reward",
                    "probability": 0.84521484375
                },
                {
                    "start": 2492.43,
                    "end": 2493.45,
                    "word": " observations",
                    "probability": 0.63916015625
                },
                {
                    "start": 2493.45,
                    "end": 2494.01,
                    "word": ",",
                    "probability": 0.27490234375
                },
                {
                    "start": 2494.01,
                    "end": 2494.41,
                    "word": " we",
                    "probability": 0.89501953125
                },
                {
                    "start": 2494.41,
                    "end": 2494.71,
                    "word": " see",
                    "probability": 0.82421875
                },
                {
                    "start": 2494.71,
                    "end": 2495.11,
                    "word": " that",
                    "probability": 0.880859375
                },
                {
                    "start": 2495.11,
                    "end": 2496.17,
                    "word": " all",
                    "probability": 0.87646484375
                },
                {
                    "start": 2496.17,
                    "end": 2496.65,
                    "word": " three",
                    "probability": 0.83935546875
                }
            ]
        },
        {
            "id": 371,
            "seek": 250830,
            "start": 2497.45,
            "end": 2506.67,
            "text": " policies are almost exactly replicated. We can see convergence to the optimal performance",
            "tokens": [
                50772,
                7657,
                366,
                1920,
                2293,
                46365,
                13,
                492,
                393,
                536,
                32181,
                281,
                264,
                16252,
                3389,
                51216
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2243303627307926,
            "compression_ratio": 1.5411764705882354,
            "no_speech_prob": 0.0012845993041992188,
            "words": [
                {
                    "start": 2497.45,
                    "end": 2498.23,
                    "word": " policies",
                    "probability": 0.81201171875
                },
                {
                    "start": 2498.23,
                    "end": 2498.75,
                    "word": " are",
                    "probability": 0.89208984375
                },
                {
                    "start": 2498.75,
                    "end": 2499.55,
                    "word": " almost",
                    "probability": 0.67333984375
                },
                {
                    "start": 2499.55,
                    "end": 2502.57,
                    "word": " exactly",
                    "probability": 0.5419921875
                },
                {
                    "start": 2502.57,
                    "end": 2503.19,
                    "word": " replicated",
                    "probability": 0.86083984375
                },
                {
                    "start": 2503.19,
                    "end": 2503.57,
                    "word": ".",
                    "probability": 0.66455078125
                },
                {
                    "start": 2503.57,
                    "end": 2503.65,
                    "word": " We",
                    "probability": 0.89990234375
                },
                {
                    "start": 2503.65,
                    "end": 2503.81,
                    "word": " can",
                    "probability": 0.89599609375
                },
                {
                    "start": 2503.81,
                    "end": 2504.07,
                    "word": " see",
                    "probability": 0.81494140625
                },
                {
                    "start": 2504.07,
                    "end": 2504.65,
                    "word": " convergence",
                    "probability": 0.7919921875
                },
                {
                    "start": 2504.65,
                    "end": 2505.27,
                    "word": " to",
                    "probability": 0.90185546875
                },
                {
                    "start": 2505.27,
                    "end": 2505.67,
                    "word": " the",
                    "probability": 0.814453125
                },
                {
                    "start": 2505.67,
                    "end": 2505.99,
                    "word": " optimal",
                    "probability": 0.72119140625
                },
                {
                    "start": 2505.99,
                    "end": 2506.67,
                    "word": " performance",
                    "probability": 0.853515625
                }
            ]
        },
        {
            "id": 372,
            "seek": 250830,
            "start": 2507.35,
            "end": 2514.43,
            "text": " of those policies in the environment. And then we perform adversarial training. We",
            "tokens": [
                51252,
                295,
                729,
                7657,
                294,
                264,
                2823,
                13,
                400,
                550,
                321,
                2042,
                17641,
                44745,
                3097,
                13,
                492,
                51600
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2243303627307926,
            "compression_ratio": 1.5411764705882354,
            "no_speech_prob": 0.0012845993041992188,
            "words": [
                {
                    "start": 2507.35,
                    "end": 2507.89,
                    "word": " of",
                    "probability": 0.90576171875
                },
                {
                    "start": 2507.89,
                    "end": 2509.23,
                    "word": " those",
                    "probability": 0.83154296875
                },
                {
                    "start": 2509.23,
                    "end": 2509.75,
                    "word": " policies",
                    "probability": 0.861328125
                },
                {
                    "start": 2509.75,
                    "end": 2510.29,
                    "word": " in",
                    "probability": 0.896484375
                },
                {
                    "start": 2510.29,
                    "end": 2510.39,
                    "word": " the",
                    "probability": 0.806640625
                },
                {
                    "start": 2510.39,
                    "end": 2510.75,
                    "word": " environment",
                    "probability": 0.74853515625
                },
                {
                    "start": 2510.75,
                    "end": 2511.53,
                    "word": ".",
                    "probability": 0.76416015625
                },
                {
                    "start": 2512.03,
                    "end": 2512.29,
                    "word": " And",
                    "probability": 0.82177734375
                },
                {
                    "start": 2512.29,
                    "end": 2512.55,
                    "word": " then",
                    "probability": 0.7255859375
                },
                {
                    "start": 2512.55,
                    "end": 2512.85,
                    "word": " we",
                    "probability": 0.798828125
                },
                {
                    "start": 2512.85,
                    "end": 2513.17,
                    "word": " perform",
                    "probability": 0.77392578125
                },
                {
                    "start": 2513.17,
                    "end": 2513.65,
                    "word": " adversarial",
                    "probability": 0.9150390625
                },
                {
                    "start": 2513.65,
                    "end": 2514.03,
                    "word": " training",
                    "probability": 0.8076171875
                },
                {
                    "start": 2514.03,
                    "end": 2514.25,
                    "word": ".",
                    "probability": 0.68115234375
                },
                {
                    "start": 2514.25,
                    "end": 2514.43,
                    "word": " We",
                    "probability": 0.89990234375
                }
            ]
        },
        {
            "id": 373,
            "seek": 253242,
            "start": 2514.43,
            "end": 2521.61,
            "text": " train an adversarial RL agent to attack and maximize the regret of those policies,",
            "tokens": [
                50388,
                3847,
                364,
                17641,
                44745,
                497,
                43,
                9461,
                281,
                2690,
                293,
                19874,
                264,
                10879,
                295,
                729,
                7657,
                11,
                50720
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2402935673793157,
            "compression_ratio": 1.425414364640884,
            "no_speech_prob": 0.0330810546875,
            "words": [
                {
                    "start": 2514.43,
                    "end": 2515.35,
                    "word": " train",
                    "probability": 0.2393798828125
                },
                {
                    "start": 2515.35,
                    "end": 2515.61,
                    "word": " an",
                    "probability": 0.79345703125
                },
                {
                    "start": 2515.61,
                    "end": 2516.23,
                    "word": " adversarial",
                    "probability": 0.92333984375
                },
                {
                    "start": 2516.23,
                    "end": 2516.61,
                    "word": " RL",
                    "probability": 0.90576171875
                },
                {
                    "start": 2516.61,
                    "end": 2516.95,
                    "word": " agent",
                    "probability": 0.75439453125
                },
                {
                    "start": 2516.95,
                    "end": 2517.17,
                    "word": " to",
                    "probability": 0.861328125
                },
                {
                    "start": 2517.17,
                    "end": 2517.65,
                    "word": " attack",
                    "probability": 0.83056640625
                },
                {
                    "start": 2517.65,
                    "end": 2518.05,
                    "word": " and",
                    "probability": 0.87890625
                },
                {
                    "start": 2518.05,
                    "end": 2518.45,
                    "word": " maximize",
                    "probability": 0.69091796875
                },
                {
                    "start": 2518.45,
                    "end": 2518.77,
                    "word": " the",
                    "probability": 0.66064453125
                },
                {
                    "start": 2518.77,
                    "end": 2519.27,
                    "word": " regret",
                    "probability": 0.4033203125
                },
                {
                    "start": 2519.27,
                    "end": 2520.27,
                    "word": " of",
                    "probability": 0.73828125
                },
                {
                    "start": 2520.27,
                    "end": 2520.55,
                    "word": " those",
                    "probability": 0.82373046875
                },
                {
                    "start": 2520.55,
                    "end": 2521.01,
                    "word": " policies",
                    "probability": 0.8515625
                },
                {
                    "start": 2521.01,
                    "end": 2521.61,
                    "word": ",",
                    "probability": 0.200439453125
                }
            ]
        },
        {
            "id": 374,
            "seek": 253242,
            "start": 2521.61,
            "end": 2529.33,
            "text": " and you can see that this can also be easily achieved within very few iterations of training",
            "tokens": [
                50720,
                293,
                291,
                393,
                536,
                300,
                341,
                393,
                611,
                312,
                3612,
                11042,
                1951,
                588,
                1326,
                36540,
                295,
                3097,
                51116
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2402935673793157,
            "compression_ratio": 1.425414364640884,
            "no_speech_prob": 0.0330810546875,
            "words": [
                {
                    "start": 2521.61,
                    "end": 2521.81,
                    "word": " and",
                    "probability": 0.89013671875
                },
                {
                    "start": 2521.81,
                    "end": 2521.95,
                    "word": " you",
                    "probability": 0.89697265625
                },
                {
                    "start": 2521.95,
                    "end": 2522.11,
                    "word": " can",
                    "probability": 0.89794921875
                },
                {
                    "start": 2522.11,
                    "end": 2522.37,
                    "word": " see",
                    "probability": 0.82275390625
                },
                {
                    "start": 2522.37,
                    "end": 2522.69,
                    "word": " that",
                    "probability": 0.88037109375
                },
                {
                    "start": 2522.69,
                    "end": 2523.11,
                    "word": " this",
                    "probability": 0.8759765625
                },
                {
                    "start": 2523.11,
                    "end": 2523.27,
                    "word": " can",
                    "probability": 0.89501953125
                },
                {
                    "start": 2523.27,
                    "end": 2523.51,
                    "word": " also",
                    "probability": 0.81396484375
                },
                {
                    "start": 2523.51,
                    "end": 2523.69,
                    "word": " be",
                    "probability": 0.9072265625
                },
                {
                    "start": 2523.69,
                    "end": 2523.99,
                    "word": " easily",
                    "probability": 0.8359375
                },
                {
                    "start": 2523.99,
                    "end": 2524.37,
                    "word": " achieved",
                    "probability": 0.736328125
                },
                {
                    "start": 2524.37,
                    "end": 2524.87,
                    "word": " within",
                    "probability": 0.77001953125
                },
                {
                    "start": 2524.87,
                    "end": 2526.25,
                    "word": " very",
                    "probability": 0.65869140625
                },
                {
                    "start": 2526.25,
                    "end": 2526.73,
                    "word": " few",
                    "probability": 0.8095703125
                },
                {
                    "start": 2526.73,
                    "end": 2527.49,
                    "word": " iterations",
                    "probability": 0.705078125
                },
                {
                    "start": 2527.49,
                    "end": 2528.75,
                    "word": " of",
                    "probability": 0.90087890625
                },
                {
                    "start": 2528.75,
                    "end": 2529.33,
                    "word": " training",
                    "probability": 0.77392578125
                }
            ]
        },
        {
            "id": 375,
            "seek": 253242,
            "start": 2529.33,
            "end": 2538.55,
            "text": " in CardPol. I believe we can see that for PPO2, which is a somewhat robust deep RL",
            "tokens": [
                51116,
                294,
                11877,
                47,
                401,
                13,
                286,
                1697,
                321,
                393,
                536,
                300,
                337,
                430,
                34885,
                17,
                11,
                597,
                307,
                257,
                8344,
                13956,
                2452,
                497,
                43,
                51560
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2402935673793157,
            "compression_ratio": 1.425414364640884,
            "no_speech_prob": 0.0330810546875,
            "words": [
                {
                    "start": 2529.33,
                    "end": 2529.79,
                    "word": " in",
                    "probability": 0.83740234375
                },
                {
                    "start": 2529.79,
                    "end": 2530.29,
                    "word": " CardPol",
                    "probability": 0.4755859375
                },
                {
                    "start": 2530.29,
                    "end": 2530.63,
                    "word": ".",
                    "probability": 0.83154296875
                },
                {
                    "start": 2530.69,
                    "end": 2531.45,
                    "word": " I",
                    "probability": 0.96533203125
                },
                {
                    "start": 2531.45,
                    "end": 2531.79,
                    "word": " believe",
                    "probability": 0.701171875
                },
                {
                    "start": 2531.79,
                    "end": 2532.43,
                    "word": " we",
                    "probability": 0.873046875
                },
                {
                    "start": 2532.43,
                    "end": 2532.57,
                    "word": " can",
                    "probability": 0.8984375
                },
                {
                    "start": 2532.57,
                    "end": 2532.75,
                    "word": " see",
                    "probability": 0.82470703125
                },
                {
                    "start": 2532.75,
                    "end": 2532.91,
                    "word": " that",
                    "probability": 0.87939453125
                },
                {
                    "start": 2532.91,
                    "end": 2533.13,
                    "word": " for",
                    "probability": 0.8916015625
                },
                {
                    "start": 2533.13,
                    "end": 2533.91,
                    "word": " PPO2",
                    "probability": 0.8155924479166666
                },
                {
                    "start": 2533.91,
                    "end": 2533.99,
                    "word": ",",
                    "probability": 0.7197265625
                },
                {
                    "start": 2533.99,
                    "end": 2534.17,
                    "word": " which",
                    "probability": 0.91796875
                },
                {
                    "start": 2534.17,
                    "end": 2534.43,
                    "word": " is",
                    "probability": 0.89892578125
                },
                {
                    "start": 2534.85,
                    "end": 2535.19,
                    "word": " a somewhat",
                    "probability": 0.7138671875
                },
                {
                    "start": 2535.19,
                    "end": 2535.97,
                    "word": " robust",
                    "probability": 0.76171875
                },
                {
                    "start": 2535.97,
                    "end": 2538.09,
                    "word": " deep",
                    "probability": 0.6435546875
                },
                {
                    "start": 2538.09,
                    "end": 2538.55,
                    "word": " RL",
                    "probability": 0.969970703125
                }
            ]
        },
        {
            "id": 376,
            "seek": 255808,
            "start": 2539.51,
            "end": 2545.15,
            "text": " algorithm or approach, it's possible to incur maximum damage or find a policy that incurs",
            "tokens": [
                50412,
                9284,
                420,
                3109,
                11,
                309,
                311,
                1944,
                281,
                35774,
                6674,
                4344,
                420,
                915,
                257,
                3897,
                300,
                834,
                2156,
                50692
            ],
            "temperature": 0.0,
            "avg_logprob": -0.190625,
            "compression_ratio": 1.439153439153439,
            "no_speech_prob": 0.0006165504455566406,
            "words": [
                {
                    "start": 2539.51,
                    "end": 2539.93,
                    "word": " algorithm",
                    "probability": 0.27880859375
                },
                {
                    "start": 2539.93,
                    "end": 2540.33,
                    "word": " or",
                    "probability": 0.783203125
                },
                {
                    "start": 2540.33,
                    "end": 2540.79,
                    "word": " approach",
                    "probability": 0.79248046875
                },
                {
                    "start": 2540.79,
                    "end": 2540.99,
                    "word": ",",
                    "probability": 0.5439453125
                },
                {
                    "start": 2540.99,
                    "end": 2541.43,
                    "word": " it's",
                    "probability": 0.912841796875
                },
                {
                    "start": 2541.43,
                    "end": 2541.85,
                    "word": " possible",
                    "probability": 0.85986328125
                },
                {
                    "start": 2541.85,
                    "end": 2542.35,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 2542.35,
                    "end": 2542.59,
                    "word": " incur",
                    "probability": 0.8916015625
                },
                {
                    "start": 2542.59,
                    "end": 2543.09,
                    "word": " maximum",
                    "probability": 0.83154296875
                },
                {
                    "start": 2543.09,
                    "end": 2543.53,
                    "word": " damage",
                    "probability": 0.77783203125
                },
                {
                    "start": 2543.53,
                    "end": 2543.71,
                    "word": " or",
                    "probability": 0.8271484375
                },
                {
                    "start": 2543.71,
                    "end": 2544.11,
                    "word": " find",
                    "probability": 0.78271484375
                },
                {
                    "start": 2544.49,
                    "end": 2544.75,
                    "word": " a policy",
                    "probability": 0.794921875
                },
                {
                    "start": 2544.75,
                    "end": 2544.91,
                    "word": " that",
                    "probability": 0.88720703125
                },
                {
                    "start": 2544.91,
                    "end": 2545.15,
                    "word": " incurs",
                    "probability": 0.863525390625
                }
            ]
        },
        {
            "id": 377,
            "seek": 255808,
            "start": 2545.15,
            "end": 2553.31,
            "text": " maximum regret on the target within 60,000 iterations, which is a relatively low amount.",
            "tokens": [
                50692,
                6674,
                10879,
                322,
                264,
                3779,
                1951,
                4060,
                11,
                1360,
                36540,
                11,
                597,
                307,
                257,
                7226,
                2295,
                2372,
                13,
                51060
            ],
            "temperature": 0.0,
            "avg_logprob": -0.190625,
            "compression_ratio": 1.439153439153439,
            "no_speech_prob": 0.0006165504455566406,
            "words": [
                {
                    "start": 2545.15,
                    "end": 2545.59,
                    "word": " maximum",
                    "probability": 0.76025390625
                },
                {
                    "start": 2545.59,
                    "end": 2546.07,
                    "word": " regret",
                    "probability": 0.814453125
                },
                {
                    "start": 2546.07,
                    "end": 2546.47,
                    "word": " on",
                    "probability": 0.89697265625
                },
                {
                    "start": 2546.47,
                    "end": 2547.17,
                    "word": " the",
                    "probability": 0.78662109375
                },
                {
                    "start": 2547.17,
                    "end": 2547.53,
                    "word": " target",
                    "probability": 0.50439453125
                },
                {
                    "start": 2547.53,
                    "end": 2548.01,
                    "word": " within",
                    "probability": 0.73974609375
                },
                {
                    "start": 2548.01,
                    "end": 2548.41,
                    "word": " 60",
                    "probability": 0.78076171875
                },
                {
                    "start": 2548.41,
                    "end": 2548.67,
                    "word": ",000",
                    "probability": 0.8076171875
                },
                {
                    "start": 2548.67,
                    "end": 2549.09,
                    "word": " iterations",
                    "probability": 0.712890625
                },
                {
                    "start": 2549.09,
                    "end": 2549.29,
                    "word": ",",
                    "probability": 0.7744140625
                },
                {
                    "start": 2549.29,
                    "end": 2549.49,
                    "word": " which",
                    "probability": 0.91259765625
                },
                {
                    "start": 2549.49,
                    "end": 2549.65,
                    "word": " is",
                    "probability": 0.9052734375
                },
                {
                    "start": 2549.81,
                    "end": 2550.07,
                    "word": " a relatively",
                    "probability": 0.76806640625
                },
                {
                    "start": 2550.07,
                    "end": 2550.63,
                    "word": " low",
                    "probability": 0.904296875
                },
                {
                    "start": 2550.63,
                    "end": 2552.51,
                    "word": " amount",
                    "probability": 0.8232421875
                },
                {
                    "start": 2552.51,
                    "end": 2553.31,
                    "word": ".",
                    "probability": 0.96728515625
                }
            ]
        },
        {
            "id": 378,
            "seek": 255808,
            "start": 2553.31,
            "end": 2564.21,
            "text": " Now, what about defenses? So, of course, similar to adversarial examples, one approach or one",
            "tokens": [
                51096,
                823,
                11,
                437,
                466,
                35989,
                30,
                407,
                11,
                295,
                1164,
                11,
                2531,
                281,
                17641,
                44745,
                5110,
                11,
                472,
                3109,
                420,
                472,
                51656
            ],
            "temperature": 0.0,
            "avg_logprob": -0.190625,
            "compression_ratio": 1.439153439153439,
            "no_speech_prob": 0.0006165504455566406,
            "words": [
                {
                    "start": 2553.31,
                    "end": 2553.51,
                    "word": " Now",
                    "probability": 0.865234375
                },
                {
                    "start": 2553.51,
                    "end": 2553.55,
                    "word": ",",
                    "probability": 0.6044921875
                },
                {
                    "start": 2553.55,
                    "end": 2553.65,
                    "word": " what",
                    "probability": 0.88232421875
                },
                {
                    "start": 2553.65,
                    "end": 2553.79,
                    "word": " about",
                    "probability": 0.8203125
                },
                {
                    "start": 2553.79,
                    "end": 2554.19,
                    "word": " defenses",
                    "probability": 0.68115234375
                },
                {
                    "start": 2554.19,
                    "end": 2554.83,
                    "word": "?",
                    "probability": 0.96826171875
                },
                {
                    "start": 2554.83,
                    "end": 2554.95,
                    "word": " So",
                    "probability": 0.6103515625
                },
                {
                    "start": 2554.95,
                    "end": 2555.93,
                    "word": ",",
                    "probability": 0.486083984375
                },
                {
                    "start": 2555.93,
                    "end": 2556.55,
                    "word": " of",
                    "probability": 0.845703125
                },
                {
                    "start": 2556.55,
                    "end": 2559.27,
                    "word": " course",
                    "probability": 0.86181640625
                },
                {
                    "start": 2559.27,
                    "end": 2559.65,
                    "word": ",",
                    "probability": 0.91259765625
                },
                {
                    "start": 2559.65,
                    "end": 2560.07,
                    "word": " similar",
                    "probability": 0.919921875
                },
                {
                    "start": 2560.07,
                    "end": 2561.05,
                    "word": " to",
                    "probability": 0.91162109375
                },
                {
                    "start": 2561.05,
                    "end": 2561.59,
                    "word": " adversarial",
                    "probability": 0.91845703125
                },
                {
                    "start": 2561.59,
                    "end": 2562.11,
                    "word": " examples",
                    "probability": 0.79833984375
                },
                {
                    "start": 2562.11,
                    "end": 2562.37,
                    "word": ",",
                    "probability": 0.912109375
                },
                {
                    "start": 2562.37,
                    "end": 2562.55,
                    "word": " one",
                    "probability": 0.8125
                },
                {
                    "start": 2562.55,
                    "end": 2563.19,
                    "word": " approach",
                    "probability": 0.8115234375
                },
                {
                    "start": 2563.19,
                    "end": 2563.67,
                    "word": " or",
                    "probability": 0.857421875
                },
                {
                    "start": 2563.67,
                    "end": 2564.21,
                    "word": " one",
                    "probability": 0.8193359375
                }
            ]
        },
        {
            "id": 379,
            "seek": 258374,
            "start": 2564.21,
            "end": 2571.55,
            "text": " technique for reducing the impact of adversarial example attacks is through regularization. And",
            "tokens": [
                50372,
                6532,
                337,
                12245,
                264,
                2712,
                295,
                17641,
                44745,
                1365,
                8122,
                307,
                807,
                3890,
                2144,
                13,
                400,
                50740
            ],
            "temperature": 0.0,
            "avg_logprob": -0.18780438002053793,
            "compression_ratio": 1.8048780487804879,
            "no_speech_prob": 0.004070281982421875,
            "words": [
                {
                    "start": 2564.21,
                    "end": 2564.77,
                    "word": " technique",
                    "probability": 0.26708984375
                },
                {
                    "start": 2564.77,
                    "end": 2565.53,
                    "word": " for",
                    "probability": 0.87158203125
                },
                {
                    "start": 2565.53,
                    "end": 2566.67,
                    "word": " reducing",
                    "probability": 0.79345703125
                },
                {
                    "start": 2566.67,
                    "end": 2566.91,
                    "word": " the",
                    "probability": 0.79638671875
                },
                {
                    "start": 2566.91,
                    "end": 2567.33,
                    "word": " impact",
                    "probability": 0.6064453125
                },
                {
                    "start": 2567.33,
                    "end": 2567.61,
                    "word": " of",
                    "probability": 0.89111328125
                },
                {
                    "start": 2567.61,
                    "end": 2568.03,
                    "word": " adversarial",
                    "probability": 0.90771484375
                },
                {
                    "start": 2568.03,
                    "end": 2568.43,
                    "word": " example",
                    "probability": 0.89599609375
                },
                {
                    "start": 2568.43,
                    "end": 2568.83,
                    "word": " attacks",
                    "probability": 0.7451171875
                },
                {
                    "start": 2568.83,
                    "end": 2569.23,
                    "word": " is",
                    "probability": 0.81494140625
                },
                {
                    "start": 2569.23,
                    "end": 2569.61,
                    "word": " through",
                    "probability": 0.80078125
                },
                {
                    "start": 2569.61,
                    "end": 2570.51,
                    "word": " regularization",
                    "probability": 0.919677734375
                },
                {
                    "start": 2570.51,
                    "end": 2571.35,
                    "word": ".",
                    "probability": 0.65673828125
                },
                {
                    "start": 2571.39,
                    "end": 2571.55,
                    "word": " And",
                    "probability": 0.73095703125
                }
            ]
        },
        {
            "id": 380,
            "seek": 258374,
            "start": 2571.55,
            "end": 2577.13,
            "text": " one common type of regularization in supervised learning for mitigating adversarial example",
            "tokens": [
                50740,
                472,
                2689,
                2010,
                295,
                3890,
                2144,
                294,
                46533,
                2539,
                337,
                15699,
                990,
                17641,
                44745,
                1365,
                51020
            ],
            "temperature": 0.0,
            "avg_logprob": -0.18780438002053793,
            "compression_ratio": 1.8048780487804879,
            "no_speech_prob": 0.004070281982421875,
            "words": [
                {
                    "start": 2571.55,
                    "end": 2572.15,
                    "word": " one",
                    "probability": 0.80078125
                },
                {
                    "start": 2572.15,
                    "end": 2572.65,
                    "word": " common",
                    "probability": 0.7109375
                },
                {
                    "start": 2572.65,
                    "end": 2572.81,
                    "word": " type",
                    "probability": 0.8857421875
                },
                {
                    "start": 2572.81,
                    "end": 2572.91,
                    "word": " of",
                    "probability": 0.9033203125
                },
                {
                    "start": 2572.91,
                    "end": 2573.49,
                    "word": " regularization",
                    "probability": 0.9677734375
                },
                {
                    "start": 2573.49,
                    "end": 2573.81,
                    "word": " in",
                    "probability": 0.82568359375
                },
                {
                    "start": 2573.81,
                    "end": 2574.15,
                    "word": " supervised",
                    "probability": 0.779296875
                },
                {
                    "start": 2574.15,
                    "end": 2574.65,
                    "word": " learning",
                    "probability": 0.84375
                },
                {
                    "start": 2574.65,
                    "end": 2575.65,
                    "word": " for",
                    "probability": 0.78466796875
                },
                {
                    "start": 2575.65,
                    "end": 2576.15,
                    "word": " mitigating",
                    "probability": 0.946533203125
                },
                {
                    "start": 2576.15,
                    "end": 2576.57,
                    "word": " adversarial",
                    "probability": 0.93994140625
                },
                {
                    "start": 2576.57,
                    "end": 2577.13,
                    "word": " example",
                    "probability": 0.82958984375
                }
            ]
        },
        {
            "id": 381,
            "seek": 258374,
            "start": 2577.13,
            "end": 2583.41,
            "text": " attacks is adversarial training, training the model on adversarially perturbed samples to",
            "tokens": [
                51020,
                8122,
                307,
                17641,
                44745,
                3097,
                11,
                3097,
                264,
                2316,
                322,
                17641,
                289,
                2270,
                13269,
                374,
                2883,
                10938,
                281,
                51316
            ],
            "temperature": 0.0,
            "avg_logprob": -0.18780438002053793,
            "compression_ratio": 1.8048780487804879,
            "no_speech_prob": 0.004070281982421875,
            "words": [
                {
                    "start": 2577.13,
                    "end": 2577.83,
                    "word": " attacks",
                    "probability": 0.72998046875
                },
                {
                    "start": 2577.83,
                    "end": 2578.37,
                    "word": " is",
                    "probability": 0.8779296875
                },
                {
                    "start": 2578.37,
                    "end": 2578.87,
                    "word": " adversarial",
                    "probability": 0.92578125
                },
                {
                    "start": 2578.87,
                    "end": 2579.31,
                    "word": " training",
                    "probability": 0.8037109375
                },
                {
                    "start": 2579.31,
                    "end": 2579.45,
                    "word": ",",
                    "probability": 0.4375
                },
                {
                    "start": 2579.45,
                    "end": 2579.71,
                    "word": " training",
                    "probability": 0.81396484375
                },
                {
                    "start": 2579.71,
                    "end": 2579.89,
                    "word": " the",
                    "probability": 0.80126953125
                },
                {
                    "start": 2579.89,
                    "end": 2580.19,
                    "word": " model",
                    "probability": 0.82177734375
                },
                {
                    "start": 2580.19,
                    "end": 2580.67,
                    "word": " on",
                    "probability": 0.8642578125
                },
                {
                    "start": 2580.67,
                    "end": 2581.31,
                    "word": " adversarially",
                    "probability": 0.8271484375
                },
                {
                    "start": 2581.31,
                    "end": 2581.99,
                    "word": " perturbed",
                    "probability": 0.9070638020833334
                },
                {
                    "start": 2581.99,
                    "end": 2582.95,
                    "word": " samples",
                    "probability": 0.76416015625
                },
                {
                    "start": 2582.95,
                    "end": 2583.41,
                    "word": " to",
                    "probability": 0.7744140625
                }
            ]
        },
        {
            "id": 382,
            "seek": 258374,
            "start": 2583.41,
            "end": 2589.87,
            "text": " make sure that it sees different perturbations of the same image and knows that all of those",
            "tokens": [
                51316,
                652,
                988,
                300,
                309,
                8194,
                819,
                40468,
                763,
                295,
                264,
                912,
                3256,
                293,
                3255,
                300,
                439,
                295,
                729,
                51644
            ],
            "temperature": 0.0,
            "avg_logprob": -0.18780438002053793,
            "compression_ratio": 1.8048780487804879,
            "no_speech_prob": 0.004070281982421875,
            "words": [
                {
                    "start": 2583.41,
                    "end": 2583.57,
                    "word": " make",
                    "probability": 0.88037109375
                },
                {
                    "start": 2583.57,
                    "end": 2583.75,
                    "word": " sure",
                    "probability": 0.8173828125
                },
                {
                    "start": 2583.75,
                    "end": 2584.11,
                    "word": " that",
                    "probability": 0.873046875
                },
                {
                    "start": 2584.11,
                    "end": 2584.57,
                    "word": " it",
                    "probability": 0.9052734375
                },
                {
                    "start": 2584.57,
                    "end": 2585.03,
                    "word": " sees",
                    "probability": 0.88232421875
                },
                {
                    "start": 2585.03,
                    "end": 2585.59,
                    "word": " different",
                    "probability": 0.68310546875
                },
                {
                    "start": 2585.59,
                    "end": 2587.33,
                    "word": " perturbations",
                    "probability": 0.89013671875
                },
                {
                    "start": 2587.33,
                    "end": 2587.65,
                    "word": " of",
                    "probability": 0.90576171875
                },
                {
                    "start": 2587.65,
                    "end": 2587.79,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 2587.79,
                    "end": 2587.99,
                    "word": " same",
                    "probability": 0.79248046875
                },
                {
                    "start": 2587.99,
                    "end": 2588.35,
                    "word": " image",
                    "probability": 0.853515625
                },
                {
                    "start": 2588.35,
                    "end": 2588.55,
                    "word": " and",
                    "probability": 0.8427734375
                },
                {
                    "start": 2588.55,
                    "end": 2589.13,
                    "word": " knows",
                    "probability": 0.720703125
                },
                {
                    "start": 2589.13,
                    "end": 2589.35,
                    "word": " that",
                    "probability": 0.8798828125
                },
                {
                    "start": 2589.35,
                    "end": 2589.53,
                    "word": " all",
                    "probability": 0.89404296875
                },
                {
                    "start": 2589.53,
                    "end": 2589.63,
                    "word": " of",
                    "probability": 0.86865234375
                },
                {
                    "start": 2589.63,
                    "end": 2589.87,
                    "word": " those",
                    "probability": 0.85107421875
                }
            ]
        },
        {
            "id": 383,
            "seek": 261036,
            "start": 2589.87,
            "end": 2596.11,
            "text": " results in the same label, in the correct label. This is called essentially data augmentation in",
            "tokens": [
                50364,
                3542,
                294,
                264,
                912,
                7645,
                11,
                294,
                264,
                3006,
                7645,
                13,
                639,
                307,
                1219,
                4476,
                1412,
                14501,
                19631,
                294,
                50676
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20628511637784122,
            "compression_ratio": 1.5403225806451613,
            "no_speech_prob": 0.00284576416015625,
            "words": [
                {
                    "start": 2589.87,
                    "end": 2590.25,
                    "word": " results",
                    "probability": 0.1549072265625
                },
                {
                    "start": 2590.25,
                    "end": 2590.65,
                    "word": " in",
                    "probability": 0.853515625
                },
                {
                    "start": 2590.65,
                    "end": 2590.77,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 2590.77,
                    "end": 2591.09,
                    "word": " same",
                    "probability": 0.7177734375
                },
                {
                    "start": 2591.09,
                    "end": 2591.59,
                    "word": " label",
                    "probability": 0.775390625
                },
                {
                    "start": 2591.59,
                    "end": 2591.75,
                    "word": ",",
                    "probability": 0.55712890625
                },
                {
                    "start": 2591.75,
                    "end": 2591.87,
                    "word": " in",
                    "probability": 0.8330078125
                },
                {
                    "start": 2591.87,
                    "end": 2592.07,
                    "word": " the",
                    "probability": 0.80810546875
                },
                {
                    "start": 2592.07,
                    "end": 2592.45,
                    "word": " correct",
                    "probability": 0.875
                },
                {
                    "start": 2592.45,
                    "end": 2592.77,
                    "word": " label",
                    "probability": 0.8076171875
                },
                {
                    "start": 2592.77,
                    "end": 2593.23,
                    "word": ".",
                    "probability": 0.92626953125
                },
                {
                    "start": 2593.23,
                    "end": 2593.45,
                    "word": " This",
                    "probability": 0.81298828125
                },
                {
                    "start": 2593.45,
                    "end": 2593.57,
                    "word": " is",
                    "probability": 0.90576171875
                },
                {
                    "start": 2593.57,
                    "end": 2593.89,
                    "word": " called",
                    "probability": 0.78466796875
                },
                {
                    "start": 2593.89,
                    "end": 2594.69,
                    "word": " essentially",
                    "probability": 0.56005859375
                },
                {
                    "start": 2594.69,
                    "end": 2594.99,
                    "word": " data",
                    "probability": 0.64892578125
                },
                {
                    "start": 2594.99,
                    "end": 2595.51,
                    "word": " augmentation",
                    "probability": 0.88427734375
                },
                {
                    "start": 2595.51,
                    "end": 2596.11,
                    "word": " in",
                    "probability": 0.82373046875
                }
            ]
        },
        {
            "id": 384,
            "seek": 261036,
            "start": 2596.11,
            "end": 2602.99,
            "text": " general as a regularization technique. So in 2017, when I was just starting to look into this",
            "tokens": [
                50676,
                2674,
                382,
                257,
                3890,
                2144,
                6532,
                13,
                407,
                294,
                6591,
                11,
                562,
                286,
                390,
                445,
                2891,
                281,
                574,
                666,
                341,
                51012
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20628511637784122,
            "compression_ratio": 1.5403225806451613,
            "no_speech_prob": 0.00284576416015625,
            "words": [
                {
                    "start": 2596.11,
                    "end": 2596.55,
                    "word": " general",
                    "probability": 0.80322265625
                },
                {
                    "start": 2596.55,
                    "end": 2596.93,
                    "word": " as",
                    "probability": 0.654296875
                },
                {
                    "start": 2597.13,
                    "end": 2597.67,
                    "word": " a regularization",
                    "probability": 0.84375
                },
                {
                    "start": 2597.67,
                    "end": 2598.15,
                    "word": " technique",
                    "probability": 0.8974609375
                },
                {
                    "start": 2598.15,
                    "end": 2598.79,
                    "word": ".",
                    "probability": 0.9384765625
                },
                {
                    "start": 2598.85,
                    "end": 2598.99,
                    "word": " So",
                    "probability": 0.82177734375
                },
                {
                    "start": 2598.99,
                    "end": 2599.31,
                    "word": " in",
                    "probability": 0.74658203125
                },
                {
                    "start": 2599.31,
                    "end": 2600.15,
                    "word": " 2017",
                    "probability": 0.75439453125
                },
                {
                    "start": 2600.15,
                    "end": 2600.15,
                    "word": ",",
                    "probability": 0.5966796875
                },
                {
                    "start": 2600.15,
                    "end": 2600.69,
                    "word": " when",
                    "probability": 0.87890625
                },
                {
                    "start": 2600.69,
                    "end": 2600.87,
                    "word": " I",
                    "probability": 0.9951171875
                },
                {
                    "start": 2600.87,
                    "end": 2601.13,
                    "word": " was",
                    "probability": 0.892578125
                },
                {
                    "start": 2601.13,
                    "end": 2601.71,
                    "word": " just",
                    "probability": 0.88037109375
                },
                {
                    "start": 2601.71,
                    "end": 2601.99,
                    "word": " starting",
                    "probability": 0.83154296875
                },
                {
                    "start": 2601.99,
                    "end": 2602.15,
                    "word": " to",
                    "probability": 0.90673828125
                },
                {
                    "start": 2602.15,
                    "end": 2602.31,
                    "word": " look",
                    "probability": 0.9189453125
                },
                {
                    "start": 2602.31,
                    "end": 2602.51,
                    "word": " into",
                    "probability": 0.7353515625
                },
                {
                    "start": 2602.51,
                    "end": 2602.99,
                    "word": " this",
                    "probability": 0.88623046875
                }
            ]
        },
        {
            "id": 385,
            "seek": 261036,
            "start": 2603.79,
            "end": 2610.45,
            "text": " problem or this domain, I had gone through the adversarial training literature and thought that",
            "tokens": [
                51060,
                1154,
                420,
                341,
                9274,
                11,
                286,
                632,
                2780,
                807,
                264,
                17641,
                44745,
                3097,
                10394,
                293,
                1194,
                300,
                51392
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20628511637784122,
            "compression_ratio": 1.5403225806451613,
            "no_speech_prob": 0.00284576416015625,
            "words": [
                {
                    "start": 2603.79,
                    "end": 2604.21,
                    "word": " problem",
                    "probability": 0.73291015625
                },
                {
                    "start": 2604.21,
                    "end": 2604.69,
                    "word": " or",
                    "probability": 0.55908203125
                },
                {
                    "start": 2604.69,
                    "end": 2605.01,
                    "word": " this",
                    "probability": 0.87841796875
                },
                {
                    "start": 2605.01,
                    "end": 2605.81,
                    "word": " domain",
                    "probability": 0.85205078125
                },
                {
                    "start": 2605.81,
                    "end": 2606.67,
                    "word": ",",
                    "probability": 0.92724609375
                },
                {
                    "start": 2606.67,
                    "end": 2607.07,
                    "word": " I",
                    "probability": 0.99755859375
                },
                {
                    "start": 2607.07,
                    "end": 2607.93,
                    "word": " had",
                    "probability": 0.81591796875
                },
                {
                    "start": 2607.93,
                    "end": 2608.35,
                    "word": " gone",
                    "probability": 0.87060546875
                },
                {
                    "start": 2608.35,
                    "end": 2608.55,
                    "word": " through",
                    "probability": 0.79296875
                },
                {
                    "start": 2608.55,
                    "end": 2608.71,
                    "word": " the",
                    "probability": 0.80712890625
                },
                {
                    "start": 2608.71,
                    "end": 2609.11,
                    "word": " adversarial",
                    "probability": 0.918212890625
                },
                {
                    "start": 2609.11,
                    "end": 2609.47,
                    "word": " training",
                    "probability": 0.796875
                },
                {
                    "start": 2609.47,
                    "end": 2609.83,
                    "word": " literature",
                    "probability": 0.875
                },
                {
                    "start": 2609.83,
                    "end": 2610.11,
                    "word": " and",
                    "probability": 0.8603515625
                },
                {
                    "start": 2610.11,
                    "end": 2610.27,
                    "word": " thought",
                    "probability": 0.63330078125
                },
                {
                    "start": 2610.27,
                    "end": 2610.45,
                    "word": " that",
                    "probability": 0.87646484375
                }
            ]
        },
        {
            "id": 386,
            "seek": 261036,
            "start": 2610.45,
            "end": 2616.49,
            "text": " the same may also hold true for deep RL. I made a hypothesis, actually two. One is with regards",
            "tokens": [
                51392,
                264,
                912,
                815,
                611,
                1797,
                2074,
                337,
                2452,
                497,
                43,
                13,
                286,
                1027,
                257,
                17291,
                11,
                767,
                732,
                13,
                1485,
                307,
                365,
                14258,
                51692
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20628511637784122,
            "compression_ratio": 1.5403225806451613,
            "no_speech_prob": 0.00284576416015625,
            "words": [
                {
                    "start": 2610.45,
                    "end": 2610.63,
                    "word": " the",
                    "probability": 0.81640625
                },
                {
                    "start": 2610.63,
                    "end": 2610.81,
                    "word": " same",
                    "probability": 0.81005859375
                },
                {
                    "start": 2610.81,
                    "end": 2610.99,
                    "word": " may",
                    "probability": 0.89794921875
                },
                {
                    "start": 2610.99,
                    "end": 2611.25,
                    "word": " also",
                    "probability": 0.82275390625
                },
                {
                    "start": 2611.25,
                    "end": 2611.45,
                    "word": " hold",
                    "probability": 0.822265625
                },
                {
                    "start": 2611.45,
                    "end": 2611.71,
                    "word": " true",
                    "probability": 0.9091796875
                },
                {
                    "start": 2611.71,
                    "end": 2612.01,
                    "word": " for",
                    "probability": 0.8984375
                },
                {
                    "start": 2612.01,
                    "end": 2612.27,
                    "word": " deep",
                    "probability": 0.11083984375
                },
                {
                    "start": 2612.27,
                    "end": 2612.59,
                    "word": " RL",
                    "probability": 0.946044921875
                },
                {
                    "start": 2612.59,
                    "end": 2612.75,
                    "word": ".",
                    "probability": 0.8740234375
                },
                {
                    "start": 2612.75,
                    "end": 2612.91,
                    "word": " I",
                    "probability": 0.99560546875
                },
                {
                    "start": 2612.91,
                    "end": 2613.33,
                    "word": " made",
                    "probability": 0.8125
                },
                {
                    "start": 2613.43,
                    "end": 2613.91,
                    "word": " a hypothesis",
                    "probability": 0.7490234375
                },
                {
                    "start": 2613.91,
                    "end": 2614.31,
                    "word": ",",
                    "probability": 0.52978515625
                },
                {
                    "start": 2614.31,
                    "end": 2614.55,
                    "word": " actually",
                    "probability": 0.77978515625
                },
                {
                    "start": 2614.55,
                    "end": 2614.87,
                    "word": " two",
                    "probability": 0.73583984375
                },
                {
                    "start": 2614.87,
                    "end": 2615.51,
                    "word": ".",
                    "probability": 0.88818359375
                },
                {
                    "start": 2615.57,
                    "end": 2615.87,
                    "word": " One",
                    "probability": 0.88916015625
                },
                {
                    "start": 2615.87,
                    "end": 2616.13,
                    "word": " is",
                    "probability": 0.89208984375
                },
                {
                    "start": 2616.13,
                    "end": 2616.27,
                    "word": " with",
                    "probability": 0.7822265625
                },
                {
                    "start": 2616.27,
                    "end": 2616.49,
                    "word": " regards",
                    "probability": 0.81396484375
                }
            ]
        },
        {
            "id": 387,
            "seek": 263852,
            "start": 2616.49,
            "end": 2622.19,
            "text": " to recovery. If training time attacks are not contiguous, if not all of the observations are",
            "tokens": [
                50364,
                281,
                8597,
                13,
                759,
                3097,
                565,
                8122,
                366,
                406,
                660,
                30525,
                11,
                498,
                406,
                439,
                295,
                264,
                18163,
                366,
                50664
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2285610494572063,
            "compression_ratio": 1.6488888888888888,
            "no_speech_prob": 0.0018100738525390625,
            "words": [
                {
                    "start": 2616.49,
                    "end": 2616.63,
                    "word": " to",
                    "probability": 0.06317138671875
                },
                {
                    "start": 2616.63,
                    "end": 2616.95,
                    "word": " recovery",
                    "probability": 0.68603515625
                },
                {
                    "start": 2616.95,
                    "end": 2617.13,
                    "word": ".",
                    "probability": 0.56787109375
                },
                {
                    "start": 2617.13,
                    "end": 2617.21,
                    "word": " If",
                    "probability": 0.8837890625
                },
                {
                    "start": 2617.21,
                    "end": 2617.49,
                    "word": " training",
                    "probability": 0.78759765625
                },
                {
                    "start": 2617.49,
                    "end": 2617.73,
                    "word": " time",
                    "probability": 0.732421875
                },
                {
                    "start": 2617.73,
                    "end": 2617.99,
                    "word": " attacks",
                    "probability": 0.72216796875
                },
                {
                    "start": 2617.99,
                    "end": 2618.15,
                    "word": " are",
                    "probability": 0.89404296875
                },
                {
                    "start": 2618.15,
                    "end": 2618.53,
                    "word": " not",
                    "probability": 0.89404296875
                },
                {
                    "start": 2618.53,
                    "end": 2619.19,
                    "word": " contiguous",
                    "probability": 0.792236328125
                },
                {
                    "start": 2619.19,
                    "end": 2619.41,
                    "word": ",",
                    "probability": 0.85009765625
                },
                {
                    "start": 2619.41,
                    "end": 2619.53,
                    "word": " if",
                    "probability": 0.89990234375
                },
                {
                    "start": 2619.53,
                    "end": 2619.89,
                    "word": " not",
                    "probability": 0.880859375
                },
                {
                    "start": 2619.89,
                    "end": 2620.23,
                    "word": " all",
                    "probability": 0.86572265625
                },
                {
                    "start": 2620.23,
                    "end": 2620.37,
                    "word": " of",
                    "probability": 0.8876953125
                },
                {
                    "start": 2620.37,
                    "end": 2620.47,
                    "word": " the",
                    "probability": 0.8154296875
                },
                {
                    "start": 2620.47,
                    "end": 2621.03,
                    "word": " observations",
                    "probability": 0.69384765625
                },
                {
                    "start": 2621.03,
                    "end": 2622.19,
                    "word": " are",
                    "probability": 0.89990234375
                }
            ]
        },
        {
            "id": 388,
            "seek": 263852,
            "start": 2622.19,
            "end": 2629.01,
            "text": " perturbed, then DRL adapts to the environment and adjusts the policy to overcome the attacks.",
            "tokens": [
                50664,
                13269,
                374,
                2883,
                11,
                550,
                413,
                10740,
                23169,
                1373,
                281,
                264,
                2823,
                293,
                4369,
                82,
                264,
                3897,
                281,
                10473,
                264,
                8122,
                13,
                50960
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2285610494572063,
            "compression_ratio": 1.6488888888888888,
            "no_speech_prob": 0.0018100738525390625,
            "words": [
                {
                    "start": 2622.19,
                    "end": 2623.31,
                    "word": " perturbed",
                    "probability": 0.9231770833333334
                },
                {
                    "start": 2623.31,
                    "end": 2624.07,
                    "word": ",",
                    "probability": 0.8623046875
                },
                {
                    "start": 2624.07,
                    "end": 2624.37,
                    "word": " then",
                    "probability": 0.74365234375
                },
                {
                    "start": 2624.37,
                    "end": 2624.79,
                    "word": " DRL",
                    "probability": 0.89990234375
                },
                {
                    "start": 2624.79,
                    "end": 2625.51,
                    "word": " adapts",
                    "probability": 0.861572265625
                },
                {
                    "start": 2625.51,
                    "end": 2625.69,
                    "word": " to",
                    "probability": 0.90478515625
                },
                {
                    "start": 2625.69,
                    "end": 2625.87,
                    "word": " the",
                    "probability": 0.80810546875
                },
                {
                    "start": 2625.87,
                    "end": 2626.19,
                    "word": " environment",
                    "probability": 0.751953125
                },
                {
                    "start": 2626.19,
                    "end": 2626.35,
                    "word": " and",
                    "probability": 0.8720703125
                },
                {
                    "start": 2626.35,
                    "end": 2626.87,
                    "word": " adjusts",
                    "probability": 0.85791015625
                },
                {
                    "start": 2626.87,
                    "end": 2627.03,
                    "word": " the",
                    "probability": 0.80712890625
                },
                {
                    "start": 2627.03,
                    "end": 2627.35,
                    "word": " policy",
                    "probability": 0.806640625
                },
                {
                    "start": 2627.35,
                    "end": 2627.53,
                    "word": " to",
                    "probability": 0.89697265625
                },
                {
                    "start": 2627.53,
                    "end": 2627.83,
                    "word": " overcome",
                    "probability": 0.853515625
                },
                {
                    "start": 2627.83,
                    "end": 2628.03,
                    "word": " the",
                    "probability": 0.81201171875
                },
                {
                    "start": 2628.03,
                    "end": 2628.47,
                    "word": " attacks",
                    "probability": 0.74072265625
                },
                {
                    "start": 2628.47,
                    "end": 2629.01,
                    "word": ".",
                    "probability": 0.9541015625
                }
            ]
        },
        {
            "id": 389,
            "seek": 263852,
            "start": 2629.01,
            "end": 2635.75,
            "text": " This is training time. And with regards to robustness, I made another hypothesis. Such policies,",
            "tokens": [
                50988,
                639,
                307,
                3097,
                565,
                13,
                400,
                365,
                14258,
                281,
                13956,
                1287,
                11,
                286,
                1027,
                1071,
                17291,
                13,
                9653,
                7657,
                11,
                51292
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2285610494572063,
            "compression_ratio": 1.6488888888888888,
            "no_speech_prob": 0.0018100738525390625,
            "words": [
                {
                    "start": 2629.01,
                    "end": 2629.23,
                    "word": " This",
                    "probability": 0.79833984375
                },
                {
                    "start": 2629.23,
                    "end": 2629.41,
                    "word": " is",
                    "probability": 0.90673828125
                },
                {
                    "start": 2629.41,
                    "end": 2630.27,
                    "word": " training",
                    "probability": 0.6767578125
                },
                {
                    "start": 2630.27,
                    "end": 2630.59,
                    "word": " time",
                    "probability": 0.79931640625
                },
                {
                    "start": 2630.59,
                    "end": 2631.23,
                    "word": ".",
                    "probability": 0.77685546875
                },
                {
                    "start": 2631.33,
                    "end": 2631.37,
                    "word": " And",
                    "probability": 0.75341796875
                },
                {
                    "start": 2631.37,
                    "end": 2631.47,
                    "word": " with",
                    "probability": 0.73681640625
                },
                {
                    "start": 2631.47,
                    "end": 2631.69,
                    "word": " regards",
                    "probability": 0.68359375
                },
                {
                    "start": 2631.69,
                    "end": 2631.81,
                    "word": " to",
                    "probability": 0.88330078125
                },
                {
                    "start": 2631.81,
                    "end": 2632.51,
                    "word": " robustness",
                    "probability": 0.886474609375
                },
                {
                    "start": 2632.51,
                    "end": 2632.89,
                    "word": ",",
                    "probability": 0.88916015625
                },
                {
                    "start": 2632.89,
                    "end": 2633.07,
                    "word": " I",
                    "probability": 0.99462890625
                },
                {
                    "start": 2633.07,
                    "end": 2633.23,
                    "word": " made",
                    "probability": 0.81884765625
                },
                {
                    "start": 2633.23,
                    "end": 2633.51,
                    "word": " another",
                    "probability": 0.8271484375
                },
                {
                    "start": 2633.51,
                    "end": 2633.97,
                    "word": " hypothesis",
                    "probability": 0.787109375
                },
                {
                    "start": 2633.97,
                    "end": 2634.37,
                    "word": ".",
                    "probability": 0.471923828125
                },
                {
                    "start": 2634.37,
                    "end": 2634.53,
                    "word": " Such",
                    "probability": 0.8154296875
                },
                {
                    "start": 2634.53,
                    "end": 2635.15,
                    "word": " policies",
                    "probability": 0.80029296875
                },
                {
                    "start": 2635.15,
                    "end": 2635.75,
                    "word": ",",
                    "probability": 0.76806640625
                }
            ]
        },
        {
            "id": 390,
            "seek": 263852,
            "start": 2635.75,
            "end": 2644.65,
            "text": " policies trained under attack are more robust to test time attacks. And this particular",
            "tokens": [
                51324,
                7657,
                8895,
                833,
                2690,
                366,
                544,
                13956,
                281,
                1500,
                565,
                8122,
                13,
                400,
                341,
                1729,
                51780
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2285610494572063,
            "compression_ratio": 1.6488888888888888,
            "no_speech_prob": 0.0018100738525390625,
            "words": [
                {
                    "start": 2635.75,
                    "end": 2636.11,
                    "word": " policies",
                    "probability": 0.75146484375
                },
                {
                    "start": 2636.11,
                    "end": 2636.65,
                    "word": " trained",
                    "probability": 0.7734375
                },
                {
                    "start": 2636.65,
                    "end": 2637.97,
                    "word": " under",
                    "probability": 0.63037109375
                },
                {
                    "start": 2637.97,
                    "end": 2638.55,
                    "word": " attack",
                    "probability": 0.8212890625
                },
                {
                    "start": 2638.55,
                    "end": 2638.89,
                    "word": " are",
                    "probability": 0.49609375
                },
                {
                    "start": 2638.89,
                    "end": 2639.07,
                    "word": " more",
                    "probability": 0.8759765625
                },
                {
                    "start": 2639.07,
                    "end": 2639.43,
                    "word": " robust",
                    "probability": 0.77001953125
                },
                {
                    "start": 2639.43,
                    "end": 2639.59,
                    "word": " to",
                    "probability": 0.88818359375
                },
                {
                    "start": 2639.59,
                    "end": 2639.79,
                    "word": " test",
                    "probability": 0.7431640625
                },
                {
                    "start": 2639.79,
                    "end": 2639.93,
                    "word": " time",
                    "probability": 0.5791015625
                },
                {
                    "start": 2639.93,
                    "end": 2640.41,
                    "word": " attacks",
                    "probability": 0.70751953125
                },
                {
                    "start": 2640.41,
                    "end": 2641.05,
                    "word": ".",
                    "probability": 0.90087890625
                },
                {
                    "start": 2641.13,
                    "end": 2641.43,
                    "word": " And",
                    "probability": 0.8251953125
                },
                {
                    "start": 2641.43,
                    "end": 2643.15,
                    "word": " this",
                    "probability": 0.85498046875
                },
                {
                    "start": 2643.15,
                    "end": 2644.65,
                    "word": " particular",
                    "probability": 0.79296875
                }
            ]
        },
        {
            "id": 391,
            "seek": 266696,
            "start": 2644.65,
            "end": 2648.91,
            "text": " investigation is published in a paper titled whatever does not kill deep reinforcement",
            "tokens": [
                50372,
                9627,
                307,
                6572,
                294,
                257,
                3035,
                19841,
                2035,
                775,
                406,
                1961,
                2452,
                29280,
                50576
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22034801508892665,
            "compression_ratio": 1.5541666666666667,
            "no_speech_prob": 0.0013246536254882812,
            "words": [
                {
                    "start": 2644.65,
                    "end": 2645.47,
                    "word": " investigation",
                    "probability": 0.279052734375
                },
                {
                    "start": 2645.47,
                    "end": 2645.67,
                    "word": " is",
                    "probability": 0.34912109375
                },
                {
                    "start": 2645.67,
                    "end": 2645.93,
                    "word": " published",
                    "probability": 0.7216796875
                },
                {
                    "start": 2645.93,
                    "end": 2646.11,
                    "word": " in",
                    "probability": 0.9052734375
                },
                {
                    "start": 2646.19,
                    "end": 2646.51,
                    "word": " a paper",
                    "probability": 0.81591796875
                },
                {
                    "start": 2646.51,
                    "end": 2646.99,
                    "word": " titled",
                    "probability": 0.7890625
                },
                {
                    "start": 2646.99,
                    "end": 2647.25,
                    "word": " whatever",
                    "probability": 0.10882568359375
                },
                {
                    "start": 2647.25,
                    "end": 2647.49,
                    "word": " does",
                    "probability": 0.7509765625
                },
                {
                    "start": 2647.49,
                    "end": 2647.65,
                    "word": " not",
                    "probability": 0.89697265625
                },
                {
                    "start": 2647.65,
                    "end": 2647.93,
                    "word": " kill",
                    "probability": 0.8837890625
                },
                {
                    "start": 2647.93,
                    "end": 2648.51,
                    "word": " deep",
                    "probability": 0.435302734375
                },
                {
                    "start": 2648.51,
                    "end": 2648.91,
                    "word": " reinforcement",
                    "probability": 0.85693359375
                }
            ]
        },
        {
            "id": 392,
            "seek": 266696,
            "start": 2648.91,
            "end": 2656.37,
            "text": " learning makes it stronger. So similar as before, we are looking at DQN in Atari games, Breakout,",
            "tokens": [
                50576,
                2539,
                1669,
                309,
                7249,
                13,
                407,
                2531,
                382,
                949,
                11,
                321,
                366,
                1237,
                412,
                413,
                48,
                45,
                294,
                41381,
                2813,
                11,
                16925,
                346,
                11,
                50940
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22034801508892665,
            "compression_ratio": 1.5541666666666667,
            "no_speech_prob": 0.0013246536254882812,
            "words": [
                {
                    "start": 2648.91,
                    "end": 2649.25,
                    "word": " learning",
                    "probability": 0.82275390625
                },
                {
                    "start": 2649.25,
                    "end": 2649.51,
                    "word": " makes",
                    "probability": 0.67333984375
                },
                {
                    "start": 2649.51,
                    "end": 2649.59,
                    "word": " it",
                    "probability": 0.8984375
                },
                {
                    "start": 2649.59,
                    "end": 2650.01,
                    "word": " stronger",
                    "probability": 0.63671875
                },
                {
                    "start": 2650.01,
                    "end": 2650.65,
                    "word": ".",
                    "probability": 0.64892578125
                },
                {
                    "start": 2650.85,
                    "end": 2651.43,
                    "word": " So",
                    "probability": 0.8173828125
                },
                {
                    "start": 2651.43,
                    "end": 2652.27,
                    "word": " similar",
                    "probability": 0.765625
                },
                {
                    "start": 2652.27,
                    "end": 2652.43,
                    "word": " as",
                    "probability": 0.85595703125
                },
                {
                    "start": 2652.43,
                    "end": 2652.69,
                    "word": " before",
                    "probability": 0.794921875
                },
                {
                    "start": 2652.69,
                    "end": 2652.73,
                    "word": ",",
                    "probability": 0.335693359375
                },
                {
                    "start": 2652.73,
                    "end": 2652.83,
                    "word": " we",
                    "probability": 0.87646484375
                },
                {
                    "start": 2652.83,
                    "end": 2652.89,
                    "word": " are",
                    "probability": 0.748046875
                },
                {
                    "start": 2652.89,
                    "end": 2653.11,
                    "word": " looking",
                    "probability": 0.81005859375
                },
                {
                    "start": 2653.11,
                    "end": 2653.59,
                    "word": " at",
                    "probability": 0.90869140625
                },
                {
                    "start": 2653.59,
                    "end": 2654.43,
                    "word": " DQN",
                    "probability": 0.8546549479166666
                },
                {
                    "start": 2654.43,
                    "end": 2654.67,
                    "word": " in",
                    "probability": 0.76513671875
                },
                {
                    "start": 2654.67,
                    "end": 2654.89,
                    "word": " Atari",
                    "probability": 0.87158203125
                },
                {
                    "start": 2654.89,
                    "end": 2655.25,
                    "word": " games",
                    "probability": 0.7392578125
                },
                {
                    "start": 2655.25,
                    "end": 2655.45,
                    "word": ",",
                    "probability": 0.71142578125
                },
                {
                    "start": 2655.45,
                    "end": 2656.09,
                    "word": " Breakout",
                    "probability": 0.8017578125
                },
                {
                    "start": 2656.09,
                    "end": 2656.37,
                    "word": ",",
                    "probability": 0.90087890625
                }
            ]
        },
        {
            "id": 393,
            "seek": 266696,
            "start": 2656.37,
            "end": 2666.23,
            "text": " Enduro, and Pong. Now, the way I designed the experiment was based on a probability of attack.",
            "tokens": [
                50940,
                6967,
                7052,
                11,
                293,
                430,
                556,
                13,
                823,
                11,
                264,
                636,
                286,
                4761,
                264,
                5120,
                390,
                2361,
                322,
                257,
                8482,
                295,
                2690,
                13,
                51448
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22034801508892665,
            "compression_ratio": 1.5541666666666667,
            "no_speech_prob": 0.0013246536254882812,
            "words": [
                {
                    "start": 2656.37,
                    "end": 2656.81,
                    "word": " Enduro",
                    "probability": 0.882080078125
                },
                {
                    "start": 2656.81,
                    "end": 2657.45,
                    "word": ",",
                    "probability": 0.493408203125
                },
                {
                    "start": 2657.45,
                    "end": 2657.95,
                    "word": " and",
                    "probability": 0.88623046875
                },
                {
                    "start": 2657.95,
                    "end": 2658.51,
                    "word": " Pong",
                    "probability": 0.934326171875
                },
                {
                    "start": 2658.51,
                    "end": 2659.15,
                    "word": ".",
                    "probability": 0.94921875
                },
                {
                    "start": 2659.61,
                    "end": 2660.25,
                    "word": " Now",
                    "probability": 0.8720703125
                },
                {
                    "start": 2660.25,
                    "end": 2661.11,
                    "word": ",",
                    "probability": 0.54541015625
                },
                {
                    "start": 2661.11,
                    "end": 2662.09,
                    "word": " the",
                    "probability": 0.7763671875
                },
                {
                    "start": 2662.09,
                    "end": 2662.37,
                    "word": " way",
                    "probability": 0.888671875
                },
                {
                    "start": 2662.37,
                    "end": 2662.95,
                    "word": " I",
                    "probability": 0.98681640625
                },
                {
                    "start": 2662.95,
                    "end": 2663.49,
                    "word": " designed",
                    "probability": 0.697265625
                },
                {
                    "start": 2663.49,
                    "end": 2663.61,
                    "word": " the",
                    "probability": 0.4990234375
                },
                {
                    "start": 2663.61,
                    "end": 2663.87,
                    "word": " experiment",
                    "probability": 0.8623046875
                },
                {
                    "start": 2663.87,
                    "end": 2664.15,
                    "word": " was",
                    "probability": 0.888671875
                },
                {
                    "start": 2664.15,
                    "end": 2664.39,
                    "word": " based",
                    "probability": 0.81396484375
                },
                {
                    "start": 2664.39,
                    "end": 2664.55,
                    "word": " on",
                    "probability": 0.91162109375
                },
                {
                    "start": 2664.67,
                    "end": 2665.11,
                    "word": " a probability",
                    "probability": 0.89453125
                },
                {
                    "start": 2665.11,
                    "end": 2665.53,
                    "word": " of",
                    "probability": 0.900390625
                },
                {
                    "start": 2665.53,
                    "end": 2665.99,
                    "word": " attack",
                    "probability": 0.82080078125
                },
                {
                    "start": 2665.99,
                    "end": 2666.23,
                    "word": ".",
                    "probability": 0.91845703125
                }
            ]
        },
        {
            "id": 394,
            "seek": 266696,
            "start": 2666.23,
            "end": 2673.09,
            "text": " So as an attacker, I assign a certain probability for each state, for each observation during",
            "tokens": [
                51448,
                407,
                382,
                364,
                35871,
                11,
                286,
                6269,
                257,
                1629,
                8482,
                337,
                1184,
                1785,
                11,
                337,
                1184,
                14816,
                1830,
                51784
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22034801508892665,
            "compression_ratio": 1.5541666666666667,
            "no_speech_prob": 0.0013246536254882812,
            "words": [
                {
                    "start": 2666.23,
                    "end": 2666.73,
                    "word": " So",
                    "probability": 0.89111328125
                },
                {
                    "start": 2666.73,
                    "end": 2667.35,
                    "word": " as",
                    "probability": 0.71484375
                },
                {
                    "start": 2667.35,
                    "end": 2667.51,
                    "word": " an",
                    "probability": 0.91162109375
                },
                {
                    "start": 2667.51,
                    "end": 2667.87,
                    "word": " attacker",
                    "probability": 0.7880859375
                },
                {
                    "start": 2667.87,
                    "end": 2668.19,
                    "word": ",",
                    "probability": 0.95068359375
                },
                {
                    "start": 2668.19,
                    "end": 2668.39,
                    "word": " I",
                    "probability": 0.9990234375
                },
                {
                    "start": 2668.39,
                    "end": 2668.87,
                    "word": " assign",
                    "probability": 0.400634765625
                },
                {
                    "start": 2669.01,
                    "end": 2669.19,
                    "word": " a certain",
                    "probability": 0.80224609375
                },
                {
                    "start": 2669.19,
                    "end": 2669.87,
                    "word": " probability",
                    "probability": 0.900390625
                },
                {
                    "start": 2669.87,
                    "end": 2671.03,
                    "word": " for",
                    "probability": 0.85888671875
                },
                {
                    "start": 2671.03,
                    "end": 2671.41,
                    "word": " each",
                    "probability": 0.86474609375
                },
                {
                    "start": 2671.41,
                    "end": 2671.93,
                    "word": " state",
                    "probability": 0.8759765625
                },
                {
                    "start": 2671.93,
                    "end": 2672.03,
                    "word": ",",
                    "probability": 0.215087890625
                },
                {
                    "start": 2672.03,
                    "end": 2672.09,
                    "word": " for",
                    "probability": 0.88671875
                },
                {
                    "start": 2672.09,
                    "end": 2672.23,
                    "word": " each",
                    "probability": 0.857421875
                },
                {
                    "start": 2672.23,
                    "end": 2672.81,
                    "word": " observation",
                    "probability": 0.822265625
                },
                {
                    "start": 2672.81,
                    "end": 2673.09,
                    "word": " during",
                    "probability": 0.74609375
                }
            ]
        },
        {
            "id": 395,
            "seek": 268829,
            "start": 2673.09,
            "end": 2679.63,
            "text": " training time to be perturbed. I perform experiments, different experiments with",
            "tokens": [
                50364,
                3097,
                565,
                281,
                312,
                13269,
                374,
                2883,
                13,
                286,
                2042,
                12050,
                11,
                819,
                12050,
                365,
                50688
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20928030890045743,
            "compression_ratio": 1.4914285714285713,
            "no_speech_prob": 0.0009546279907226562,
            "words": [
                {
                    "start": 2673.09,
                    "end": 2673.35,
                    "word": " training",
                    "probability": 0.424560546875
                },
                {
                    "start": 2673.35,
                    "end": 2673.67,
                    "word": " time",
                    "probability": 0.7919921875
                },
                {
                    "start": 2673.67,
                    "end": 2673.81,
                    "word": " to",
                    "probability": 0.81591796875
                },
                {
                    "start": 2673.81,
                    "end": 2673.95,
                    "word": " be",
                    "probability": 0.90771484375
                },
                {
                    "start": 2673.95,
                    "end": 2674.55,
                    "word": " perturbed",
                    "probability": 0.9420572916666666
                },
                {
                    "start": 2674.55,
                    "end": 2675.21,
                    "word": ".",
                    "probability": 0.91796875
                },
                {
                    "start": 2675.61,
                    "end": 2675.61,
                    "word": " I",
                    "probability": 0.880859375
                },
                {
                    "start": 2675.61,
                    "end": 2677.99,
                    "word": " perform",
                    "probability": 0.42138671875
                },
                {
                    "start": 2677.99,
                    "end": 2678.73,
                    "word": " experiments",
                    "probability": 0.454833984375
                },
                {
                    "start": 2678.73,
                    "end": 2678.89,
                    "word": ",",
                    "probability": 0.284912109375
                },
                {
                    "start": 2678.89,
                    "end": 2679.15,
                    "word": " different",
                    "probability": 0.6328125
                },
                {
                    "start": 2679.15,
                    "end": 2679.49,
                    "word": " experiments",
                    "probability": 0.8505859375
                },
                {
                    "start": 2679.49,
                    "end": 2679.63,
                    "word": " with",
                    "probability": 0.405029296875
                }
            ]
        },
        {
            "id": 396,
            "seek": 268829,
            "start": 2679.63,
            "end": 2685.67,
            "text": " different values of this P attack, 20%, 40%, 80%, and 1, which means contiguous attack.",
            "tokens": [
                50688,
                819,
                4190,
                295,
                341,
                430,
                2690,
                11,
                945,
                8923,
                3356,
                8923,
                4688,
                8923,
                293,
                502,
                11,
                597,
                1355,
                660,
                30525,
                2690,
                13,
                50956
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20928030890045743,
            "compression_ratio": 1.4914285714285713,
            "no_speech_prob": 0.0009546279907226562,
            "words": [
                {
                    "start": 2679.63,
                    "end": 2679.83,
                    "word": " different",
                    "probability": 0.6884765625
                },
                {
                    "start": 2679.83,
                    "end": 2680.25,
                    "word": " values",
                    "probability": 0.9111328125
                },
                {
                    "start": 2680.25,
                    "end": 2680.47,
                    "word": " of",
                    "probability": 0.8857421875
                },
                {
                    "start": 2680.47,
                    "end": 2680.63,
                    "word": " this",
                    "probability": 0.8486328125
                },
                {
                    "start": 2680.63,
                    "end": 2680.79,
                    "word": " P",
                    "probability": 0.50341796875
                },
                {
                    "start": 2680.79,
                    "end": 2681.15,
                    "word": " attack",
                    "probability": 0.7744140625
                },
                {
                    "start": 2681.15,
                    "end": 2681.39,
                    "word": ",",
                    "probability": 0.39501953125
                },
                {
                    "start": 2681.39,
                    "end": 2681.95,
                    "word": " 20%,",
                    "probability": 0.78125
                },
                {
                    "start": 2681.95,
                    "end": 2682.47,
                    "word": " 40%,",
                    "probability": 0.894287109375
                },
                {
                    "start": 2682.47,
                    "end": 2683.01,
                    "word": " 80%,",
                    "probability": 0.69189453125
                },
                {
                    "start": 2683.01,
                    "end": 2683.21,
                    "word": " and",
                    "probability": 0.86181640625
                },
                {
                    "start": 2683.21,
                    "end": 2683.43,
                    "word": " 1",
                    "probability": 0.7548828125
                },
                {
                    "start": 2683.43,
                    "end": 2683.51,
                    "word": ",",
                    "probability": 0.7021484375
                },
                {
                    "start": 2683.51,
                    "end": 2683.63,
                    "word": " which",
                    "probability": 0.89697265625
                },
                {
                    "start": 2683.63,
                    "end": 2683.89,
                    "word": " means",
                    "probability": 0.876953125
                },
                {
                    "start": 2683.89,
                    "end": 2684.57,
                    "word": " contiguous",
                    "probability": 0.784423828125
                },
                {
                    "start": 2684.57,
                    "end": 2685.01,
                    "word": " attack",
                    "probability": 0.857421875
                },
                {
                    "start": 2685.01,
                    "end": 2685.67,
                    "word": ".",
                    "probability": 0.58935546875
                }
            ]
        },
        {
            "id": 397,
            "seek": 268829,
            "start": 2685.67,
            "end": 2694.42,
            "text": " And it's interesting to see that for values of P less than 50%, the agent actually recovers.",
            "tokens": [
                50984,
                400,
                309,
                311,
                1880,
                281,
                536,
                300,
                337,
                4190,
                295,
                430,
                1570,
                813,
                2625,
                8923,
                264,
                9461,
                767,
                7759,
                840,
                13,
                51416
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20928030890045743,
            "compression_ratio": 1.4914285714285713,
            "no_speech_prob": 0.0009546279907226562,
            "words": [
                {
                    "start": 2685.67,
                    "end": 2685.73,
                    "word": " And",
                    "probability": 0.697265625
                },
                {
                    "start": 2685.73,
                    "end": 2685.93,
                    "word": " it's",
                    "probability": 0.8876953125
                },
                {
                    "start": 2685.93,
                    "end": 2686.35,
                    "word": " interesting",
                    "probability": 0.65234375
                },
                {
                    "start": 2686.35,
                    "end": 2686.57,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 2686.57,
                    "end": 2686.89,
                    "word": " see",
                    "probability": 0.818359375
                },
                {
                    "start": 2686.89,
                    "end": 2687.75,
                    "word": " that",
                    "probability": 0.84814453125
                },
                {
                    "start": 2687.75,
                    "end": 2688.13,
                    "word": " for",
                    "probability": 0.8759765625
                },
                {
                    "start": 2688.13,
                    "end": 2689.49,
                    "word": " values",
                    "probability": 0.919921875
                },
                {
                    "start": 2689.49,
                    "end": 2689.75,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 2689.75,
                    "end": 2690.13,
                    "word": " P",
                    "probability": 0.943359375
                },
                {
                    "start": 2690.13,
                    "end": 2690.73,
                    "word": " less",
                    "probability": 0.77587890625
                },
                {
                    "start": 2690.73,
                    "end": 2690.93,
                    "word": " than",
                    "probability": 0.87158203125
                },
                {
                    "start": 2690.93,
                    "end": 2691.79,
                    "word": " 50%,",
                    "probability": 0.8076171875
                },
                {
                    "start": 2691.79,
                    "end": 2692.83,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 2692.83,
                    "end": 2693.05,
                    "word": " agent",
                    "probability": 0.80908203125
                },
                {
                    "start": 2693.05,
                    "end": 2693.49,
                    "word": " actually",
                    "probability": 0.794921875
                },
                {
                    "start": 2693.49,
                    "end": 2694.09,
                    "word": " recovers",
                    "probability": 0.877685546875
                },
                {
                    "start": 2694.09,
                    "end": 2694.42,
                    "word": ".",
                    "probability": 0.83203125
                }
            ]
        },
        {
            "id": 398,
            "seek": 271591,
            "start": 2695.26,
            "end": 2703.46,
            "text": " But for values greater than 50%, the training process plummets and either does not converge or",
            "tokens": [
                50412,
                583,
                337,
                4190,
                5044,
                813,
                2625,
                8923,
                264,
                3097,
                1399,
                25854,
                76,
                1385,
                293,
                2139,
                775,
                406,
                41881,
                420,
                50804
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24813432657896584,
            "compression_ratio": 1.4739583333333333,
            "no_speech_prob": 0.00048041343688964844,
            "words": [
                {
                    "start": 2695.26,
                    "end": 2696.06,
                    "word": " But",
                    "probability": 0.39208984375
                },
                {
                    "start": 2696.06,
                    "end": 2696.42,
                    "word": " for",
                    "probability": 0.85791015625
                },
                {
                    "start": 2696.42,
                    "end": 2697.38,
                    "word": " values",
                    "probability": 0.9013671875
                },
                {
                    "start": 2697.38,
                    "end": 2697.64,
                    "word": " greater",
                    "probability": 0.81591796875
                },
                {
                    "start": 2697.64,
                    "end": 2697.88,
                    "word": " than",
                    "probability": 0.87841796875
                },
                {
                    "start": 2697.88,
                    "end": 2698.66,
                    "word": " 50%,",
                    "probability": 0.6943359375
                },
                {
                    "start": 2698.66,
                    "end": 2699.74,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 2699.74,
                    "end": 2700.3,
                    "word": " training",
                    "probability": 0.78564453125
                },
                {
                    "start": 2700.3,
                    "end": 2700.94,
                    "word": " process",
                    "probability": 0.869140625
                },
                {
                    "start": 2700.94,
                    "end": 2701.74,
                    "word": " plummets",
                    "probability": 0.9091796875
                },
                {
                    "start": 2701.74,
                    "end": 2702.22,
                    "word": " and",
                    "probability": 0.75
                },
                {
                    "start": 2702.22,
                    "end": 2702.48,
                    "word": " either",
                    "probability": 0.75439453125
                },
                {
                    "start": 2702.48,
                    "end": 2702.7,
                    "word": " does",
                    "probability": 0.89892578125
                },
                {
                    "start": 2702.7,
                    "end": 2702.88,
                    "word": " not",
                    "probability": 0.90576171875
                },
                {
                    "start": 2702.88,
                    "end": 2703.24,
                    "word": " converge",
                    "probability": 0.7001953125
                },
                {
                    "start": 2703.24,
                    "end": 2703.46,
                    "word": " or",
                    "probability": 0.79248046875
                }
            ]
        },
        {
            "id": 399,
            "seek": 271591,
            "start": 2703.46,
            "end": 2716.34,
            "text": " converges to a very, very low mean return value or mean total reward value. I later on publish a",
            "tokens": [
                50804,
                9652,
                2880,
                281,
                257,
                588,
                11,
                588,
                2295,
                914,
                2736,
                2158,
                420,
                914,
                3217,
                7782,
                2158,
                13,
                286,
                1780,
                322,
                11374,
                257,
                51420
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24813432657896584,
            "compression_ratio": 1.4739583333333333,
            "no_speech_prob": 0.00048041343688964844,
            "words": [
                {
                    "start": 2703.46,
                    "end": 2704.2,
                    "word": " converges",
                    "probability": 0.93115234375
                },
                {
                    "start": 2704.2,
                    "end": 2704.7,
                    "word": " to",
                    "probability": 0.89794921875
                },
                {
                    "start": 2706.52,
                    "end": 2707.02,
                    "word": " a very",
                    "probability": 0.73388671875
                },
                {
                    "start": 2707.02,
                    "end": 2707.3,
                    "word": ",",
                    "probability": 0.5517578125
                },
                {
                    "start": 2707.3,
                    "end": 2707.92,
                    "word": " very",
                    "probability": 0.7412109375
                },
                {
                    "start": 2707.92,
                    "end": 2708.36,
                    "word": " low",
                    "probability": 0.904296875
                },
                {
                    "start": 2708.36,
                    "end": 2708.74,
                    "word": " mean",
                    "probability": 0.9208984375
                },
                {
                    "start": 2708.74,
                    "end": 2709.18,
                    "word": " return",
                    "probability": 0.6943359375
                },
                {
                    "start": 2709.18,
                    "end": 2709.86,
                    "word": " value",
                    "probability": 0.8935546875
                },
                {
                    "start": 2709.86,
                    "end": 2710.08,
                    "word": " or",
                    "probability": 0.73681640625
                },
                {
                    "start": 2710.08,
                    "end": 2710.38,
                    "word": " mean",
                    "probability": 0.900390625
                },
                {
                    "start": 2710.38,
                    "end": 2711.02,
                    "word": " total",
                    "probability": 0.748046875
                },
                {
                    "start": 2711.02,
                    "end": 2711.34,
                    "word": " reward",
                    "probability": 0.4423828125
                },
                {
                    "start": 2711.34,
                    "end": 2711.8,
                    "word": " value",
                    "probability": 0.9189453125
                },
                {
                    "start": 2711.8,
                    "end": 2712.06,
                    "word": ".",
                    "probability": 0.9755859375
                },
                {
                    "start": 2712.06,
                    "end": 2712.54,
                    "word": " I",
                    "probability": 0.9365234375
                },
                {
                    "start": 2712.54,
                    "end": 2712.8,
                    "word": " later",
                    "probability": 0.74169921875
                },
                {
                    "start": 2712.8,
                    "end": 2713.16,
                    "word": " on",
                    "probability": 0.87841796875
                },
                {
                    "start": 2713.16,
                    "end": 2715.24,
                    "word": " publish",
                    "probability": 0.59814453125
                },
                {
                    "start": 2715.78,
                    "end": 2716.34,
                    "word": " a theoretical",
                    "probability": 0.7509765625
                }
            ]
        },
        {
            "id": 400,
            "seek": 271591,
            "start": 2716.34,
            "end": 2722.04,
            "text": " theoretical analysis of why this happens. It's available in my PhD dissertation, which I'll",
            "tokens": [
                51420,
                20864,
                5215,
                295,
                983,
                341,
                2314,
                13,
                467,
                311,
                2435,
                294,
                452,
                14476,
                39555,
                11,
                597,
                286,
                603,
                51728
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24813432657896584,
            "compression_ratio": 1.4739583333333333,
            "no_speech_prob": 0.00048041343688964844,
            "words": [
                {
                    "start": 2716.34,
                    "end": 2716.78,
                    "word": " analysis",
                    "probability": 0.7880859375
                },
                {
                    "start": 2716.78,
                    "end": 2717.0,
                    "word": " of",
                    "probability": 0.89306640625
                },
                {
                    "start": 2717.0,
                    "end": 2717.14,
                    "word": " why",
                    "probability": 0.80859375
                },
                {
                    "start": 2717.14,
                    "end": 2717.32,
                    "word": " this",
                    "probability": 0.87353515625
                },
                {
                    "start": 2717.32,
                    "end": 2717.76,
                    "word": " happens",
                    "probability": 0.90966796875
                },
                {
                    "start": 2717.76,
                    "end": 2717.96,
                    "word": ".",
                    "probability": 0.7119140625
                },
                {
                    "start": 2717.96,
                    "end": 2718.28,
                    "word": " It's",
                    "probability": 0.948486328125
                },
                {
                    "start": 2718.28,
                    "end": 2718.6,
                    "word": " available",
                    "probability": 0.75634765625
                },
                {
                    "start": 2718.6,
                    "end": 2719.34,
                    "word": " in",
                    "probability": 0.900390625
                },
                {
                    "start": 2719.34,
                    "end": 2720.54,
                    "word": " my",
                    "probability": 0.8994140625
                },
                {
                    "start": 2720.54,
                    "end": 2720.94,
                    "word": " PhD",
                    "probability": 0.78759765625
                },
                {
                    "start": 2720.94,
                    "end": 2721.34,
                    "word": " dissertation",
                    "probability": 0.87109375
                },
                {
                    "start": 2721.34,
                    "end": 2721.7,
                    "word": ",",
                    "probability": 0.7685546875
                },
                {
                    "start": 2721.7,
                    "end": 2721.88,
                    "word": " which",
                    "probability": 0.9150390625
                },
                {
                    "start": 2721.88,
                    "end": 2722.04,
                    "word": " I'll",
                    "probability": 0.95068359375
                }
            ]
        },
        {
            "id": 401,
            "seek": 274243,
            "start": 2722.04,
            "end": 2728.66,
            "text": " reference in the final slide. Also, it was interesting to see that the robustness is also,",
            "tokens": [
                50364,
                6408,
                294,
                264,
                2572,
                4137,
                13,
                2743,
                11,
                309,
                390,
                1880,
                281,
                536,
                300,
                264,
                13956,
                1287,
                307,
                611,
                11,
                50664
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23666158045937374,
            "compression_ratio": 1.6763285024154588,
            "no_speech_prob": 0.005641937255859375,
            "words": [
                {
                    "start": 2722.04,
                    "end": 2722.32,
                    "word": " reference",
                    "probability": 0.32177734375
                },
                {
                    "start": 2722.32,
                    "end": 2722.46,
                    "word": " in",
                    "probability": 0.5029296875
                },
                {
                    "start": 2722.46,
                    "end": 2722.54,
                    "word": " the",
                    "probability": 0.423828125
                },
                {
                    "start": 2722.54,
                    "end": 2722.94,
                    "word": " final",
                    "probability": 0.7607421875
                },
                {
                    "start": 2722.94,
                    "end": 2723.36,
                    "word": " slide",
                    "probability": 0.865234375
                },
                {
                    "start": 2723.36,
                    "end": 2724.04,
                    "word": ".",
                    "probability": 0.94140625
                },
                {
                    "start": 2724.36,
                    "end": 2724.66,
                    "word": " Also",
                    "probability": 0.9189453125
                },
                {
                    "start": 2724.66,
                    "end": 2724.92,
                    "word": ",",
                    "probability": 0.85595703125
                },
                {
                    "start": 2724.92,
                    "end": 2725.02,
                    "word": " it",
                    "probability": 0.89501953125
                },
                {
                    "start": 2725.02,
                    "end": 2725.18,
                    "word": " was",
                    "probability": 0.8935546875
                },
                {
                    "start": 2725.18,
                    "end": 2725.6,
                    "word": " interesting",
                    "probability": 0.63720703125
                },
                {
                    "start": 2725.6,
                    "end": 2725.8,
                    "word": " to",
                    "probability": 0.904296875
                },
                {
                    "start": 2725.8,
                    "end": 2726.14,
                    "word": " see",
                    "probability": 0.82080078125
                },
                {
                    "start": 2726.14,
                    "end": 2726.52,
                    "word": " that",
                    "probability": 0.86767578125
                },
                {
                    "start": 2726.52,
                    "end": 2726.72,
                    "word": " the",
                    "probability": 0.7802734375
                },
                {
                    "start": 2726.72,
                    "end": 2727.36,
                    "word": " robustness",
                    "probability": 0.854248046875
                },
                {
                    "start": 2727.36,
                    "end": 2727.6,
                    "word": " is",
                    "probability": 0.468017578125
                },
                {
                    "start": 2727.6,
                    "end": 2728.1,
                    "word": " also",
                    "probability": 0.8076171875
                },
                {
                    "start": 2728.1,
                    "end": 2728.66,
                    "word": ",",
                    "probability": 0.248291015625
                }
            ]
        },
        {
            "id": 402,
            "seek": 274243,
            "start": 2728.66,
            "end": 2733.3,
            "text": " the robustness hypothesis is also true. You can see here that after training,",
            "tokens": [
                50692,
                264,
                13956,
                1287,
                17291,
                307,
                611,
                2074,
                13,
                509,
                393,
                536,
                510,
                300,
                934,
                3097,
                11,
                50868
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23666158045937374,
            "compression_ratio": 1.6763285024154588,
            "no_speech_prob": 0.005641937255859375,
            "words": [
                {
                    "start": 2728.66,
                    "end": 2728.86,
                    "word": " the",
                    "probability": 0.76171875
                },
                {
                    "start": 2728.86,
                    "end": 2729.36,
                    "word": " robustness",
                    "probability": 0.824462890625
                },
                {
                    "start": 2729.36,
                    "end": 2729.7,
                    "word": " hypothesis",
                    "probability": 0.7861328125
                },
                {
                    "start": 2729.7,
                    "end": 2729.98,
                    "word": " is",
                    "probability": 0.87841796875
                },
                {
                    "start": 2729.98,
                    "end": 2730.22,
                    "word": " also",
                    "probability": 0.81982421875
                },
                {
                    "start": 2730.22,
                    "end": 2730.48,
                    "word": " true",
                    "probability": 0.91259765625
                },
                {
                    "start": 2730.48,
                    "end": 2730.58,
                    "word": ".",
                    "probability": 0.80419921875
                },
                {
                    "start": 2730.58,
                    "end": 2730.62,
                    "word": " You",
                    "probability": 0.89208984375
                },
                {
                    "start": 2730.62,
                    "end": 2730.78,
                    "word": " can",
                    "probability": 0.896484375
                },
                {
                    "start": 2730.78,
                    "end": 2730.94,
                    "word": " see",
                    "probability": 0.8251953125
                },
                {
                    "start": 2730.94,
                    "end": 2731.16,
                    "word": " here",
                    "probability": 0.8173828125
                },
                {
                    "start": 2731.16,
                    "end": 2731.32,
                    "word": " that",
                    "probability": 0.87109375
                },
                {
                    "start": 2731.32,
                    "end": 2731.6,
                    "word": " after",
                    "probability": 0.7275390625
                },
                {
                    "start": 2731.6,
                    "end": 2732.14,
                    "word": " training",
                    "probability": 0.8134765625
                },
                {
                    "start": 2732.14,
                    "end": 2733.3,
                    "word": ",",
                    "probability": 0.83447265625
                }
            ]
        },
        {
            "id": 403,
            "seek": 274243,
            "start": 2733.3,
            "end": 2740.78,
            "text": " if we attack the test time policy with probability of 1, the plane or vanilla policy,",
            "tokens": [
                50952,
                498,
                321,
                2690,
                264,
                1500,
                565,
                3897,
                365,
                8482,
                295,
                502,
                11,
                264,
                5720,
                420,
                17528,
                3897,
                11,
                51300
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23666158045937374,
            "compression_ratio": 1.6763285024154588,
            "no_speech_prob": 0.005641937255859375,
            "words": [
                {
                    "start": 2733.3,
                    "end": 2734.06,
                    "word": " if",
                    "probability": 0.8173828125
                },
                {
                    "start": 2734.06,
                    "end": 2734.26,
                    "word": " we",
                    "probability": 0.779296875
                },
                {
                    "start": 2734.26,
                    "end": 2734.76,
                    "word": " attack",
                    "probability": 0.712890625
                },
                {
                    "start": 2734.76,
                    "end": 2734.98,
                    "word": " the",
                    "probability": 0.81103515625
                },
                {
                    "start": 2734.98,
                    "end": 2735.2,
                    "word": " test",
                    "probability": 0.767578125
                },
                {
                    "start": 2735.2,
                    "end": 2735.34,
                    "word": " time",
                    "probability": 0.640625
                },
                {
                    "start": 2735.34,
                    "end": 2735.8,
                    "word": " policy",
                    "probability": 0.82373046875
                },
                {
                    "start": 2735.8,
                    "end": 2736.14,
                    "word": " with",
                    "probability": 0.78271484375
                },
                {
                    "start": 2736.14,
                    "end": 2736.66,
                    "word": " probability",
                    "probability": 0.86328125
                },
                {
                    "start": 2736.66,
                    "end": 2737.0,
                    "word": " of",
                    "probability": 0.85986328125
                },
                {
                    "start": 2737.0,
                    "end": 2737.34,
                    "word": " 1",
                    "probability": 0.1983642578125
                },
                {
                    "start": 2737.34,
                    "end": 2737.94,
                    "word": ",",
                    "probability": 0.8828125
                },
                {
                    "start": 2737.94,
                    "end": 2738.1,
                    "word": " the",
                    "probability": 0.80322265625
                },
                {
                    "start": 2738.1,
                    "end": 2738.46,
                    "word": " plane",
                    "probability": 0.4501953125
                },
                {
                    "start": 2738.46,
                    "end": 2739.26,
                    "word": " or",
                    "probability": 0.6005859375
                },
                {
                    "start": 2739.26,
                    "end": 2739.92,
                    "word": " vanilla",
                    "probability": 0.74462890625
                },
                {
                    "start": 2739.92,
                    "end": 2740.58,
                    "word": " policy",
                    "probability": 0.80712890625
                },
                {
                    "start": 2740.58,
                    "end": 2740.78,
                    "word": ",",
                    "probability": 0.87548828125
                }
            ]
        },
        {
            "id": 404,
            "seek": 274243,
            "start": 2741.0,
            "end": 2748.56,
            "text": " a policy that was not trained adversarially, performs very poorly. However, policies trained",
            "tokens": [
                51300,
                257,
                3897,
                300,
                390,
                406,
                8895,
                17641,
                289,
                2270,
                11,
                26213,
                588,
                22271,
                13,
                2908,
                11,
                7657,
                8895,
                51692
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23666158045937374,
            "compression_ratio": 1.6763285024154588,
            "no_speech_prob": 0.005641937255859375,
            "words": [
                {
                    "start": 2741.0,
                    "end": 2741.22,
                    "word": " a policy",
                    "probability": 0.8251953125
                },
                {
                    "start": 2741.22,
                    "end": 2741.36,
                    "word": " that",
                    "probability": 0.86376953125
                },
                {
                    "start": 2741.36,
                    "end": 2741.46,
                    "word": " was",
                    "probability": 0.8623046875
                },
                {
                    "start": 2741.46,
                    "end": 2741.76,
                    "word": " not",
                    "probability": 0.89501953125
                },
                {
                    "start": 2741.76,
                    "end": 2742.38,
                    "word": " trained",
                    "probability": 0.8115234375
                },
                {
                    "start": 2742.38,
                    "end": 2743.16,
                    "word": " adversarially",
                    "probability": 0.84716796875
                },
                {
                    "start": 2743.16,
                    "end": 2744.04,
                    "word": ",",
                    "probability": 0.7177734375
                },
                {
                    "start": 2744.04,
                    "end": 2744.58,
                    "word": " performs",
                    "probability": 0.434326171875
                },
                {
                    "start": 2744.58,
                    "end": 2746.0,
                    "word": " very",
                    "probability": 0.74658203125
                },
                {
                    "start": 2746.0,
                    "end": 2746.4,
                    "word": " poorly",
                    "probability": 0.8974609375
                },
                {
                    "start": 2746.4,
                    "end": 2746.84,
                    "word": ".",
                    "probability": 0.9541015625
                },
                {
                    "start": 2746.84,
                    "end": 2747.2,
                    "word": " However",
                    "probability": 0.8525390625
                },
                {
                    "start": 2747.2,
                    "end": 2747.72,
                    "word": ",",
                    "probability": 0.9755859375
                },
                {
                    "start": 2747.72,
                    "end": 2748.06,
                    "word": " policies",
                    "probability": 0.83935546875
                },
                {
                    "start": 2748.06,
                    "end": 2748.56,
                    "word": " trained",
                    "probability": 0.80712890625
                }
            ]
        },
        {
            "id": 405,
            "seek": 277219,
            "start": 2748.56,
            "end": 2755.7,
            "text": " under adversarial attacks with probability 0.2 or 20% and 0.4 perform really well. For 80%,",
            "tokens": [
                50364,
                833,
                17641,
                44745,
                8122,
                365,
                8482,
                1958,
                13,
                17,
                420,
                945,
                4,
                293,
                1958,
                13,
                19,
                2042,
                534,
                731,
                13,
                1171,
                4688,
                8923,
                50720
            ],
            "temperature": 0.0,
            "avg_logprob": -0.254273511405684,
            "compression_ratio": 1.5777027027027026,
            "no_speech_prob": 0.0022869110107421875,
            "words": [
                {
                    "start": 2748.56,
                    "end": 2748.86,
                    "word": " under",
                    "probability": 0.44384765625
                },
                {
                    "start": 2748.86,
                    "end": 2750.04,
                    "word": " adversarial",
                    "probability": 0.89013671875
                },
                {
                    "start": 2750.04,
                    "end": 2750.4,
                    "word": " attacks",
                    "probability": 0.61376953125
                },
                {
                    "start": 2750.4,
                    "end": 2750.54,
                    "word": " with",
                    "probability": 0.7431640625
                },
                {
                    "start": 2750.54,
                    "end": 2751.06,
                    "word": " probability",
                    "probability": 0.87255859375
                },
                {
                    "start": 2751.06,
                    "end": 2751.72,
                    "word": " 0",
                    "probability": 0.705078125
                },
                {
                    "start": 2751.72,
                    "end": 2751.86,
                    "word": ".2",
                    "probability": 0.93701171875
                },
                {
                    "start": 2751.86,
                    "end": 2751.96,
                    "word": " or",
                    "probability": 0.6962890625
                },
                {
                    "start": 2751.96,
                    "end": 2752.18,
                    "word": " 20",
                    "probability": 0.73779296875
                },
                {
                    "start": 2752.18,
                    "end": 2752.5,
                    "word": "%",
                    "probability": 0.517578125
                },
                {
                    "start": 2752.5,
                    "end": 2752.84,
                    "word": " and",
                    "probability": 0.8447265625
                },
                {
                    "start": 2752.84,
                    "end": 2753.02,
                    "word": " 0",
                    "probability": 0.96630859375
                },
                {
                    "start": 2753.02,
                    "end": 2753.42,
                    "word": ".4",
                    "probability": 0.99755859375
                },
                {
                    "start": 2753.42,
                    "end": 2754.16,
                    "word": " perform",
                    "probability": 0.44921875
                },
                {
                    "start": 2754.16,
                    "end": 2754.52,
                    "word": " really",
                    "probability": 0.7822265625
                },
                {
                    "start": 2754.52,
                    "end": 2754.86,
                    "word": " well",
                    "probability": 0.88232421875
                },
                {
                    "start": 2754.86,
                    "end": 2755.32,
                    "word": ".",
                    "probability": 0.85791015625
                },
                {
                    "start": 2755.32,
                    "end": 2755.4,
                    "word": " For",
                    "probability": 0.8662109375
                },
                {
                    "start": 2755.4,
                    "end": 2755.7,
                    "word": " 80%,",
                    "probability": 0.597412109375
                }
            ]
        },
        {
            "id": 406,
            "seek": 277219,
            "start": 2756.32,
            "end": 2761.36,
            "text": " it's even for policies trained at 80%, as you can see, the policy itself is already performing",
            "tokens": [
                50760,
                309,
                311,
                754,
                337,
                7657,
                8895,
                412,
                4688,
                8923,
                382,
                291,
                393,
                536,
                11,
                264,
                3897,
                2564,
                307,
                1217,
                10205,
                51004
            ],
            "temperature": 0.0,
            "avg_logprob": -0.254273511405684,
            "compression_ratio": 1.5777027027027026,
            "no_speech_prob": 0.0022869110107421875,
            "words": [
                {
                    "start": 2756.32,
                    "end": 2756.9,
                    "word": " it's",
                    "probability": 0.599365234375
                },
                {
                    "start": 2756.9,
                    "end": 2757.24,
                    "word": " even",
                    "probability": 0.6220703125
                },
                {
                    "start": 2757.24,
                    "end": 2757.92,
                    "word": " for",
                    "probability": 0.53076171875
                },
                {
                    "start": 2757.92,
                    "end": 2758.22,
                    "word": " policies",
                    "probability": 0.84326171875
                },
                {
                    "start": 2758.22,
                    "end": 2758.62,
                    "word": " trained",
                    "probability": 0.76806640625
                },
                {
                    "start": 2758.62,
                    "end": 2759.16,
                    "word": " at",
                    "probability": 0.884765625
                },
                {
                    "start": 2759.16,
                    "end": 2759.62,
                    "word": " 80%,",
                    "probability": 0.674560546875
                },
                {
                    "start": 2759.62,
                    "end": 2759.78,
                    "word": " as",
                    "probability": 0.861328125
                },
                {
                    "start": 2759.78,
                    "end": 2759.86,
                    "word": " you",
                    "probability": 0.89599609375
                },
                {
                    "start": 2759.86,
                    "end": 2759.98,
                    "word": " can",
                    "probability": 0.89697265625
                },
                {
                    "start": 2759.98,
                    "end": 2760.1,
                    "word": " see",
                    "probability": 0.8125
                },
                {
                    "start": 2760.1,
                    "end": 2760.16,
                    "word": ",",
                    "probability": 0.54296875
                },
                {
                    "start": 2760.16,
                    "end": 2760.22,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 2760.22,
                    "end": 2760.44,
                    "word": " policy",
                    "probability": 0.810546875
                },
                {
                    "start": 2760.44,
                    "end": 2760.68,
                    "word": " itself",
                    "probability": 0.71142578125
                },
                {
                    "start": 2760.68,
                    "end": 2760.82,
                    "word": " is",
                    "probability": 0.89794921875
                },
                {
                    "start": 2760.82,
                    "end": 2761.02,
                    "word": " already",
                    "probability": 0.84521484375
                },
                {
                    "start": 2761.02,
                    "end": 2761.36,
                    "word": " performing",
                    "probability": 0.69580078125
                }
            ]
        },
        {
            "id": 407,
            "seek": 277219,
            "start": 2761.36,
            "end": 2768.88,
            "text": " very poorly. It's surprising to see that at P equals 1, the performance gets slightly better.",
            "tokens": [
                51004,
                588,
                22271,
                13,
                467,
                311,
                8830,
                281,
                536,
                300,
                412,
                430,
                6915,
                502,
                11,
                264,
                3389,
                2170,
                4748,
                1101,
                13,
                51372
            ],
            "temperature": 0.0,
            "avg_logprob": -0.254273511405684,
            "compression_ratio": 1.5777027027027026,
            "no_speech_prob": 0.0022869110107421875,
            "words": [
                {
                    "start": 2761.36,
                    "end": 2761.62,
                    "word": " very",
                    "probability": 0.751953125
                },
                {
                    "start": 2761.62,
                    "end": 2761.94,
                    "word": " poorly",
                    "probability": 0.8837890625
                },
                {
                    "start": 2761.94,
                    "end": 2762.54,
                    "word": ".",
                    "probability": 0.9296875
                },
                {
                    "start": 2762.7,
                    "end": 2762.98,
                    "word": " It's",
                    "probability": 0.94140625
                },
                {
                    "start": 2762.98,
                    "end": 2763.54,
                    "word": " surprising",
                    "probability": 0.8251953125
                },
                {
                    "start": 2763.54,
                    "end": 2763.78,
                    "word": " to",
                    "probability": 0.90185546875
                },
                {
                    "start": 2763.78,
                    "end": 2763.98,
                    "word": " see",
                    "probability": 0.8251953125
                },
                {
                    "start": 2763.98,
                    "end": 2764.34,
                    "word": " that",
                    "probability": 0.8828125
                },
                {
                    "start": 2764.34,
                    "end": 2764.8,
                    "word": " at",
                    "probability": 0.85791015625
                },
                {
                    "start": 2764.8,
                    "end": 2765.1,
                    "word": " P",
                    "probability": 0.3583984375
                },
                {
                    "start": 2765.1,
                    "end": 2765.54,
                    "word": " equals",
                    "probability": 0.277587890625
                },
                {
                    "start": 2765.54,
                    "end": 2766.1,
                    "word": " 1",
                    "probability": 0.580078125
                },
                {
                    "start": 2766.1,
                    "end": 2766.76,
                    "word": ",",
                    "probability": 0.80078125
                },
                {
                    "start": 2766.76,
                    "end": 2767.1,
                    "word": " the",
                    "probability": 0.8017578125
                },
                {
                    "start": 2767.1,
                    "end": 2767.52,
                    "word": " performance",
                    "probability": 0.84228515625
                },
                {
                    "start": 2767.52,
                    "end": 2767.92,
                    "word": " gets",
                    "probability": 0.814453125
                },
                {
                    "start": 2767.92,
                    "end": 2768.28,
                    "word": " slightly",
                    "probability": 0.79443359375
                },
                {
                    "start": 2768.28,
                    "end": 2768.64,
                    "word": " better",
                    "probability": 0.826171875
                },
                {
                    "start": 2768.64,
                    "end": 2768.88,
                    "word": ".",
                    "probability": 0.603515625
                }
            ]
        },
        {
            "id": 408,
            "seek": 277219,
            "start": 2768.88,
            "end": 2774.96,
            "text": " It's comparable with 40%. To this day, I'm not entirely sure why this happened. I've repeated",
            "tokens": [
                51372,
                467,
                311,
                25323,
                365,
                3356,
                6856,
                1407,
                341,
                786,
                11,
                286,
                478,
                406,
                7696,
                988,
                983,
                341,
                2011,
                13,
                286,
                600,
                10477,
                51684
            ],
            "temperature": 0.0,
            "avg_logprob": -0.254273511405684,
            "compression_ratio": 1.5777027027027026,
            "no_speech_prob": 0.0022869110107421875,
            "words": [
                {
                    "start": 2768.88,
                    "end": 2768.98,
                    "word": " It's",
                    "probability": 0.94384765625
                },
                {
                    "start": 2768.98,
                    "end": 2769.32,
                    "word": " comparable",
                    "probability": 0.751953125
                },
                {
                    "start": 2769.32,
                    "end": 2769.66,
                    "word": " with",
                    "probability": 0.7685546875
                },
                {
                    "start": 2769.66,
                    "end": 2770.6,
                    "word": " 40%.",
                    "probability": 0.908447265625
                },
                {
                    "start": 2770.6,
                    "end": 2771.18,
                    "word": " To",
                    "probability": 0.89404296875
                },
                {
                    "start": 2771.18,
                    "end": 2771.5,
                    "word": " this",
                    "probability": 0.8798828125
                },
                {
                    "start": 2771.5,
                    "end": 2771.76,
                    "word": " day",
                    "probability": 0.890625
                },
                {
                    "start": 2771.76,
                    "end": 2771.78,
                    "word": ",",
                    "probability": 0.92578125
                },
                {
                    "start": 2771.78,
                    "end": 2771.92,
                    "word": " I'm",
                    "probability": 0.9951171875
                },
                {
                    "start": 2771.92,
                    "end": 2772.08,
                    "word": " not",
                    "probability": 0.89794921875
                },
                {
                    "start": 2772.08,
                    "end": 2772.42,
                    "word": " entirely",
                    "probability": 0.75146484375
                },
                {
                    "start": 2772.42,
                    "end": 2772.68,
                    "word": " sure",
                    "probability": 0.8134765625
                },
                {
                    "start": 2772.68,
                    "end": 2773.2,
                    "word": " why",
                    "probability": 0.56396484375
                },
                {
                    "start": 2773.2,
                    "end": 2773.46,
                    "word": " this",
                    "probability": 0.8544921875
                },
                {
                    "start": 2773.46,
                    "end": 2773.8,
                    "word": " happened",
                    "probability": 0.603515625
                },
                {
                    "start": 2773.8,
                    "end": 2774.4,
                    "word": ".",
                    "probability": 0.98681640625
                },
                {
                    "start": 2774.44,
                    "end": 2774.74,
                    "word": " I've",
                    "probability": 0.962890625
                },
                {
                    "start": 2774.74,
                    "end": 2774.96,
                    "word": " repeated",
                    "probability": 0.849609375
                }
            ]
        },
        {
            "id": 409,
            "seek": 277219,
            "start": 2774.96,
            "end": 2778.32,
            "text": " the experiments a number of times and still get the same result. I still don't know why this",
            "tokens": [
                51684,
                264,
                12050,
                257,
                1230,
                295,
                1413,
                293,
                920,
                483,
                264,
                912,
                1874,
                13,
                286,
                920,
                500,
                380,
                458,
                983,
                341,
                51848
            ],
            "temperature": 0.0,
            "avg_logprob": -0.254273511405684,
            "compression_ratio": 1.5777027027027026,
            "no_speech_prob": 0.0022869110107421875,
            "words": [
                {
                    "start": 2774.96,
                    "end": 2775.14,
                    "word": " the",
                    "probability": 0.80419921875
                },
                {
                    "start": 2775.14,
                    "end": 2775.52,
                    "word": " experiments",
                    "probability": 0.492919921875
                },
                {
                    "start": 2775.68,
                    "end": 2775.88,
                    "word": " a number",
                    "probability": 0.875
                },
                {
                    "start": 2775.88,
                    "end": 2776.02,
                    "word": " of",
                    "probability": 0.90380859375
                },
                {
                    "start": 2776.02,
                    "end": 2776.22,
                    "word": " times",
                    "probability": 0.8408203125
                },
                {
                    "start": 2776.22,
                    "end": 2776.32,
                    "word": " and",
                    "probability": 0.468994140625
                },
                {
                    "start": 2776.32,
                    "end": 2776.46,
                    "word": " still",
                    "probability": 0.86376953125
                },
                {
                    "start": 2776.46,
                    "end": 2776.6,
                    "word": " get",
                    "probability": 0.8720703125
                },
                {
                    "start": 2776.6,
                    "end": 2776.72,
                    "word": " the",
                    "probability": 0.80419921875
                },
                {
                    "start": 2776.72,
                    "end": 2776.88,
                    "word": " same",
                    "probability": 0.81884765625
                },
                {
                    "start": 2776.88,
                    "end": 2777.22,
                    "word": " result",
                    "probability": 0.56396484375
                },
                {
                    "start": 2777.22,
                    "end": 2777.32,
                    "word": ".",
                    "probability": 0.95458984375
                },
                {
                    "start": 2777.32,
                    "end": 2777.46,
                    "word": " I",
                    "probability": 0.99853515625
                },
                {
                    "start": 2777.46,
                    "end": 2777.64,
                    "word": " still",
                    "probability": 0.91259765625
                },
                {
                    "start": 2777.64,
                    "end": 2777.9,
                    "word": " don't",
                    "probability": 0.9462890625
                },
                {
                    "start": 2777.9,
                    "end": 2778.02,
                    "word": " know",
                    "probability": 0.82275390625
                },
                {
                    "start": 2778.02,
                    "end": 2778.14,
                    "word": " why",
                    "probability": 0.79931640625
                },
                {
                    "start": 2778.14,
                    "end": 2778.32,
                    "word": " this",
                    "probability": 0.82373046875
                }
            ]
        },
        {
            "id": 410,
            "seek": 280165,
            "start": 2778.32,
            "end": 2784.24,
            "text": " happened. This is one of the interesting problems that we are looking at right now",
            "tokens": [
                50364,
                2011,
                13,
                639,
                307,
                472,
                295,
                264,
                1880,
                2740,
                300,
                321,
                366,
                1237,
                412,
                558,
                586,
                50656
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2578855162469026,
            "compression_ratio": 1.7099236641221374,
            "no_speech_prob": 0.0021495819091796875,
            "words": [
                {
                    "start": 2778.32,
                    "end": 2778.48,
                    "word": " happened",
                    "probability": 0.033355712890625
                },
                {
                    "start": 2778.48,
                    "end": 2778.54,
                    "word": ".",
                    "probability": 0.59326171875
                },
                {
                    "start": 2778.54,
                    "end": 2778.74,
                    "word": " This",
                    "probability": 0.1851806640625
                },
                {
                    "start": 2778.74,
                    "end": 2778.86,
                    "word": " is",
                    "probability": 0.8896484375
                },
                {
                    "start": 2778.86,
                    "end": 2779.02,
                    "word": " one",
                    "probability": 0.81640625
                },
                {
                    "start": 2779.02,
                    "end": 2779.24,
                    "word": " of",
                    "probability": 0.90234375
                },
                {
                    "start": 2779.24,
                    "end": 2779.92,
                    "word": " the",
                    "probability": 0.82373046875
                },
                {
                    "start": 2779.92,
                    "end": 2781.56,
                    "word": " interesting",
                    "probability": 0.6396484375
                },
                {
                    "start": 2781.56,
                    "end": 2782.14,
                    "word": " problems",
                    "probability": 0.7080078125
                },
                {
                    "start": 2782.14,
                    "end": 2782.56,
                    "word": " that",
                    "probability": 0.8603515625
                },
                {
                    "start": 2782.56,
                    "end": 2782.8,
                    "word": " we",
                    "probability": 0.7861328125
                },
                {
                    "start": 2782.8,
                    "end": 2783.0,
                    "word": " are",
                    "probability": 0.79248046875
                },
                {
                    "start": 2783.0,
                    "end": 2783.44,
                    "word": " looking",
                    "probability": 0.80419921875
                },
                {
                    "start": 2783.44,
                    "end": 2783.68,
                    "word": " at",
                    "probability": 0.908203125
                },
                {
                    "start": 2783.68,
                    "end": 2783.9,
                    "word": " right",
                    "probability": 0.87451171875
                },
                {
                    "start": 2783.9,
                    "end": 2784.24,
                    "word": " now",
                    "probability": 0.8974609375
                }
            ]
        },
        {
            "id": 411,
            "seek": 280165,
            "start": 2784.8,
            "end": 2790.16,
            "text": " as a research group. Another defense that we introduce is based on parameter space noise.",
            "tokens": [
                50688,
                382,
                257,
                2132,
                1594,
                13,
                3996,
                7654,
                300,
                321,
                5366,
                307,
                2361,
                322,
                13075,
                1901,
                5658,
                13,
                50956
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2578855162469026,
            "compression_ratio": 1.7099236641221374,
            "no_speech_prob": 0.0021495819091796875,
            "words": [
                {
                    "start": 2784.8,
                    "end": 2785.12,
                    "word": " as",
                    "probability": 0.65771484375
                },
                {
                    "start": 2785.36,
                    "end": 2785.62,
                    "word": " a research",
                    "probability": 0.76513671875
                },
                {
                    "start": 2785.62,
                    "end": 2785.92,
                    "word": " group",
                    "probability": 0.83984375
                },
                {
                    "start": 2785.92,
                    "end": 2786.5,
                    "word": ".",
                    "probability": 0.9736328125
                },
                {
                    "start": 2786.68,
                    "end": 2787.26,
                    "word": " Another",
                    "probability": 0.81591796875
                },
                {
                    "start": 2787.26,
                    "end": 2787.56,
                    "word": " defense",
                    "probability": 0.70263671875
                },
                {
                    "start": 2787.56,
                    "end": 2787.68,
                    "word": " that",
                    "probability": 0.7939453125
                },
                {
                    "start": 2787.68,
                    "end": 2787.76,
                    "word": " we",
                    "probability": 0.859375
                },
                {
                    "start": 2787.76,
                    "end": 2788.1,
                    "word": " introduce",
                    "probability": 0.435791015625
                },
                {
                    "start": 2788.1,
                    "end": 2788.28,
                    "word": " is",
                    "probability": 0.8896484375
                },
                {
                    "start": 2788.28,
                    "end": 2788.48,
                    "word": " based",
                    "probability": 0.79296875
                },
                {
                    "start": 2788.48,
                    "end": 2788.68,
                    "word": " on",
                    "probability": 0.89501953125
                },
                {
                    "start": 2788.68,
                    "end": 2789.1,
                    "word": " parameter",
                    "probability": 0.74560546875
                },
                {
                    "start": 2789.1,
                    "end": 2789.42,
                    "word": " space",
                    "probability": 0.58447265625
                },
                {
                    "start": 2789.42,
                    "end": 2789.76,
                    "word": " noise",
                    "probability": 0.87060546875
                },
                {
                    "start": 2789.76,
                    "end": 2790.16,
                    "word": ".",
                    "probability": 0.89404296875
                }
            ]
        },
        {
            "id": 412,
            "seek": 280165,
            "start": 2790.16,
            "end": 2796.96,
            "text": " This is very much like dropout. The idea of parameter space noise was introduced in 2017,",
            "tokens": [
                50956,
                639,
                307,
                588,
                709,
                411,
                3270,
                346,
                13,
                440,
                1558,
                295,
                13075,
                1901,
                5658,
                390,
                7268,
                294,
                6591,
                11,
                51272
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2578855162469026,
            "compression_ratio": 1.7099236641221374,
            "no_speech_prob": 0.0021495819091796875,
            "words": [
                {
                    "start": 2790.16,
                    "end": 2790.42,
                    "word": " This",
                    "probability": 0.7958984375
                },
                {
                    "start": 2790.42,
                    "end": 2790.52,
                    "word": " is",
                    "probability": 0.908203125
                },
                {
                    "start": 2790.52,
                    "end": 2790.64,
                    "word": " very",
                    "probability": 0.73583984375
                },
                {
                    "start": 2790.64,
                    "end": 2790.78,
                    "word": " much",
                    "probability": 0.81640625
                },
                {
                    "start": 2790.78,
                    "end": 2790.94,
                    "word": " like",
                    "probability": 0.87841796875
                },
                {
                    "start": 2790.94,
                    "end": 2791.44,
                    "word": " dropout",
                    "probability": 0.658203125
                },
                {
                    "start": 2791.44,
                    "end": 2792.02,
                    "word": ".",
                    "probability": 0.8330078125
                },
                {
                    "start": 2792.26,
                    "end": 2792.4,
                    "word": " The",
                    "probability": 0.8134765625
                },
                {
                    "start": 2792.4,
                    "end": 2792.82,
                    "word": " idea",
                    "probability": 0.67138671875
                },
                {
                    "start": 2792.82,
                    "end": 2792.94,
                    "word": " of",
                    "probability": 0.70166015625
                },
                {
                    "start": 2792.94,
                    "end": 2793.16,
                    "word": " parameter",
                    "probability": 0.85693359375
                },
                {
                    "start": 2793.16,
                    "end": 2793.48,
                    "word": " space",
                    "probability": 0.78173828125
                },
                {
                    "start": 2793.48,
                    "end": 2793.76,
                    "word": " noise",
                    "probability": 0.88916015625
                },
                {
                    "start": 2793.76,
                    "end": 2794.18,
                    "word": " was",
                    "probability": 0.9013671875
                },
                {
                    "start": 2794.18,
                    "end": 2794.88,
                    "word": " introduced",
                    "probability": 0.76611328125
                },
                {
                    "start": 2794.88,
                    "end": 2795.5,
                    "word": " in",
                    "probability": 0.904296875
                },
                {
                    "start": 2795.5,
                    "end": 2796.32,
                    "word": " 2017",
                    "probability": 0.73974609375
                },
                {
                    "start": 2796.32,
                    "end": 2796.96,
                    "word": ",",
                    "probability": 0.7041015625
                }
            ]
        },
        {
            "id": 413,
            "seek": 280165,
            "start": 2796.96,
            "end": 2803.08,
            "text": " I believe, independently by Plamper et al and Fortunato et al. And the idea here is, again,",
            "tokens": [
                51308,
                286,
                1697,
                11,
                21761,
                538,
                2149,
                335,
                610,
                1030,
                419,
                293,
                479,
                1584,
                2513,
                1030,
                419,
                13,
                400,
                264,
                1558,
                510,
                307,
                11,
                797,
                11,
                51596
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2578855162469026,
            "compression_ratio": 1.7099236641221374,
            "no_speech_prob": 0.0021495819091796875,
            "words": [
                {
                    "start": 2796.96,
                    "end": 2797.46,
                    "word": " I",
                    "probability": 0.97998046875
                },
                {
                    "start": 2797.46,
                    "end": 2797.64,
                    "word": " believe",
                    "probability": 0.7294921875
                },
                {
                    "start": 2797.64,
                    "end": 2797.7,
                    "word": ",",
                    "probability": 0.379638671875
                },
                {
                    "start": 2797.7,
                    "end": 2798.08,
                    "word": " independently",
                    "probability": 0.83349609375
                },
                {
                    "start": 2798.08,
                    "end": 2798.52,
                    "word": " by",
                    "probability": 0.8740234375
                },
                {
                    "start": 2798.52,
                    "end": 2799.02,
                    "word": " Plamper",
                    "probability": 0.669189453125
                },
                {
                    "start": 2799.02,
                    "end": 2799.34,
                    "word": " et",
                    "probability": 0.78369140625
                },
                {
                    "start": 2799.34,
                    "end": 2799.5,
                    "word": " al",
                    "probability": 0.87060546875
                },
                {
                    "start": 2799.5,
                    "end": 2799.86,
                    "word": " and",
                    "probability": 0.5263671875
                },
                {
                    "start": 2799.86,
                    "end": 2800.34,
                    "word": " Fortunato",
                    "probability": 0.9070638020833334
                },
                {
                    "start": 2800.34,
                    "end": 2800.54,
                    "word": " et",
                    "probability": 0.92431640625
                },
                {
                    "start": 2800.54,
                    "end": 2800.7,
                    "word": " al",
                    "probability": 0.90234375
                },
                {
                    "start": 2800.7,
                    "end": 2801.2,
                    "word": ".",
                    "probability": 0.71044921875
                },
                {
                    "start": 2801.2,
                    "end": 2801.42,
                    "word": " And",
                    "probability": 0.6845703125
                },
                {
                    "start": 2801.42,
                    "end": 2801.56,
                    "word": " the",
                    "probability": 0.8193359375
                },
                {
                    "start": 2801.56,
                    "end": 2801.88,
                    "word": " idea",
                    "probability": 0.77978515625
                },
                {
                    "start": 2801.88,
                    "end": 2802.18,
                    "word": " here",
                    "probability": 0.81103515625
                },
                {
                    "start": 2802.18,
                    "end": 2802.72,
                    "word": " is",
                    "probability": 0.890625
                },
                {
                    "start": 2802.72,
                    "end": 2802.92,
                    "word": ",",
                    "probability": 0.57666015625
                },
                {
                    "start": 2802.92,
                    "end": 2803.0,
                    "word": " again",
                    "probability": 0.90869140625
                },
                {
                    "start": 2803.0,
                    "end": 2803.08,
                    "word": ",",
                    "probability": 0.8525390625
                }
            ]
        },
        {
            "id": 414,
            "seek": 280165,
            "start": 2803.08,
            "end": 2807.78,
            "text": " similar to dropout, to introduce zero mean random noise to the learnable parameters of neural",
            "tokens": [
                51596,
                2531,
                281,
                3270,
                346,
                11,
                281,
                5366,
                4018,
                914,
                4974,
                5658,
                281,
                264,
                1466,
                712,
                9834,
                295,
                18161,
                51840
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2578855162469026,
            "compression_ratio": 1.7099236641221374,
            "no_speech_prob": 0.0021495819091796875,
            "words": [
                {
                    "start": 2803.08,
                    "end": 2803.26,
                    "word": " similar",
                    "probability": 0.9248046875
                },
                {
                    "start": 2803.26,
                    "end": 2803.4,
                    "word": " to",
                    "probability": 0.9052734375
                },
                {
                    "start": 2803.4,
                    "end": 2803.86,
                    "word": " dropout",
                    "probability": 0.794189453125
                },
                {
                    "start": 2803.86,
                    "end": 2804.26,
                    "word": ",",
                    "probability": 0.72265625
                },
                {
                    "start": 2804.26,
                    "end": 2804.4,
                    "word": " to",
                    "probability": 0.888671875
                },
                {
                    "start": 2804.4,
                    "end": 2804.74,
                    "word": " introduce",
                    "probability": 0.6875
                },
                {
                    "start": 2804.74,
                    "end": 2805.12,
                    "word": " zero",
                    "probability": 0.79150390625
                },
                {
                    "start": 2805.12,
                    "end": 2805.34,
                    "word": " mean",
                    "probability": 0.61181640625
                },
                {
                    "start": 2805.34,
                    "end": 2805.66,
                    "word": " random",
                    "probability": 0.732421875
                },
                {
                    "start": 2805.66,
                    "end": 2805.98,
                    "word": " noise",
                    "probability": 0.8720703125
                },
                {
                    "start": 2805.98,
                    "end": 2806.2,
                    "word": " to",
                    "probability": 0.8994140625
                },
                {
                    "start": 2806.2,
                    "end": 2806.36,
                    "word": " the",
                    "probability": 0.787109375
                },
                {
                    "start": 2806.36,
                    "end": 2806.78,
                    "word": " learnable",
                    "probability": 0.898193359375
                },
                {
                    "start": 2806.78,
                    "end": 2807.38,
                    "word": " parameters",
                    "probability": 0.8642578125
                },
                {
                    "start": 2807.38,
                    "end": 2807.6,
                    "word": " of",
                    "probability": 0.89306640625
                },
                {
                    "start": 2807.6,
                    "end": 2807.78,
                    "word": " neural",
                    "probability": 0.90185546875
                }
            ]
        },
        {
            "id": 415,
            "seek": 282415,
            "start": 2807.78,
            "end": 2813.12,
            "text": " network in deep RL to enhance exploration and convergence in deep RL benchmarks.",
            "tokens": [
                50364,
                3209,
                294,
                2452,
                497,
                43,
                281,
                11985,
                16197,
                293,
                32181,
                294,
                2452,
                497,
                43,
                43751,
                13,
                50612
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23271003707510526,
            "compression_ratio": 1.430939226519337,
            "no_speech_prob": 0.00400543212890625,
            "words": [
                {
                    "start": 2807.78,
                    "end": 2808.2,
                    "word": " network",
                    "probability": 0.4765625
                },
                {
                    "start": 2808.2,
                    "end": 2808.42,
                    "word": " in",
                    "probability": 0.8017578125
                },
                {
                    "start": 2808.42,
                    "end": 2808.62,
                    "word": " deep",
                    "probability": 0.23388671875
                },
                {
                    "start": 2808.62,
                    "end": 2809.0,
                    "word": " RL",
                    "probability": 0.933349609375
                },
                {
                    "start": 2809.0,
                    "end": 2809.7,
                    "word": " to",
                    "probability": 0.7109375
                },
                {
                    "start": 2809.7,
                    "end": 2810.14,
                    "word": " enhance",
                    "probability": 0.77880859375
                },
                {
                    "start": 2810.14,
                    "end": 2810.74,
                    "word": " exploration",
                    "probability": 0.751953125
                },
                {
                    "start": 2810.74,
                    "end": 2810.96,
                    "word": " and",
                    "probability": 0.8798828125
                },
                {
                    "start": 2810.96,
                    "end": 2811.42,
                    "word": " convergence",
                    "probability": 0.853515625
                },
                {
                    "start": 2811.42,
                    "end": 2811.72,
                    "word": " in",
                    "probability": 0.8828125
                },
                {
                    "start": 2811.72,
                    "end": 2811.92,
                    "word": " deep",
                    "probability": 0.7705078125
                },
                {
                    "start": 2811.92,
                    "end": 2812.16,
                    "word": " RL",
                    "probability": 0.964599609375
                },
                {
                    "start": 2812.16,
                    "end": 2812.7,
                    "word": " benchmarks",
                    "probability": 0.66845703125
                },
                {
                    "start": 2812.7,
                    "end": 2813.12,
                    "word": ".",
                    "probability": 0.9482421875
                }
            ]
        },
        {
            "id": 416,
            "seek": 282415,
            "start": 2813.48,
            "end": 2822.88,
            "text": " Now, in another paper in 2018, we investigate whether this approach can be used to mitigate",
            "tokens": [
                50644,
                823,
                11,
                294,
                1071,
                3035,
                294,
                6096,
                11,
                321,
                15013,
                1968,
                341,
                3109,
                393,
                312,
                1143,
                281,
                27336,
                51124
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23271003707510526,
            "compression_ratio": 1.430939226519337,
            "no_speech_prob": 0.00400543212890625,
            "words": [
                {
                    "start": 2813.48,
                    "end": 2813.74,
                    "word": " Now",
                    "probability": 0.52001953125
                },
                {
                    "start": 2813.74,
                    "end": 2814.12,
                    "word": ",",
                    "probability": 0.50732421875
                },
                {
                    "start": 2814.12,
                    "end": 2814.8,
                    "word": " in",
                    "probability": 0.88427734375
                },
                {
                    "start": 2814.8,
                    "end": 2815.52,
                    "word": " another",
                    "probability": 0.8232421875
                },
                {
                    "start": 2815.52,
                    "end": 2815.76,
                    "word": " paper",
                    "probability": 0.794921875
                },
                {
                    "start": 2815.76,
                    "end": 2815.9,
                    "word": " in",
                    "probability": 0.876953125
                },
                {
                    "start": 2815.9,
                    "end": 2816.6,
                    "word": " 2018",
                    "probability": 0.7548828125
                },
                {
                    "start": 2816.6,
                    "end": 2817.18,
                    "word": ",",
                    "probability": 0.75341796875
                },
                {
                    "start": 2817.18,
                    "end": 2818.04,
                    "word": " we",
                    "probability": 0.9033203125
                },
                {
                    "start": 2818.04,
                    "end": 2818.9,
                    "word": " investigate",
                    "probability": 0.7314453125
                },
                {
                    "start": 2818.9,
                    "end": 2819.34,
                    "word": " whether",
                    "probability": 0.8369140625
                },
                {
                    "start": 2819.34,
                    "end": 2819.8,
                    "word": " this",
                    "probability": 0.87451171875
                },
                {
                    "start": 2819.8,
                    "end": 2820.16,
                    "word": " approach",
                    "probability": 0.78466796875
                },
                {
                    "start": 2820.16,
                    "end": 2820.36,
                    "word": " can",
                    "probability": 0.88916015625
                },
                {
                    "start": 2820.36,
                    "end": 2820.52,
                    "word": " be",
                    "probability": 0.904296875
                },
                {
                    "start": 2820.52,
                    "end": 2820.92,
                    "word": " used",
                    "probability": 0.8056640625
                },
                {
                    "start": 2820.92,
                    "end": 2821.6,
                    "word": " to",
                    "probability": 0.90869140625
                },
                {
                    "start": 2821.6,
                    "end": 2822.88,
                    "word": " mitigate",
                    "probability": 0.908203125
                }
            ]
        },
        {
            "id": 417,
            "seek": 282415,
            "start": 2822.88,
            "end": 2830.28,
            "text": " the impact or severity of policy manipulation attacks on DQN, and it was shown that it",
            "tokens": [
                51124,
                264,
                2712,
                420,
                35179,
                295,
                3897,
                26475,
                8122,
                322,
                413,
                48,
                45,
                11,
                293,
                309,
                390,
                4898,
                300,
                309,
                51496
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23271003707510526,
            "compression_ratio": 1.430939226519337,
            "no_speech_prob": 0.00400543212890625,
            "words": [
                {
                    "start": 2822.88,
                    "end": 2823.22,
                    "word": " the",
                    "probability": 0.76220703125
                },
                {
                    "start": 2823.22,
                    "end": 2823.54,
                    "word": " impact",
                    "probability": 0.8779296875
                },
                {
                    "start": 2823.54,
                    "end": 2823.76,
                    "word": " or",
                    "probability": 0.8720703125
                },
                {
                    "start": 2823.76,
                    "end": 2824.26,
                    "word": " severity",
                    "probability": 0.90869140625
                },
                {
                    "start": 2824.26,
                    "end": 2824.64,
                    "word": " of",
                    "probability": 0.89697265625
                },
                {
                    "start": 2824.64,
                    "end": 2824.9,
                    "word": " policy",
                    "probability": 0.8212890625
                },
                {
                    "start": 2824.9,
                    "end": 2825.32,
                    "word": " manipulation",
                    "probability": 0.9228515625
                },
                {
                    "start": 2825.32,
                    "end": 2825.74,
                    "word": " attacks",
                    "probability": 0.76904296875
                },
                {
                    "start": 2825.74,
                    "end": 2825.9,
                    "word": " on",
                    "probability": 0.80615234375
                },
                {
                    "start": 2825.9,
                    "end": 2826.54,
                    "word": " DQN",
                    "probability": 0.7752278645833334
                },
                {
                    "start": 2826.54,
                    "end": 2827.0,
                    "word": ",",
                    "probability": 0.4013671875
                },
                {
                    "start": 2827.0,
                    "end": 2827.58,
                    "word": " and",
                    "probability": 0.8896484375
                },
                {
                    "start": 2827.58,
                    "end": 2827.76,
                    "word": " it",
                    "probability": 0.9033203125
                },
                {
                    "start": 2827.76,
                    "end": 2827.9,
                    "word": " was",
                    "probability": 0.90234375
                },
                {
                    "start": 2827.9,
                    "end": 2828.28,
                    "word": " shown",
                    "probability": 0.87841796875
                },
                {
                    "start": 2828.28,
                    "end": 2828.82,
                    "word": " that",
                    "probability": 0.88232421875
                },
                {
                    "start": 2828.82,
                    "end": 2830.28,
                    "word": " it",
                    "probability": 0.88916015625
                }
            ]
        },
        {
            "id": 418,
            "seek": 284919,
            "start": 2830.28,
            "end": 2843.38,
            "text": " actually performs very well compared to vanilla or classical DQN. And these are the training time",
            "tokens": [
                50368,
                767,
                26213,
                588,
                731,
                5347,
                281,
                17528,
                420,
                13735,
                413,
                48,
                45,
                13,
                400,
                613,
                366,
                264,
                3097,
                565,
                51012
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20999052420710074,
            "compression_ratio": 1.5519125683060109,
            "no_speech_prob": 0.02886962890625,
            "words": [
                {
                    "start": 2830.28,
                    "end": 2830.72,
                    "word": " actually",
                    "probability": 0.40673828125
                },
                {
                    "start": 2830.72,
                    "end": 2831.48,
                    "word": " performs",
                    "probability": 0.666015625
                },
                {
                    "start": 2831.48,
                    "end": 2832.66,
                    "word": " very",
                    "probability": 0.5078125
                },
                {
                    "start": 2832.66,
                    "end": 2833.06,
                    "word": " well",
                    "probability": 0.87744140625
                },
                {
                    "start": 2833.06,
                    "end": 2833.5,
                    "word": " compared",
                    "probability": 0.79638671875
                },
                {
                    "start": 2833.5,
                    "end": 2834.54,
                    "word": " to",
                    "probability": 0.9033203125
                },
                {
                    "start": 2834.54,
                    "end": 2839.0,
                    "word": " vanilla",
                    "probability": 0.457763671875
                },
                {
                    "start": 2839.0,
                    "end": 2839.38,
                    "word": " or",
                    "probability": 0.8681640625
                },
                {
                    "start": 2839.38,
                    "end": 2839.96,
                    "word": " classical",
                    "probability": 0.763671875
                },
                {
                    "start": 2839.96,
                    "end": 2840.96,
                    "word": " DQN",
                    "probability": 0.6570638020833334
                },
                {
                    "start": 2840.96,
                    "end": 2841.6,
                    "word": ".",
                    "probability": 0.76904296875
                },
                {
                    "start": 2841.84,
                    "end": 2842.16,
                    "word": " And",
                    "probability": 0.662109375
                },
                {
                    "start": 2842.16,
                    "end": 2842.32,
                    "word": " these",
                    "probability": 0.7099609375
                },
                {
                    "start": 2842.32,
                    "end": 2842.48,
                    "word": " are",
                    "probability": 0.90087890625
                },
                {
                    "start": 2842.48,
                    "end": 2842.72,
                    "word": " the",
                    "probability": 0.80712890625
                },
                {
                    "start": 2842.72,
                    "end": 2843.04,
                    "word": " training",
                    "probability": 0.8046875
                },
                {
                    "start": 2843.04,
                    "end": 2843.38,
                    "word": " time",
                    "probability": 0.7939453125
                }
            ]
        },
        {
            "id": 419,
            "seek": 284919,
            "start": 2843.38,
            "end": 2849.84,
            "text": " results. You can see that if noisy net is used, if parameter space noise is used, the performance",
            "tokens": [
                51012,
                3542,
                13,
                509,
                393,
                536,
                300,
                498,
                24518,
                2533,
                307,
                1143,
                11,
                498,
                13075,
                1901,
                5658,
                307,
                1143,
                11,
                264,
                3389,
                51340
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20999052420710074,
            "compression_ratio": 1.5519125683060109,
            "no_speech_prob": 0.02886962890625,
            "words": [
                {
                    "start": 2843.38,
                    "end": 2843.66,
                    "word": " results",
                    "probability": 0.7080078125
                },
                {
                    "start": 2843.66,
                    "end": 2843.74,
                    "word": ".",
                    "probability": 0.548828125
                },
                {
                    "start": 2843.74,
                    "end": 2843.8,
                    "word": " You",
                    "probability": 0.88525390625
                },
                {
                    "start": 2843.8,
                    "end": 2843.92,
                    "word": " can",
                    "probability": 0.8955078125
                },
                {
                    "start": 2843.92,
                    "end": 2844.16,
                    "word": " see",
                    "probability": 0.82275390625
                },
                {
                    "start": 2844.16,
                    "end": 2844.56,
                    "word": " that",
                    "probability": 0.876953125
                },
                {
                    "start": 2844.56,
                    "end": 2845.62,
                    "word": " if",
                    "probability": 0.865234375
                },
                {
                    "start": 2845.62,
                    "end": 2846.1,
                    "word": " noisy",
                    "probability": 0.342529296875
                },
                {
                    "start": 2846.1,
                    "end": 2846.38,
                    "word": " net",
                    "probability": 0.498779296875
                },
                {
                    "start": 2846.38,
                    "end": 2846.58,
                    "word": " is",
                    "probability": 0.8896484375
                },
                {
                    "start": 2846.58,
                    "end": 2846.8,
                    "word": " used",
                    "probability": 0.828125
                },
                {
                    "start": 2846.8,
                    "end": 2846.9,
                    "word": ",",
                    "probability": 0.8330078125
                },
                {
                    "start": 2846.9,
                    "end": 2847.02,
                    "word": " if",
                    "probability": 0.85498046875
                },
                {
                    "start": 2847.02,
                    "end": 2847.34,
                    "word": " parameter",
                    "probability": 0.65087890625
                },
                {
                    "start": 2847.34,
                    "end": 2847.58,
                    "word": " space",
                    "probability": 0.6708984375
                },
                {
                    "start": 2847.58,
                    "end": 2847.84,
                    "word": " noise",
                    "probability": 0.8603515625
                },
                {
                    "start": 2847.84,
                    "end": 2848.24,
                    "word": " is",
                    "probability": 0.90234375
                },
                {
                    "start": 2848.24,
                    "end": 2848.64,
                    "word": " used",
                    "probability": 0.80810546875
                },
                {
                    "start": 2848.64,
                    "end": 2849.34,
                    "word": ",",
                    "probability": 0.91845703125
                },
                {
                    "start": 2849.34,
                    "end": 2849.52,
                    "word": " the",
                    "probability": 0.79541015625
                },
                {
                    "start": 2849.52,
                    "end": 2849.84,
                    "word": " performance",
                    "probability": 0.8515625
                }
            ]
        },
        {
            "id": 420,
            "seek": 284919,
            "start": 2849.84,
            "end": 2855.32,
            "text": " degradation for all environments is at a much lower slope, at a much lower rate than the",
            "tokens": [
                51340,
                40519,
                337,
                439,
                12388,
                307,
                412,
                257,
                709,
                3126,
                13525,
                11,
                412,
                257,
                709,
                3126,
                3314,
                813,
                264,
                51616
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20999052420710074,
            "compression_ratio": 1.5519125683060109,
            "no_speech_prob": 0.02886962890625,
            "words": [
                {
                    "start": 2849.84,
                    "end": 2850.22,
                    "word": " degradation",
                    "probability": 0.806640625
                },
                {
                    "start": 2850.22,
                    "end": 2851.04,
                    "word": " for",
                    "probability": 0.89404296875
                },
                {
                    "start": 2851.04,
                    "end": 2851.64,
                    "word": " all",
                    "probability": 0.9052734375
                },
                {
                    "start": 2851.64,
                    "end": 2852.26,
                    "word": " environments",
                    "probability": 0.6455078125
                },
                {
                    "start": 2852.26,
                    "end": 2852.72,
                    "word": " is",
                    "probability": 0.88720703125
                },
                {
                    "start": 2852.72,
                    "end": 2852.96,
                    "word": " at",
                    "probability": 0.90234375
                },
                {
                    "start": 2853.08,
                    "end": 2853.24,
                    "word": " a much",
                    "probability": 0.8076171875
                },
                {
                    "start": 2853.24,
                    "end": 2853.46,
                    "word": " lower",
                    "probability": 0.80029296875
                },
                {
                    "start": 2853.46,
                    "end": 2853.78,
                    "word": " slope",
                    "probability": 0.78271484375
                },
                {
                    "start": 2853.78,
                    "end": 2853.84,
                    "word": ",",
                    "probability": 0.671875
                },
                {
                    "start": 2853.84,
                    "end": 2853.9,
                    "word": " at",
                    "probability": 0.89501953125
                },
                {
                    "start": 2853.98,
                    "end": 2854.14,
                    "word": " a much",
                    "probability": 0.798828125
                },
                {
                    "start": 2854.14,
                    "end": 2854.34,
                    "word": " lower",
                    "probability": 0.7841796875
                },
                {
                    "start": 2854.34,
                    "end": 2854.66,
                    "word": " rate",
                    "probability": 0.80224609375
                },
                {
                    "start": 2854.66,
                    "end": 2854.86,
                    "word": " than",
                    "probability": 0.82080078125
                },
                {
                    "start": 2854.86,
                    "end": 2855.32,
                    "word": " the",
                    "probability": 0.80712890625
                }
            ]
        },
        {
            "id": 421,
            "seek": 287373,
            "start": 2855.32,
            "end": 2866.1,
            "text": " vanilla architecture. Finally, to have proposed a solution for the policy extraction problem,",
            "tokens": [
                50364,
                17528,
                9482,
                13,
                6288,
                11,
                281,
                362,
                10348,
                257,
                3827,
                337,
                264,
                3897,
                30197,
                1154,
                11,
                50816
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22120535948446818,
            "compression_ratio": 1.4789473684210526,
            "no_speech_prob": 0.00572967529296875,
            "words": [
                {
                    "start": 2855.32,
                    "end": 2855.78,
                    "word": " vanilla",
                    "probability": 0.271728515625
                },
                {
                    "start": 2855.78,
                    "end": 2856.9,
                    "word": " architecture",
                    "probability": 0.837890625
                },
                {
                    "start": 2856.9,
                    "end": 2857.66,
                    "word": ".",
                    "probability": 0.9248046875
                },
                {
                    "start": 2857.94,
                    "end": 2858.52,
                    "word": " Finally",
                    "probability": 0.7158203125
                },
                {
                    "start": 2858.52,
                    "end": 2859.2,
                    "word": ",",
                    "probability": 0.73876953125
                },
                {
                    "start": 2859.2,
                    "end": 2859.5,
                    "word": " to",
                    "probability": 0.89208984375
                },
                {
                    "start": 2859.5,
                    "end": 2861.32,
                    "word": " have",
                    "probability": 0.30078125
                },
                {
                    "start": 2861.32,
                    "end": 2861.78,
                    "word": " proposed",
                    "probability": 0.79638671875
                },
                {
                    "start": 2861.98,
                    "end": 2862.32,
                    "word": " a solution",
                    "probability": 0.83203125
                },
                {
                    "start": 2862.32,
                    "end": 2862.84,
                    "word": " for",
                    "probability": 0.888671875
                },
                {
                    "start": 2862.84,
                    "end": 2863.18,
                    "word": " the",
                    "probability": 0.81201171875
                },
                {
                    "start": 2863.18,
                    "end": 2863.46,
                    "word": " policy",
                    "probability": 0.78857421875
                },
                {
                    "start": 2863.46,
                    "end": 2863.88,
                    "word": " extraction",
                    "probability": 0.74462890625
                },
                {
                    "start": 2863.88,
                    "end": 2864.48,
                    "word": " problem",
                    "probability": 0.78076171875
                },
                {
                    "start": 2864.48,
                    "end": 2866.1,
                    "word": ",",
                    "probability": 0.7041015625
                }
            ]
        },
        {
            "id": 422,
            "seek": 287373,
            "start": 2866.98,
            "end": 2873.94,
            "text": " I, along with William Hsu at K-State, came up with the idea of watermarking DRL policies.",
            "tokens": [
                50952,
                286,
                11,
                2051,
                365,
                6740,
                389,
                15091,
                412,
                591,
                12,
                4520,
                473,
                11,
                1361,
                493,
                365,
                264,
                1558,
                295,
                1281,
                5638,
                278,
                413,
                10740,
                7657,
                13,
                51288
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22120535948446818,
            "compression_ratio": 1.4789473684210526,
            "no_speech_prob": 0.00572967529296875,
            "words": [
                {
                    "start": 2866.98,
                    "end": 2867.36,
                    "word": " I",
                    "probability": 0.92529296875
                },
                {
                    "start": 2867.36,
                    "end": 2867.6,
                    "word": ",",
                    "probability": 0.283203125
                },
                {
                    "start": 2867.6,
                    "end": 2867.6,
                    "word": " along",
                    "probability": 0.9111328125
                },
                {
                    "start": 2867.6,
                    "end": 2868.24,
                    "word": " with",
                    "probability": 0.8251953125
                },
                {
                    "start": 2868.24,
                    "end": 2868.96,
                    "word": " William",
                    "probability": 0.9111328125
                },
                {
                    "start": 2868.96,
                    "end": 2869.4,
                    "word": " Hsu",
                    "probability": 0.6246337890625
                },
                {
                    "start": 2869.4,
                    "end": 2869.64,
                    "word": " at",
                    "probability": 0.7724609375
                },
                {
                    "start": 2869.64,
                    "end": 2869.82,
                    "word": " K",
                    "probability": 0.8046875
                },
                {
                    "start": 2869.82,
                    "end": 2870.1,
                    "word": "-State",
                    "probability": 0.78857421875
                },
                {
                    "start": 2870.1,
                    "end": 2870.2,
                    "word": ",",
                    "probability": 0.66943359375
                },
                {
                    "start": 2870.2,
                    "end": 2870.36,
                    "word": " came",
                    "probability": 0.8251953125
                },
                {
                    "start": 2870.36,
                    "end": 2870.46,
                    "word": " up",
                    "probability": 0.9013671875
                },
                {
                    "start": 2870.46,
                    "end": 2870.56,
                    "word": " with",
                    "probability": 0.81494140625
                },
                {
                    "start": 2870.56,
                    "end": 2870.74,
                    "word": " the",
                    "probability": 0.81640625
                },
                {
                    "start": 2870.74,
                    "end": 2871.12,
                    "word": " idea",
                    "probability": 0.767578125
                },
                {
                    "start": 2871.12,
                    "end": 2871.68,
                    "word": " of",
                    "probability": 0.90576171875
                },
                {
                    "start": 2871.68,
                    "end": 2872.66,
                    "word": " watermarking",
                    "probability": 0.9119466145833334
                },
                {
                    "start": 2872.66,
                    "end": 2873.22,
                    "word": " DRL",
                    "probability": 0.823486328125
                },
                {
                    "start": 2873.22,
                    "end": 2873.76,
                    "word": " policies",
                    "probability": 0.7919921875
                },
                {
                    "start": 2873.76,
                    "end": 2873.94,
                    "word": ".",
                    "probability": 0.97265625
                }
            ]
        },
        {
            "id": 423,
            "seek": 287373,
            "start": 2873.94,
            "end": 2879.86,
            "text": " Watermarking has already been introduced in deep learning in general. The idea is to come up with",
            "tokens": [
                51288,
                8772,
                5638,
                278,
                575,
                1217,
                668,
                7268,
                294,
                2452,
                2539,
                294,
                2674,
                13,
                440,
                1558,
                307,
                281,
                808,
                493,
                365,
                51588
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22120535948446818,
            "compression_ratio": 1.4789473684210526,
            "no_speech_prob": 0.00572967529296875,
            "words": [
                {
                    "start": 2873.94,
                    "end": 2874.44,
                    "word": " Watermarking",
                    "probability": 0.9420572916666666
                },
                {
                    "start": 2874.44,
                    "end": 2874.56,
                    "word": " has",
                    "probability": 0.8994140625
                },
                {
                    "start": 2874.56,
                    "end": 2874.76,
                    "word": " already",
                    "probability": 0.8388671875
                },
                {
                    "start": 2874.76,
                    "end": 2874.92,
                    "word": " been",
                    "probability": 0.814453125
                },
                {
                    "start": 2874.92,
                    "end": 2875.34,
                    "word": " introduced",
                    "probability": 0.75390625
                },
                {
                    "start": 2875.34,
                    "end": 2875.78,
                    "word": " in",
                    "probability": 0.90234375
                },
                {
                    "start": 2875.78,
                    "end": 2876.42,
                    "word": " deep",
                    "probability": 0.828125
                },
                {
                    "start": 2876.42,
                    "end": 2876.62,
                    "word": " learning",
                    "probability": 0.81884765625
                },
                {
                    "start": 2876.62,
                    "end": 2876.84,
                    "word": " in",
                    "probability": 0.8330078125
                },
                {
                    "start": 2876.84,
                    "end": 2877.12,
                    "word": " general",
                    "probability": 0.7919921875
                },
                {
                    "start": 2877.12,
                    "end": 2877.32,
                    "word": ".",
                    "probability": 0.8251953125
                },
                {
                    "start": 2877.32,
                    "end": 2877.44,
                    "word": " The",
                    "probability": 0.814453125
                },
                {
                    "start": 2877.44,
                    "end": 2877.78,
                    "word": " idea",
                    "probability": 0.783203125
                },
                {
                    "start": 2877.78,
                    "end": 2878.36,
                    "word": " is",
                    "probability": 0.9091796875
                },
                {
                    "start": 2878.36,
                    "end": 2878.96,
                    "word": " to",
                    "probability": 0.90185546875
                },
                {
                    "start": 2878.96,
                    "end": 2879.7,
                    "word": " come",
                    "probability": 0.81494140625
                },
                {
                    "start": 2879.7,
                    "end": 2879.8,
                    "word": " up",
                    "probability": 0.8994140625
                },
                {
                    "start": 2879.8,
                    "end": 2879.86,
                    "word": " with",
                    "probability": 0.81494140625
                }
            ]
        },
        {
            "id": 424,
            "seek": 290143,
            "start": 2879.96,
            "end": 2885.98,
            "text": " a unique signature that is both difficult to remove and does not impact the performance of",
            "tokens": [
                50364,
                257,
                3845,
                13397,
                300,
                307,
                1293,
                2252,
                281,
                4159,
                293,
                775,
                406,
                2712,
                264,
                3389,
                295,
                50664
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2095905103902707,
            "compression_ratio": 1.6785714285714286,
            "no_speech_prob": 0.0179901123046875,
            "words": [
                {
                    "start": 2879.96,
                    "end": 2880.3,
                    "word": " a unique",
                    "probability": 0.65576171875
                },
                {
                    "start": 2880.3,
                    "end": 2880.8,
                    "word": " signature",
                    "probability": 0.890625
                },
                {
                    "start": 2880.8,
                    "end": 2881.02,
                    "word": " that",
                    "probability": 0.7783203125
                },
                {
                    "start": 2881.02,
                    "end": 2881.22,
                    "word": " is",
                    "probability": 0.873046875
                },
                {
                    "start": 2881.22,
                    "end": 2881.62,
                    "word": " both",
                    "probability": 0.67626953125
                },
                {
                    "start": 2881.62,
                    "end": 2882.0,
                    "word": " difficult",
                    "probability": 0.83056640625
                },
                {
                    "start": 2882.0,
                    "end": 2882.2,
                    "word": " to",
                    "probability": 0.90283203125
                },
                {
                    "start": 2882.2,
                    "end": 2882.74,
                    "word": " remove",
                    "probability": 0.8681640625
                },
                {
                    "start": 2882.74,
                    "end": 2883.44,
                    "word": " and",
                    "probability": 0.70263671875
                },
                {
                    "start": 2883.44,
                    "end": 2883.7,
                    "word": " does",
                    "probability": 0.90234375
                },
                {
                    "start": 2883.7,
                    "end": 2884.06,
                    "word": " not",
                    "probability": 0.8955078125
                },
                {
                    "start": 2884.06,
                    "end": 2884.6,
                    "word": " impact",
                    "probability": 0.80908203125
                },
                {
                    "start": 2884.6,
                    "end": 2884.9,
                    "word": " the",
                    "probability": 0.79931640625
                },
                {
                    "start": 2884.9,
                    "end": 2885.56,
                    "word": " performance",
                    "probability": 0.87158203125
                },
                {
                    "start": 2885.56,
                    "end": 2885.98,
                    "word": " of",
                    "probability": 0.89111328125
                }
            ]
        },
        {
            "id": 425,
            "seek": 290143,
            "start": 2885.98,
            "end": 2893.18,
            "text": " the policy itself, or the model itself, but it still provides a unique signature, proof that",
            "tokens": [
                50664,
                264,
                3897,
                2564,
                11,
                420,
                264,
                2316,
                2564,
                11,
                457,
                309,
                920,
                6417,
                257,
                3845,
                13397,
                11,
                8177,
                300,
                51040
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2095905103902707,
            "compression_ratio": 1.6785714285714286,
            "no_speech_prob": 0.0179901123046875,
            "words": [
                {
                    "start": 2885.98,
                    "end": 2886.1,
                    "word": " the",
                    "probability": 0.8154296875
                },
                {
                    "start": 2886.1,
                    "end": 2886.44,
                    "word": " policy",
                    "probability": 0.8212890625
                },
                {
                    "start": 2886.44,
                    "end": 2887.08,
                    "word": " itself",
                    "probability": 0.685546875
                },
                {
                    "start": 2887.08,
                    "end": 2887.38,
                    "word": ",",
                    "probability": 0.31005859375
                },
                {
                    "start": 2887.38,
                    "end": 2887.52,
                    "word": " or",
                    "probability": 0.76806640625
                },
                {
                    "start": 2887.52,
                    "end": 2887.68,
                    "word": " the",
                    "probability": 0.79736328125
                },
                {
                    "start": 2887.68,
                    "end": 2887.88,
                    "word": " model",
                    "probability": 0.81298828125
                },
                {
                    "start": 2887.88,
                    "end": 2888.44,
                    "word": " itself",
                    "probability": 0.77783203125
                },
                {
                    "start": 2888.44,
                    "end": 2889.24,
                    "word": ",",
                    "probability": 0.4716796875
                },
                {
                    "start": 2889.24,
                    "end": 2890.0,
                    "word": " but",
                    "probability": 0.87158203125
                },
                {
                    "start": 2890.0,
                    "end": 2890.26,
                    "word": " it",
                    "probability": 0.8662109375
                },
                {
                    "start": 2890.26,
                    "end": 2890.46,
                    "word": " still",
                    "probability": 0.87255859375
                },
                {
                    "start": 2890.46,
                    "end": 2890.76,
                    "word": " provides",
                    "probability": 0.80029296875
                },
                {
                    "start": 2890.88,
                    "end": 2891.08,
                    "word": " a unique",
                    "probability": 0.8193359375
                },
                {
                    "start": 2891.08,
                    "end": 2891.38,
                    "word": " signature",
                    "probability": 0.876953125
                },
                {
                    "start": 2891.38,
                    "end": 2891.6,
                    "word": ",",
                    "probability": 0.325439453125
                },
                {
                    "start": 2891.6,
                    "end": 2891.92,
                    "word": " proof",
                    "probability": 0.52783203125
                },
                {
                    "start": 2891.92,
                    "end": 2893.18,
                    "word": " that",
                    "probability": 0.748046875
                }
            ]
        },
        {
            "id": 426,
            "seek": 290143,
            "start": 2893.64,
            "end": 2902.98,
            "text": " a model is the same as another model, or is a replica of the suspected model. So we introduce",
            "tokens": [
                51040,
                257,
                2316,
                307,
                264,
                912,
                382,
                1071,
                2316,
                11,
                420,
                307,
                257,
                35456,
                295,
                264,
                26439,
                2316,
                13,
                407,
                321,
                5366,
                51520
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2095905103902707,
            "compression_ratio": 1.6785714285714286,
            "no_speech_prob": 0.0179901123046875,
            "words": [
                {
                    "start": 2893.64,
                    "end": 2893.98,
                    "word": " a model",
                    "probability": 0.8134765625
                },
                {
                    "start": 2893.98,
                    "end": 2894.14,
                    "word": " is",
                    "probability": 0.90478515625
                },
                {
                    "start": 2894.14,
                    "end": 2894.3,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 2894.3,
                    "end": 2894.52,
                    "word": " same",
                    "probability": 0.8125
                },
                {
                    "start": 2894.52,
                    "end": 2895.02,
                    "word": " as",
                    "probability": 0.90185546875
                },
                {
                    "start": 2895.02,
                    "end": 2896.56,
                    "word": " another",
                    "probability": 0.81787109375
                },
                {
                    "start": 2896.56,
                    "end": 2896.94,
                    "word": " model",
                    "probability": 0.81298828125
                },
                {
                    "start": 2896.94,
                    "end": 2897.46,
                    "word": ",",
                    "probability": 0.4384765625
                },
                {
                    "start": 2897.46,
                    "end": 2897.72,
                    "word": " or",
                    "probability": 0.9052734375
                },
                {
                    "start": 2897.72,
                    "end": 2898.2,
                    "word": " is",
                    "probability": 0.87744140625
                },
                {
                    "start": 2898.44,
                    "end": 2898.7,
                    "word": " a replica",
                    "probability": 0.71875
                },
                {
                    "start": 2898.7,
                    "end": 2899.52,
                    "word": " of",
                    "probability": 0.90283203125
                },
                {
                    "start": 2899.52,
                    "end": 2900.42,
                    "word": " the",
                    "probability": 0.73291015625
                },
                {
                    "start": 2900.42,
                    "end": 2901.52,
                    "word": " suspected",
                    "probability": 0.7490234375
                },
                {
                    "start": 2901.52,
                    "end": 2901.86,
                    "word": " model",
                    "probability": 0.814453125
                },
                {
                    "start": 2901.86,
                    "end": 2902.36,
                    "word": ".",
                    "probability": 0.98583984375
                },
                {
                    "start": 2902.36,
                    "end": 2902.56,
                    "word": " So",
                    "probability": 0.77685546875
                },
                {
                    "start": 2902.56,
                    "end": 2902.68,
                    "word": " we",
                    "probability": 0.701171875
                },
                {
                    "start": 2902.68,
                    "end": 2902.98,
                    "word": " introduce",
                    "probability": 0.5791015625
                }
            ]
        },
        {
            "id": 427,
            "seek": 290143,
            "start": 2902.98,
            "end": 2907.56,
            "text": " an interesting watermarking procedure. I know it's arrogant of me to call my own work interesting,",
            "tokens": [
                51520,
                364,
                1880,
                1281,
                5638,
                278,
                10747,
                13,
                286,
                458,
                309,
                311,
                30467,
                295,
                385,
                281,
                818,
                452,
                1065,
                589,
                1880,
                11,
                51744
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2095905103902707,
            "compression_ratio": 1.6785714285714286,
            "no_speech_prob": 0.0179901123046875,
            "words": [
                {
                    "start": 2902.98,
                    "end": 2903.16,
                    "word": " an",
                    "probability": 0.89892578125
                },
                {
                    "start": 2903.16,
                    "end": 2903.66,
                    "word": " interesting",
                    "probability": 0.6376953125
                },
                {
                    "start": 2903.66,
                    "end": 2904.18,
                    "word": " watermarking",
                    "probability": 0.9537760416666666
                },
                {
                    "start": 2904.18,
                    "end": 2904.5,
                    "word": " procedure",
                    "probability": 0.79248046875
                },
                {
                    "start": 2904.5,
                    "end": 2904.7,
                    "word": ".",
                    "probability": 0.83251953125
                },
                {
                    "start": 2904.7,
                    "end": 2904.72,
                    "word": " I",
                    "probability": 0.98876953125
                },
                {
                    "start": 2904.72,
                    "end": 2904.9,
                    "word": " know",
                    "probability": 0.8203125
                },
                {
                    "start": 2904.9,
                    "end": 2905.44,
                    "word": " it's",
                    "probability": 0.917236328125
                },
                {
                    "start": 2905.44,
                    "end": 2905.72,
                    "word": " arrogant",
                    "probability": 0.7783203125
                },
                {
                    "start": 2905.72,
                    "end": 2905.9,
                    "word": " of",
                    "probability": 0.86669921875
                },
                {
                    "start": 2905.9,
                    "end": 2906.0,
                    "word": " me",
                    "probability": 0.8984375
                },
                {
                    "start": 2906.0,
                    "end": 2906.26,
                    "word": " to",
                    "probability": 0.90283203125
                },
                {
                    "start": 2906.26,
                    "end": 2906.5,
                    "word": " call",
                    "probability": 0.8359375
                },
                {
                    "start": 2906.5,
                    "end": 2906.68,
                    "word": " my",
                    "probability": 0.9033203125
                },
                {
                    "start": 2906.68,
                    "end": 2906.84,
                    "word": " own",
                    "probability": 0.818359375
                },
                {
                    "start": 2906.84,
                    "end": 2907.02,
                    "word": " work",
                    "probability": 0.76806640625
                },
                {
                    "start": 2907.02,
                    "end": 2907.4,
                    "word": " interesting",
                    "probability": 0.6298828125
                },
                {
                    "start": 2907.4,
                    "end": 2907.56,
                    "word": ",",
                    "probability": 0.88623046875
                }
            ]
        },
        {
            "id": 428,
            "seek": 293089,
            "start": 2907.56,
            "end": 2911.42,
            "text": " but I still get excited when I think about the moment I came up with this idea. The idea is to",
            "tokens": [
                50364,
                457,
                286,
                920,
                483,
                2919,
                562,
                286,
                519,
                466,
                264,
                1623,
                286,
                1361,
                493,
                365,
                341,
                1558,
                13,
                440,
                1558,
                307,
                281,
                50556
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2313646821800722,
            "compression_ratio": 1.9076305220883534,
            "no_speech_prob": 0.0005526542663574219,
            "words": [
                {
                    "start": 2907.56,
                    "end": 2907.66,
                    "word": " but",
                    "probability": 0.2734375
                },
                {
                    "start": 2907.66,
                    "end": 2907.9,
                    "word": " I",
                    "probability": 0.978515625
                },
                {
                    "start": 2907.9,
                    "end": 2908.08,
                    "word": " still",
                    "probability": 0.859375
                },
                {
                    "start": 2908.08,
                    "end": 2908.24,
                    "word": " get",
                    "probability": 0.8828125
                },
                {
                    "start": 2908.24,
                    "end": 2908.62,
                    "word": " excited",
                    "probability": 0.892578125
                },
                {
                    "start": 2908.62,
                    "end": 2908.8,
                    "word": " when",
                    "probability": 0.703125
                },
                {
                    "start": 2908.8,
                    "end": 2908.84,
                    "word": " I",
                    "probability": 0.9921875
                },
                {
                    "start": 2908.84,
                    "end": 2908.98,
                    "word": " think",
                    "probability": 0.8017578125
                },
                {
                    "start": 2908.98,
                    "end": 2909.14,
                    "word": " about",
                    "probability": 0.80224609375
                },
                {
                    "start": 2909.14,
                    "end": 2909.24,
                    "word": " the",
                    "probability": 0.7958984375
                },
                {
                    "start": 2909.24,
                    "end": 2909.38,
                    "word": " moment",
                    "probability": 0.91064453125
                },
                {
                    "start": 2909.38,
                    "end": 2909.54,
                    "word": " I",
                    "probability": 0.94873046875
                },
                {
                    "start": 2909.54,
                    "end": 2909.72,
                    "word": " came",
                    "probability": 0.79296875
                },
                {
                    "start": 2909.72,
                    "end": 2909.82,
                    "word": " up",
                    "probability": 0.90380859375
                },
                {
                    "start": 2909.82,
                    "end": 2909.94,
                    "word": " with",
                    "probability": 0.81494140625
                },
                {
                    "start": 2909.94,
                    "end": 2910.14,
                    "word": " this",
                    "probability": 0.8701171875
                },
                {
                    "start": 2910.14,
                    "end": 2910.52,
                    "word": " idea",
                    "probability": 0.75732421875
                },
                {
                    "start": 2910.52,
                    "end": 2910.7,
                    "word": ".",
                    "probability": 0.919921875
                },
                {
                    "start": 2910.7,
                    "end": 2910.88,
                    "word": " The",
                    "probability": 0.7978515625
                },
                {
                    "start": 2910.88,
                    "end": 2911.24,
                    "word": " idea",
                    "probability": 0.77880859375
                },
                {
                    "start": 2911.24,
                    "end": 2911.36,
                    "word": " is",
                    "probability": 0.869140625
                },
                {
                    "start": 2911.36,
                    "end": 2911.42,
                    "word": " to",
                    "probability": 0.89599609375
                }
            ]
        },
        {
            "id": 429,
            "seek": 293089,
            "start": 2911.42,
            "end": 2917.46,
            "text": " create a second environment whose state space is this drawing from the main environment. So,",
            "tokens": [
                50556,
                1884,
                257,
                1150,
                2823,
                6104,
                1785,
                1901,
                307,
                341,
                6316,
                490,
                264,
                2135,
                2823,
                13,
                407,
                11,
                50828
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2313646821800722,
            "compression_ratio": 1.9076305220883534,
            "no_speech_prob": 0.0005526542663574219,
            "words": [
                {
                    "start": 2911.42,
                    "end": 2911.6,
                    "word": " create",
                    "probability": 0.85302734375
                },
                {
                    "start": 2911.74,
                    "end": 2911.96,
                    "word": " a second",
                    "probability": 0.78076171875
                },
                {
                    "start": 2911.96,
                    "end": 2912.52,
                    "word": " environment",
                    "probability": 0.76611328125
                },
                {
                    "start": 2912.52,
                    "end": 2913.3,
                    "word": " whose",
                    "probability": 0.54638671875
                },
                {
                    "start": 2913.3,
                    "end": 2913.62,
                    "word": " state",
                    "probability": 0.634765625
                },
                {
                    "start": 2913.62,
                    "end": 2914.06,
                    "word": " space",
                    "probability": 0.74560546875
                },
                {
                    "start": 2914.06,
                    "end": 2914.34,
                    "word": " is",
                    "probability": 0.89697265625
                },
                {
                    "start": 2914.34,
                    "end": 2914.68,
                    "word": " this",
                    "probability": 0.55712890625
                },
                {
                    "start": 2914.68,
                    "end": 2915.02,
                    "word": " drawing",
                    "probability": 0.68408203125
                },
                {
                    "start": 2915.02,
                    "end": 2915.26,
                    "word": " from",
                    "probability": 0.82080078125
                },
                {
                    "start": 2915.26,
                    "end": 2915.46,
                    "word": " the",
                    "probability": 0.81640625
                },
                {
                    "start": 2915.46,
                    "end": 2915.9,
                    "word": " main",
                    "probability": 0.81494140625
                },
                {
                    "start": 2915.9,
                    "end": 2916.44,
                    "word": " environment",
                    "probability": 0.74755859375
                },
                {
                    "start": 2916.44,
                    "end": 2916.76,
                    "word": ".",
                    "probability": 0.9013671875
                },
                {
                    "start": 2916.76,
                    "end": 2916.92,
                    "word": " So",
                    "probability": 0.5419921875
                },
                {
                    "start": 2916.92,
                    "end": 2917.46,
                    "word": ",",
                    "probability": 0.3779296875
                }
            ]
        },
        {
            "id": 430,
            "seek": 293089,
            "start": 2917.46,
            "end": 2923.14,
            "text": " create an environment. If you're training an agent to play a game, let's create another environment",
            "tokens": [
                50856,
                1884,
                364,
                2823,
                13,
                759,
                291,
                434,
                3097,
                364,
                9461,
                281,
                862,
                257,
                1216,
                11,
                718,
                311,
                1884,
                1071,
                2823,
                51168
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2313646821800722,
            "compression_ratio": 1.9076305220883534,
            "no_speech_prob": 0.0005526542663574219,
            "words": [
                {
                    "start": 2917.46,
                    "end": 2917.76,
                    "word": " create",
                    "probability": 0.67236328125
                },
                {
                    "start": 2917.76,
                    "end": 2917.94,
                    "word": " an",
                    "probability": 0.90380859375
                },
                {
                    "start": 2917.94,
                    "end": 2918.34,
                    "word": " environment",
                    "probability": 0.76513671875
                },
                {
                    "start": 2918.34,
                    "end": 2918.46,
                    "word": ".",
                    "probability": 0.5693359375
                },
                {
                    "start": 2918.46,
                    "end": 2918.54,
                    "word": " If",
                    "probability": 0.892578125
                },
                {
                    "start": 2918.54,
                    "end": 2918.66,
                    "word": " you're",
                    "probability": 0.9072265625
                },
                {
                    "start": 2918.66,
                    "end": 2918.98,
                    "word": " training",
                    "probability": 0.81884765625
                },
                {
                    "start": 2918.98,
                    "end": 2919.6,
                    "word": " an",
                    "probability": 0.88525390625
                },
                {
                    "start": 2919.6,
                    "end": 2919.82,
                    "word": " agent",
                    "probability": 0.8017578125
                },
                {
                    "start": 2919.82,
                    "end": 2920.42,
                    "word": " to",
                    "probability": 0.8984375
                },
                {
                    "start": 2920.42,
                    "end": 2920.94,
                    "word": " play",
                    "probability": 0.87158203125
                },
                {
                    "start": 2921.16,
                    "end": 2921.5,
                    "word": " a game",
                    "probability": 0.8125
                },
                {
                    "start": 2921.5,
                    "end": 2921.94,
                    "word": ",",
                    "probability": 0.953125
                },
                {
                    "start": 2921.94,
                    "end": 2922.08,
                    "word": " let's",
                    "probability": 0.9248046875
                },
                {
                    "start": 2922.08,
                    "end": 2922.26,
                    "word": " create",
                    "probability": 0.85205078125
                },
                {
                    "start": 2922.26,
                    "end": 2922.58,
                    "word": " another",
                    "probability": 0.8291015625
                },
                {
                    "start": 2922.58,
                    "end": 2923.14,
                    "word": " environment",
                    "probability": 0.76025390625
                }
            ]
        },
        {
            "id": 431,
            "seek": 293089,
            "start": 2923.14,
            "end": 2931.5,
            "text": " which has no states. None of these states are the same as the original training environment,",
            "tokens": [
                51168,
                597,
                575,
                572,
                4368,
                13,
                14492,
                295,
                613,
                4368,
                366,
                264,
                912,
                382,
                264,
                3380,
                3097,
                2823,
                11,
                51556
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2313646821800722,
            "compression_ratio": 1.9076305220883534,
            "no_speech_prob": 0.0005526542663574219,
            "words": [
                {
                    "start": 2923.14,
                    "end": 2923.94,
                    "word": " which",
                    "probability": 0.736328125
                },
                {
                    "start": 2923.94,
                    "end": 2924.2,
                    "word": " has",
                    "probability": 0.90478515625
                },
                {
                    "start": 2924.2,
                    "end": 2924.82,
                    "word": " no",
                    "probability": 0.87451171875
                },
                {
                    "start": 2924.82,
                    "end": 2925.3,
                    "word": " states",
                    "probability": 0.86572265625
                },
                {
                    "start": 2925.3,
                    "end": 2925.56,
                    "word": ".",
                    "probability": 0.71142578125
                },
                {
                    "start": 2925.56,
                    "end": 2925.96,
                    "word": " None",
                    "probability": 0.892578125
                },
                {
                    "start": 2925.96,
                    "end": 2926.1,
                    "word": " of",
                    "probability": 0.9091796875
                },
                {
                    "start": 2926.1,
                    "end": 2926.24,
                    "word": " these",
                    "probability": 0.6611328125
                },
                {
                    "start": 2926.24,
                    "end": 2926.58,
                    "word": " states",
                    "probability": 0.87646484375
                },
                {
                    "start": 2926.58,
                    "end": 2927.28,
                    "word": " are",
                    "probability": 0.890625
                },
                {
                    "start": 2927.28,
                    "end": 2927.5,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 2927.5,
                    "end": 2927.74,
                    "word": " same",
                    "probability": 0.8095703125
                },
                {
                    "start": 2927.74,
                    "end": 2928.16,
                    "word": " as",
                    "probability": 0.90625
                },
                {
                    "start": 2928.16,
                    "end": 2928.82,
                    "word": " the",
                    "probability": 0.818359375
                },
                {
                    "start": 2928.82,
                    "end": 2929.44,
                    "word": " original",
                    "probability": 0.833984375
                },
                {
                    "start": 2929.44,
                    "end": 2930.84,
                    "word": " training",
                    "probability": 0.8251953125
                },
                {
                    "start": 2930.84,
                    "end": 2931.34,
                    "word": " environment",
                    "probability": 0.77587890625
                },
                {
                    "start": 2931.34,
                    "end": 2931.5,
                    "word": ",",
                    "probability": 0.1207275390625
                }
            ]
        },
        {
            "id": 432,
            "seek": 293089,
            "start": 2931.5,
            "end": 2937.02,
            "text": " or deployment environment of the agent. But the dimensionality of the states are the same. So,",
            "tokens": [
                51556,
                420,
                19317,
                2823,
                295,
                264,
                9461,
                13,
                583,
                264,
                10139,
                1860,
                295,
                264,
                4368,
                366,
                264,
                912,
                13,
                407,
                11,
                51832
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2313646821800722,
            "compression_ratio": 1.9076305220883534,
            "no_speech_prob": 0.0005526542663574219,
            "words": [
                {
                    "start": 2931.5,
                    "end": 2931.64,
                    "word": " or",
                    "probability": 0.89306640625
                },
                {
                    "start": 2931.64,
                    "end": 2931.96,
                    "word": " deployment",
                    "probability": 0.6943359375
                },
                {
                    "start": 2931.96,
                    "end": 2932.42,
                    "word": " environment",
                    "probability": 0.74755859375
                },
                {
                    "start": 2932.42,
                    "end": 2932.62,
                    "word": " of",
                    "probability": 0.8017578125
                },
                {
                    "start": 2932.62,
                    "end": 2932.72,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 2932.72,
                    "end": 2932.92,
                    "word": " agent",
                    "probability": 0.81201171875
                },
                {
                    "start": 2932.92,
                    "end": 2933.38,
                    "word": ".",
                    "probability": 0.73583984375
                },
                {
                    "start": 2933.72,
                    "end": 2934.08,
                    "word": " But",
                    "probability": 0.8759765625
                },
                {
                    "start": 2934.08,
                    "end": 2934.3,
                    "word": " the",
                    "probability": 0.78759765625
                },
                {
                    "start": 2934.3,
                    "end": 2935.1,
                    "word": " dimensionality",
                    "probability": 0.90771484375
                },
                {
                    "start": 2935.1,
                    "end": 2935.28,
                    "word": " of",
                    "probability": 0.9091796875
                },
                {
                    "start": 2935.28,
                    "end": 2935.44,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 2935.44,
                    "end": 2935.76,
                    "word": " states",
                    "probability": 0.880859375
                },
                {
                    "start": 2935.76,
                    "end": 2936.26,
                    "word": " are",
                    "probability": 0.88623046875
                },
                {
                    "start": 2936.26,
                    "end": 2936.46,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 2936.46,
                    "end": 2936.66,
                    "word": " same",
                    "probability": 0.8154296875
                },
                {
                    "start": 2936.66,
                    "end": 2936.8,
                    "word": ".",
                    "probability": 0.79931640625
                },
                {
                    "start": 2936.8,
                    "end": 2936.92,
                    "word": " So",
                    "probability": 0.87548828125
                },
                {
                    "start": 2936.92,
                    "end": 2937.02,
                    "word": ",",
                    "probability": 0.67724609375
                }
            ]
        },
        {
            "id": 433,
            "seek": 295995,
            "start": 2937.02,
            "end": 2942.64,
            "text": " if each state in the original environment is represented by, let's say, three values,",
            "tokens": [
                50364,
                498,
                1184,
                1785,
                294,
                264,
                3380,
                2823,
                307,
                10379,
                538,
                11,
                718,
                311,
                584,
                11,
                1045,
                4190,
                11,
                50604
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23646907708079545,
            "compression_ratio": 1.9567099567099566,
            "no_speech_prob": 2.7120113372802734e-05,
            "words": [
                {
                    "start": 2937.02,
                    "end": 2937.1,
                    "word": " if",
                    "probability": 0.22021484375
                },
                {
                    "start": 2937.1,
                    "end": 2937.42,
                    "word": " each",
                    "probability": 0.84033203125
                },
                {
                    "start": 2937.42,
                    "end": 2937.72,
                    "word": " state",
                    "probability": 0.88037109375
                },
                {
                    "start": 2937.72,
                    "end": 2937.84,
                    "word": " in",
                    "probability": 0.892578125
                },
                {
                    "start": 2937.84,
                    "end": 2937.96,
                    "word": " the",
                    "probability": 0.7734375
                },
                {
                    "start": 2937.96,
                    "end": 2938.4,
                    "word": " original",
                    "probability": 0.8173828125
                },
                {
                    "start": 2938.4,
                    "end": 2939.28,
                    "word": " environment",
                    "probability": 0.76708984375
                },
                {
                    "start": 2939.28,
                    "end": 2939.76,
                    "word": " is",
                    "probability": 0.8994140625
                },
                {
                    "start": 2939.76,
                    "end": 2940.32,
                    "word": " represented",
                    "probability": 0.71240234375
                },
                {
                    "start": 2940.32,
                    "end": 2940.58,
                    "word": " by",
                    "probability": 0.8955078125
                },
                {
                    "start": 2940.58,
                    "end": 2940.66,
                    "word": ",",
                    "probability": 0.5498046875
                },
                {
                    "start": 2940.66,
                    "end": 2940.8,
                    "word": " let's",
                    "probability": 0.934814453125
                },
                {
                    "start": 2940.8,
                    "end": 2940.94,
                    "word": " say",
                    "probability": 0.8994140625
                },
                {
                    "start": 2940.94,
                    "end": 2941.02,
                    "word": ",",
                    "probability": 0.82568359375
                },
                {
                    "start": 2941.02,
                    "end": 2941.26,
                    "word": " three",
                    "probability": 0.8046875
                },
                {
                    "start": 2941.26,
                    "end": 2941.88,
                    "word": " values",
                    "probability": 0.830078125
                },
                {
                    "start": 2941.88,
                    "end": 2942.64,
                    "word": ",",
                    "probability": 0.81884765625
                }
            ]
        },
        {
            "id": 434,
            "seek": 295995,
            "start": 2943.76,
            "end": 2948.74,
            "text": " three features, then each state in the second environment is also represented by three",
            "tokens": [
                50720,
                1045,
                4122,
                11,
                550,
                1184,
                1785,
                294,
                264,
                1150,
                2823,
                307,
                611,
                2556,
                6003,
                538,
                1045,
                50948
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23646907708079545,
            "compression_ratio": 1.9567099567099566,
            "no_speech_prob": 2.7120113372802734e-05,
            "words": [
                {
                    "start": 2943.76,
                    "end": 2944.36,
                    "word": " three",
                    "probability": 0.76416015625
                },
                {
                    "start": 2944.36,
                    "end": 2944.82,
                    "word": " features",
                    "probability": 0.7236328125
                },
                {
                    "start": 2944.82,
                    "end": 2945.3,
                    "word": ",",
                    "probability": 0.9453125
                },
                {
                    "start": 2945.3,
                    "end": 2945.54,
                    "word": " then",
                    "probability": 0.72998046875
                },
                {
                    "start": 2945.54,
                    "end": 2946.24,
                    "word": " each",
                    "probability": 0.84521484375
                },
                {
                    "start": 2946.24,
                    "end": 2946.54,
                    "word": " state",
                    "probability": 0.88818359375
                },
                {
                    "start": 2946.54,
                    "end": 2946.66,
                    "word": " in",
                    "probability": 0.904296875
                },
                {
                    "start": 2946.66,
                    "end": 2946.78,
                    "word": " the",
                    "probability": 0.75439453125
                },
                {
                    "start": 2946.78,
                    "end": 2946.98,
                    "word": " second",
                    "probability": 0.78515625
                },
                {
                    "start": 2946.98,
                    "end": 2947.36,
                    "word": " environment",
                    "probability": 0.77001953125
                },
                {
                    "start": 2947.36,
                    "end": 2947.72,
                    "word": " is",
                    "probability": 0.896484375
                },
                {
                    "start": 2947.72,
                    "end": 2947.94,
                    "word": " also",
                    "probability": 0.80517578125
                },
                {
                    "start": 2947.94,
                    "end": 2948.32,
                    "word": " represented",
                    "probability": 0.60894775390625
                },
                {
                    "start": 2948.32,
                    "end": 2948.52,
                    "word": " by",
                    "probability": 0.9091796875
                },
                {
                    "start": 2948.52,
                    "end": 2948.74,
                    "word": " three",
                    "probability": 0.833984375
                }
            ]
        },
        {
            "id": 435,
            "seek": 295995,
            "start": 2948.74,
            "end": 2954.4,
            "text": " features. It doesn't really matter what the second environment looks like. It's just some",
            "tokens": [
                50948,
                4122,
                13,
                467,
                1177,
                380,
                534,
                1871,
                437,
                264,
                1150,
                2823,
                1542,
                411,
                13,
                467,
                311,
                445,
                512,
                51236
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23646907708079545,
            "compression_ratio": 1.9567099567099566,
            "no_speech_prob": 2.7120113372802734e-05,
            "words": [
                {
                    "start": 2948.74,
                    "end": 2949.08,
                    "word": " features",
                    "probability": 0.7353515625
                },
                {
                    "start": 2949.08,
                    "end": 2949.58,
                    "word": ".",
                    "probability": 0.98681640625
                },
                {
                    "start": 2949.58,
                    "end": 2949.94,
                    "word": " It",
                    "probability": 0.9052734375
                },
                {
                    "start": 2949.94,
                    "end": 2950.22,
                    "word": " doesn't",
                    "probability": 0.933349609375
                },
                {
                    "start": 2950.22,
                    "end": 2950.5,
                    "word": " really",
                    "probability": 0.763671875
                },
                {
                    "start": 2950.5,
                    "end": 2950.74,
                    "word": " matter",
                    "probability": 0.798828125
                },
                {
                    "start": 2950.74,
                    "end": 2950.98,
                    "word": " what",
                    "probability": 0.884765625
                },
                {
                    "start": 2950.98,
                    "end": 2951.3,
                    "word": " the",
                    "probability": 0.80908203125
                },
                {
                    "start": 2951.3,
                    "end": 2951.82,
                    "word": " second",
                    "probability": 0.79443359375
                },
                {
                    "start": 2951.82,
                    "end": 2952.28,
                    "word": " environment",
                    "probability": 0.7724609375
                },
                {
                    "start": 2952.28,
                    "end": 2952.58,
                    "word": " looks",
                    "probability": 0.79931640625
                },
                {
                    "start": 2952.58,
                    "end": 2952.8,
                    "word": " like",
                    "probability": 0.8798828125
                },
                {
                    "start": 2952.8,
                    "end": 2952.86,
                    "word": ".",
                    "probability": 0.7099609375
                },
                {
                    "start": 2952.86,
                    "end": 2953.06,
                    "word": " It's",
                    "probability": 0.9501953125
                },
                {
                    "start": 2953.06,
                    "end": 2953.56,
                    "word": " just",
                    "probability": 0.88623046875
                },
                {
                    "start": 2953.56,
                    "end": 2954.4,
                    "word": " some",
                    "probability": 0.81689453125
                }
            ]
        },
        {
            "id": 436,
            "seek": 295995,
            "start": 2954.4,
            "end": 2961.38,
            "text": " other environment that the agent may interact with. And then, we craft the transition dynamics",
            "tokens": [
                51236,
                661,
                2823,
                300,
                264,
                9461,
                815,
                4648,
                365,
                13,
                400,
                550,
                11,
                321,
                8448,
                264,
                6034,
                15679,
                51580
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23646907708079545,
            "compression_ratio": 1.9567099567099566,
            "no_speech_prob": 2.7120113372802734e-05,
            "words": [
                {
                    "start": 2954.4,
                    "end": 2954.72,
                    "word": " other",
                    "probability": 0.765625
                },
                {
                    "start": 2954.72,
                    "end": 2955.34,
                    "word": " environment",
                    "probability": 0.71240234375
                },
                {
                    "start": 2955.34,
                    "end": 2956.14,
                    "word": " that",
                    "probability": 0.87353515625
                },
                {
                    "start": 2956.14,
                    "end": 2956.26,
                    "word": " the",
                    "probability": 0.8193359375
                },
                {
                    "start": 2956.26,
                    "end": 2956.48,
                    "word": " agent",
                    "probability": 0.27880859375
                },
                {
                    "start": 2956.48,
                    "end": 2956.72,
                    "word": " may",
                    "probability": 0.65673828125
                },
                {
                    "start": 2956.72,
                    "end": 2957.12,
                    "word": " interact",
                    "probability": 0.771484375
                },
                {
                    "start": 2957.12,
                    "end": 2957.34,
                    "word": " with",
                    "probability": 0.81298828125
                },
                {
                    "start": 2957.34,
                    "end": 2957.94,
                    "word": ".",
                    "probability": 0.96435546875
                },
                {
                    "start": 2958.76,
                    "end": 2958.96,
                    "word": " And",
                    "probability": 0.57763671875
                },
                {
                    "start": 2958.96,
                    "end": 2959.14,
                    "word": " then",
                    "probability": 0.71826171875
                },
                {
                    "start": 2959.14,
                    "end": 2959.42,
                    "word": ",",
                    "probability": 0.3056640625
                },
                {
                    "start": 2959.42,
                    "end": 2959.82,
                    "word": " we",
                    "probability": 0.90283203125
                },
                {
                    "start": 2959.82,
                    "end": 2960.28,
                    "word": " craft",
                    "probability": 0.91162109375
                },
                {
                    "start": 2960.28,
                    "end": 2960.44,
                    "word": " the",
                    "probability": 0.75
                },
                {
                    "start": 2960.44,
                    "end": 2960.88,
                    "word": " transition",
                    "probability": 0.9208984375
                },
                {
                    "start": 2960.88,
                    "end": 2961.38,
                    "word": " dynamics",
                    "probability": 0.71826171875
                }
            ]
        },
        {
            "id": 437,
            "seek": 295995,
            "start": 2961.38,
            "end": 2966.08,
            "text": " and reward procedure for the second environment such that the optimal policy follows a looping",
            "tokens": [
                51580,
                293,
                7782,
                10747,
                337,
                264,
                1150,
                2823,
                1270,
                300,
                264,
                16252,
                3897,
                10002,
                257,
                6367,
                278,
                51812
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23646907708079545,
            "compression_ratio": 1.9567099567099566,
            "no_speech_prob": 2.7120113372802734e-05,
            "words": [
                {
                    "start": 2961.38,
                    "end": 2961.52,
                    "word": " and",
                    "probability": 0.80712890625
                },
                {
                    "start": 2961.52,
                    "end": 2961.82,
                    "word": " reward",
                    "probability": 0.8623046875
                },
                {
                    "start": 2961.82,
                    "end": 2962.18,
                    "word": " procedure",
                    "probability": 0.85009765625
                },
                {
                    "start": 2962.18,
                    "end": 2962.32,
                    "word": " for",
                    "probability": 0.89306640625
                },
                {
                    "start": 2962.32,
                    "end": 2962.46,
                    "word": " the",
                    "probability": 0.78076171875
                },
                {
                    "start": 2962.46,
                    "end": 2962.7,
                    "word": " second",
                    "probability": 0.78515625
                },
                {
                    "start": 2962.7,
                    "end": 2963.12,
                    "word": " environment",
                    "probability": 0.70703125
                },
                {
                    "start": 2963.12,
                    "end": 2963.58,
                    "word": " such",
                    "probability": 0.5986328125
                },
                {
                    "start": 2963.58,
                    "end": 2963.76,
                    "word": " that",
                    "probability": 0.876953125
                },
                {
                    "start": 2963.76,
                    "end": 2963.96,
                    "word": " the",
                    "probability": 0.798828125
                },
                {
                    "start": 2963.96,
                    "end": 2964.3,
                    "word": " optimal",
                    "probability": 0.76611328125
                },
                {
                    "start": 2964.3,
                    "end": 2964.9,
                    "word": " policy",
                    "probability": 0.8291015625
                },
                {
                    "start": 2964.9,
                    "end": 2965.4,
                    "word": " follows",
                    "probability": 0.79931640625
                },
                {
                    "start": 2965.7,
                    "end": 2966.08,
                    "word": " a looping",
                    "probability": 0.86865234375
                }
            ]
        },
        {
            "id": 438,
            "seek": 298941,
            "start": 2966.08,
            "end": 2973.24,
            "text": " trajectory. So, an optimal policy for an agent trained in the second environment is going to be",
            "tokens": [
                50364,
                21512,
                13,
                407,
                11,
                364,
                16252,
                3897,
                337,
                364,
                9461,
                8895,
                294,
                264,
                1150,
                2823,
                307,
                516,
                281,
                312,
                50732
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24085022441975706,
            "compression_ratio": 1.8253968253968254,
            "no_speech_prob": 0.00154876708984375,
            "words": [
                {
                    "start": 2966.08,
                    "end": 2966.46,
                    "word": " trajectory",
                    "probability": 0.054443359375
                },
                {
                    "start": 2966.46,
                    "end": 2966.7,
                    "word": ".",
                    "probability": 0.7158203125
                },
                {
                    "start": 2966.7,
                    "end": 2966.84,
                    "word": " So",
                    "probability": 0.64453125
                },
                {
                    "start": 2966.84,
                    "end": 2967.72,
                    "word": ",",
                    "probability": 0.35107421875
                },
                {
                    "start": 2967.72,
                    "end": 2967.82,
                    "word": " an",
                    "probability": 0.8779296875
                },
                {
                    "start": 2967.82,
                    "end": 2968.2,
                    "word": " optimal",
                    "probability": 0.7734375
                },
                {
                    "start": 2968.2,
                    "end": 2968.74,
                    "word": " policy",
                    "probability": 0.80615234375
                },
                {
                    "start": 2968.74,
                    "end": 2969.02,
                    "word": " for",
                    "probability": 0.8876953125
                },
                {
                    "start": 2969.02,
                    "end": 2969.3,
                    "word": " an",
                    "probability": 0.88671875
                },
                {
                    "start": 2969.3,
                    "end": 2969.56,
                    "word": " agent",
                    "probability": 0.82861328125
                },
                {
                    "start": 2969.56,
                    "end": 2970.02,
                    "word": " trained",
                    "probability": 0.787109375
                },
                {
                    "start": 2970.02,
                    "end": 2970.48,
                    "word": " in",
                    "probability": 0.8994140625
                },
                {
                    "start": 2970.48,
                    "end": 2970.8,
                    "word": " the",
                    "probability": 0.80126953125
                },
                {
                    "start": 2970.8,
                    "end": 2971.1,
                    "word": " second",
                    "probability": 0.77685546875
                },
                {
                    "start": 2971.1,
                    "end": 2971.62,
                    "word": " environment",
                    "probability": 0.748046875
                },
                {
                    "start": 2971.62,
                    "end": 2972.44,
                    "word": " is",
                    "probability": 0.81103515625
                },
                {
                    "start": 2972.44,
                    "end": 2972.7,
                    "word": " going",
                    "probability": 0.8388671875
                },
                {
                    "start": 2972.7,
                    "end": 2972.92,
                    "word": " to",
                    "probability": 0.90576171875
                },
                {
                    "start": 2972.92,
                    "end": 2973.24,
                    "word": " be",
                    "probability": 0.89990234375
                }
            ]
        },
        {
            "id": 439,
            "seek": 298941,
            "start": 2973.24,
            "end": 2977.12,
            "text": " one that follows a loop, goes through, let's say, state one, then state two, state three,",
            "tokens": [
                50732,
                472,
                300,
                10002,
                257,
                6367,
                11,
                1709,
                807,
                11,
                718,
                311,
                584,
                11,
                1785,
                472,
                11,
                550,
                1785,
                732,
                11,
                1785,
                1045,
                11,
                50912
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24085022441975706,
            "compression_ratio": 1.8253968253968254,
            "no_speech_prob": 0.00154876708984375,
            "words": [
                {
                    "start": 2973.24,
                    "end": 2973.76,
                    "word": " one",
                    "probability": 0.82275390625
                },
                {
                    "start": 2973.76,
                    "end": 2974.02,
                    "word": " that",
                    "probability": 0.87548828125
                },
                {
                    "start": 2974.02,
                    "end": 2974.34,
                    "word": " follows",
                    "probability": 0.76806640625
                },
                {
                    "start": 2974.48,
                    "end": 2974.72,
                    "word": " a loop",
                    "probability": 0.8681640625
                },
                {
                    "start": 2974.72,
                    "end": 2974.86,
                    "word": ",",
                    "probability": 0.5576171875
                },
                {
                    "start": 2974.86,
                    "end": 2975.08,
                    "word": " goes",
                    "probability": 0.734375
                },
                {
                    "start": 2975.08,
                    "end": 2975.36,
                    "word": " through",
                    "probability": 0.333984375
                },
                {
                    "start": 2975.36,
                    "end": 2975.46,
                    "word": ",",
                    "probability": 0.400390625
                },
                {
                    "start": 2975.46,
                    "end": 2975.6,
                    "word": " let's",
                    "probability": 0.931396484375
                },
                {
                    "start": 2975.6,
                    "end": 2975.72,
                    "word": " say",
                    "probability": 0.88671875
                },
                {
                    "start": 2975.72,
                    "end": 2975.76,
                    "word": ",",
                    "probability": 0.60400390625
                },
                {
                    "start": 2975.76,
                    "end": 2975.92,
                    "word": " state",
                    "probability": 0.80517578125
                },
                {
                    "start": 2975.92,
                    "end": 2976.22,
                    "word": " one",
                    "probability": 0.52880859375
                },
                {
                    "start": 2976.22,
                    "end": 2976.34,
                    "word": ",",
                    "probability": 0.85107421875
                },
                {
                    "start": 2976.34,
                    "end": 2976.42,
                    "word": " then",
                    "probability": 0.64990234375
                },
                {
                    "start": 2976.42,
                    "end": 2976.6,
                    "word": " state",
                    "probability": 0.83203125
                },
                {
                    "start": 2976.6,
                    "end": 2976.74,
                    "word": " two",
                    "probability": 0.7919921875
                },
                {
                    "start": 2976.74,
                    "end": 2976.78,
                    "word": ",",
                    "probability": 0.90625
                },
                {
                    "start": 2976.78,
                    "end": 2976.92,
                    "word": " state",
                    "probability": 0.71484375
                },
                {
                    "start": 2976.92,
                    "end": 2977.08,
                    "word": " three",
                    "probability": 0.8486328125
                },
                {
                    "start": 2977.08,
                    "end": 2977.12,
                    "word": ",",
                    "probability": 0.80615234375
                }
            ]
        },
        {
            "id": 440,
            "seek": 298941,
            "start": 2977.12,
            "end": 2983.16,
            "text": " and then goes back to state one. During training, what happens is, we periodically alternate",
            "tokens": [
                50912,
                293,
                550,
                1709,
                646,
                281,
                1785,
                472,
                13,
                6842,
                3097,
                11,
                437,
                2314,
                307,
                11,
                321,
                38916,
                18873,
                51236
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24085022441975706,
            "compression_ratio": 1.8253968253968254,
            "no_speech_prob": 0.00154876708984375,
            "words": [
                {
                    "start": 2977.12,
                    "end": 2977.18,
                    "word": " and",
                    "probability": 0.85107421875
                },
                {
                    "start": 2977.18,
                    "end": 2977.3,
                    "word": " then",
                    "probability": 0.72021484375
                },
                {
                    "start": 2977.3,
                    "end": 2978.04,
                    "word": " goes",
                    "probability": 0.76708984375
                },
                {
                    "start": 2978.04,
                    "end": 2978.2,
                    "word": " back",
                    "probability": 0.814453125
                },
                {
                    "start": 2978.2,
                    "end": 2978.32,
                    "word": " to",
                    "probability": 0.88623046875
                },
                {
                    "start": 2978.32,
                    "end": 2978.56,
                    "word": " state",
                    "probability": 0.8740234375
                },
                {
                    "start": 2978.56,
                    "end": 2978.84,
                    "word": " one",
                    "probability": 0.80712890625
                },
                {
                    "start": 2978.84,
                    "end": 2979.32,
                    "word": ".",
                    "probability": 0.994140625
                },
                {
                    "start": 2979.34,
                    "end": 2979.68,
                    "word": " During",
                    "probability": 0.84130859375
                },
                {
                    "start": 2979.68,
                    "end": 2980.14,
                    "word": " training",
                    "probability": 0.79541015625
                },
                {
                    "start": 2980.14,
                    "end": 2980.6,
                    "word": ",",
                    "probability": 0.904296875
                },
                {
                    "start": 2980.6,
                    "end": 2980.76,
                    "word": " what",
                    "probability": 0.8759765625
                },
                {
                    "start": 2980.76,
                    "end": 2981.06,
                    "word": " happens",
                    "probability": 0.9228515625
                },
                {
                    "start": 2981.06,
                    "end": 2981.36,
                    "word": " is",
                    "probability": 0.90380859375
                },
                {
                    "start": 2981.36,
                    "end": 2981.56,
                    "word": ",",
                    "probability": 0.343505859375
                },
                {
                    "start": 2981.56,
                    "end": 2981.7,
                    "word": " we",
                    "probability": 0.89453125
                },
                {
                    "start": 2981.7,
                    "end": 2982.32,
                    "word": " periodically",
                    "probability": 0.86767578125
                },
                {
                    "start": 2982.32,
                    "end": 2983.16,
                    "word": " alternate",
                    "probability": 0.806640625
                }
            ]
        },
        {
            "id": 441,
            "seek": 298941,
            "start": 2983.16,
            "end": 2990.04,
            "text": " between the two environments. So, let's say, at every N iterations of the training process,",
            "tokens": [
                51236,
                1296,
                264,
                732,
                12388,
                13,
                407,
                11,
                718,
                311,
                584,
                11,
                412,
                633,
                426,
                36540,
                295,
                264,
                3097,
                1399,
                11,
                51532
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24085022441975706,
            "compression_ratio": 1.8253968253968254,
            "no_speech_prob": 0.00154876708984375,
            "words": [
                {
                    "start": 2983.16,
                    "end": 2983.88,
                    "word": " between",
                    "probability": 0.78515625
                },
                {
                    "start": 2983.88,
                    "end": 2984.06,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 2984.06,
                    "end": 2984.2,
                    "word": " two",
                    "probability": 0.80615234375
                },
                {
                    "start": 2984.2,
                    "end": 2984.62,
                    "word": " environments",
                    "probability": 0.66357421875
                },
                {
                    "start": 2984.62,
                    "end": 2984.8,
                    "word": ".",
                    "probability": 0.9638671875
                },
                {
                    "start": 2984.8,
                    "end": 2984.96,
                    "word": " So",
                    "probability": 0.86865234375
                },
                {
                    "start": 2984.96,
                    "end": 2985.04,
                    "word": ",",
                    "probability": 0.6533203125
                },
                {
                    "start": 2985.04,
                    "end": 2985.22,
                    "word": " let's",
                    "probability": 0.939697265625
                },
                {
                    "start": 2985.22,
                    "end": 2985.46,
                    "word": " say",
                    "probability": 0.896484375
                },
                {
                    "start": 2985.46,
                    "end": 2985.66,
                    "word": ",",
                    "probability": 0.431884765625
                },
                {
                    "start": 2985.66,
                    "end": 2985.78,
                    "word": " at",
                    "probability": 0.8974609375
                },
                {
                    "start": 2985.78,
                    "end": 2986.24,
                    "word": " every",
                    "probability": 0.74169921875
                },
                {
                    "start": 2986.24,
                    "end": 2987.1,
                    "word": " N",
                    "probability": 0.3525390625
                },
                {
                    "start": 2987.1,
                    "end": 2988.26,
                    "word": " iterations",
                    "probability": 0.72216796875
                },
                {
                    "start": 2988.26,
                    "end": 2988.64,
                    "word": " of",
                    "probability": 0.90185546875
                },
                {
                    "start": 2988.64,
                    "end": 2988.76,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 2988.76,
                    "end": 2988.98,
                    "word": " training",
                    "probability": 0.79443359375
                },
                {
                    "start": 2988.98,
                    "end": 2989.48,
                    "word": " process",
                    "probability": 0.86572265625
                },
                {
                    "start": 2989.48,
                    "end": 2990.04,
                    "word": ",",
                    "probability": 0.9609375
                }
            ]
        },
        {
            "id": 442,
            "seek": 298941,
            "start": 2990.04,
            "end": 2995.54,
            "text": " we take our RL agent from the original environment, we drop it in the second environment,",
            "tokens": [
                51564,
                321,
                747,
                527,
                497,
                43,
                9461,
                490,
                264,
                3380,
                2823,
                11,
                321,
                3270,
                309,
                294,
                264,
                1150,
                2823,
                11,
                51832
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24085022441975706,
            "compression_ratio": 1.8253968253968254,
            "no_speech_prob": 0.00154876708984375,
            "words": [
                {
                    "start": 2990.04,
                    "end": 2990.42,
                    "word": " we",
                    "probability": 0.89794921875
                },
                {
                    "start": 2990.42,
                    "end": 2990.96,
                    "word": " take",
                    "probability": 0.82177734375
                },
                {
                    "start": 2990.96,
                    "end": 2991.74,
                    "word": " our",
                    "probability": 0.806640625
                },
                {
                    "start": 2991.74,
                    "end": 2992.24,
                    "word": " RL",
                    "probability": 0.87646484375
                },
                {
                    "start": 2992.24,
                    "end": 2992.68,
                    "word": " agent",
                    "probability": 0.76708984375
                },
                {
                    "start": 2992.68,
                    "end": 2993.02,
                    "word": " from",
                    "probability": 0.81640625
                },
                {
                    "start": 2993.02,
                    "end": 2993.14,
                    "word": " the",
                    "probability": 0.7705078125
                },
                {
                    "start": 2993.14,
                    "end": 2993.36,
                    "word": " original",
                    "probability": 0.80126953125
                },
                {
                    "start": 2993.36,
                    "end": 2993.74,
                    "word": " environment",
                    "probability": 0.75732421875
                },
                {
                    "start": 2993.74,
                    "end": 2993.86,
                    "word": ",",
                    "probability": 0.8515625
                },
                {
                    "start": 2993.86,
                    "end": 2993.96,
                    "word": " we",
                    "probability": 0.8896484375
                },
                {
                    "start": 2993.96,
                    "end": 2994.26,
                    "word": " drop",
                    "probability": 0.82421875
                },
                {
                    "start": 2994.26,
                    "end": 2994.5,
                    "word": " it",
                    "probability": 0.90478515625
                },
                {
                    "start": 2994.5,
                    "end": 2994.7,
                    "word": " in",
                    "probability": 0.88232421875
                },
                {
                    "start": 2994.7,
                    "end": 2994.82,
                    "word": " the",
                    "probability": 0.80322265625
                },
                {
                    "start": 2994.82,
                    "end": 2995.04,
                    "word": " second",
                    "probability": 0.79443359375
                },
                {
                    "start": 2995.04,
                    "end": 2995.44,
                    "word": " environment",
                    "probability": 0.7578125
                },
                {
                    "start": 2995.44,
                    "end": 2995.54,
                    "word": ",",
                    "probability": 0.962890625
                }
            ]
        },
        {
            "id": 443,
            "seek": 301821,
            "start": 2995.54,
            "end": 3000.22,
            "text": " we train it for a few iterations, and then bring it back to the original environment.",
            "tokens": [
                50364,
                321,
                3847,
                309,
                337,
                257,
                1326,
                36540,
                11,
                293,
                550,
                1565,
                309,
                646,
                281,
                264,
                3380,
                2823,
                13,
                50588
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2852879347426168,
            "compression_ratio": 1.6403508771929824,
            "no_speech_prob": 0.0005440711975097656,
            "words": [
                {
                    "start": 2995.54,
                    "end": 2995.68,
                    "word": " we",
                    "probability": 0.11956787109375
                },
                {
                    "start": 2995.68,
                    "end": 2995.7,
                    "word": " train",
                    "probability": 0.56982421875
                },
                {
                    "start": 2995.7,
                    "end": 2995.86,
                    "word": " it",
                    "probability": 0.8671875
                },
                {
                    "start": 2995.86,
                    "end": 2996.18,
                    "word": " for",
                    "probability": 0.8974609375
                },
                {
                    "start": 2996.68,
                    "end": 2996.86,
                    "word": " a few",
                    "probability": 0.810546875
                },
                {
                    "start": 2996.86,
                    "end": 2997.3,
                    "word": " iterations",
                    "probability": 0.669921875
                },
                {
                    "start": 2997.3,
                    "end": 2997.4,
                    "word": ",",
                    "probability": 0.2188720703125
                },
                {
                    "start": 2997.4,
                    "end": 2997.6,
                    "word": " and",
                    "probability": 0.89013671875
                },
                {
                    "start": 2997.6,
                    "end": 2997.78,
                    "word": " then",
                    "probability": 0.7392578125
                },
                {
                    "start": 2997.78,
                    "end": 2997.98,
                    "word": " bring",
                    "probability": 0.876953125
                },
                {
                    "start": 2997.98,
                    "end": 2998.12,
                    "word": " it",
                    "probability": 0.9013671875
                },
                {
                    "start": 2998.12,
                    "end": 2998.36,
                    "word": " back",
                    "probability": 0.8173828125
                },
                {
                    "start": 2998.36,
                    "end": 2998.5,
                    "word": " to",
                    "probability": 0.90576171875
                },
                {
                    "start": 2998.5,
                    "end": 2998.62,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 2998.62,
                    "end": 2998.9,
                    "word": " original",
                    "probability": 0.81494140625
                },
                {
                    "start": 2998.9,
                    "end": 2999.66,
                    "word": " environment",
                    "probability": 0.7548828125
                },
                {
                    "start": 2999.66,
                    "end": 3000.22,
                    "word": ".",
                    "probability": 0.99609375
                }
            ]
        },
        {
            "id": 444,
            "seek": 301821,
            "start": 3000.46,
            "end": 3006.92,
            "text": " Now, once trained, if we want to examine the authenticity or whether a policy is copied or not,",
            "tokens": [
                50620,
                823,
                11,
                1564,
                8895,
                11,
                498,
                321,
                528,
                281,
                17496,
                264,
                34215,
                420,
                1968,
                257,
                3897,
                307,
                25365,
                420,
                406,
                11,
                50944
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2852879347426168,
            "compression_ratio": 1.6403508771929824,
            "no_speech_prob": 0.0005440711975097656,
            "words": [
                {
                    "start": 3000.46,
                    "end": 3000.74,
                    "word": " Now",
                    "probability": 0.8662109375
                },
                {
                    "start": 3000.74,
                    "end": 3001.32,
                    "word": ",",
                    "probability": 0.82080078125
                },
                {
                    "start": 3001.32,
                    "end": 3001.7,
                    "word": " once",
                    "probability": 0.8134765625
                },
                {
                    "start": 3001.7,
                    "end": 3002.14,
                    "word": " trained",
                    "probability": 0.8125
                },
                {
                    "start": 3002.14,
                    "end": 3002.38,
                    "word": ",",
                    "probability": 0.95751953125
                },
                {
                    "start": 3002.38,
                    "end": 3002.54,
                    "word": " if",
                    "probability": 0.90185546875
                },
                {
                    "start": 3002.54,
                    "end": 3002.6,
                    "word": " we",
                    "probability": 0.475341796875
                },
                {
                    "start": 3002.6,
                    "end": 3002.72,
                    "word": " want",
                    "probability": 0.8173828125
                },
                {
                    "start": 3002.72,
                    "end": 3002.82,
                    "word": " to",
                    "probability": 0.90283203125
                },
                {
                    "start": 3002.82,
                    "end": 3003.1,
                    "word": " examine",
                    "probability": 0.88623046875
                },
                {
                    "start": 3003.1,
                    "end": 3003.28,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 3003.28,
                    "end": 3003.88,
                    "word": " authenticity",
                    "probability": 0.57275390625
                },
                {
                    "start": 3003.88,
                    "end": 3004.9,
                    "word": " or",
                    "probability": 0.38916015625
                },
                {
                    "start": 3004.9,
                    "end": 3005.2,
                    "word": " whether",
                    "probability": 0.83349609375
                },
                {
                    "start": 3005.74,
                    "end": 3005.98,
                    "word": " a policy",
                    "probability": 0.787109375
                },
                {
                    "start": 3005.98,
                    "end": 3006.26,
                    "word": " is",
                    "probability": 0.90185546875
                },
                {
                    "start": 3006.26,
                    "end": 3006.56,
                    "word": " copied",
                    "probability": 0.765625
                },
                {
                    "start": 3006.56,
                    "end": 3006.7,
                    "word": " or",
                    "probability": 0.8955078125
                },
                {
                    "start": 3006.7,
                    "end": 3006.88,
                    "word": " not",
                    "probability": 0.8935546875
                },
                {
                    "start": 3006.88,
                    "end": 3006.92,
                    "word": ",",
                    "probability": 0.96044921875
                }
            ]
        },
        {
            "id": 445,
            "seek": 301821,
            "start": 3006.92,
            "end": 3013.26,
            "text": " we apply the policy in the second environment and measure the total reward. Here's the experimental",
            "tokens": [
                50944,
                321,
                3079,
                264,
                3897,
                294,
                264,
                1150,
                2823,
                293,
                3481,
                264,
                3217,
                7782,
                13,
                1692,
                311,
                264,
                17069,
                51268
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2852879347426168,
            "compression_ratio": 1.6403508771929824,
            "no_speech_prob": 0.0005440711975097656,
            "words": [
                {
                    "start": 3006.92,
                    "end": 3007.02,
                    "word": " we",
                    "probability": 0.90380859375
                },
                {
                    "start": 3007.02,
                    "end": 3007.38,
                    "word": " apply",
                    "probability": 0.7255859375
                },
                {
                    "start": 3007.38,
                    "end": 3007.58,
                    "word": " the",
                    "probability": 0.68505859375
                },
                {
                    "start": 3007.58,
                    "end": 3008.16,
                    "word": " policy",
                    "probability": 0.81396484375
                },
                {
                    "start": 3008.16,
                    "end": 3009.32,
                    "word": " in",
                    "probability": 0.88623046875
                },
                {
                    "start": 3009.32,
                    "end": 3009.48,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 3009.48,
                    "end": 3009.8,
                    "word": " second",
                    "probability": 0.79052734375
                },
                {
                    "start": 3009.8,
                    "end": 3010.26,
                    "word": " environment",
                    "probability": 0.76318359375
                },
                {
                    "start": 3010.26,
                    "end": 3010.8,
                    "word": " and",
                    "probability": 0.7568359375
                },
                {
                    "start": 3010.8,
                    "end": 3011.06,
                    "word": " measure",
                    "probability": 0.78466796875
                },
                {
                    "start": 3011.06,
                    "end": 3011.26,
                    "word": " the",
                    "probability": 0.806640625
                },
                {
                    "start": 3011.26,
                    "end": 3011.5,
                    "word": " total",
                    "probability": 0.8076171875
                },
                {
                    "start": 3011.5,
                    "end": 3011.96,
                    "word": " reward",
                    "probability": 0.84814453125
                },
                {
                    "start": 3011.96,
                    "end": 3012.52,
                    "word": ".",
                    "probability": 0.99365234375
                },
                {
                    "start": 3012.64,
                    "end": 3012.92,
                    "word": " Here's",
                    "probability": 0.892333984375
                },
                {
                    "start": 3012.92,
                    "end": 3012.98,
                    "word": " the",
                    "probability": 0.8037109375
                },
                {
                    "start": 3012.98,
                    "end": 3013.26,
                    "word": " experimental",
                    "probability": 0.7265625
                }
            ]
        },
        {
            "id": 446,
            "seek": 301821,
            "start": 3013.26,
            "end": 3024.34,
            "text": " setup. Again, we are working with CardPool. So, the watermarking environment is defined with",
            "tokens": [
                51268,
                8657,
                13,
                3764,
                11,
                321,
                366,
                1364,
                365,
                11877,
                47,
                1092,
                13,
                407,
                11,
                264,
                1281,
                5638,
                278,
                2823,
                307,
                7642,
                365,
                51820
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2852879347426168,
            "compression_ratio": 1.6403508771929824,
            "no_speech_prob": 0.0005440711975097656,
            "words": [
                {
                    "start": 3013.26,
                    "end": 3013.54,
                    "word": " setup",
                    "probability": 0.736328125
                },
                {
                    "start": 3013.54,
                    "end": 3013.66,
                    "word": ".",
                    "probability": 0.459716796875
                },
                {
                    "start": 3013.66,
                    "end": 3013.74,
                    "word": " Again",
                    "probability": 0.90771484375
                },
                {
                    "start": 3013.74,
                    "end": 3013.78,
                    "word": ",",
                    "probability": 0.943359375
                },
                {
                    "start": 3013.78,
                    "end": 3013.86,
                    "word": " we",
                    "probability": 0.90185546875
                },
                {
                    "start": 3013.86,
                    "end": 3013.92,
                    "word": " are",
                    "probability": 0.78857421875
                },
                {
                    "start": 3013.92,
                    "end": 3015.28,
                    "word": " working",
                    "probability": 0.441162109375
                },
                {
                    "start": 3015.28,
                    "end": 3015.66,
                    "word": " with",
                    "probability": 0.80810546875
                },
                {
                    "start": 3015.66,
                    "end": 3016.48,
                    "word": " CardPool",
                    "probability": 0.5348307291666666
                },
                {
                    "start": 3016.48,
                    "end": 3017.04,
                    "word": ".",
                    "probability": 0.9599609375
                },
                {
                    "start": 3018.3,
                    "end": 3018.5,
                    "word": " So",
                    "probability": 0.69873046875
                },
                {
                    "start": 3018.5,
                    "end": 3019.64,
                    "word": ",",
                    "probability": 0.25927734375
                },
                {
                    "start": 3019.64,
                    "end": 3020.42,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 3020.42,
                    "end": 3021.2,
                    "word": " watermarking",
                    "probability": 0.9298502604166666
                },
                {
                    "start": 3021.2,
                    "end": 3021.72,
                    "word": " environment",
                    "probability": 0.7294921875
                },
                {
                    "start": 3021.72,
                    "end": 3022.86,
                    "word": " is",
                    "probability": 0.90087890625
                },
                {
                    "start": 3022.86,
                    "end": 3024.04,
                    "word": " defined",
                    "probability": 0.71923828125
                },
                {
                    "start": 3024.04,
                    "end": 3024.34,
                    "word": " with",
                    "probability": 0.82080078125
                }
            ]
        },
        {
            "id": 447,
            "seek": 304365,
            "start": 3024.34,
            "end": 3032.88,
            "text": " five states, states one to four, plus a terminal state, which should never be reached if the",
            "tokens": [
                50364,
                1732,
                4368,
                11,
                4368,
                472,
                281,
                1451,
                11,
                1804,
                257,
                14709,
                1785,
                11,
                597,
                820,
                1128,
                312,
                6488,
                498,
                264,
                50792
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21913109574376083,
            "compression_ratio": 1.7621359223300972,
            "no_speech_prob": 3.319978713989258e-05,
            "words": [
                {
                    "start": 3024.34,
                    "end": 3024.52,
                    "word": " five",
                    "probability": 0.06878662109375
                },
                {
                    "start": 3024.52,
                    "end": 3025.1,
                    "word": " states",
                    "probability": 0.86328125
                },
                {
                    "start": 3025.1,
                    "end": 3025.32,
                    "word": ",",
                    "probability": 0.52880859375
                },
                {
                    "start": 3025.32,
                    "end": 3025.54,
                    "word": " states",
                    "probability": 0.81689453125
                },
                {
                    "start": 3025.54,
                    "end": 3026.62,
                    "word": " one",
                    "probability": 0.464599609375
                },
                {
                    "start": 3026.62,
                    "end": 3027.32,
                    "word": " to",
                    "probability": 0.8720703125
                },
                {
                    "start": 3027.32,
                    "end": 3027.66,
                    "word": " four",
                    "probability": 0.88427734375
                },
                {
                    "start": 3027.66,
                    "end": 3028.04,
                    "word": ",",
                    "probability": 0.266845703125
                },
                {
                    "start": 3028.04,
                    "end": 3028.46,
                    "word": " plus",
                    "probability": 0.87841796875
                },
                {
                    "start": 3028.7,
                    "end": 3029.06,
                    "word": " a terminal",
                    "probability": 0.73291015625
                },
                {
                    "start": 3029.06,
                    "end": 3029.6,
                    "word": " state",
                    "probability": 0.8623046875
                },
                {
                    "start": 3029.6,
                    "end": 3030.34,
                    "word": ",",
                    "probability": 0.6904296875
                },
                {
                    "start": 3030.34,
                    "end": 3030.96,
                    "word": " which",
                    "probability": 0.90478515625
                },
                {
                    "start": 3030.96,
                    "end": 3031.12,
                    "word": " should",
                    "probability": 0.884765625
                },
                {
                    "start": 3031.12,
                    "end": 3031.26,
                    "word": " never",
                    "probability": 0.87353515625
                },
                {
                    "start": 3031.26,
                    "end": 3031.56,
                    "word": " be",
                    "probability": 0.90478515625
                },
                {
                    "start": 3031.56,
                    "end": 3031.94,
                    "word": " reached",
                    "probability": 0.7646484375
                },
                {
                    "start": 3031.94,
                    "end": 3032.32,
                    "word": " if",
                    "probability": 0.8671875
                },
                {
                    "start": 3032.32,
                    "end": 3032.88,
                    "word": " the",
                    "probability": 0.80712890625
                }
            ]
        },
        {
            "id": 448,
            "seek": 304365,
            "start": 3032.88,
            "end": 3039.98,
            "text": " policy is optimal in this environment. And none of these states, as represented here,",
            "tokens": [
                50792,
                3897,
                307,
                16252,
                294,
                341,
                2823,
                13,
                400,
                6022,
                295,
                613,
                4368,
                11,
                382,
                10379,
                510,
                11,
                51112
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21913109574376083,
            "compression_ratio": 1.7621359223300972,
            "no_speech_prob": 3.319978713989258e-05,
            "words": [
                {
                    "start": 3032.88,
                    "end": 3033.36,
                    "word": " policy",
                    "probability": 0.74951171875
                },
                {
                    "start": 3033.36,
                    "end": 3033.76,
                    "word": " is",
                    "probability": 0.7314453125
                },
                {
                    "start": 3033.76,
                    "end": 3034.62,
                    "word": " optimal",
                    "probability": 0.6162109375
                },
                {
                    "start": 3034.62,
                    "end": 3034.82,
                    "word": " in",
                    "probability": 0.9013671875
                },
                {
                    "start": 3034.82,
                    "end": 3034.96,
                    "word": " this",
                    "probability": 0.869140625
                },
                {
                    "start": 3034.96,
                    "end": 3035.34,
                    "word": " environment",
                    "probability": 0.76513671875
                },
                {
                    "start": 3035.34,
                    "end": 3036.02,
                    "word": ".",
                    "probability": 0.64892578125
                },
                {
                    "start": 3036.2,
                    "end": 3036.5,
                    "word": " And",
                    "probability": 0.72265625
                },
                {
                    "start": 3036.5,
                    "end": 3036.94,
                    "word": " none",
                    "probability": 0.85498046875
                },
                {
                    "start": 3036.94,
                    "end": 3037.08,
                    "word": " of",
                    "probability": 0.9072265625
                },
                {
                    "start": 3037.08,
                    "end": 3037.36,
                    "word": " these",
                    "probability": 0.70166015625
                },
                {
                    "start": 3037.36,
                    "end": 3038.2,
                    "word": " states",
                    "probability": 0.88525390625
                },
                {
                    "start": 3038.2,
                    "end": 3038.3,
                    "word": ",",
                    "probability": 0.591796875
                },
                {
                    "start": 3038.3,
                    "end": 3038.54,
                    "word": " as",
                    "probability": 0.90283203125
                },
                {
                    "start": 3038.54,
                    "end": 3038.9,
                    "word": " represented",
                    "probability": 0.62841796875
                },
                {
                    "start": 3038.9,
                    "end": 3039.36,
                    "word": " here",
                    "probability": 0.81494140625
                },
                {
                    "start": 3039.36,
                    "end": 3039.98,
                    "word": ",",
                    "probability": 0.97119140625
                }
            ]
        },
        {
            "id": 449,
            "seek": 304365,
            "start": 3039.98,
            "end": 3044.64,
            "text": " can be found in the original, can occur in the original CardPool environment. These are",
            "tokens": [
                51144,
                393,
                312,
                1352,
                294,
                264,
                3380,
                11,
                393,
                5160,
                294,
                264,
                3380,
                11877,
                47,
                1092,
                2823,
                13,
                1981,
                366,
                51376
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21913109574376083,
            "compression_ratio": 1.7621359223300972,
            "no_speech_prob": 3.319978713989258e-05,
            "words": [
                {
                    "start": 3039.98,
                    "end": 3040.36,
                    "word": " can",
                    "probability": 0.9052734375
                },
                {
                    "start": 3040.36,
                    "end": 3040.78,
                    "word": " be",
                    "probability": 0.68505859375
                },
                {
                    "start": 3040.78,
                    "end": 3041.12,
                    "word": " found",
                    "probability": 0.85205078125
                },
                {
                    "start": 3041.12,
                    "end": 3041.28,
                    "word": " in",
                    "probability": 0.87353515625
                },
                {
                    "start": 3041.28,
                    "end": 3041.4,
                    "word": " the",
                    "probability": 0.8095703125
                },
                {
                    "start": 3041.4,
                    "end": 3041.74,
                    "word": " original",
                    "probability": 0.79345703125
                },
                {
                    "start": 3041.74,
                    "end": 3041.82,
                    "word": ",",
                    "probability": 0.712890625
                },
                {
                    "start": 3041.82,
                    "end": 3041.98,
                    "word": " can",
                    "probability": 0.86865234375
                },
                {
                    "start": 3041.98,
                    "end": 3042.34,
                    "word": " occur",
                    "probability": 0.83544921875
                },
                {
                    "start": 3042.34,
                    "end": 3042.54,
                    "word": " in",
                    "probability": 0.89306640625
                },
                {
                    "start": 3042.54,
                    "end": 3042.72,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 3042.72,
                    "end": 3043.22,
                    "word": " original",
                    "probability": 0.8193359375
                },
                {
                    "start": 3043.22,
                    "end": 3043.76,
                    "word": " CardPool",
                    "probability": 0.12350972493489583
                },
                {
                    "start": 3043.76,
                    "end": 3044.18,
                    "word": " environment",
                    "probability": 0.7529296875
                },
                {
                    "start": 3044.18,
                    "end": 3044.36,
                    "word": ".",
                    "probability": 0.9189453125
                },
                {
                    "start": 3044.36,
                    "end": 3044.5,
                    "word": " These",
                    "probability": 0.78466796875
                },
                {
                    "start": 3044.5,
                    "end": 3044.64,
                    "word": " are",
                    "probability": 0.896484375
                }
            ]
        },
        {
            "id": 450,
            "seek": 304365,
            "start": 3044.64,
            "end": 3049.78,
            "text": " all highly impossible, not highly, definitely and absolutely impossible to occur in the original",
            "tokens": [
                51376,
                439,
                5405,
                6243,
                11,
                406,
                5405,
                11,
                2138,
                293,
                3122,
                6243,
                281,
                5160,
                294,
                264,
                3380,
                51636
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21913109574376083,
            "compression_ratio": 1.7621359223300972,
            "no_speech_prob": 3.319978713989258e-05,
            "words": [
                {
                    "start": 3044.64,
                    "end": 3044.88,
                    "word": " all",
                    "probability": 0.8896484375
                },
                {
                    "start": 3044.88,
                    "end": 3045.18,
                    "word": " highly",
                    "probability": 0.88818359375
                },
                {
                    "start": 3045.18,
                    "end": 3046.0,
                    "word": " impossible",
                    "probability": 0.63671875
                },
                {
                    "start": 3046.0,
                    "end": 3046.44,
                    "word": ",",
                    "probability": 0.6513671875
                },
                {
                    "start": 3046.44,
                    "end": 3046.7,
                    "word": " not",
                    "probability": 0.763671875
                },
                {
                    "start": 3046.7,
                    "end": 3046.94,
                    "word": " highly",
                    "probability": 0.89990234375
                },
                {
                    "start": 3046.94,
                    "end": 3047.2,
                    "word": ",",
                    "probability": 0.890625
                },
                {
                    "start": 3047.2,
                    "end": 3047.72,
                    "word": " definitely",
                    "probability": 0.75244140625
                },
                {
                    "start": 3047.72,
                    "end": 3047.88,
                    "word": " and",
                    "probability": 0.6875
                },
                {
                    "start": 3047.88,
                    "end": 3048.14,
                    "word": " absolutely",
                    "probability": 0.705078125
                },
                {
                    "start": 3048.14,
                    "end": 3048.52,
                    "word": " impossible",
                    "probability": 0.779296875
                },
                {
                    "start": 3048.52,
                    "end": 3048.7,
                    "word": " to",
                    "probability": 0.8642578125
                },
                {
                    "start": 3048.7,
                    "end": 3048.94,
                    "word": " occur",
                    "probability": 0.84814453125
                },
                {
                    "start": 3048.94,
                    "end": 3049.44,
                    "word": " in",
                    "probability": 0.89794921875
                },
                {
                    "start": 3049.44,
                    "end": 3049.54,
                    "word": " the",
                    "probability": 0.8134765625
                },
                {
                    "start": 3049.54,
                    "end": 3049.78,
                    "word": " original",
                    "probability": 0.81396484375
                }
            ]
        },
        {
            "id": 451,
            "seek": 306427,
            "start": 3049.78,
            "end": 3058.06,
            "text": " environment. As for the transition dynamics, this is how we've defined it. Let a particular state be",
            "tokens": [
                50364,
                2823,
                13,
                1018,
                337,
                264,
                6034,
                15679,
                11,
                341,
                307,
                577,
                321,
                600,
                7642,
                309,
                13,
                961,
                257,
                1729,
                1785,
                312,
                50780
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23012610067401015,
            "compression_ratio": 1.3676470588235294,
            "no_speech_prob": 0.01666259765625,
            "words": [
                {
                    "start": 3049.78,
                    "end": 3050.18,
                    "word": " environment",
                    "probability": 0.52294921875
                },
                {
                    "start": 3050.18,
                    "end": 3050.7,
                    "word": ".",
                    "probability": 0.9111328125
                },
                {
                    "start": 3050.7,
                    "end": 3051.0,
                    "word": " As",
                    "probability": 0.818359375
                },
                {
                    "start": 3051.0,
                    "end": 3051.22,
                    "word": " for",
                    "probability": 0.884765625
                },
                {
                    "start": 3051.22,
                    "end": 3051.38,
                    "word": " the",
                    "probability": 0.771484375
                },
                {
                    "start": 3051.38,
                    "end": 3051.82,
                    "word": " transition",
                    "probability": 0.91455078125
                },
                {
                    "start": 3051.82,
                    "end": 3052.44,
                    "word": " dynamics",
                    "probability": 0.74609375
                },
                {
                    "start": 3052.44,
                    "end": 3052.74,
                    "word": ",",
                    "probability": 0.818359375
                },
                {
                    "start": 3052.74,
                    "end": 3052.9,
                    "word": " this",
                    "probability": 0.75
                },
                {
                    "start": 3052.9,
                    "end": 3053.0,
                    "word": " is",
                    "probability": 0.90380859375
                },
                {
                    "start": 3053.0,
                    "end": 3053.14,
                    "word": " how",
                    "probability": 0.904296875
                },
                {
                    "start": 3053.14,
                    "end": 3053.3,
                    "word": " we've",
                    "probability": 0.77197265625
                },
                {
                    "start": 3053.3,
                    "end": 3053.54,
                    "word": " defined",
                    "probability": 0.76025390625
                },
                {
                    "start": 3053.54,
                    "end": 3053.8,
                    "word": " it",
                    "probability": 0.88330078125
                },
                {
                    "start": 3053.8,
                    "end": 3054.12,
                    "word": ".",
                    "probability": 0.95166015625
                },
                {
                    "start": 3054.12,
                    "end": 3054.38,
                    "word": " Let",
                    "probability": 0.8486328125
                },
                {
                    "start": 3054.7,
                    "end": 3055.16,
                    "word": " a particular",
                    "probability": 0.8349609375
                },
                {
                    "start": 3055.16,
                    "end": 3055.72,
                    "word": " state",
                    "probability": 0.85107421875
                },
                {
                    "start": 3055.72,
                    "end": 3058.06,
                    "word": " be",
                    "probability": 0.75927734375
                }
            ]
        },
        {
            "id": 452,
            "seek": 306427,
            "start": 3058.06,
            "end": 3070.4,
            "text": " A0 and A1. If we are, I'm sorry, actions. Actions A0 and A1. If the agent is in state",
            "tokens": [
                50780,
                316,
                15,
                293,
                316,
                16,
                13,
                759,
                321,
                366,
                11,
                286,
                478,
                2597,
                11,
                5909,
                13,
                3251,
                626,
                316,
                15,
                293,
                316,
                16,
                13,
                759,
                264,
                9461,
                307,
                294,
                1785,
                51388
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23012610067401015,
            "compression_ratio": 1.3676470588235294,
            "no_speech_prob": 0.01666259765625,
            "words": [
                {
                    "start": 3058.06,
                    "end": 3058.72,
                    "word": " A0",
                    "probability": 0.583251953125
                },
                {
                    "start": 3058.72,
                    "end": 3059.38,
                    "word": " and",
                    "probability": 0.85693359375
                },
                {
                    "start": 3059.38,
                    "end": 3061.26,
                    "word": " A1",
                    "probability": 0.99169921875
                },
                {
                    "start": 3061.26,
                    "end": 3061.96,
                    "word": ".",
                    "probability": 0.499267578125
                },
                {
                    "start": 3062.76,
                    "end": 3063.46,
                    "word": " If",
                    "probability": 0.54638671875
                },
                {
                    "start": 3063.46,
                    "end": 3063.96,
                    "word": " we",
                    "probability": 0.2069091796875
                },
                {
                    "start": 3063.96,
                    "end": 3064.22,
                    "word": " are",
                    "probability": 0.8681640625
                },
                {
                    "start": 3064.22,
                    "end": 3064.72,
                    "word": ",",
                    "probability": 0.2125244140625
                },
                {
                    "start": 3064.72,
                    "end": 3065.08,
                    "word": " I'm",
                    "probability": 0.871826171875
                },
                {
                    "start": 3065.08,
                    "end": 3065.22,
                    "word": " sorry",
                    "probability": 0.751953125
                },
                {
                    "start": 3065.22,
                    "end": 3065.36,
                    "word": ",",
                    "probability": 0.7578125
                },
                {
                    "start": 3065.36,
                    "end": 3065.7,
                    "word": " actions",
                    "probability": 0.8193359375
                },
                {
                    "start": 3065.7,
                    "end": 3065.74,
                    "word": ".",
                    "probability": 0.30615234375
                },
                {
                    "start": 3065.74,
                    "end": 3066.16,
                    "word": " Actions",
                    "probability": 0.88671875
                },
                {
                    "start": 3066.16,
                    "end": 3066.92,
                    "word": " A0",
                    "probability": 0.859130859375
                },
                {
                    "start": 3066.92,
                    "end": 3067.18,
                    "word": " and",
                    "probability": 0.89306640625
                },
                {
                    "start": 3067.18,
                    "end": 3067.62,
                    "word": " A1",
                    "probability": 0.996826171875
                },
                {
                    "start": 3067.62,
                    "end": 3068.32,
                    "word": ".",
                    "probability": 0.76708984375
                },
                {
                    "start": 3068.56,
                    "end": 3069.08,
                    "word": " If",
                    "probability": 0.8798828125
                },
                {
                    "start": 3069.08,
                    "end": 3069.48,
                    "word": " the",
                    "probability": 0.80126953125
                },
                {
                    "start": 3069.48,
                    "end": 3069.68,
                    "word": " agent",
                    "probability": 0.80810546875
                },
                {
                    "start": 3069.68,
                    "end": 3069.84,
                    "word": " is",
                    "probability": 0.89794921875
                },
                {
                    "start": 3069.84,
                    "end": 3069.98,
                    "word": " in",
                    "probability": 0.89013671875
                },
                {
                    "start": 3069.98,
                    "end": 3070.4,
                    "word": " state",
                    "probability": 0.84765625
                }
            ]
        },
        {
            "id": 453,
            "seek": 309133,
            "start": 3071.36,
            "end": 3082.22,
            "text": " I and performs action I modulo two, if I is even, this is going to be action zero. If I is odd,",
            "tokens": [
                50392,
                286,
                293,
                26213,
                3069,
                286,
                1072,
                13455,
                732,
                11,
                498,
                286,
                307,
                754,
                11,
                341,
                307,
                516,
                281,
                312,
                3069,
                4018,
                13,
                759,
                286,
                307,
                7401,
                11,
                50948
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2382812501862645,
            "compression_ratio": 1.7721518987341771,
            "no_speech_prob": 0.268798828125,
            "words": [
                {
                    "start": 3071.36,
                    "end": 3071.96,
                    "word": " I",
                    "probability": 0.03790283203125
                },
                {
                    "start": 3071.96,
                    "end": 3072.3,
                    "word": " and",
                    "probability": 0.3505859375
                },
                {
                    "start": 3072.3,
                    "end": 3072.86,
                    "word": " performs",
                    "probability": 0.7783203125
                },
                {
                    "start": 3072.86,
                    "end": 3073.84,
                    "word": " action",
                    "probability": 0.78173828125
                },
                {
                    "start": 3073.84,
                    "end": 3074.98,
                    "word": " I",
                    "probability": 0.658203125
                },
                {
                    "start": 3074.98,
                    "end": 3075.64,
                    "word": " modulo",
                    "probability": 0.797607421875
                },
                {
                    "start": 3075.64,
                    "end": 3076.0,
                    "word": " two",
                    "probability": 0.327880859375
                },
                {
                    "start": 3076.0,
                    "end": 3076.68,
                    "word": ",",
                    "probability": 0.355224609375
                },
                {
                    "start": 3076.68,
                    "end": 3077.22,
                    "word": " if",
                    "probability": 0.79833984375
                },
                {
                    "start": 3077.22,
                    "end": 3077.58,
                    "word": " I",
                    "probability": 0.93115234375
                },
                {
                    "start": 3077.58,
                    "end": 3078.04,
                    "word": " is",
                    "probability": 0.896484375
                },
                {
                    "start": 3078.04,
                    "end": 3078.6,
                    "word": " even",
                    "probability": 0.7802734375
                },
                {
                    "start": 3078.6,
                    "end": 3078.82,
                    "word": ",",
                    "probability": 0.591796875
                },
                {
                    "start": 3078.82,
                    "end": 3078.98,
                    "word": " this",
                    "probability": 0.84912109375
                },
                {
                    "start": 3078.98,
                    "end": 3079.14,
                    "word": " is",
                    "probability": 0.89501953125
                },
                {
                    "start": 3079.14,
                    "end": 3079.32,
                    "word": " going",
                    "probability": 0.853515625
                },
                {
                    "start": 3079.32,
                    "end": 3079.44,
                    "word": " to",
                    "probability": 0.89794921875
                },
                {
                    "start": 3079.44,
                    "end": 3079.54,
                    "word": " be",
                    "probability": 0.8857421875
                },
                {
                    "start": 3079.54,
                    "end": 3079.74,
                    "word": " action",
                    "probability": 0.787109375
                },
                {
                    "start": 3079.74,
                    "end": 3080.02,
                    "word": " zero",
                    "probability": 0.8076171875
                },
                {
                    "start": 3080.02,
                    "end": 3080.22,
                    "word": ".",
                    "probability": 0.43505859375
                },
                {
                    "start": 3080.22,
                    "end": 3080.38,
                    "word": " If",
                    "probability": 0.87109375
                },
                {
                    "start": 3080.38,
                    "end": 3081.06,
                    "word": " I",
                    "probability": 0.95263671875
                },
                {
                    "start": 3081.06,
                    "end": 3081.42,
                    "word": " is",
                    "probability": 0.908203125
                },
                {
                    "start": 3081.42,
                    "end": 3082.14,
                    "word": " odd",
                    "probability": 0.80126953125
                },
                {
                    "start": 3082.14,
                    "end": 3082.22,
                    "word": ",",
                    "probability": 0.9013671875
                }
            ]
        },
        {
            "id": 454,
            "seek": 309133,
            "start": 3082.22,
            "end": 3090.52,
            "text": " this is going to be action one. Then the next state is going to be state I star or I modulo",
            "tokens": [
                50948,
                341,
                307,
                516,
                281,
                312,
                3069,
                472,
                13,
                1396,
                264,
                958,
                1785,
                307,
                516,
                281,
                312,
                1785,
                286,
                3543,
                420,
                286,
                1072,
                13455,
                51388
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2382812501862645,
            "compression_ratio": 1.7721518987341771,
            "no_speech_prob": 0.268798828125,
            "words": [
                {
                    "start": 3082.22,
                    "end": 3082.38,
                    "word": " this",
                    "probability": 0.8740234375
                },
                {
                    "start": 3082.38,
                    "end": 3082.52,
                    "word": " is",
                    "probability": 0.9013671875
                },
                {
                    "start": 3082.52,
                    "end": 3082.66,
                    "word": " going",
                    "probability": 0.85791015625
                },
                {
                    "start": 3082.66,
                    "end": 3082.78,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 3082.78,
                    "end": 3082.88,
                    "word": " be",
                    "probability": 0.89404296875
                },
                {
                    "start": 3082.88,
                    "end": 3083.14,
                    "word": " action",
                    "probability": 0.78466796875
                },
                {
                    "start": 3083.14,
                    "end": 3083.56,
                    "word": " one",
                    "probability": 0.8046875
                },
                {
                    "start": 3083.56,
                    "end": 3084.04,
                    "word": ".",
                    "probability": 0.814453125
                },
                {
                    "start": 3084.04,
                    "end": 3084.5,
                    "word": " Then",
                    "probability": 0.7900390625
                },
                {
                    "start": 3084.5,
                    "end": 3084.76,
                    "word": " the",
                    "probability": 0.68603515625
                },
                {
                    "start": 3084.76,
                    "end": 3085.02,
                    "word": " next",
                    "probability": 0.85546875
                },
                {
                    "start": 3085.02,
                    "end": 3085.32,
                    "word": " state",
                    "probability": 0.86376953125
                },
                {
                    "start": 3085.32,
                    "end": 3085.58,
                    "word": " is",
                    "probability": 0.89208984375
                },
                {
                    "start": 3085.58,
                    "end": 3085.86,
                    "word": " going",
                    "probability": 0.85888671875
                },
                {
                    "start": 3085.86,
                    "end": 3086.08,
                    "word": " to",
                    "probability": 0.89794921875
                },
                {
                    "start": 3086.08,
                    "end": 3086.82,
                    "word": " be",
                    "probability": 0.89013671875
                },
                {
                    "start": 3086.82,
                    "end": 3087.38,
                    "word": " state",
                    "probability": 0.67529296875
                },
                {
                    "start": 3087.38,
                    "end": 3088.08,
                    "word": " I",
                    "probability": 0.89501953125
                },
                {
                    "start": 3088.08,
                    "end": 3088.5,
                    "word": " star",
                    "probability": 0.477294921875
                },
                {
                    "start": 3088.5,
                    "end": 3088.82,
                    "word": " or",
                    "probability": 0.406494140625
                },
                {
                    "start": 3088.82,
                    "end": 3089.92,
                    "word": " I",
                    "probability": 0.7236328125
                },
                {
                    "start": 3089.92,
                    "end": 3090.52,
                    "word": " modulo",
                    "probability": 0.851318359375
                }
            ]
        },
        {
            "id": 455,
            "seek": 309133,
            "start": 3090.52,
            "end": 3097.46,
            "text": " four plus one. So, and it receives a reward of one. If the agent performs any other actions,",
            "tokens": [
                51388,
                1451,
                1804,
                472,
                13,
                407,
                11,
                293,
                309,
                20717,
                257,
                7782,
                295,
                472,
                13,
                759,
                264,
                9461,
                26213,
                604,
                661,
                5909,
                11,
                51712
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2382812501862645,
            "compression_ratio": 1.7721518987341771,
            "no_speech_prob": 0.268798828125,
            "words": [
                {
                    "start": 3090.52,
                    "end": 3091.18,
                    "word": " four",
                    "probability": 0.8505859375
                },
                {
                    "start": 3091.18,
                    "end": 3091.54,
                    "word": " plus",
                    "probability": 0.8369140625
                },
                {
                    "start": 3091.54,
                    "end": 3091.94,
                    "word": " one",
                    "probability": 0.8125
                },
                {
                    "start": 3091.94,
                    "end": 3092.2,
                    "word": ".",
                    "probability": 0.60986328125
                },
                {
                    "start": 3092.2,
                    "end": 3092.38,
                    "word": " So",
                    "probability": 0.64892578125
                },
                {
                    "start": 3092.38,
                    "end": 3093.14,
                    "word": ",",
                    "probability": 0.65087890625
                },
                {
                    "start": 3093.14,
                    "end": 3093.9,
                    "word": " and",
                    "probability": 0.859375
                },
                {
                    "start": 3093.9,
                    "end": 3094.06,
                    "word": " it",
                    "probability": 0.88427734375
                },
                {
                    "start": 3094.06,
                    "end": 3094.34,
                    "word": " receives",
                    "probability": 0.78173828125
                },
                {
                    "start": 3094.46,
                    "end": 3094.68,
                    "word": " a reward",
                    "probability": 0.89892578125
                },
                {
                    "start": 3094.68,
                    "end": 3094.8,
                    "word": " of",
                    "probability": 0.853515625
                },
                {
                    "start": 3094.8,
                    "end": 3095.0,
                    "word": " one",
                    "probability": 0.81591796875
                },
                {
                    "start": 3095.0,
                    "end": 3095.18,
                    "word": ".",
                    "probability": 0.78173828125
                },
                {
                    "start": 3095.18,
                    "end": 3095.48,
                    "word": " If",
                    "probability": 0.88916015625
                },
                {
                    "start": 3095.48,
                    "end": 3095.78,
                    "word": " the",
                    "probability": 0.8115234375
                },
                {
                    "start": 3095.78,
                    "end": 3095.92,
                    "word": " agent",
                    "probability": 0.849609375
                },
                {
                    "start": 3095.92,
                    "end": 3096.22,
                    "word": " performs",
                    "probability": 0.73974609375
                },
                {
                    "start": 3096.22,
                    "end": 3096.56,
                    "word": " any",
                    "probability": 0.80712890625
                },
                {
                    "start": 3096.56,
                    "end": 3096.82,
                    "word": " other",
                    "probability": 0.78369140625
                },
                {
                    "start": 3096.82,
                    "end": 3097.32,
                    "word": " actions",
                    "probability": 0.873046875
                },
                {
                    "start": 3097.32,
                    "end": 3097.46,
                    "word": ",",
                    "probability": 0.611328125
                }
            ]
        },
        {
            "id": 456,
            "seek": 312133,
            "start": 3097.46,
            "end": 3103.08,
            "text": " instead of going from one to two, it performs any other action, it will immediately go to the",
            "tokens": [
                50368,
                2602,
                295,
                516,
                490,
                472,
                281,
                732,
                11,
                309,
                26213,
                604,
                661,
                3069,
                11,
                309,
                486,
                4258,
                352,
                281,
                264,
                50636
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25712720056374866,
            "compression_ratio": 1.8188976377952757,
            "no_speech_prob": 0.0009698867797851562,
            "words": [
                {
                    "start": 3097.46,
                    "end": 3097.78,
                    "word": " instead",
                    "probability": 0.0186614990234375
                },
                {
                    "start": 3097.78,
                    "end": 3098.04,
                    "word": " of",
                    "probability": 0.88330078125
                },
                {
                    "start": 3098.04,
                    "end": 3098.2,
                    "word": " going",
                    "probability": 0.828125
                },
                {
                    "start": 3098.2,
                    "end": 3098.4,
                    "word": " from",
                    "probability": 0.79248046875
                },
                {
                    "start": 3098.4,
                    "end": 3098.64,
                    "word": " one",
                    "probability": 0.345703125
                },
                {
                    "start": 3098.64,
                    "end": 3098.8,
                    "word": " to",
                    "probability": 0.89453125
                },
                {
                    "start": 3098.8,
                    "end": 3099.14,
                    "word": " two",
                    "probability": 0.8154296875
                },
                {
                    "start": 3099.14,
                    "end": 3099.78,
                    "word": ",",
                    "probability": 0.6416015625
                },
                {
                    "start": 3099.78,
                    "end": 3099.96,
                    "word": " it",
                    "probability": 0.880859375
                },
                {
                    "start": 3099.96,
                    "end": 3100.46,
                    "word": " performs",
                    "probability": 0.767578125
                },
                {
                    "start": 3100.46,
                    "end": 3100.86,
                    "word": " any",
                    "probability": 0.8056640625
                },
                {
                    "start": 3100.86,
                    "end": 3101.14,
                    "word": " other",
                    "probability": 0.7802734375
                },
                {
                    "start": 3101.14,
                    "end": 3101.56,
                    "word": " action",
                    "probability": 0.81201171875
                },
                {
                    "start": 3101.56,
                    "end": 3101.76,
                    "word": ",",
                    "probability": 0.60546875
                },
                {
                    "start": 3101.76,
                    "end": 3101.78,
                    "word": " it",
                    "probability": 0.8349609375
                },
                {
                    "start": 3101.78,
                    "end": 3102.02,
                    "word": " will",
                    "probability": 0.7255859375
                },
                {
                    "start": 3102.02,
                    "end": 3102.64,
                    "word": " immediately",
                    "probability": 0.681640625
                },
                {
                    "start": 3102.64,
                    "end": 3102.86,
                    "word": " go",
                    "probability": 0.89599609375
                },
                {
                    "start": 3102.86,
                    "end": 3102.96,
                    "word": " to",
                    "probability": 0.8935546875
                },
                {
                    "start": 3102.96,
                    "end": 3103.08,
                    "word": " the",
                    "probability": 0.8076171875
                }
            ]
        },
        {
            "id": 457,
            "seek": 312133,
            "start": 3103.08,
            "end": 3107.56,
            "text": " terminal state and receives a reward of zero. So, the optimal trajectory is state one to two to three",
            "tokens": [
                50636,
                14709,
                1785,
                293,
                20717,
                257,
                7782,
                295,
                4018,
                13,
                407,
                11,
                264,
                16252,
                21512,
                307,
                1785,
                472,
                281,
                732,
                281,
                1045,
                50868
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25712720056374866,
            "compression_ratio": 1.8188976377952757,
            "no_speech_prob": 0.0009698867797851562,
            "words": [
                {
                    "start": 3103.08,
                    "end": 3103.32,
                    "word": " terminal",
                    "probability": 0.71728515625
                },
                {
                    "start": 3103.32,
                    "end": 3103.72,
                    "word": " state",
                    "probability": 0.8505859375
                },
                {
                    "start": 3103.72,
                    "end": 3103.88,
                    "word": " and",
                    "probability": 0.845703125
                },
                {
                    "start": 3103.88,
                    "end": 3104.1,
                    "word": " receives",
                    "probability": 0.62890625
                },
                {
                    "start": 3104.22,
                    "end": 3104.48,
                    "word": " a reward",
                    "probability": 0.84912109375
                },
                {
                    "start": 3104.48,
                    "end": 3104.6,
                    "word": " of",
                    "probability": 0.61962890625
                },
                {
                    "start": 3104.6,
                    "end": 3104.84,
                    "word": " zero",
                    "probability": 0.78759765625
                },
                {
                    "start": 3104.84,
                    "end": 3104.98,
                    "word": ".",
                    "probability": 0.8583984375
                },
                {
                    "start": 3104.98,
                    "end": 3105.06,
                    "word": " So",
                    "probability": 0.58349609375
                },
                {
                    "start": 3105.06,
                    "end": 3105.12,
                    "word": ",",
                    "probability": 0.322509765625
                },
                {
                    "start": 3105.12,
                    "end": 3105.2,
                    "word": " the",
                    "probability": 0.8037109375
                },
                {
                    "start": 3105.2,
                    "end": 3105.5,
                    "word": " optimal",
                    "probability": 0.74951171875
                },
                {
                    "start": 3105.5,
                    "end": 3105.92,
                    "word": " trajectory",
                    "probability": 0.771484375
                },
                {
                    "start": 3105.92,
                    "end": 3106.14,
                    "word": " is",
                    "probability": 0.85498046875
                },
                {
                    "start": 3106.14,
                    "end": 3106.4,
                    "word": " state",
                    "probability": 0.8203125
                },
                {
                    "start": 3106.4,
                    "end": 3106.76,
                    "word": " one",
                    "probability": 0.6904296875
                },
                {
                    "start": 3106.76,
                    "end": 3106.9,
                    "word": " to",
                    "probability": 0.734375
                },
                {
                    "start": 3106.9,
                    "end": 3107.16,
                    "word": " two",
                    "probability": 0.78271484375
                },
                {
                    "start": 3107.16,
                    "end": 3107.32,
                    "word": " to",
                    "probability": 0.6748046875
                },
                {
                    "start": 3107.32,
                    "end": 3107.56,
                    "word": " three",
                    "probability": 0.84716796875
                }
            ]
        },
        {
            "id": 458,
            "seek": 312133,
            "start": 3107.56,
            "end": 3113.7,
            "text": " to four back to state one and so on. All right. Let's see how it works. Let's look at the test",
            "tokens": [
                50868,
                281,
                1451,
                646,
                281,
                1785,
                472,
                293,
                370,
                322,
                13,
                1057,
                558,
                13,
                961,
                311,
                536,
                577,
                309,
                1985,
                13,
                961,
                311,
                574,
                412,
                264,
                1500,
                51168
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25712720056374866,
            "compression_ratio": 1.8188976377952757,
            "no_speech_prob": 0.0009698867797851562,
            "words": [
                {
                    "start": 3107.56,
                    "end": 3107.74,
                    "word": " to",
                    "probability": 0.89013671875
                },
                {
                    "start": 3107.74,
                    "end": 3108.06,
                    "word": " four",
                    "probability": 0.86767578125
                },
                {
                    "start": 3108.06,
                    "end": 3108.68,
                    "word": " back",
                    "probability": 0.398681640625
                },
                {
                    "start": 3108.68,
                    "end": 3108.88,
                    "word": " to",
                    "probability": 0.90185546875
                },
                {
                    "start": 3108.88,
                    "end": 3109.18,
                    "word": " state",
                    "probability": 0.87353515625
                },
                {
                    "start": 3109.18,
                    "end": 3109.44,
                    "word": " one",
                    "probability": 0.802734375
                },
                {
                    "start": 3109.44,
                    "end": 3109.64,
                    "word": " and",
                    "probability": 0.68310546875
                },
                {
                    "start": 3109.64,
                    "end": 3109.86,
                    "word": " so",
                    "probability": 0.8955078125
                },
                {
                    "start": 3109.86,
                    "end": 3110.1,
                    "word": " on",
                    "probability": 0.9052734375
                },
                {
                    "start": 3110.1,
                    "end": 3110.58,
                    "word": ".",
                    "probability": 0.986328125
                },
                {
                    "start": 3110.76,
                    "end": 3110.9,
                    "word": " All",
                    "probability": 0.58740234375
                },
                {
                    "start": 3110.9,
                    "end": 3111.02,
                    "word": " right",
                    "probability": 0.8779296875
                },
                {
                    "start": 3111.02,
                    "end": 3111.06,
                    "word": ".",
                    "probability": 0.35400390625
                },
                {
                    "start": 3111.06,
                    "end": 3111.28,
                    "word": " Let's",
                    "probability": 0.940185546875
                },
                {
                    "start": 3111.28,
                    "end": 3111.46,
                    "word": " see",
                    "probability": 0.81494140625
                },
                {
                    "start": 3111.46,
                    "end": 3111.8,
                    "word": " how",
                    "probability": 0.89990234375
                },
                {
                    "start": 3111.8,
                    "end": 3111.94,
                    "word": " it",
                    "probability": 0.9013671875
                },
                {
                    "start": 3111.94,
                    "end": 3112.24,
                    "word": " works",
                    "probability": 0.72607421875
                },
                {
                    "start": 3112.24,
                    "end": 3112.72,
                    "word": ".",
                    "probability": 0.990234375
                },
                {
                    "start": 3113.08,
                    "end": 3113.26,
                    "word": " Let's",
                    "probability": 0.876708984375
                },
                {
                    "start": 3113.26,
                    "end": 3113.34,
                    "word": " look",
                    "probability": 0.86669921875
                },
                {
                    "start": 3113.34,
                    "end": 3113.4,
                    "word": " at",
                    "probability": 0.9013671875
                },
                {
                    "start": 3113.4,
                    "end": 3113.5,
                    "word": " the",
                    "probability": 0.79345703125
                },
                {
                    "start": 3113.5,
                    "end": 3113.7,
                    "word": " test",
                    "probability": 0.75830078125
                }
            ]
        },
        {
            "id": 459,
            "seek": 312133,
            "start": 3113.7,
            "end": 3118.6,
            "text": " time performance comparison of watermarked and nominal, non-watermarked policies. The",
            "tokens": [
                51168,
                565,
                3389,
                9660,
                295,
                1281,
                5638,
                292,
                293,
                41641,
                11,
                2107,
                12,
                8002,
                5638,
                292,
                7657,
                13,
                440,
                51420
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25712720056374866,
            "compression_ratio": 1.8188976377952757,
            "no_speech_prob": 0.0009698867797851562,
            "words": [
                {
                    "start": 3113.7,
                    "end": 3113.86,
                    "word": " time",
                    "probability": 0.5751953125
                },
                {
                    "start": 3113.86,
                    "end": 3114.28,
                    "word": " performance",
                    "probability": 0.77197265625
                },
                {
                    "start": 3114.28,
                    "end": 3114.66,
                    "word": " comparison",
                    "probability": 0.72802734375
                },
                {
                    "start": 3114.66,
                    "end": 3114.88,
                    "word": " of",
                    "probability": 0.87451171875
                },
                {
                    "start": 3114.88,
                    "end": 3115.38,
                    "word": " watermarked",
                    "probability": 0.6991373697916666
                },
                {
                    "start": 3115.38,
                    "end": 3115.48,
                    "word": " and",
                    "probability": 0.89208984375
                },
                {
                    "start": 3115.48,
                    "end": 3115.9,
                    "word": " nominal",
                    "probability": 0.802734375
                },
                {
                    "start": 3115.9,
                    "end": 3116.2,
                    "word": ",",
                    "probability": 0.72900390625
                },
                {
                    "start": 3116.2,
                    "end": 3116.46,
                    "word": " non",
                    "probability": 0.89697265625
                },
                {
                    "start": 3116.46,
                    "end": 3117.32,
                    "word": "-watermarked",
                    "probability": 0.8441162109375
                },
                {
                    "start": 3117.32,
                    "end": 3117.8,
                    "word": " policies",
                    "probability": 0.7744140625
                },
                {
                    "start": 3117.8,
                    "end": 3118.28,
                    "word": ".",
                    "probability": 0.97607421875
                },
                {
                    "start": 3118.4,
                    "end": 3118.6,
                    "word": " The",
                    "probability": 0.81103515625
                }
            ]
        },
        {
            "id": 460,
            "seek": 312133,
            "start": 3118.6,
            "end": 3126.62,
            "text": " watermarked policy performs exactly as well as unwatermarked policies. It reaches the",
            "tokens": [
                51420,
                1281,
                5638,
                292,
                3897,
                26213,
                2293,
                382,
                731,
                382,
                14853,
                771,
                5638,
                292,
                7657,
                13,
                467,
                14235,
                264,
                51812
            ],
            "temperature": 0.0,
            "avg_logprob": -0.25712720056374866,
            "compression_ratio": 1.8188976377952757,
            "no_speech_prob": 0.0009698867797851562,
            "words": [
                {
                    "start": 3118.6,
                    "end": 3119.42,
                    "word": " watermarked",
                    "probability": 0.8370768229166666
                },
                {
                    "start": 3119.42,
                    "end": 3119.82,
                    "word": " policy",
                    "probability": 0.80029296875
                },
                {
                    "start": 3119.82,
                    "end": 3120.7,
                    "word": " performs",
                    "probability": 0.72998046875
                },
                {
                    "start": 3120.7,
                    "end": 3121.82,
                    "word": " exactly",
                    "probability": 0.76513671875
                },
                {
                    "start": 3121.82,
                    "end": 3122.28,
                    "word": " as",
                    "probability": 0.90087890625
                },
                {
                    "start": 3122.28,
                    "end": 3122.62,
                    "word": " well",
                    "probability": 0.876953125
                },
                {
                    "start": 3122.62,
                    "end": 3123.46,
                    "word": " as",
                    "probability": 0.892578125
                },
                {
                    "start": 3123.46,
                    "end": 3125.16,
                    "word": " unwatermarked",
                    "probability": 0.787353515625
                },
                {
                    "start": 3125.16,
                    "end": 3125.56,
                    "word": " policies",
                    "probability": 0.84228515625
                },
                {
                    "start": 3125.56,
                    "end": 3125.76,
                    "word": ".",
                    "probability": 0.837890625
                },
                {
                    "start": 3125.76,
                    "end": 3125.92,
                    "word": " It",
                    "probability": 0.90771484375
                },
                {
                    "start": 3125.92,
                    "end": 3126.22,
                    "word": " reaches",
                    "probability": 0.72216796875
                },
                {
                    "start": 3126.22,
                    "end": 3126.62,
                    "word": " the",
                    "probability": 0.81005859375
                }
            ]
        },
        {
            "id": 461,
            "seek": 314649,
            "start": 3127.5,
            "end": 3137.84,
            "text": " optimal or best performance of 500. And when it's applied to the watermark environment,",
            "tokens": [
                50376,
                16252,
                420,
                1151,
                3389,
                295,
                5923,
                13,
                400,
                562,
                309,
                311,
                6456,
                281,
                264,
                1281,
                5638,
                2823,
                11,
                50876
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20348557692307692,
            "compression_ratio": 1.5344827586206897,
            "no_speech_prob": 0.00051116943359375,
            "words": [
                {
                    "start": 3127.5,
                    "end": 3128.22,
                    "word": " optimal",
                    "probability": 0.60546875
                },
                {
                    "start": 3128.22,
                    "end": 3128.82,
                    "word": " or",
                    "probability": 0.765625
                },
                {
                    "start": 3128.82,
                    "end": 3130.24,
                    "word": " best",
                    "probability": 0.783203125
                },
                {
                    "start": 3130.24,
                    "end": 3130.62,
                    "word": " performance",
                    "probability": 0.8193359375
                },
                {
                    "start": 3130.62,
                    "end": 3130.88,
                    "word": " of",
                    "probability": 0.8955078125
                },
                {
                    "start": 3130.88,
                    "end": 3132.12,
                    "word": " 500",
                    "probability": 0.74169921875
                },
                {
                    "start": 3132.12,
                    "end": 3132.84,
                    "word": ".",
                    "probability": 0.477783203125
                },
                {
                    "start": 3133.38,
                    "end": 3134.0,
                    "word": " And",
                    "probability": 0.759765625
                },
                {
                    "start": 3134.0,
                    "end": 3135.52,
                    "word": " when",
                    "probability": 0.8076171875
                },
                {
                    "start": 3135.52,
                    "end": 3135.74,
                    "word": " it's",
                    "probability": 0.942138671875
                },
                {
                    "start": 3135.74,
                    "end": 3136.12,
                    "word": " applied",
                    "probability": 0.68798828125
                },
                {
                    "start": 3136.12,
                    "end": 3136.34,
                    "word": " to",
                    "probability": 0.90625
                },
                {
                    "start": 3136.34,
                    "end": 3136.64,
                    "word": " the",
                    "probability": 0.81201171875
                },
                {
                    "start": 3136.64,
                    "end": 3137.3,
                    "word": " watermark",
                    "probability": 0.896484375
                },
                {
                    "start": 3137.3,
                    "end": 3137.76,
                    "word": " environment",
                    "probability": 0.7138671875
                },
                {
                    "start": 3137.76,
                    "end": 3137.84,
                    "word": ",",
                    "probability": 0.9189453125
                }
            ]
        },
        {
            "id": 462,
            "seek": 314649,
            "start": 3137.84,
            "end": 3143.8,
            "text": " the second environment, it also performs optimally. It gets the maximum reward possible.",
            "tokens": [
                50876,
                264,
                1150,
                2823,
                11,
                309,
                611,
                26213,
                5028,
                379,
                13,
                467,
                2170,
                264,
                6674,
                7782,
                1944,
                13,
                51144
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20348557692307692,
            "compression_ratio": 1.5344827586206897,
            "no_speech_prob": 0.00051116943359375,
            "words": [
                {
                    "start": 3137.84,
                    "end": 3137.94,
                    "word": " the",
                    "probability": 0.771484375
                },
                {
                    "start": 3137.94,
                    "end": 3138.16,
                    "word": " second",
                    "probability": 0.7998046875
                },
                {
                    "start": 3138.16,
                    "end": 3138.6,
                    "word": " environment",
                    "probability": 0.7529296875
                },
                {
                    "start": 3138.6,
                    "end": 3138.78,
                    "word": ",",
                    "probability": 0.93212890625
                },
                {
                    "start": 3138.78,
                    "end": 3138.84,
                    "word": " it",
                    "probability": 0.90185546875
                },
                {
                    "start": 3138.84,
                    "end": 3139.2,
                    "word": " also",
                    "probability": 0.83251953125
                },
                {
                    "start": 3139.2,
                    "end": 3139.76,
                    "word": " performs",
                    "probability": 0.7265625
                },
                {
                    "start": 3139.76,
                    "end": 3141.46,
                    "word": " optimally",
                    "probability": 0.80517578125
                },
                {
                    "start": 3141.46,
                    "end": 3141.62,
                    "word": ".",
                    "probability": 0.296142578125
                },
                {
                    "start": 3141.62,
                    "end": 3141.62,
                    "word": " It",
                    "probability": 0.875
                },
                {
                    "start": 3141.62,
                    "end": 3141.78,
                    "word": " gets",
                    "probability": 0.78076171875
                },
                {
                    "start": 3141.78,
                    "end": 3141.92,
                    "word": " the",
                    "probability": 0.67236328125
                },
                {
                    "start": 3141.92,
                    "end": 3142.24,
                    "word": " maximum",
                    "probability": 0.76513671875
                },
                {
                    "start": 3142.24,
                    "end": 3142.6,
                    "word": " reward",
                    "probability": 0.8037109375
                },
                {
                    "start": 3142.6,
                    "end": 3143.08,
                    "word": " possible",
                    "probability": 0.83349609375
                },
                {
                    "start": 3143.08,
                    "end": 3143.8,
                    "word": ".",
                    "probability": 0.99267578125
                }
            ]
        },
        {
            "id": 463,
            "seek": 314649,
            "start": 3144.18,
            "end": 3152.62,
            "text": " However, you can see that if we try to apply the non-watermarked policies to the watermark",
            "tokens": [
                51212,
                2908,
                11,
                291,
                393,
                536,
                300,
                498,
                321,
                853,
                281,
                3079,
                264,
                2107,
                12,
                8002,
                5638,
                292,
                7657,
                281,
                264,
                1281,
                5638,
                51620
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20348557692307692,
            "compression_ratio": 1.5344827586206897,
            "no_speech_prob": 0.00051116943359375,
            "words": [
                {
                    "start": 3144.18,
                    "end": 3144.9,
                    "word": " However",
                    "probability": 0.822265625
                },
                {
                    "start": 3144.9,
                    "end": 3145.08,
                    "word": ",",
                    "probability": 0.97314453125
                },
                {
                    "start": 3145.08,
                    "end": 3145.34,
                    "word": " you",
                    "probability": 0.88818359375
                },
                {
                    "start": 3145.34,
                    "end": 3145.46,
                    "word": " can",
                    "probability": 0.8955078125
                },
                {
                    "start": 3145.46,
                    "end": 3145.68,
                    "word": " see",
                    "probability": 0.82275390625
                },
                {
                    "start": 3145.68,
                    "end": 3145.92,
                    "word": " that",
                    "probability": 0.8779296875
                },
                {
                    "start": 3145.92,
                    "end": 3146.24,
                    "word": " if",
                    "probability": 0.88671875
                },
                {
                    "start": 3146.24,
                    "end": 3146.6,
                    "word": " we",
                    "probability": 0.89794921875
                },
                {
                    "start": 3146.6,
                    "end": 3147.06,
                    "word": " try",
                    "probability": 0.794921875
                },
                {
                    "start": 3147.06,
                    "end": 3148.94,
                    "word": " to",
                    "probability": 0.732421875
                },
                {
                    "start": 3148.94,
                    "end": 3149.24,
                    "word": " apply",
                    "probability": 0.849609375
                },
                {
                    "start": 3149.24,
                    "end": 3149.46,
                    "word": " the",
                    "probability": 0.7783203125
                },
                {
                    "start": 3149.46,
                    "end": 3149.78,
                    "word": " non",
                    "probability": 0.90234375
                },
                {
                    "start": 3149.78,
                    "end": 3150.68,
                    "word": "-watermarked",
                    "probability": 0.7784423828125
                },
                {
                    "start": 3150.68,
                    "end": 3151.68,
                    "word": " policies",
                    "probability": 0.8408203125
                },
                {
                    "start": 3151.68,
                    "end": 3151.98,
                    "word": " to",
                    "probability": 0.90087890625
                },
                {
                    "start": 3151.98,
                    "end": 3152.16,
                    "word": " the",
                    "probability": 0.810546875
                },
                {
                    "start": 3152.16,
                    "end": 3152.62,
                    "word": " watermark",
                    "probability": 0.94287109375
                }
            ]
        },
        {
            "id": 464,
            "seek": 317413,
            "start": 3152.62,
            "end": 3160.56,
            "text": " environment, we'll see very, very small values of score, the total reward. So, you can see that",
            "tokens": [
                50364,
                2823,
                11,
                321,
                603,
                536,
                588,
                11,
                588,
                1359,
                4190,
                295,
                6175,
                11,
                264,
                3217,
                7782,
                13,
                407,
                11,
                291,
                393,
                536,
                300,
                50752
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23489583995607163,
            "compression_ratio": 1.6666666666666667,
            "no_speech_prob": 0.00600433349609375,
            "words": [
                {
                    "start": 3152.62,
                    "end": 3153.0,
                    "word": " environment",
                    "probability": 0.7158203125
                },
                {
                    "start": 3153.0,
                    "end": 3153.24,
                    "word": ",",
                    "probability": 0.62744140625
                },
                {
                    "start": 3153.24,
                    "end": 3153.52,
                    "word": " we'll",
                    "probability": 0.5980224609375
                },
                {
                    "start": 3153.52,
                    "end": 3153.78,
                    "word": " see",
                    "probability": 0.81982421875
                },
                {
                    "start": 3153.78,
                    "end": 3154.44,
                    "word": " very",
                    "probability": 0.72607421875
                },
                {
                    "start": 3154.44,
                    "end": 3154.64,
                    "word": ",",
                    "probability": 0.66064453125
                },
                {
                    "start": 3154.64,
                    "end": 3154.74,
                    "word": " very",
                    "probability": 0.74267578125
                },
                {
                    "start": 3154.74,
                    "end": 3155.18,
                    "word": " small",
                    "probability": 0.8525390625
                },
                {
                    "start": 3155.18,
                    "end": 3156.18,
                    "word": " values",
                    "probability": 0.8603515625
                },
                {
                    "start": 3156.18,
                    "end": 3156.6,
                    "word": " of",
                    "probability": 0.89208984375
                },
                {
                    "start": 3156.6,
                    "end": 3157.06,
                    "word": " score",
                    "probability": 0.5146484375
                },
                {
                    "start": 3157.06,
                    "end": 3157.5,
                    "word": ",",
                    "probability": 0.51123046875
                },
                {
                    "start": 3157.5,
                    "end": 3157.8,
                    "word": " the",
                    "probability": 0.41259765625
                },
                {
                    "start": 3157.8,
                    "end": 3158.16,
                    "word": " total",
                    "probability": 0.80908203125
                },
                {
                    "start": 3158.16,
                    "end": 3158.62,
                    "word": " reward",
                    "probability": 0.810546875
                },
                {
                    "start": 3158.62,
                    "end": 3159.1,
                    "word": ".",
                    "probability": 0.916015625
                },
                {
                    "start": 3159.1,
                    "end": 3159.18,
                    "word": " So",
                    "probability": 0.794921875
                },
                {
                    "start": 3159.18,
                    "end": 3159.26,
                    "word": ",",
                    "probability": 0.29541015625
                },
                {
                    "start": 3159.26,
                    "end": 3159.28,
                    "word": " you",
                    "probability": 0.8818359375
                },
                {
                    "start": 3159.28,
                    "end": 3159.52,
                    "word": " can",
                    "probability": 0.89794921875
                },
                {
                    "start": 3159.52,
                    "end": 3160.2,
                    "word": " see",
                    "probability": 0.81982421875
                },
                {
                    "start": 3160.2,
                    "end": 3160.56,
                    "word": " that",
                    "probability": 0.8701171875
                }
            ]
        },
        {
            "id": 465,
            "seek": 317413,
            "start": 3160.56,
            "end": 3166.52,
            "text": " it's possible to determine whether a policy is authentic or not or whether a policy is an exact",
            "tokens": [
                50784,
                309,
                311,
                1944,
                281,
                6997,
                1968,
                257,
                3897,
                307,
                12466,
                420,
                406,
                420,
                1968,
                257,
                3897,
                307,
                364,
                1900,
                51056
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23489583995607163,
            "compression_ratio": 1.6666666666666667,
            "no_speech_prob": 0.00600433349609375,
            "words": [
                {
                    "start": 3160.56,
                    "end": 3161.4,
                    "word": " it's",
                    "probability": 0.917236328125
                },
                {
                    "start": 3161.4,
                    "end": 3161.98,
                    "word": " possible",
                    "probability": 0.84765625
                },
                {
                    "start": 3161.98,
                    "end": 3162.3,
                    "word": " to",
                    "probability": 0.89990234375
                },
                {
                    "start": 3162.3,
                    "end": 3162.7,
                    "word": " determine",
                    "probability": 0.7548828125
                },
                {
                    "start": 3162.7,
                    "end": 3163.26,
                    "word": " whether",
                    "probability": 0.82958984375
                },
                {
                    "start": 3163.52,
                    "end": 3163.92,
                    "word": " a policy",
                    "probability": 0.8134765625
                },
                {
                    "start": 3163.92,
                    "end": 3164.5,
                    "word": " is",
                    "probability": 0.90380859375
                },
                {
                    "start": 3164.5,
                    "end": 3164.94,
                    "word": " authentic",
                    "probability": 0.7138671875
                },
                {
                    "start": 3164.94,
                    "end": 3165.12,
                    "word": " or",
                    "probability": 0.88818359375
                },
                {
                    "start": 3165.12,
                    "end": 3165.34,
                    "word": " not",
                    "probability": 0.89794921875
                },
                {
                    "start": 3165.34,
                    "end": 3165.48,
                    "word": " or",
                    "probability": 0.2841796875
                },
                {
                    "start": 3165.48,
                    "end": 3165.66,
                    "word": " whether",
                    "probability": 0.81884765625
                },
                {
                    "start": 3165.8,
                    "end": 3166.04,
                    "word": " a policy",
                    "probability": 0.8173828125
                },
                {
                    "start": 3166.04,
                    "end": 3166.2,
                    "word": " is",
                    "probability": 0.880859375
                },
                {
                    "start": 3166.2,
                    "end": 3166.3,
                    "word": " an",
                    "probability": 0.90576171875
                },
                {
                    "start": 3166.3,
                    "end": 3166.52,
                    "word": " exact",
                    "probability": 0.853515625
                }
            ]
        },
        {
            "id": 466,
            "seek": 317413,
            "start": 3166.52,
            "end": 3175.0,
            "text": " copy of another policy by just applying it to a second environment and see whether it performs",
            "tokens": [
                51056,
                5055,
                295,
                1071,
                3897,
                538,
                445,
                9275,
                309,
                281,
                257,
                1150,
                2823,
                293,
                536,
                1968,
                309,
                26213,
                51484
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23489583995607163,
            "compression_ratio": 1.6666666666666667,
            "no_speech_prob": 0.00600433349609375,
            "words": [
                {
                    "start": 3166.52,
                    "end": 3166.82,
                    "word": " copy",
                    "probability": 0.8359375
                },
                {
                    "start": 3166.82,
                    "end": 3166.94,
                    "word": " of",
                    "probability": 0.908203125
                },
                {
                    "start": 3166.94,
                    "end": 3167.16,
                    "word": " another",
                    "probability": 0.80126953125
                },
                {
                    "start": 3167.16,
                    "end": 3167.62,
                    "word": " policy",
                    "probability": 0.81396484375
                },
                {
                    "start": 3167.62,
                    "end": 3168.48,
                    "word": " by",
                    "probability": 0.77880859375
                },
                {
                    "start": 3168.48,
                    "end": 3168.78,
                    "word": " just",
                    "probability": 0.88037109375
                },
                {
                    "start": 3168.78,
                    "end": 3169.62,
                    "word": " applying",
                    "probability": 0.69677734375
                },
                {
                    "start": 3169.62,
                    "end": 3169.86,
                    "word": " it",
                    "probability": 0.90380859375
                },
                {
                    "start": 3169.86,
                    "end": 3170.22,
                    "word": " to",
                    "probability": 0.90576171875
                },
                {
                    "start": 3170.86,
                    "end": 3171.64,
                    "word": " a second",
                    "probability": 0.80078125
                },
                {
                    "start": 3171.64,
                    "end": 3172.72,
                    "word": " environment",
                    "probability": 0.763671875
                },
                {
                    "start": 3172.72,
                    "end": 3173.86,
                    "word": " and",
                    "probability": 0.8203125
                },
                {
                    "start": 3173.86,
                    "end": 3174.16,
                    "word": " see",
                    "probability": 0.7841796875
                },
                {
                    "start": 3174.16,
                    "end": 3174.56,
                    "word": " whether",
                    "probability": 0.82421875
                },
                {
                    "start": 3174.56,
                    "end": 3174.7,
                    "word": " it",
                    "probability": 0.9111328125
                },
                {
                    "start": 3174.7,
                    "end": 3175.0,
                    "word": " performs",
                    "probability": 0.7275390625
                }
            ]
        },
        {
            "id": 467,
            "seek": 317413,
            "start": 3175.0,
            "end": 3180.26,
            "text": " optimally or not. Now, there's there are many other things that I really wanted to touch upon",
            "tokens": [
                51484,
                5028,
                379,
                420,
                406,
                13,
                823,
                11,
                456,
                311,
                456,
                366,
                867,
                661,
                721,
                300,
                286,
                534,
                1415,
                281,
                2557,
                3564,
                51744
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23489583995607163,
            "compression_ratio": 1.6666666666666667,
            "no_speech_prob": 0.00600433349609375,
            "words": [
                {
                    "start": 3175.0,
                    "end": 3175.5,
                    "word": " optimally",
                    "probability": 0.860595703125
                },
                {
                    "start": 3175.5,
                    "end": 3175.66,
                    "word": " or",
                    "probability": 0.9033203125
                },
                {
                    "start": 3175.66,
                    "end": 3175.92,
                    "word": " not",
                    "probability": 0.89306640625
                },
                {
                    "start": 3175.92,
                    "end": 3176.52,
                    "word": ".",
                    "probability": 0.9970703125
                },
                {
                    "start": 3176.98,
                    "end": 3177.24,
                    "word": " Now",
                    "probability": 0.8271484375
                },
                {
                    "start": 3177.24,
                    "end": 3177.32,
                    "word": ",",
                    "probability": 0.888671875
                },
                {
                    "start": 3177.32,
                    "end": 3177.62,
                    "word": " there's",
                    "probability": 0.5538330078125
                },
                {
                    "start": 3177.62,
                    "end": 3177.92,
                    "word": " there",
                    "probability": 0.251708984375
                },
                {
                    "start": 3177.92,
                    "end": 3178.24,
                    "word": " are",
                    "probability": 0.86181640625
                },
                {
                    "start": 3178.24,
                    "end": 3178.6,
                    "word": " many",
                    "probability": 0.81787109375
                },
                {
                    "start": 3178.6,
                    "end": 3178.76,
                    "word": " other",
                    "probability": 0.78466796875
                },
                {
                    "start": 3178.76,
                    "end": 3178.94,
                    "word": " things",
                    "probability": 0.7353515625
                },
                {
                    "start": 3178.94,
                    "end": 3179.1,
                    "word": " that",
                    "probability": 0.8701171875
                },
                {
                    "start": 3179.1,
                    "end": 3179.3,
                    "word": " I",
                    "probability": 0.98046875
                },
                {
                    "start": 3179.3,
                    "end": 3179.44,
                    "word": " really",
                    "probability": 0.7626953125
                },
                {
                    "start": 3179.44,
                    "end": 3179.6,
                    "word": " wanted",
                    "probability": 0.79443359375
                },
                {
                    "start": 3179.6,
                    "end": 3179.74,
                    "word": " to",
                    "probability": 0.904296875
                },
                {
                    "start": 3179.74,
                    "end": 3180.06,
                    "word": " touch",
                    "probability": 0.74072265625
                },
                {
                    "start": 3180.06,
                    "end": 3180.26,
                    "word": " upon",
                    "probability": 0.8623046875
                }
            ]
        },
        {
            "id": 468,
            "seek": 319665,
            "start": 3180.26,
            "end": 3186.36,
            "text": " in this talk, but unfortunately we are a little short in time. For practitioners, it may be",
            "tokens": [
                50364,
                294,
                341,
                751,
                11,
                457,
                7015,
                321,
                366,
                257,
                707,
                2099,
                294,
                565,
                13,
                1171,
                25742,
                11,
                309,
                815,
                312,
                50676
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20733173688252768,
            "compression_ratio": 1.625,
            "no_speech_prob": 0.0009255409240722656,
            "words": [
                {
                    "start": 3180.26,
                    "end": 3180.36,
                    "word": " in",
                    "probability": 0.265625
                },
                {
                    "start": 3180.36,
                    "end": 3180.52,
                    "word": " this",
                    "probability": 0.8544921875
                },
                {
                    "start": 3180.52,
                    "end": 3180.72,
                    "word": " talk",
                    "probability": 0.82958984375
                },
                {
                    "start": 3180.72,
                    "end": 3180.74,
                    "word": ",",
                    "probability": 0.5830078125
                },
                {
                    "start": 3180.74,
                    "end": 3180.84,
                    "word": " but",
                    "probability": 0.89013671875
                },
                {
                    "start": 3180.84,
                    "end": 3181.26,
                    "word": " unfortunately",
                    "probability": 0.794921875
                },
                {
                    "start": 3181.26,
                    "end": 3181.5,
                    "word": " we",
                    "probability": 0.634765625
                },
                {
                    "start": 3181.5,
                    "end": 3181.6,
                    "word": " are",
                    "probability": 0.87890625
                },
                {
                    "start": 3181.76,
                    "end": 3181.92,
                    "word": " a little",
                    "probability": 0.80126953125
                },
                {
                    "start": 3181.92,
                    "end": 3182.28,
                    "word": " short",
                    "probability": 0.91845703125
                },
                {
                    "start": 3182.28,
                    "end": 3182.42,
                    "word": " in",
                    "probability": 0.87841796875
                },
                {
                    "start": 3182.42,
                    "end": 3182.8,
                    "word": " time",
                    "probability": 0.80126953125
                },
                {
                    "start": 3182.8,
                    "end": 3183.36,
                    "word": ".",
                    "probability": 0.89501953125
                },
                {
                    "start": 3184.2,
                    "end": 3184.76,
                    "word": " For",
                    "probability": 0.8720703125
                },
                {
                    "start": 3184.76,
                    "end": 3185.34,
                    "word": " practitioners",
                    "probability": 0.66357421875
                },
                {
                    "start": 3185.34,
                    "end": 3185.68,
                    "word": ",",
                    "probability": 0.744140625
                },
                {
                    "start": 3185.68,
                    "end": 3185.8,
                    "word": " it",
                    "probability": 0.89697265625
                },
                {
                    "start": 3185.8,
                    "end": 3186.08,
                    "word": " may",
                    "probability": 0.89501953125
                },
                {
                    "start": 3186.08,
                    "end": 3186.36,
                    "word": " be",
                    "probability": 0.90283203125
                }
            ]
        },
        {
            "id": 469,
            "seek": 319665,
            "start": 3186.36,
            "end": 3191.1,
            "text": " of interest to have some way of benchmarking or evaluating the resilience and robustness",
            "tokens": [
                50676,
                295,
                1179,
                281,
                362,
                512,
                636,
                295,
                18927,
                278,
                420,
                27479,
                264,
                19980,
                293,
                13956,
                1287,
                50912
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20733173688252768,
            "compression_ratio": 1.625,
            "no_speech_prob": 0.0009255409240722656,
            "words": [
                {
                    "start": 3186.36,
                    "end": 3186.74,
                    "word": " of",
                    "probability": 0.9052734375
                },
                {
                    "start": 3186.74,
                    "end": 3187.16,
                    "word": " interest",
                    "probability": 0.7646484375
                },
                {
                    "start": 3187.16,
                    "end": 3187.54,
                    "word": " to",
                    "probability": 0.90234375
                },
                {
                    "start": 3187.54,
                    "end": 3188.06,
                    "word": " have",
                    "probability": 0.87646484375
                },
                {
                    "start": 3188.06,
                    "end": 3188.4,
                    "word": " some",
                    "probability": 0.826171875
                },
                {
                    "start": 3188.4,
                    "end": 3188.54,
                    "word": " way",
                    "probability": 0.90625
                },
                {
                    "start": 3188.54,
                    "end": 3188.66,
                    "word": " of",
                    "probability": 0.8955078125
                },
                {
                    "start": 3188.66,
                    "end": 3189.18,
                    "word": " benchmarking",
                    "probability": 0.912109375
                },
                {
                    "start": 3189.18,
                    "end": 3189.28,
                    "word": " or",
                    "probability": 0.853515625
                },
                {
                    "start": 3189.28,
                    "end": 3189.78,
                    "word": " evaluating",
                    "probability": 0.77001953125
                },
                {
                    "start": 3189.78,
                    "end": 3189.96,
                    "word": " the",
                    "probability": 0.7890625
                },
                {
                    "start": 3189.96,
                    "end": 3190.28,
                    "word": " resilience",
                    "probability": 0.78173828125
                },
                {
                    "start": 3190.28,
                    "end": 3190.48,
                    "word": " and",
                    "probability": 0.87548828125
                },
                {
                    "start": 3190.48,
                    "end": 3191.1,
                    "word": " robustness",
                    "probability": 0.866943359375
                }
            ]
        },
        {
            "id": 470,
            "seek": 319665,
            "start": 3191.1,
            "end": 3196.26,
            "text": " of policies and compare different policies, different approaches with regards to the",
            "tokens": [
                50912,
                295,
                7657,
                293,
                6794,
                819,
                7657,
                11,
                819,
                11587,
                365,
                14258,
                281,
                264,
                51160
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20733173688252768,
            "compression_ratio": 1.625,
            "no_speech_prob": 0.0009255409240722656,
            "words": [
                {
                    "start": 3191.1,
                    "end": 3191.56,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 3191.56,
                    "end": 3192.12,
                    "word": " policies",
                    "probability": 0.81640625
                },
                {
                    "start": 3192.12,
                    "end": 3192.96,
                    "word": " and",
                    "probability": 0.5498046875
                },
                {
                    "start": 3192.96,
                    "end": 3193.34,
                    "word": " compare",
                    "probability": 0.8740234375
                },
                {
                    "start": 3193.34,
                    "end": 3193.7,
                    "word": " different",
                    "probability": 0.6767578125
                },
                {
                    "start": 3193.7,
                    "end": 3194.06,
                    "word": " policies",
                    "probability": 0.83056640625
                },
                {
                    "start": 3194.06,
                    "end": 3194.18,
                    "word": ",",
                    "probability": 0.865234375
                },
                {
                    "start": 3194.18,
                    "end": 3194.46,
                    "word": " different",
                    "probability": 0.67919921875
                },
                {
                    "start": 3194.46,
                    "end": 3195.16,
                    "word": " approaches",
                    "probability": 0.68115234375
                },
                {
                    "start": 3195.16,
                    "end": 3195.9,
                    "word": " with",
                    "probability": 0.63623046875
                },
                {
                    "start": 3195.9,
                    "end": 3196.08,
                    "word": " regards",
                    "probability": 0.7421875
                },
                {
                    "start": 3196.08,
                    "end": 3196.18,
                    "word": " to",
                    "probability": 0.89599609375
                },
                {
                    "start": 3196.18,
                    "end": 3196.26,
                    "word": " the",
                    "probability": 0.32177734375
                }
            ]
        },
        {
            "id": 471,
            "seek": 319665,
            "start": 3196.26,
            "end": 3202.78,
            "text": " resilience and robustness. Some of my work already introduces or proposes an RL-based",
            "tokens": [
                51160,
                19980,
                293,
                13956,
                1287,
                13,
                2188,
                295,
                452,
                589,
                1217,
                31472,
                420,
                2365,
                4201,
                364,
                497,
                43,
                12,
                6032,
                51492
            ],
            "temperature": 0.0,
            "avg_logprob": -0.20733173688252768,
            "compression_ratio": 1.625,
            "no_speech_prob": 0.0009255409240722656,
            "words": [
                {
                    "start": 3196.26,
                    "end": 3196.56,
                    "word": " resilience",
                    "probability": 0.71484375
                },
                {
                    "start": 3196.56,
                    "end": 3196.82,
                    "word": " and",
                    "probability": 0.90087890625
                },
                {
                    "start": 3196.82,
                    "end": 3197.48,
                    "word": " robustness",
                    "probability": 0.843505859375
                },
                {
                    "start": 3197.48,
                    "end": 3198.04,
                    "word": ".",
                    "probability": 0.99072265625
                },
                {
                    "start": 3199.18,
                    "end": 3199.74,
                    "word": " Some",
                    "probability": 0.8701171875
                },
                {
                    "start": 3199.74,
                    "end": 3199.84,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 3199.84,
                    "end": 3199.96,
                    "word": " my",
                    "probability": 0.90380859375
                },
                {
                    "start": 3199.96,
                    "end": 3200.18,
                    "word": " work",
                    "probability": 0.80810546875
                },
                {
                    "start": 3200.18,
                    "end": 3200.46,
                    "word": " already",
                    "probability": 0.85595703125
                },
                {
                    "start": 3200.46,
                    "end": 3201.48,
                    "word": " introduces",
                    "probability": 0.76318359375
                },
                {
                    "start": 3201.48,
                    "end": 3201.7,
                    "word": " or",
                    "probability": 0.88330078125
                },
                {
                    "start": 3201.7,
                    "end": 3202.08,
                    "word": " proposes",
                    "probability": 0.866943359375
                },
                {
                    "start": 3202.08,
                    "end": 3202.34,
                    "word": " an",
                    "probability": 0.89794921875
                },
                {
                    "start": 3202.34,
                    "end": 3202.62,
                    "word": " RL",
                    "probability": 0.91796875
                },
                {
                    "start": 3202.62,
                    "end": 3202.78,
                    "word": "-based",
                    "probability": 0.7841796875
                }
            ]
        },
        {
            "id": 472,
            "seek": 322073,
            "start": 3202.78,
            "end": 3210.28,
            "text": " approach to perform this evaluation and benchmarking. I've also done some work on",
            "tokens": [
                50364,
                3109,
                281,
                2042,
                341,
                13344,
                293,
                18927,
                278,
                13,
                286,
                600,
                611,
                1096,
                512,
                589,
                322,
                50728
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22676282070386103,
            "compression_ratio": 1.462962962962963,
            "no_speech_prob": 0.005138397216796875,
            "words": [
                {
                    "start": 3202.78,
                    "end": 3203.2,
                    "word": " approach",
                    "probability": 0.213134765625
                },
                {
                    "start": 3203.2,
                    "end": 3203.62,
                    "word": " to",
                    "probability": 0.8623046875
                },
                {
                    "start": 3203.62,
                    "end": 3204.0,
                    "word": " perform",
                    "probability": 0.79833984375
                },
                {
                    "start": 3204.0,
                    "end": 3204.16,
                    "word": " this",
                    "probability": 0.880859375
                },
                {
                    "start": 3204.16,
                    "end": 3204.54,
                    "word": " evaluation",
                    "probability": 0.6416015625
                },
                {
                    "start": 3204.54,
                    "end": 3204.76,
                    "word": " and",
                    "probability": 0.861328125
                },
                {
                    "start": 3204.76,
                    "end": 3205.46,
                    "word": " benchmarking",
                    "probability": 0.9091796875
                },
                {
                    "start": 3205.46,
                    "end": 3206.18,
                    "word": ".",
                    "probability": 0.9189453125
                },
                {
                    "start": 3207.3,
                    "end": 3208.02,
                    "word": " I've",
                    "probability": 0.924072265625
                },
                {
                    "start": 3208.02,
                    "end": 3208.44,
                    "word": " also",
                    "probability": 0.81982421875
                },
                {
                    "start": 3208.44,
                    "end": 3209.1,
                    "word": " done",
                    "probability": 0.82373046875
                },
                {
                    "start": 3209.1,
                    "end": 3209.4,
                    "word": " some",
                    "probability": 0.818359375
                },
                {
                    "start": 3209.4,
                    "end": 3209.74,
                    "word": " work",
                    "probability": 0.7978515625
                },
                {
                    "start": 3209.74,
                    "end": 3210.28,
                    "word": " on",
                    "probability": 0.88037109375
                }
            ]
        },
        {
            "id": 473,
            "seek": 322073,
            "start": 3211.18,
            "end": 3215.48,
            "text": " investigating the impact of hyperparameter choices on resilience and robustness of",
            "tokens": [
                50784,
                22858,
                264,
                2712,
                295,
                9848,
                2181,
                335,
                2398,
                7994,
                322,
                19980,
                293,
                13956,
                1287,
                295,
                51012
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22676282070386103,
            "compression_ratio": 1.462962962962963,
            "no_speech_prob": 0.005138397216796875,
            "words": [
                {
                    "start": 3211.18,
                    "end": 3211.64,
                    "word": " investigating",
                    "probability": 0.734375
                },
                {
                    "start": 3211.64,
                    "end": 3211.98,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 3211.98,
                    "end": 3212.36,
                    "word": " impact",
                    "probability": 0.86767578125
                },
                {
                    "start": 3212.36,
                    "end": 3212.5,
                    "word": " of",
                    "probability": 0.90771484375
                },
                {
                    "start": 3212.5,
                    "end": 3213.08,
                    "word": " hyperparameter",
                    "probability": 0.857177734375
                },
                {
                    "start": 3213.08,
                    "end": 3213.54,
                    "word": " choices",
                    "probability": 0.82275390625
                },
                {
                    "start": 3213.54,
                    "end": 3213.92,
                    "word": " on",
                    "probability": 0.88134765625
                },
                {
                    "start": 3213.92,
                    "end": 3214.28,
                    "word": " resilience",
                    "probability": 0.77490234375
                },
                {
                    "start": 3214.28,
                    "end": 3214.48,
                    "word": " and",
                    "probability": 0.8837890625
                },
                {
                    "start": 3214.48,
                    "end": 3215.04,
                    "word": " robustness",
                    "probability": 0.864013671875
                },
                {
                    "start": 3215.04,
                    "end": 3215.48,
                    "word": " of",
                    "probability": 0.90087890625
                }
            ]
        },
        {
            "id": 474,
            "seek": 322073,
            "start": 3215.48,
            "end": 3220.72,
            "text": " DQNs in particular, but also other model-free and actor-critic approaches.",
            "tokens": [
                51012,
                413,
                48,
                45,
                82,
                294,
                1729,
                11,
                457,
                611,
                661,
                2316,
                12,
                10792,
                293,
                8747,
                12,
                32255,
                299,
                11587,
                13,
                51248
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22676282070386103,
            "compression_ratio": 1.462962962962963,
            "no_speech_prob": 0.005138397216796875,
            "words": [
                {
                    "start": 3215.48,
                    "end": 3216.42,
                    "word": " DQNs",
                    "probability": 0.7537841796875
                },
                {
                    "start": 3216.42,
                    "end": 3216.72,
                    "word": " in",
                    "probability": 0.8173828125
                },
                {
                    "start": 3216.72,
                    "end": 3217.1,
                    "word": " particular",
                    "probability": 0.82861328125
                },
                {
                    "start": 3217.1,
                    "end": 3217.24,
                    "word": ",",
                    "probability": 0.6357421875
                },
                {
                    "start": 3217.24,
                    "end": 3217.36,
                    "word": " but",
                    "probability": 0.88818359375
                },
                {
                    "start": 3217.36,
                    "end": 3217.66,
                    "word": " also",
                    "probability": 0.818359375
                },
                {
                    "start": 3217.66,
                    "end": 3218.02,
                    "word": " other",
                    "probability": 0.78662109375
                },
                {
                    "start": 3218.02,
                    "end": 3218.76,
                    "word": " model",
                    "probability": 0.83349609375
                },
                {
                    "start": 3218.76,
                    "end": 3219.12,
                    "word": "-free",
                    "probability": 0.764892578125
                },
                {
                    "start": 3219.12,
                    "end": 3219.34,
                    "word": " and",
                    "probability": 0.888671875
                },
                {
                    "start": 3219.34,
                    "end": 3219.6,
                    "word": " actor",
                    "probability": 0.3349609375
                },
                {
                    "start": 3219.6,
                    "end": 3219.94,
                    "word": "-critic",
                    "probability": 0.8763020833333334
                },
                {
                    "start": 3219.94,
                    "end": 3220.36,
                    "word": " approaches",
                    "probability": 0.68896484375
                },
                {
                    "start": 3220.36,
                    "end": 3220.72,
                    "word": ".",
                    "probability": 0.9736328125
                }
            ]
        },
        {
            "id": 475,
            "seek": 322073,
            "start": 3221.92,
            "end": 3226.86,
            "text": " This can be very helpful to those who want to engineer and design the new RL",
            "tokens": [
                51300,
                639,
                393,
                312,
                588,
                4961,
                281,
                729,
                567,
                528,
                281,
                11403,
                293,
                1715,
                264,
                777,
                497,
                43,
                51568
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22676282070386103,
            "compression_ratio": 1.462962962962963,
            "no_speech_prob": 0.005138397216796875,
            "words": [
                {
                    "start": 3221.92,
                    "end": 3222.64,
                    "word": " This",
                    "probability": 0.80517578125
                },
                {
                    "start": 3222.64,
                    "end": 3223.36,
                    "word": " can",
                    "probability": 0.72802734375
                },
                {
                    "start": 3223.36,
                    "end": 3223.52,
                    "word": " be",
                    "probability": 0.90673828125
                },
                {
                    "start": 3223.52,
                    "end": 3223.7,
                    "word": " very",
                    "probability": 0.71875
                },
                {
                    "start": 3223.7,
                    "end": 3223.92,
                    "word": " helpful",
                    "probability": 0.82080078125
                },
                {
                    "start": 3223.92,
                    "end": 3224.1,
                    "word": " to",
                    "probability": 0.8896484375
                },
                {
                    "start": 3224.1,
                    "end": 3224.32,
                    "word": " those",
                    "probability": 0.84521484375
                },
                {
                    "start": 3224.32,
                    "end": 3224.42,
                    "word": " who",
                    "probability": 0.79931640625
                },
                {
                    "start": 3224.42,
                    "end": 3224.54,
                    "word": " want",
                    "probability": 0.8212890625
                },
                {
                    "start": 3224.54,
                    "end": 3225.1,
                    "word": " to",
                    "probability": 0.904296875
                },
                {
                    "start": 3225.1,
                    "end": 3225.42,
                    "word": " engineer",
                    "probability": 0.68115234375
                },
                {
                    "start": 3225.42,
                    "end": 3225.64,
                    "word": " and",
                    "probability": 0.890625
                },
                {
                    "start": 3225.64,
                    "end": 3226.06,
                    "word": " design",
                    "probability": 0.8984375
                },
                {
                    "start": 3226.06,
                    "end": 3226.26,
                    "word": " the",
                    "probability": 0.496826171875
                },
                {
                    "start": 3226.26,
                    "end": 3226.48,
                    "word": " new",
                    "probability": 0.8349609375
                },
                {
                    "start": 3226.48,
                    "end": 3226.86,
                    "word": " RL",
                    "probability": 0.884033203125
                }
            ]
        },
        {
            "id": 476,
            "seek": 324821,
            "start": 3226.86,
            "end": 3233.68,
            "text": " agents to be deployed in critical environments. Also, something that I wanted to mention, but",
            "tokens": [
                50364,
                12554,
                281,
                312,
                17826,
                294,
                4924,
                12388,
                13,
                2743,
                11,
                746,
                300,
                286,
                1415,
                281,
                2152,
                11,
                457,
                50704
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1979779369690839,
            "compression_ratio": 1.6063348416289593,
            "no_speech_prob": 0.0003514289855957031,
            "words": [
                {
                    "start": 3226.86,
                    "end": 3227.22,
                    "word": " agents",
                    "probability": 0.357421875
                },
                {
                    "start": 3227.22,
                    "end": 3227.34,
                    "word": " to",
                    "probability": 0.869140625
                },
                {
                    "start": 3227.34,
                    "end": 3227.44,
                    "word": " be",
                    "probability": 0.89794921875
                },
                {
                    "start": 3227.44,
                    "end": 3227.78,
                    "word": " deployed",
                    "probability": 0.724609375
                },
                {
                    "start": 3227.78,
                    "end": 3228.14,
                    "word": " in",
                    "probability": 0.9013671875
                },
                {
                    "start": 3228.14,
                    "end": 3228.58,
                    "word": " critical",
                    "probability": 0.80908203125
                },
                {
                    "start": 3228.58,
                    "end": 3229.28,
                    "word": " environments",
                    "probability": 0.6806640625
                },
                {
                    "start": 3229.28,
                    "end": 3229.96,
                    "word": ".",
                    "probability": 0.92822265625
                },
                {
                    "start": 3230.88,
                    "end": 3231.56,
                    "word": " Also",
                    "probability": 0.8486328125
                },
                {
                    "start": 3231.56,
                    "end": 3231.98,
                    "word": ",",
                    "probability": 0.62744140625
                },
                {
                    "start": 3231.98,
                    "end": 3232.44,
                    "word": " something",
                    "probability": 0.56396484375
                },
                {
                    "start": 3232.44,
                    "end": 3232.7,
                    "word": " that",
                    "probability": 0.86328125
                },
                {
                    "start": 3232.7,
                    "end": 3232.86,
                    "word": " I",
                    "probability": 0.99072265625
                },
                {
                    "start": 3232.86,
                    "end": 3233.12,
                    "word": " wanted",
                    "probability": 0.79541015625
                },
                {
                    "start": 3233.12,
                    "end": 3233.3,
                    "word": " to",
                    "probability": 0.904296875
                },
                {
                    "start": 3233.3,
                    "end": 3233.52,
                    "word": " mention",
                    "probability": 0.78369140625
                },
                {
                    "start": 3233.52,
                    "end": 3233.62,
                    "word": ",",
                    "probability": 0.55615234375
                },
                {
                    "start": 3233.62,
                    "end": 3233.68,
                    "word": " but",
                    "probability": 0.8818359375
                }
            ]
        },
        {
            "id": 477,
            "seek": 324821,
            "start": 3233.68,
            "end": 3242.02,
            "text": " unfortunately I don't have time to do so, is that adversarial training is not a silver bullet. It's",
            "tokens": [
                50704,
                7015,
                286,
                500,
                380,
                362,
                565,
                281,
                360,
                370,
                11,
                307,
                300,
                17641,
                44745,
                3097,
                307,
                406,
                257,
                8753,
                11632,
                13,
                467,
                311,
                51124
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1979779369690839,
            "compression_ratio": 1.6063348416289593,
            "no_speech_prob": 0.0003514289855957031,
            "words": [
                {
                    "start": 3233.68,
                    "end": 3234.12,
                    "word": " unfortunately",
                    "probability": 0.80615234375
                },
                {
                    "start": 3234.12,
                    "end": 3234.66,
                    "word": " I",
                    "probability": 0.3544921875
                },
                {
                    "start": 3234.66,
                    "end": 3234.78,
                    "word": " don't",
                    "probability": 0.942138671875
                },
                {
                    "start": 3234.78,
                    "end": 3234.96,
                    "word": " have",
                    "probability": 0.88720703125
                },
                {
                    "start": 3234.96,
                    "end": 3235.14,
                    "word": " time",
                    "probability": 0.8095703125
                },
                {
                    "start": 3235.14,
                    "end": 3235.28,
                    "word": " to",
                    "probability": 0.90869140625
                },
                {
                    "start": 3235.28,
                    "end": 3235.4,
                    "word": " do",
                    "probability": 0.89990234375
                },
                {
                    "start": 3235.4,
                    "end": 3235.58,
                    "word": " so",
                    "probability": 0.89306640625
                },
                {
                    "start": 3235.58,
                    "end": 3235.68,
                    "word": ",",
                    "probability": 0.37646484375
                },
                {
                    "start": 3235.68,
                    "end": 3235.7,
                    "word": " is",
                    "probability": 0.900390625
                },
                {
                    "start": 3235.7,
                    "end": 3235.86,
                    "word": " that",
                    "probability": 0.87548828125
                },
                {
                    "start": 3235.86,
                    "end": 3236.4,
                    "word": " adversarial",
                    "probability": 0.9267578125
                },
                {
                    "start": 3236.4,
                    "end": 3236.88,
                    "word": " training",
                    "probability": 0.8017578125
                },
                {
                    "start": 3236.88,
                    "end": 3238.44,
                    "word": " is",
                    "probability": 0.88427734375
                },
                {
                    "start": 3238.44,
                    "end": 3239.1,
                    "word": " not",
                    "probability": 0.89111328125
                },
                {
                    "start": 3240.7,
                    "end": 3241.54,
                    "word": " a silver",
                    "probability": 0.78955078125
                },
                {
                    "start": 3241.54,
                    "end": 3241.8,
                    "word": " bullet",
                    "probability": 0.8388671875
                },
                {
                    "start": 3241.8,
                    "end": 3241.92,
                    "word": ".",
                    "probability": 0.6318359375
                },
                {
                    "start": 3241.92,
                    "end": 3242.02,
                    "word": " It's",
                    "probability": 0.93798828125
                }
            ]
        },
        {
            "id": 478,
            "seek": 324821,
            "start": 3242.02,
            "end": 3247.16,
            "text": " not an answer to all of the problems in DQN agents. There are certain limitations",
            "tokens": [
                51124,
                406,
                364,
                1867,
                281,
                439,
                295,
                264,
                2740,
                294,
                413,
                48,
                45,
                12554,
                13,
                821,
                366,
                1629,
                15705,
                51384
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1979779369690839,
            "compression_ratio": 1.6063348416289593,
            "no_speech_prob": 0.0003514289855957031,
            "words": [
                {
                    "start": 3242.02,
                    "end": 3242.36,
                    "word": " not",
                    "probability": 0.89599609375
                },
                {
                    "start": 3242.36,
                    "end": 3242.84,
                    "word": " an",
                    "probability": 0.9091796875
                },
                {
                    "start": 3242.84,
                    "end": 3243.16,
                    "word": " answer",
                    "probability": 0.884765625
                },
                {
                    "start": 3243.16,
                    "end": 3243.66,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 3243.66,
                    "end": 3244.1,
                    "word": " all",
                    "probability": 0.90283203125
                },
                {
                    "start": 3244.1,
                    "end": 3244.22,
                    "word": " of",
                    "probability": 0.87060546875
                },
                {
                    "start": 3244.22,
                    "end": 3244.32,
                    "word": " the",
                    "probability": 0.81640625
                },
                {
                    "start": 3244.32,
                    "end": 3244.74,
                    "word": " problems",
                    "probability": 0.70263671875
                },
                {
                    "start": 3244.74,
                    "end": 3245.22,
                    "word": " in",
                    "probability": 0.876953125
                },
                {
                    "start": 3245.22,
                    "end": 3245.88,
                    "word": " DQN",
                    "probability": 0.908203125
                },
                {
                    "start": 3245.88,
                    "end": 3246.1,
                    "word": " agents",
                    "probability": 0.67626953125
                },
                {
                    "start": 3246.1,
                    "end": 3246.22,
                    "word": ".",
                    "probability": 0.70703125
                },
                {
                    "start": 3246.22,
                    "end": 3246.26,
                    "word": " There",
                    "probability": 0.76904296875
                },
                {
                    "start": 3246.26,
                    "end": 3246.34,
                    "word": " are",
                    "probability": 0.89453125
                },
                {
                    "start": 3246.34,
                    "end": 3246.62,
                    "word": " certain",
                    "probability": 0.7900390625
                },
                {
                    "start": 3246.62,
                    "end": 3247.16,
                    "word": " limitations",
                    "probability": 0.6279296875
                }
            ]
        },
        {
            "id": 479,
            "seek": 324821,
            "start": 3249.2,
            "end": 3254.34,
            "text": " for robustness and resilience obtained from adversarial training of DQN agents.",
            "tokens": [
                51476,
                337,
                13956,
                1287,
                293,
                19980,
                14879,
                490,
                17641,
                44745,
                3097,
                295,
                413,
                48,
                45,
                12554,
                13,
                51704
            ],
            "temperature": 0.0,
            "avg_logprob": -0.1979779369690839,
            "compression_ratio": 1.6063348416289593,
            "no_speech_prob": 0.0003514289855957031,
            "words": [
                {
                    "start": 3249.2,
                    "end": 3249.88,
                    "word": " for",
                    "probability": 0.87255859375
                },
                {
                    "start": 3249.88,
                    "end": 3250.56,
                    "word": " robustness",
                    "probability": 0.83056640625
                },
                {
                    "start": 3250.56,
                    "end": 3250.7,
                    "word": " and",
                    "probability": 0.8515625
                },
                {
                    "start": 3250.7,
                    "end": 3251.16,
                    "word": " resilience",
                    "probability": 0.83740234375
                },
                {
                    "start": 3251.16,
                    "end": 3251.6,
                    "word": " obtained",
                    "probability": 0.87060546875
                },
                {
                    "start": 3251.6,
                    "end": 3251.82,
                    "word": " from",
                    "probability": 0.82470703125
                },
                {
                    "start": 3251.82,
                    "end": 3252.3,
                    "word": " adversarial",
                    "probability": 0.934814453125
                },
                {
                    "start": 3252.3,
                    "end": 3252.66,
                    "word": " training",
                    "probability": 0.8154296875
                },
                {
                    "start": 3252.66,
                    "end": 3252.8,
                    "word": " of",
                    "probability": 0.8994140625
                },
                {
                    "start": 3252.8,
                    "end": 3253.22,
                    "word": " DQN",
                    "probability": 0.9723307291666666
                },
                {
                    "start": 3253.22,
                    "end": 3253.66,
                    "word": " agents",
                    "probability": 0.79736328125
                },
                {
                    "start": 3253.66,
                    "end": 3254.34,
                    "word": ".",
                    "probability": 0.6240234375
                }
            ]
        },
        {
            "id": 480,
            "seek": 327511,
            "start": 3254.34,
            "end": 3261.72,
            "text": " And also adversarial training is very costly in general, especially when it comes to real-world",
            "tokens": [
                50388,
                400,
                611,
                17641,
                44745,
                3097,
                307,
                588,
                28328,
                294,
                2674,
                11,
                2318,
                562,
                309,
                1487,
                281,
                957,
                12,
                13217,
                50728
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23663403865802718,
            "compression_ratio": 1.6846846846846846,
            "no_speech_prob": 3.534555435180664e-05,
            "words": [
                {
                    "start": 3254.34,
                    "end": 3255.06,
                    "word": " And",
                    "probability": 0.54736328125
                },
                {
                    "start": 3255.06,
                    "end": 3255.62,
                    "word": " also",
                    "probability": 0.7861328125
                },
                {
                    "start": 3255.62,
                    "end": 3256.62,
                    "word": " adversarial",
                    "probability": 0.798095703125
                },
                {
                    "start": 3256.62,
                    "end": 3257.22,
                    "word": " training",
                    "probability": 0.8046875
                },
                {
                    "start": 3257.22,
                    "end": 3259.04,
                    "word": " is",
                    "probability": 0.86572265625
                },
                {
                    "start": 3259.04,
                    "end": 3259.44,
                    "word": " very",
                    "probability": 0.74951171875
                },
                {
                    "start": 3259.44,
                    "end": 3259.84,
                    "word": " costly",
                    "probability": 0.74755859375
                },
                {
                    "start": 3259.84,
                    "end": 3260.04,
                    "word": " in",
                    "probability": 0.8486328125
                },
                {
                    "start": 3260.04,
                    "end": 3260.24,
                    "word": " general",
                    "probability": 0.79833984375
                },
                {
                    "start": 3260.24,
                    "end": 3260.34,
                    "word": ",",
                    "probability": 0.751953125
                },
                {
                    "start": 3260.34,
                    "end": 3260.56,
                    "word": " especially",
                    "probability": 0.70703125
                },
                {
                    "start": 3260.56,
                    "end": 3260.82,
                    "word": " when",
                    "probability": 0.8759765625
                },
                {
                    "start": 3260.82,
                    "end": 3260.94,
                    "word": " it",
                    "probability": 0.9033203125
                },
                {
                    "start": 3260.94,
                    "end": 3261.1,
                    "word": " comes",
                    "probability": 0.80615234375
                },
                {
                    "start": 3261.1,
                    "end": 3261.22,
                    "word": " to",
                    "probability": 0.90576171875
                },
                {
                    "start": 3261.22,
                    "end": 3261.5,
                    "word": " real",
                    "probability": 0.78857421875
                },
                {
                    "start": 3261.5,
                    "end": 3261.72,
                    "word": "-world",
                    "probability": 0.6251220703125
                }
            ]
        },
        {
            "id": 481,
            "seek": 327511,
            "start": 3261.72,
            "end": 3269.48,
            "text": " scenarios, real-world environments and actions. And some of my recent work is focused on improving",
            "tokens": [
                50728,
                15077,
                11,
                957,
                12,
                13217,
                12388,
                293,
                5909,
                13,
                400,
                512,
                295,
                452,
                5162,
                589,
                307,
                5178,
                322,
                11470,
                51140
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23663403865802718,
            "compression_ratio": 1.6846846846846846,
            "no_speech_prob": 3.534555435180664e-05,
            "words": [
                {
                    "start": 3261.72,
                    "end": 3262.22,
                    "word": " scenarios",
                    "probability": 0.8505859375
                },
                {
                    "start": 3262.22,
                    "end": 3262.46,
                    "word": ",",
                    "probability": 0.92236328125
                },
                {
                    "start": 3262.46,
                    "end": 3262.64,
                    "word": " real",
                    "probability": 0.85693359375
                },
                {
                    "start": 3262.64,
                    "end": 3262.88,
                    "word": "-world",
                    "probability": 0.865478515625
                },
                {
                    "start": 3262.88,
                    "end": 3263.3,
                    "word": " environments",
                    "probability": 0.65673828125
                },
                {
                    "start": 3263.3,
                    "end": 3263.46,
                    "word": " and",
                    "probability": 0.779296875
                },
                {
                    "start": 3263.46,
                    "end": 3263.86,
                    "word": " actions",
                    "probability": 0.85791015625
                },
                {
                    "start": 3263.86,
                    "end": 3264.58,
                    "word": ".",
                    "probability": 0.8310546875
                },
                {
                    "start": 3264.76,
                    "end": 3265.06,
                    "word": " And",
                    "probability": 0.8271484375
                },
                {
                    "start": 3265.06,
                    "end": 3265.4,
                    "word": " some",
                    "probability": 0.80810546875
                },
                {
                    "start": 3265.4,
                    "end": 3265.6,
                    "word": " of",
                    "probability": 0.9091796875
                },
                {
                    "start": 3265.6,
                    "end": 3265.96,
                    "word": " my",
                    "probability": 0.90625
                },
                {
                    "start": 3265.96,
                    "end": 3266.6,
                    "word": " recent",
                    "probability": 0.798828125
                },
                {
                    "start": 3266.6,
                    "end": 3266.92,
                    "word": " work",
                    "probability": 0.81640625
                },
                {
                    "start": 3266.92,
                    "end": 3267.58,
                    "word": " is",
                    "probability": 0.88232421875
                },
                {
                    "start": 3267.58,
                    "end": 3268.2,
                    "word": " focused",
                    "probability": 0.765625
                },
                {
                    "start": 3268.2,
                    "end": 3268.56,
                    "word": " on",
                    "probability": 0.912109375
                },
                {
                    "start": 3268.56,
                    "end": 3269.48,
                    "word": " improving",
                    "probability": 0.88330078125
                }
            ]
        },
        {
            "id": 482,
            "seek": 327511,
            "start": 3269.48,
            "end": 3275.92,
            "text": " the sample efficiency and computational cost of adversarial training via a new exploration",
            "tokens": [
                51140,
                264,
                6889,
                10493,
                293,
                28270,
                2063,
                295,
                17641,
                44745,
                3097,
                5766,
                257,
                777,
                16197,
                51440
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23663403865802718,
            "compression_ratio": 1.6846846846846846,
            "no_speech_prob": 3.534555435180664e-05,
            "words": [
                {
                    "start": 3269.48,
                    "end": 3270.2,
                    "word": " the",
                    "probability": 0.8095703125
                },
                {
                    "start": 3270.2,
                    "end": 3270.62,
                    "word": " sample",
                    "probability": 0.767578125
                },
                {
                    "start": 3270.62,
                    "end": 3271.08,
                    "word": " efficiency",
                    "probability": 0.9072265625
                },
                {
                    "start": 3271.08,
                    "end": 3271.76,
                    "word": " and",
                    "probability": 0.8876953125
                },
                {
                    "start": 3271.76,
                    "end": 3272.44,
                    "word": " computational",
                    "probability": 0.8681640625
                },
                {
                    "start": 3272.44,
                    "end": 3272.68,
                    "word": " cost",
                    "probability": 0.68896484375
                },
                {
                    "start": 3272.68,
                    "end": 3272.8,
                    "word": " of",
                    "probability": 0.88427734375
                },
                {
                    "start": 3272.8,
                    "end": 3273.16,
                    "word": " adversarial",
                    "probability": 0.942138671875
                },
                {
                    "start": 3273.16,
                    "end": 3273.52,
                    "word": " training",
                    "probability": 0.80078125
                },
                {
                    "start": 3273.52,
                    "end": 3273.84,
                    "word": " via",
                    "probability": 0.67431640625
                },
                {
                    "start": 3274.26,
                    "end": 3274.84,
                    "word": " a new",
                    "probability": 0.80908203125
                },
                {
                    "start": 3274.84,
                    "end": 3275.92,
                    "word": " exploration",
                    "probability": 0.79052734375
                }
            ]
        },
        {
            "id": 483,
            "seek": 327511,
            "start": 3275.92,
            "end": 3281.24,
            "text": " mechanism called adversarially-guided exploration, or AGE. All of this work can be found",
            "tokens": [
                51440,
                7513,
                1219,
                17641,
                289,
                2270,
                12,
                2794,
                2112,
                16197,
                11,
                420,
                316,
                9177,
                13,
                1057,
                295,
                341,
                589,
                393,
                312,
                1352,
                51704
            ],
            "temperature": 0.0,
            "avg_logprob": -0.23663403865802718,
            "compression_ratio": 1.6846846846846846,
            "no_speech_prob": 3.534555435180664e-05,
            "words": [
                {
                    "start": 3275.92,
                    "end": 3276.3,
                    "word": " mechanism",
                    "probability": 0.88720703125
                },
                {
                    "start": 3276.3,
                    "end": 3276.6,
                    "word": " called",
                    "probability": 0.78515625
                },
                {
                    "start": 3276.6,
                    "end": 3277.18,
                    "word": " adversarially",
                    "probability": 0.6930338541666666
                },
                {
                    "start": 3277.18,
                    "end": 3277.46,
                    "word": "-guided",
                    "probability": 0.7569173177083334
                },
                {
                    "start": 3277.46,
                    "end": 3277.9,
                    "word": " exploration",
                    "probability": 0.76904296875
                },
                {
                    "start": 3277.9,
                    "end": 3278.18,
                    "word": ",",
                    "probability": 0.5712890625
                },
                {
                    "start": 3278.18,
                    "end": 3278.3,
                    "word": " or",
                    "probability": 0.90966796875
                },
                {
                    "start": 3278.3,
                    "end": 3278.72,
                    "word": " AGE",
                    "probability": 0.5030517578125
                },
                {
                    "start": 3278.72,
                    "end": 3279.24,
                    "word": ".",
                    "probability": 0.9951171875
                },
                {
                    "start": 3279.48,
                    "end": 3280.2,
                    "word": " All",
                    "probability": 0.89697265625
                },
                {
                    "start": 3280.2,
                    "end": 3280.3,
                    "word": " of",
                    "probability": 0.89892578125
                },
                {
                    "start": 3280.3,
                    "end": 3280.42,
                    "word": " this",
                    "probability": 0.8720703125
                },
                {
                    "start": 3280.42,
                    "end": 3280.62,
                    "word": " work",
                    "probability": 0.82080078125
                },
                {
                    "start": 3280.62,
                    "end": 3280.72,
                    "word": " can",
                    "probability": 0.89501953125
                },
                {
                    "start": 3280.72,
                    "end": 3280.88,
                    "word": " be",
                    "probability": 0.9013671875
                },
                {
                    "start": 3280.88,
                    "end": 3281.24,
                    "word": " found",
                    "probability": 0.884765625
                }
            ]
        },
        {
            "id": 484,
            "seek": 330255,
            "start": 3281.94,
            "end": 3288.36,
            "text": " in my Ph.D. dissertation, which bears the same name as this talk, False Narcotic Star Security",
            "tokens": [
                50412,
                294,
                452,
                2623,
                13,
                35,
                13,
                39555,
                11,
                597,
                17276,
                264,
                912,
                1315,
                382,
                341,
                751,
                11,
                50040,
                13512,
                66,
                9411,
                5705,
                11164,
                50716
            ],
            "temperature": 0.0,
            "avg_logprob": -0.27428669514863385,
            "compression_ratio": 1.5120967741935485,
            "no_speech_prob": 0.0008425712585449219,
            "words": [
                {
                    "start": 3281.94,
                    "end": 3282.5,
                    "word": " in",
                    "probability": 0.0904541015625
                },
                {
                    "start": 3282.5,
                    "end": 3282.82,
                    "word": " my",
                    "probability": 0.8203125
                },
                {
                    "start": 3282.82,
                    "end": 3283.3,
                    "word": " Ph",
                    "probability": 0.08135986328125
                },
                {
                    "start": 3283.3,
                    "end": 3283.46,
                    "word": ".D",
                    "probability": 0.9443359375
                },
                {
                    "start": 3283.46,
                    "end": 3283.58,
                    "word": ".",
                    "probability": 0.70458984375
                },
                {
                    "start": 3283.58,
                    "end": 3283.98,
                    "word": " dissertation",
                    "probability": 0.84765625
                },
                {
                    "start": 3283.98,
                    "end": 3284.8,
                    "word": ",",
                    "probability": 0.59814453125
                },
                {
                    "start": 3284.8,
                    "end": 3285.72,
                    "word": " which",
                    "probability": 0.89453125
                },
                {
                    "start": 3285.72,
                    "end": 3285.92,
                    "word": " bears",
                    "probability": 0.75927734375
                },
                {
                    "start": 3285.92,
                    "end": 3286.06,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 3286.06,
                    "end": 3286.22,
                    "word": " same",
                    "probability": 0.80908203125
                },
                {
                    "start": 3286.22,
                    "end": 3286.36,
                    "word": " name",
                    "probability": 0.802734375
                },
                {
                    "start": 3286.36,
                    "end": 3286.48,
                    "word": " as",
                    "probability": 0.9140625
                },
                {
                    "start": 3286.48,
                    "end": 3286.62,
                    "word": " this",
                    "probability": 0.86865234375
                },
                {
                    "start": 3286.62,
                    "end": 3286.9,
                    "word": " talk",
                    "probability": 0.78125
                },
                {
                    "start": 3286.9,
                    "end": 3287.04,
                    "word": ",",
                    "probability": 0.755859375
                },
                {
                    "start": 3287.04,
                    "end": 3287.54,
                    "word": " False",
                    "probability": 0.34423828125
                },
                {
                    "start": 3287.54,
                    "end": 3287.84,
                    "word": " Narcotic",
                    "probability": 0.4672444661458333
                },
                {
                    "start": 3287.84,
                    "end": 3288.02,
                    "word": " Star",
                    "probability": 0.255126953125
                },
                {
                    "start": 3288.02,
                    "end": 3288.36,
                    "word": " Security",
                    "probability": 0.73046875
                }
            ]
        },
        {
            "id": 485,
            "seek": 330255,
            "start": 3288.36,
            "end": 3293.48,
            "text": " and Deep Reinforcement Learning. You can find it if you search my name on Google Scholar.",
            "tokens": [
                50716,
                293,
                14895,
                42116,
                9382,
                15205,
                13,
                509,
                393,
                915,
                309,
                498,
                291,
                3164,
                452,
                1315,
                322,
                3329,
                2065,
                15276,
                13,
                50948
            ],
            "temperature": 0.0,
            "avg_logprob": -0.27428669514863385,
            "compression_ratio": 1.5120967741935485,
            "no_speech_prob": 0.0008425712585449219,
            "words": [
                {
                    "start": 3288.36,
                    "end": 3288.52,
                    "word": " and",
                    "probability": 0.646484375
                },
                {
                    "start": 3288.52,
                    "end": 3288.66,
                    "word": " Deep",
                    "probability": 0.48876953125
                },
                {
                    "start": 3288.66,
                    "end": 3289.06,
                    "word": " Reinforcement",
                    "probability": 0.8564453125
                },
                {
                    "start": 3289.06,
                    "end": 3289.46,
                    "word": " Learning",
                    "probability": 0.8203125
                },
                {
                    "start": 3289.46,
                    "end": 3289.94,
                    "word": ".",
                    "probability": 0.7431640625
                },
                {
                    "start": 3289.94,
                    "end": 3290.06,
                    "word": " You",
                    "probability": 0.8837890625
                },
                {
                    "start": 3290.06,
                    "end": 3290.16,
                    "word": " can",
                    "probability": 0.89892578125
                },
                {
                    "start": 3290.16,
                    "end": 3290.34,
                    "word": " find",
                    "probability": 0.82666015625
                },
                {
                    "start": 3290.34,
                    "end": 3290.44,
                    "word": " it",
                    "probability": 0.8935546875
                },
                {
                    "start": 3290.44,
                    "end": 3290.58,
                    "word": " if",
                    "probability": 0.857421875
                },
                {
                    "start": 3290.58,
                    "end": 3290.84,
                    "word": " you",
                    "probability": 0.896484375
                },
                {
                    "start": 3290.84,
                    "end": 3291.92,
                    "word": " search",
                    "probability": 0.90576171875
                },
                {
                    "start": 3291.92,
                    "end": 3292.06,
                    "word": " my",
                    "probability": 0.86572265625
                },
                {
                    "start": 3292.06,
                    "end": 3292.22,
                    "word": " name",
                    "probability": 0.79052734375
                },
                {
                    "start": 3292.22,
                    "end": 3292.34,
                    "word": " on",
                    "probability": 0.90234375
                },
                {
                    "start": 3292.34,
                    "end": 3292.54,
                    "word": " Google",
                    "probability": 0.859375
                },
                {
                    "start": 3292.54,
                    "end": 3292.92,
                    "word": " Scholar",
                    "probability": 0.9208984375
                },
                {
                    "start": 3292.92,
                    "end": 3293.48,
                    "word": ".",
                    "probability": 0.95458984375
                }
            ]
        },
        {
            "id": 486,
            "seek": 330255,
            "start": 3293.94,
            "end": 3298.86,
            "text": " If you're interested, of course, all of these are published in separate papers in slightly",
            "tokens": [
                50996,
                759,
                291,
                434,
                3102,
                11,
                295,
                1164,
                11,
                439,
                295,
                613,
                366,
                6572,
                294,
                4994,
                10577,
                294,
                4748,
                51244
            ],
            "temperature": 0.0,
            "avg_logprob": -0.27428669514863385,
            "compression_ratio": 1.5120967741935485,
            "no_speech_prob": 0.0008425712585449219,
            "words": [
                {
                    "start": 3293.94,
                    "end": 3294.22,
                    "word": " If",
                    "probability": 0.87109375
                },
                {
                    "start": 3294.22,
                    "end": 3294.54,
                    "word": " you're",
                    "probability": 0.810546875
                },
                {
                    "start": 3294.54,
                    "end": 3294.88,
                    "word": " interested",
                    "probability": 0.650390625
                },
                {
                    "start": 3294.88,
                    "end": 3295.0,
                    "word": ",",
                    "probability": 0.85400390625
                },
                {
                    "start": 3295.0,
                    "end": 3295.08,
                    "word": " of",
                    "probability": 0.89892578125
                },
                {
                    "start": 3295.08,
                    "end": 3295.26,
                    "word": " course",
                    "probability": 0.86572265625
                },
                {
                    "start": 3295.26,
                    "end": 3295.42,
                    "word": ",",
                    "probability": 0.93115234375
                },
                {
                    "start": 3295.42,
                    "end": 3295.54,
                    "word": " all",
                    "probability": 0.88623046875
                },
                {
                    "start": 3295.54,
                    "end": 3295.62,
                    "word": " of",
                    "probability": 0.88623046875
                },
                {
                    "start": 3295.62,
                    "end": 3295.8,
                    "word": " these",
                    "probability": 0.7001953125
                },
                {
                    "start": 3295.8,
                    "end": 3296.34,
                    "word": " are",
                    "probability": 0.890625
                },
                {
                    "start": 3296.34,
                    "end": 3297.06,
                    "word": " published",
                    "probability": 0.74169921875
                },
                {
                    "start": 3297.06,
                    "end": 3297.22,
                    "word": " in",
                    "probability": 0.89501953125
                },
                {
                    "start": 3297.22,
                    "end": 3297.5,
                    "word": " separate",
                    "probability": 0.86474609375
                },
                {
                    "start": 3297.5,
                    "end": 3297.86,
                    "word": " papers",
                    "probability": 0.79541015625
                },
                {
                    "start": 3297.86,
                    "end": 3298.44,
                    "word": " in",
                    "probability": 0.77734375
                },
                {
                    "start": 3298.44,
                    "end": 3298.86,
                    "word": " slightly",
                    "probability": 0.779296875
                }
            ]
        },
        {
            "id": 487,
            "seek": 330255,
            "start": 3299.56,
            "end": 3308.68,
            "text": " more details under the same titles. And finally, some of the open areas of research in this domain.",
            "tokens": [
                51280,
                544,
                4365,
                833,
                264,
                912,
                12992,
                13,
                400,
                2721,
                11,
                512,
                295,
                264,
                1269,
                3179,
                295,
                2132,
                294,
                341,
                9274,
                13,
                51704
            ],
            "temperature": 0.0,
            "avg_logprob": -0.27428669514863385,
            "compression_ratio": 1.5120967741935485,
            "no_speech_prob": 0.0008425712585449219,
            "words": [
                {
                    "start": 3299.56,
                    "end": 3299.92,
                    "word": " more",
                    "probability": 0.8828125
                },
                {
                    "start": 3299.92,
                    "end": 3300.52,
                    "word": " details",
                    "probability": 0.61083984375
                },
                {
                    "start": 3300.52,
                    "end": 3301.1,
                    "word": " under",
                    "probability": 0.71240234375
                },
                {
                    "start": 3301.1,
                    "end": 3301.52,
                    "word": " the",
                    "probability": 0.80908203125
                },
                {
                    "start": 3301.52,
                    "end": 3301.72,
                    "word": " same",
                    "probability": 0.79833984375
                },
                {
                    "start": 3301.72,
                    "end": 3302.14,
                    "word": " titles",
                    "probability": 0.826171875
                },
                {
                    "start": 3302.14,
                    "end": 3302.54,
                    "word": ".",
                    "probability": 0.8955078125
                },
                {
                    "start": 3302.74,
                    "end": 3303.3,
                    "word": " And",
                    "probability": 0.85546875
                },
                {
                    "start": 3303.3,
                    "end": 3303.68,
                    "word": " finally",
                    "probability": 0.73291015625
                },
                {
                    "start": 3303.68,
                    "end": 3303.84,
                    "word": ",",
                    "probability": 0.84619140625
                },
                {
                    "start": 3303.84,
                    "end": 3304.02,
                    "word": " some",
                    "probability": 0.822265625
                },
                {
                    "start": 3304.02,
                    "end": 3304.3,
                    "word": " of",
                    "probability": 0.9052734375
                },
                {
                    "start": 3304.3,
                    "end": 3304.9,
                    "word": " the",
                    "probability": 0.81591796875
                },
                {
                    "start": 3304.9,
                    "end": 3305.58,
                    "word": " open",
                    "probability": 0.80126953125
                },
                {
                    "start": 3305.58,
                    "end": 3307.0,
                    "word": " areas",
                    "probability": 0.85302734375
                },
                {
                    "start": 3307.0,
                    "end": 3307.18,
                    "word": " of",
                    "probability": 0.90380859375
                },
                {
                    "start": 3307.18,
                    "end": 3307.48,
                    "word": " research",
                    "probability": 0.779296875
                },
                {
                    "start": 3307.48,
                    "end": 3307.66,
                    "word": " in",
                    "probability": 0.89599609375
                },
                {
                    "start": 3307.66,
                    "end": 3307.8,
                    "word": " this",
                    "probability": 0.88232421875
                },
                {
                    "start": 3307.8,
                    "end": 3308.12,
                    "word": " domain",
                    "probability": 0.876953125
                },
                {
                    "start": 3308.12,
                    "end": 3308.68,
                    "word": ".",
                    "probability": 0.9541015625
                }
            ]
        },
        {
            "id": 488,
            "seek": 333163,
            "start": 3309.4,
            "end": 3315.4,
            "text": " With regards to training time resilience and robustness, not much has been done with regards",
            "tokens": [
                50400,
                2022,
                14258,
                281,
                3097,
                565,
                19980,
                293,
                13956,
                1287,
                11,
                406,
                709,
                575,
                668,
                1096,
                365,
                14258,
                50696
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22376179203109922,
            "compression_ratio": 1.6338028169014085,
            "no_speech_prob": 0.0004801750183105469,
            "words": [
                {
                    "start": 3309.4,
                    "end": 3309.66,
                    "word": " With",
                    "probability": 0.59326171875
                },
                {
                    "start": 3309.66,
                    "end": 3309.94,
                    "word": " regards",
                    "probability": 0.77490234375
                },
                {
                    "start": 3309.94,
                    "end": 3310.12,
                    "word": " to",
                    "probability": 0.90087890625
                },
                {
                    "start": 3310.12,
                    "end": 3310.4,
                    "word": " training",
                    "probability": 0.80419921875
                },
                {
                    "start": 3310.4,
                    "end": 3310.62,
                    "word": " time",
                    "probability": 0.72998046875
                },
                {
                    "start": 3310.62,
                    "end": 3310.98,
                    "word": " resilience",
                    "probability": 0.541015625
                },
                {
                    "start": 3310.98,
                    "end": 3311.2,
                    "word": " and",
                    "probability": 0.86962890625
                },
                {
                    "start": 3311.2,
                    "end": 3312.02,
                    "word": " robustness",
                    "probability": 0.87109375
                },
                {
                    "start": 3312.02,
                    "end": 3312.72,
                    "word": ",",
                    "probability": 0.6669921875
                },
                {
                    "start": 3312.72,
                    "end": 3313.0,
                    "word": " not",
                    "probability": 0.65771484375
                },
                {
                    "start": 3313.0,
                    "end": 3314.06,
                    "word": " much",
                    "probability": 0.81494140625
                },
                {
                    "start": 3314.06,
                    "end": 3314.2,
                    "word": " has",
                    "probability": 0.89306640625
                },
                {
                    "start": 3314.2,
                    "end": 3314.32,
                    "word": " been",
                    "probability": 0.8173828125
                },
                {
                    "start": 3314.32,
                    "end": 3314.76,
                    "word": " done",
                    "probability": 0.8134765625
                },
                {
                    "start": 3314.76,
                    "end": 3315.18,
                    "word": " with",
                    "probability": 0.72216796875
                },
                {
                    "start": 3315.18,
                    "end": 3315.4,
                    "word": " regards",
                    "probability": 0.7890625
                }
            ]
        },
        {
            "id": 489,
            "seek": 333163,
            "start": 3315.4,
            "end": 3320.32,
            "text": " to policy search and actor-critic methods, as well as model-based and hybrid methods. Of course,",
            "tokens": [
                50696,
                281,
                3897,
                3164,
                293,
                8747,
                12,
                32255,
                299,
                7150,
                11,
                382,
                731,
                382,
                2316,
                12,
                6032,
                293,
                13051,
                7150,
                13,
                2720,
                1164,
                11,
                50940
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22376179203109922,
            "compression_ratio": 1.6338028169014085,
            "no_speech_prob": 0.0004801750183105469,
            "words": [
                {
                    "start": 3315.4,
                    "end": 3315.52,
                    "word": " to",
                    "probability": 0.890625
                },
                {
                    "start": 3315.52,
                    "end": 3315.8,
                    "word": " policy",
                    "probability": 0.775390625
                },
                {
                    "start": 3315.8,
                    "end": 3316.1,
                    "word": " search",
                    "probability": 0.873046875
                },
                {
                    "start": 3316.1,
                    "end": 3316.24,
                    "word": " and",
                    "probability": 0.87646484375
                },
                {
                    "start": 3316.24,
                    "end": 3316.5,
                    "word": " actor",
                    "probability": 0.59423828125
                },
                {
                    "start": 3316.5,
                    "end": 3316.8,
                    "word": "-critic",
                    "probability": 0.7486165364583334
                },
                {
                    "start": 3316.8,
                    "end": 3317.16,
                    "word": " methods",
                    "probability": 0.80517578125
                },
                {
                    "start": 3317.16,
                    "end": 3317.26,
                    "word": ",",
                    "probability": 0.51318359375
                },
                {
                    "start": 3317.26,
                    "end": 3317.42,
                    "word": " as",
                    "probability": 0.8974609375
                },
                {
                    "start": 3317.42,
                    "end": 3317.54,
                    "word": " well",
                    "probability": 0.876953125
                },
                {
                    "start": 3317.54,
                    "end": 3317.98,
                    "word": " as",
                    "probability": 0.90234375
                },
                {
                    "start": 3317.98,
                    "end": 3318.68,
                    "word": " model",
                    "probability": 0.798828125
                },
                {
                    "start": 3318.68,
                    "end": 3318.98,
                    "word": "-based",
                    "probability": 0.871337890625
                },
                {
                    "start": 3318.98,
                    "end": 3319.2,
                    "word": " and",
                    "probability": 0.88671875
                },
                {
                    "start": 3319.2,
                    "end": 3319.5,
                    "word": " hybrid",
                    "probability": 0.84375
                },
                {
                    "start": 3319.5,
                    "end": 3319.86,
                    "word": " methods",
                    "probability": 0.83984375
                },
                {
                    "start": 3319.86,
                    "end": 3320.04,
                    "word": ".",
                    "probability": 0.97900390625
                },
                {
                    "start": 3320.04,
                    "end": 3320.16,
                    "word": " Of",
                    "probability": 0.89453125
                },
                {
                    "start": 3320.16,
                    "end": 3320.28,
                    "word": " course",
                    "probability": 0.8583984375
                },
                {
                    "start": 3320.28,
                    "end": 3320.32,
                    "word": ",",
                    "probability": 0.82373046875
                }
            ]
        },
        {
            "id": 490,
            "seek": 333163,
            "start": 3320.32,
            "end": 3323.88,
            "text": " when we talk about model-based, there are some approaches from optimal control theory and",
            "tokens": [
                50940,
                562,
                321,
                751,
                466,
                2316,
                12,
                6032,
                11,
                456,
                366,
                512,
                11587,
                490,
                16252,
                1969,
                5261,
                293,
                51124
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22376179203109922,
            "compression_ratio": 1.6338028169014085,
            "no_speech_prob": 0.0004801750183105469,
            "words": [
                {
                    "start": 3320.32,
                    "end": 3320.44,
                    "word": " when",
                    "probability": 0.861328125
                },
                {
                    "start": 3320.44,
                    "end": 3320.56,
                    "word": " we",
                    "probability": 0.9013671875
                },
                {
                    "start": 3320.56,
                    "end": 3320.7,
                    "word": " talk",
                    "probability": 0.80419921875
                },
                {
                    "start": 3320.7,
                    "end": 3320.86,
                    "word": " about",
                    "probability": 0.8115234375
                },
                {
                    "start": 3320.86,
                    "end": 3321.08,
                    "word": " model",
                    "probability": 0.77783203125
                },
                {
                    "start": 3321.08,
                    "end": 3321.42,
                    "word": "-based",
                    "probability": 0.867919921875
                },
                {
                    "start": 3321.42,
                    "end": 3321.5,
                    "word": ",",
                    "probability": 0.90966796875
                },
                {
                    "start": 3321.5,
                    "end": 3321.74,
                    "word": " there",
                    "probability": 0.84716796875
                },
                {
                    "start": 3321.74,
                    "end": 3321.88,
                    "word": " are",
                    "probability": 0.89208984375
                },
                {
                    "start": 3321.88,
                    "end": 3322.06,
                    "word": " some",
                    "probability": 0.8134765625
                },
                {
                    "start": 3322.06,
                    "end": 3322.34,
                    "word": " approaches",
                    "probability": 0.66552734375
                },
                {
                    "start": 3322.34,
                    "end": 3322.64,
                    "word": " from",
                    "probability": 0.78662109375
                },
                {
                    "start": 3322.64,
                    "end": 3322.94,
                    "word": " optimal",
                    "probability": 0.60107421875
                },
                {
                    "start": 3322.94,
                    "end": 3323.28,
                    "word": " control",
                    "probability": 0.62109375
                },
                {
                    "start": 3323.28,
                    "end": 3323.6,
                    "word": " theory",
                    "probability": 0.7861328125
                },
                {
                    "start": 3323.6,
                    "end": 3323.88,
                    "word": " and",
                    "probability": 0.85595703125
                }
            ]
        },
        {
            "id": 491,
            "seek": 333163,
            "start": 3325.62,
            "end": 3333.14,
            "text": " approximate dynamic programming that may be applied here, but very few have looked at this",
            "tokens": [
                51208,
                30874,
                8546,
                9410,
                300,
                815,
                312,
                6456,
                510,
                11,
                457,
                588,
                1326,
                362,
                2956,
                412,
                341,
                51592
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22376179203109922,
            "compression_ratio": 1.6338028169014085,
            "no_speech_prob": 0.0004801750183105469,
            "words": [
                {
                    "start": 3325.62,
                    "end": 3326.14,
                    "word": " approximate",
                    "probability": 0.7685546875
                },
                {
                    "start": 3326.14,
                    "end": 3326.62,
                    "word": " dynamic",
                    "probability": 0.70361328125
                },
                {
                    "start": 3326.62,
                    "end": 3327.12,
                    "word": " programming",
                    "probability": 0.70849609375
                },
                {
                    "start": 3327.12,
                    "end": 3327.5,
                    "word": " that",
                    "probability": 0.86083984375
                },
                {
                    "start": 3327.5,
                    "end": 3327.72,
                    "word": " may",
                    "probability": 0.79345703125
                },
                {
                    "start": 3327.72,
                    "end": 3327.82,
                    "word": " be",
                    "probability": 0.892578125
                },
                {
                    "start": 3327.82,
                    "end": 3328.08,
                    "word": " applied",
                    "probability": 0.67626953125
                },
                {
                    "start": 3328.08,
                    "end": 3328.36,
                    "word": " here",
                    "probability": 0.80029296875
                },
                {
                    "start": 3328.36,
                    "end": 3328.8,
                    "word": ",",
                    "probability": 0.6103515625
                },
                {
                    "start": 3328.8,
                    "end": 3329.18,
                    "word": " but",
                    "probability": 0.89697265625
                },
                {
                    "start": 3329.18,
                    "end": 3330.5,
                    "word": " very",
                    "probability": 0.72900390625
                },
                {
                    "start": 3330.5,
                    "end": 3330.74,
                    "word": " few",
                    "probability": 0.81591796875
                },
                {
                    "start": 3330.74,
                    "end": 3331.14,
                    "word": " have",
                    "probability": 0.880859375
                },
                {
                    "start": 3331.14,
                    "end": 3331.96,
                    "word": " looked",
                    "probability": 0.7333984375
                },
                {
                    "start": 3331.96,
                    "end": 3332.26,
                    "word": " at",
                    "probability": 0.91162109375
                },
                {
                    "start": 3332.26,
                    "end": 3333.14,
                    "word": " this",
                    "probability": 0.8759765625
                }
            ]
        },
        {
            "id": 492,
            "seek": 333163,
            "start": 3333.14,
            "end": 3337.76,
            "text": " problem from a security point of view. So if you're interested, this is one of the areas that",
            "tokens": [
                51592,
                1154,
                490,
                257,
                3825,
                935,
                295,
                1910,
                13,
                407,
                498,
                291,
                434,
                3102,
                11,
                341,
                307,
                472,
                295,
                264,
                3179,
                300,
                51816
            ],
            "temperature": 0.0,
            "avg_logprob": -0.22376179203109922,
            "compression_ratio": 1.6338028169014085,
            "no_speech_prob": 0.0004801750183105469,
            "words": [
                {
                    "start": 3333.14,
                    "end": 3333.72,
                    "word": " problem",
                    "probability": 0.77587890625
                },
                {
                    "start": 3333.72,
                    "end": 3334.38,
                    "word": " from",
                    "probability": 0.82373046875
                },
                {
                    "start": 3334.54,
                    "end": 3334.86,
                    "word": " a security",
                    "probability": 0.9384765625
                },
                {
                    "start": 3334.86,
                    "end": 3335.1,
                    "word": " point",
                    "probability": 0.84716796875
                },
                {
                    "start": 3335.1,
                    "end": 3335.26,
                    "word": " of",
                    "probability": 0.90087890625
                },
                {
                    "start": 3335.26,
                    "end": 3335.38,
                    "word": " view",
                    "probability": 0.79345703125
                },
                {
                    "start": 3335.38,
                    "end": 3335.52,
                    "word": ".",
                    "probability": 0.931640625
                },
                {
                    "start": 3335.52,
                    "end": 3335.62,
                    "word": " So",
                    "probability": 0.82958984375
                },
                {
                    "start": 3335.62,
                    "end": 3335.76,
                    "word": " if",
                    "probability": 0.63671875
                },
                {
                    "start": 3335.76,
                    "end": 3335.94,
                    "word": " you're",
                    "probability": 0.90283203125
                },
                {
                    "start": 3335.94,
                    "end": 3336.2,
                    "word": " interested",
                    "probability": 0.6240234375
                },
                {
                    "start": 3336.2,
                    "end": 3336.28,
                    "word": ",",
                    "probability": 0.96875
                },
                {
                    "start": 3336.28,
                    "end": 3336.44,
                    "word": " this",
                    "probability": 0.8759765625
                },
                {
                    "start": 3336.44,
                    "end": 3336.62,
                    "word": " is",
                    "probability": 0.90625
                },
                {
                    "start": 3336.62,
                    "end": 3336.84,
                    "word": " one",
                    "probability": 0.8095703125
                },
                {
                    "start": 3336.84,
                    "end": 3336.9,
                    "word": " of",
                    "probability": 0.90087890625
                },
                {
                    "start": 3336.9,
                    "end": 3337.02,
                    "word": " the",
                    "probability": 0.81396484375
                },
                {
                    "start": 3337.02,
                    "end": 3337.38,
                    "word": " areas",
                    "probability": 0.87939453125
                },
                {
                    "start": 3337.38,
                    "end": 3337.76,
                    "word": " that",
                    "probability": 0.8828125
                }
            ]
        },
        {
            "id": 493,
            "seek": 335997,
            "start": 3337.76,
            "end": 3344.26,
            "text": " is in dire need of a security-oriented investigation. As for mitigation of policy",
            "tokens": [
                50364,
                307,
                294,
                1264,
                643,
                295,
                257,
                3825,
                12,
                27414,
                9627,
                13,
                1018,
                337,
                32649,
                295,
                3897,
                50688
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2240131588358628,
            "compression_ratio": 1.7290076335877862,
            "no_speech_prob": 8.749961853027344e-05,
            "words": [
                {
                    "start": 3337.76,
                    "end": 3337.94,
                    "word": " is",
                    "probability": 0.368896484375
                },
                {
                    "start": 3337.94,
                    "end": 3338.08,
                    "word": " in",
                    "probability": 0.87548828125
                },
                {
                    "start": 3338.08,
                    "end": 3338.32,
                    "word": " dire",
                    "probability": 0.8740234375
                },
                {
                    "start": 3338.32,
                    "end": 3338.74,
                    "word": " need",
                    "probability": 0.83349609375
                },
                {
                    "start": 3338.74,
                    "end": 3339.28,
                    "word": " of",
                    "probability": 0.90185546875
                },
                {
                    "start": 3340.18,
                    "end": 3341.18,
                    "word": " a security",
                    "probability": 0.93212890625
                },
                {
                    "start": 3341.18,
                    "end": 3341.54,
                    "word": "-oriented",
                    "probability": 0.647705078125
                },
                {
                    "start": 3341.54,
                    "end": 3342.16,
                    "word": " investigation",
                    "probability": 0.671875
                },
                {
                    "start": 3342.16,
                    "end": 3342.78,
                    "word": ".",
                    "probability": 0.94921875
                },
                {
                    "start": 3342.78,
                    "end": 3343.02,
                    "word": " As",
                    "probability": 0.8662109375
                },
                {
                    "start": 3343.02,
                    "end": 3343.26,
                    "word": " for",
                    "probability": 0.89208984375
                },
                {
                    "start": 3343.26,
                    "end": 3343.78,
                    "word": " mitigation",
                    "probability": 0.7314453125
                },
                {
                    "start": 3343.78,
                    "end": 3344.0,
                    "word": " of",
                    "probability": 0.75927734375
                },
                {
                    "start": 3344.0,
                    "end": 3344.26,
                    "word": " policy",
                    "probability": 0.80224609375
                }
            ]
        },
        {
            "id": 494,
            "seek": 335997,
            "start": 3344.26,
            "end": 3347.5,
            "text": " replication, one of the ideas that my research group is currently working on is constrained",
            "tokens": [
                50688,
                39911,
                11,
                472,
                295,
                264,
                3487,
                300,
                452,
                2132,
                1594,
                307,
                4362,
                1364,
                322,
                307,
                38901,
                50852
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2240131588358628,
            "compression_ratio": 1.7290076335877862,
            "no_speech_prob": 8.749961853027344e-05,
            "words": [
                {
                    "start": 3344.26,
                    "end": 3344.66,
                    "word": " replication",
                    "probability": 0.78662109375
                },
                {
                    "start": 3344.66,
                    "end": 3344.94,
                    "word": ",",
                    "probability": 0.8701171875
                },
                {
                    "start": 3344.94,
                    "end": 3345.04,
                    "word": " one",
                    "probability": 0.8095703125
                },
                {
                    "start": 3345.04,
                    "end": 3345.12,
                    "word": " of",
                    "probability": 0.89501953125
                },
                {
                    "start": 3345.12,
                    "end": 3345.2,
                    "word": " the",
                    "probability": 0.81298828125
                },
                {
                    "start": 3345.2,
                    "end": 3345.46,
                    "word": " ideas",
                    "probability": 0.775390625
                },
                {
                    "start": 3345.46,
                    "end": 3345.66,
                    "word": " that",
                    "probability": 0.85546875
                },
                {
                    "start": 3345.66,
                    "end": 3345.84,
                    "word": " my",
                    "probability": 0.9052734375
                },
                {
                    "start": 3345.84,
                    "end": 3346.08,
                    "word": " research",
                    "probability": 0.72021484375
                },
                {
                    "start": 3346.08,
                    "end": 3346.2,
                    "word": " group",
                    "probability": 0.833984375
                },
                {
                    "start": 3346.2,
                    "end": 3346.3,
                    "word": " is",
                    "probability": 0.90234375
                },
                {
                    "start": 3346.3,
                    "end": 3346.5,
                    "word": " currently",
                    "probability": 0.72265625
                },
                {
                    "start": 3346.5,
                    "end": 3346.8,
                    "word": " working",
                    "probability": 0.72607421875
                },
                {
                    "start": 3346.8,
                    "end": 3346.96,
                    "word": " on",
                    "probability": 0.90966796875
                },
                {
                    "start": 3346.96,
                    "end": 3347.08,
                    "word": " is",
                    "probability": 0.88330078125
                },
                {
                    "start": 3347.08,
                    "end": 3347.5,
                    "word": " constrained",
                    "probability": 0.50634765625
                }
            ]
        },
        {
            "id": 495,
            "seek": 335997,
            "start": 3347.5,
            "end": 3354.28,
            "text": " randomization of policy. So randomize the policy such that the replication through techniques like",
            "tokens": [
                50852,
                4974,
                2144,
                295,
                3897,
                13,
                407,
                4974,
                1125,
                264,
                3897,
                1270,
                300,
                264,
                39911,
                807,
                7512,
                411,
                51188
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2240131588358628,
            "compression_ratio": 1.7290076335877862,
            "no_speech_prob": 8.749961853027344e-05,
            "words": [
                {
                    "start": 3347.5,
                    "end": 3348.2,
                    "word": " randomization",
                    "probability": 0.8125
                },
                {
                    "start": 3348.2,
                    "end": 3348.46,
                    "word": " of",
                    "probability": 0.9013671875
                },
                {
                    "start": 3348.46,
                    "end": 3348.86,
                    "word": " policy",
                    "probability": 0.56103515625
                },
                {
                    "start": 3348.86,
                    "end": 3349.08,
                    "word": ".",
                    "probability": 0.8173828125
                },
                {
                    "start": 3349.08,
                    "end": 3349.18,
                    "word": " So",
                    "probability": 0.84228515625
                },
                {
                    "start": 3349.18,
                    "end": 3349.7,
                    "word": " randomize",
                    "probability": 0.689697265625
                },
                {
                    "start": 3349.7,
                    "end": 3349.82,
                    "word": " the",
                    "probability": 0.69091796875
                },
                {
                    "start": 3349.82,
                    "end": 3350.2,
                    "word": " policy",
                    "probability": 0.81005859375
                },
                {
                    "start": 3350.2,
                    "end": 3351.18,
                    "word": " such",
                    "probability": 0.71875
                },
                {
                    "start": 3351.18,
                    "end": 3351.5,
                    "word": " that",
                    "probability": 0.8759765625
                },
                {
                    "start": 3351.5,
                    "end": 3352.22,
                    "word": " the",
                    "probability": 0.7880859375
                },
                {
                    "start": 3352.22,
                    "end": 3353.2,
                    "word": " replication",
                    "probability": 0.7958984375
                },
                {
                    "start": 3353.2,
                    "end": 3353.68,
                    "word": " through",
                    "probability": 0.69677734375
                },
                {
                    "start": 3353.68,
                    "end": 3354.04,
                    "word": " techniques",
                    "probability": 0.8583984375
                },
                {
                    "start": 3354.04,
                    "end": 3354.28,
                    "word": " like",
                    "probability": 0.8642578125
                }
            ]
        },
        {
            "id": 496,
            "seek": 335997,
            "start": 3354.28,
            "end": 3358.66,
            "text": " imitation learning becomes more costly, more samples, more observations will be required,",
            "tokens": [
                51188,
                47624,
                2539,
                3643,
                544,
                28328,
                11,
                544,
                10938,
                11,
                544,
                18163,
                486,
                312,
                4739,
                11,
                51380
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2240131588358628,
            "compression_ratio": 1.7290076335877862,
            "no_speech_prob": 8.749961853027344e-05,
            "words": [
                {
                    "start": 3354.28,
                    "end": 3354.66,
                    "word": " imitation",
                    "probability": 0.8486328125
                },
                {
                    "start": 3354.66,
                    "end": 3355.0,
                    "word": " learning",
                    "probability": 0.8486328125
                },
                {
                    "start": 3355.0,
                    "end": 3355.36,
                    "word": " becomes",
                    "probability": 0.77978515625
                },
                {
                    "start": 3355.36,
                    "end": 3355.66,
                    "word": " more",
                    "probability": 0.88232421875
                },
                {
                    "start": 3355.66,
                    "end": 3356.06,
                    "word": " costly",
                    "probability": 0.845703125
                },
                {
                    "start": 3356.06,
                    "end": 3356.18,
                    "word": ",",
                    "probability": 0.67529296875
                },
                {
                    "start": 3356.18,
                    "end": 3356.3,
                    "word": " more",
                    "probability": 0.88623046875
                },
                {
                    "start": 3356.3,
                    "end": 3356.66,
                    "word": " samples",
                    "probability": 0.763671875
                },
                {
                    "start": 3356.66,
                    "end": 3356.76,
                    "word": ",",
                    "probability": 0.86865234375
                },
                {
                    "start": 3356.76,
                    "end": 3356.84,
                    "word": " more",
                    "probability": 0.88330078125
                },
                {
                    "start": 3356.84,
                    "end": 3357.24,
                    "word": " observations",
                    "probability": 0.6611328125
                },
                {
                    "start": 3357.24,
                    "end": 3357.58,
                    "word": " will",
                    "probability": 0.787109375
                },
                {
                    "start": 3357.58,
                    "end": 3357.68,
                    "word": " be",
                    "probability": 0.904296875
                },
                {
                    "start": 3357.68,
                    "end": 3358.1,
                    "word": " required",
                    "probability": 0.7236328125
                },
                {
                    "start": 3358.1,
                    "end": 3358.66,
                    "word": ",",
                    "probability": 0.468994140625
                }
            ]
        },
        {
            "id": 497,
            "seek": 335997,
            "start": 3358.66,
            "end": 3366.1,
            "text": " while preserving the performance of the policy. There's almost no work done in multi-agent",
            "tokens": [
                51412,
                1339,
                33173,
                264,
                3389,
                295,
                264,
                3897,
                13,
                821,
                311,
                1920,
                572,
                589,
                1096,
                294,
                4825,
                12,
                559,
                317,
                51780
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2240131588358628,
            "compression_ratio": 1.7290076335877862,
            "no_speech_prob": 8.749961853027344e-05,
            "words": [
                {
                    "start": 3358.66,
                    "end": 3359.12,
                    "word": " while",
                    "probability": 0.91259765625
                },
                {
                    "start": 3359.12,
                    "end": 3359.8,
                    "word": " preserving",
                    "probability": 0.8603515625
                },
                {
                    "start": 3359.8,
                    "end": 3360.32,
                    "word": " the",
                    "probability": 0.818359375
                },
                {
                    "start": 3360.32,
                    "end": 3360.72,
                    "word": " performance",
                    "probability": 0.8544921875
                },
                {
                    "start": 3360.72,
                    "end": 3361.52,
                    "word": " of",
                    "probability": 0.90673828125
                },
                {
                    "start": 3361.52,
                    "end": 3362.24,
                    "word": " the",
                    "probability": 0.806640625
                },
                {
                    "start": 3362.24,
                    "end": 3362.6,
                    "word": " policy",
                    "probability": 0.8291015625
                },
                {
                    "start": 3362.6,
                    "end": 3363.28,
                    "word": ".",
                    "probability": 0.99560546875
                },
                {
                    "start": 3363.56,
                    "end": 3364.24,
                    "word": " There's",
                    "probability": 0.6572265625
                },
                {
                    "start": 3364.24,
                    "end": 3364.76,
                    "word": " almost",
                    "probability": 0.66259765625
                },
                {
                    "start": 3364.76,
                    "end": 3365.12,
                    "word": " no",
                    "probability": 0.90673828125
                },
                {
                    "start": 3365.12,
                    "end": 3365.4,
                    "word": " work",
                    "probability": 0.80078125
                },
                {
                    "start": 3365.4,
                    "end": 3365.58,
                    "word": " done",
                    "probability": 0.80029296875
                },
                {
                    "start": 3365.58,
                    "end": 3365.72,
                    "word": " in",
                    "probability": 0.904296875
                },
                {
                    "start": 3365.72,
                    "end": 3365.94,
                    "word": " multi",
                    "probability": 0.73828125
                },
                {
                    "start": 3365.94,
                    "end": 3366.1,
                    "word": "-agent",
                    "probability": 0.8426106770833334
                }
            ]
        },
        {
            "id": 498,
            "seek": 338871,
            "start": 3366.1,
            "end": 3370.6,
            "text": " settings. Of course, adversarial reinforcement learning has been investigated in settings where",
            "tokens": [
                50364,
                6257,
                13,
                2720,
                1164,
                11,
                17641,
                44745,
                29280,
                2539,
                575,
                668,
                30070,
                294,
                6257,
                689,
                50592
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2241115182930348,
            "compression_ratio": 1.7518796992481203,
            "no_speech_prob": 0.00130462646484375,
            "words": [
                {
                    "start": 3366.1,
                    "end": 3366.38,
                    "word": " settings",
                    "probability": 0.08941650390625
                },
                {
                    "start": 3366.38,
                    "end": 3366.38,
                    "word": ".",
                    "probability": 0.5498046875
                },
                {
                    "start": 3366.38,
                    "end": 3366.56,
                    "word": " Of",
                    "probability": 0.8310546875
                },
                {
                    "start": 3366.56,
                    "end": 3366.7,
                    "word": " course",
                    "probability": 0.86279296875
                },
                {
                    "start": 3366.7,
                    "end": 3366.8,
                    "word": ",",
                    "probability": 0.5546875
                },
                {
                    "start": 3366.8,
                    "end": 3367.54,
                    "word": " adversarial",
                    "probability": 0.921875
                },
                {
                    "start": 3367.54,
                    "end": 3368.12,
                    "word": " reinforcement",
                    "probability": 0.80517578125
                },
                {
                    "start": 3368.12,
                    "end": 3368.48,
                    "word": " learning",
                    "probability": 0.83251953125
                },
                {
                    "start": 3368.48,
                    "end": 3368.74,
                    "word": " has",
                    "probability": 0.90185546875
                },
                {
                    "start": 3368.74,
                    "end": 3368.96,
                    "word": " been",
                    "probability": 0.81689453125
                },
                {
                    "start": 3368.96,
                    "end": 3369.56,
                    "word": " investigated",
                    "probability": 0.78125
                },
                {
                    "start": 3369.56,
                    "end": 3370.06,
                    "word": " in",
                    "probability": 0.89111328125
                },
                {
                    "start": 3370.06,
                    "end": 3370.34,
                    "word": " settings",
                    "probability": 0.81787109375
                },
                {
                    "start": 3370.34,
                    "end": 3370.6,
                    "word": " where",
                    "probability": 0.89404296875
                }
            ]
        },
        {
            "id": 499,
            "seek": 338871,
            "start": 3370.6,
            "end": 3375.6,
            "text": " there are zero-sum agents, but not really where there is an external adversary or adversaries",
            "tokens": [
                50592,
                456,
                366,
                4018,
                12,
                82,
                449,
                12554,
                11,
                457,
                406,
                534,
                689,
                456,
                307,
                364,
                8320,
                48222,
                420,
                17641,
                4889,
                50848
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2241115182930348,
            "compression_ratio": 1.7518796992481203,
            "no_speech_prob": 0.00130462646484375,
            "words": [
                {
                    "start": 3370.6,
                    "end": 3370.9,
                    "word": " there",
                    "probability": 0.83447265625
                },
                {
                    "start": 3370.9,
                    "end": 3371.02,
                    "word": " are",
                    "probability": 0.82275390625
                },
                {
                    "start": 3371.02,
                    "end": 3371.28,
                    "word": " zero",
                    "probability": 0.72900390625
                },
                {
                    "start": 3371.28,
                    "end": 3371.52,
                    "word": "-sum",
                    "probability": 0.81396484375
                },
                {
                    "start": 3371.52,
                    "end": 3371.88,
                    "word": " agents",
                    "probability": 0.80224609375
                },
                {
                    "start": 3371.88,
                    "end": 3372.08,
                    "word": ",",
                    "probability": 0.6025390625
                },
                {
                    "start": 3372.08,
                    "end": 3372.26,
                    "word": " but",
                    "probability": 0.89990234375
                },
                {
                    "start": 3372.26,
                    "end": 3372.78,
                    "word": " not",
                    "probability": 0.87939453125
                },
                {
                    "start": 3372.78,
                    "end": 3373.02,
                    "word": " really",
                    "probability": 0.77392578125
                },
                {
                    "start": 3373.02,
                    "end": 3373.28,
                    "word": " where",
                    "probability": 0.8955078125
                },
                {
                    "start": 3373.28,
                    "end": 3373.56,
                    "word": " there",
                    "probability": 0.85693359375
                },
                {
                    "start": 3373.56,
                    "end": 3373.68,
                    "word": " is",
                    "probability": 0.8427734375
                },
                {
                    "start": 3373.68,
                    "end": 3373.78,
                    "word": " an",
                    "probability": 0.904296875
                },
                {
                    "start": 3373.78,
                    "end": 3374.06,
                    "word": " external",
                    "probability": 0.8232421875
                },
                {
                    "start": 3374.06,
                    "end": 3374.64,
                    "word": " adversary",
                    "probability": 0.79541015625
                },
                {
                    "start": 3374.64,
                    "end": 3374.92,
                    "word": " or",
                    "probability": 0.78173828125
                },
                {
                    "start": 3374.92,
                    "end": 3375.6,
                    "word": " adversaries",
                    "probability": 0.902099609375
                }
            ]
        },
        {
            "id": 500,
            "seek": 338871,
            "start": 3375.6,
            "end": 3381.26,
            "text": " trying to exploit the inner workings of the agents, the RL components of the agent.",
            "tokens": [
                50848,
                1382,
                281,
                25924,
                264,
                7284,
                589,
                1109,
                295,
                264,
                12554,
                11,
                264,
                497,
                43,
                6677,
                295,
                264,
                9461,
                13,
                51108
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2241115182930348,
            "compression_ratio": 1.7518796992481203,
            "no_speech_prob": 0.00130462646484375,
            "words": [
                {
                    "start": 3375.6,
                    "end": 3376.02,
                    "word": " trying",
                    "probability": 0.759765625
                },
                {
                    "start": 3376.02,
                    "end": 3376.76,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 3376.76,
                    "end": 3377.32,
                    "word": " exploit",
                    "probability": 0.73291015625
                },
                {
                    "start": 3377.32,
                    "end": 3377.74,
                    "word": " the",
                    "probability": 0.81201171875
                },
                {
                    "start": 3377.74,
                    "end": 3378.14,
                    "word": " inner",
                    "probability": 0.85693359375
                },
                {
                    "start": 3378.14,
                    "end": 3378.58,
                    "word": " workings",
                    "probability": 0.927978515625
                },
                {
                    "start": 3378.58,
                    "end": 3378.92,
                    "word": " of",
                    "probability": 0.9052734375
                },
                {
                    "start": 3378.92,
                    "end": 3379.12,
                    "word": " the",
                    "probability": 0.8154296875
                },
                {
                    "start": 3379.12,
                    "end": 3379.42,
                    "word": " agents",
                    "probability": 0.56591796875
                },
                {
                    "start": 3379.42,
                    "end": 3379.6,
                    "word": ",",
                    "probability": 0.80859375
                },
                {
                    "start": 3379.6,
                    "end": 3379.7,
                    "word": " the",
                    "probability": 0.82275390625
                },
                {
                    "start": 3379.7,
                    "end": 3380.0,
                    "word": " RL",
                    "probability": 0.832275390625
                },
                {
                    "start": 3380.0,
                    "end": 3380.38,
                    "word": " components",
                    "probability": 0.81494140625
                },
                {
                    "start": 3380.38,
                    "end": 3380.56,
                    "word": " of",
                    "probability": 0.904296875
                },
                {
                    "start": 3380.56,
                    "end": 3380.66,
                    "word": " the",
                    "probability": 0.81494140625
                },
                {
                    "start": 3380.66,
                    "end": 3380.96,
                    "word": " agent",
                    "probability": 0.5146484375
                },
                {
                    "start": 3380.96,
                    "end": 3381.26,
                    "word": ".",
                    "probability": 0.955078125
                }
            ]
        },
        {
            "id": 501,
            "seek": 338871,
            "start": 3381.96,
            "end": 3387.42,
            "text": " One more thing that is of note is the importance of discounting. The addiction problem that I",
            "tokens": [
                51176,
                1485,
                544,
                551,
                300,
                307,
                295,
                3637,
                307,
                264,
                7379,
                295,
                11635,
                278,
                13,
                440,
                16835,
                1154,
                300,
                286,
                51424
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2241115182930348,
            "compression_ratio": 1.7518796992481203,
            "no_speech_prob": 0.00130462646484375,
            "words": [
                {
                    "start": 3381.96,
                    "end": 3382.56,
                    "word": " One",
                    "probability": 0.87060546875
                },
                {
                    "start": 3382.56,
                    "end": 3382.74,
                    "word": " more",
                    "probability": 0.87060546875
                },
                {
                    "start": 3382.74,
                    "end": 3382.88,
                    "word": " thing",
                    "probability": 0.8251953125
                },
                {
                    "start": 3382.88,
                    "end": 3383.02,
                    "word": " that",
                    "probability": 0.76318359375
                },
                {
                    "start": 3383.02,
                    "end": 3383.34,
                    "word": " is",
                    "probability": 0.880859375
                },
                {
                    "start": 3383.34,
                    "end": 3383.58,
                    "word": " of",
                    "probability": 0.8935546875
                },
                {
                    "start": 3383.58,
                    "end": 3383.96,
                    "word": " note",
                    "probability": 0.81787109375
                },
                {
                    "start": 3383.96,
                    "end": 3384.42,
                    "word": " is",
                    "probability": 0.8896484375
                },
                {
                    "start": 3384.42,
                    "end": 3384.6,
                    "word": " the",
                    "probability": 0.80712890625
                },
                {
                    "start": 3384.6,
                    "end": 3384.86,
                    "word": " importance",
                    "probability": 0.8984375
                },
                {
                    "start": 3384.86,
                    "end": 3385.1,
                    "word": " of",
                    "probability": 0.90380859375
                },
                {
                    "start": 3385.1,
                    "end": 3385.76,
                    "word": " discounting",
                    "probability": 0.888427734375
                },
                {
                    "start": 3385.76,
                    "end": 3385.8,
                    "word": ".",
                    "probability": 0.77978515625
                },
                {
                    "start": 3385.8,
                    "end": 3385.9,
                    "word": " The",
                    "probability": 0.80126953125
                },
                {
                    "start": 3385.9,
                    "end": 3386.24,
                    "word": " addiction",
                    "probability": 0.71826171875
                },
                {
                    "start": 3386.24,
                    "end": 3386.7,
                    "word": " problem",
                    "probability": 0.78466796875
                },
                {
                    "start": 3386.7,
                    "end": 3387.14,
                    "word": " that",
                    "probability": 0.88330078125
                },
                {
                    "start": 3387.14,
                    "end": 3387.42,
                    "word": " I",
                    "probability": 0.9951171875
                }
            ]
        },
        {
            "id": 502,
            "seek": 338871,
            "start": 3387.42,
            "end": 3394.84,
            "text": " demonstrated earlier in the snake agent is mostly due to the constrained discounting solution. For",
            "tokens": [
                51424,
                18772,
                3071,
                294,
                264,
                12650,
                9461,
                307,
                5240,
                3462,
                281,
                264,
                38901,
                11635,
                278,
                3827,
                13,
                1171,
                51796
            ],
            "temperature": 0.0,
            "avg_logprob": -0.2241115182930348,
            "compression_ratio": 1.7518796992481203,
            "no_speech_prob": 0.00130462646484375,
            "words": [
                {
                    "start": 3387.42,
                    "end": 3387.86,
                    "word": " demonstrated",
                    "probability": 0.83203125
                },
                {
                    "start": 3387.86,
                    "end": 3388.28,
                    "word": " earlier",
                    "probability": 0.91650390625
                },
                {
                    "start": 3388.28,
                    "end": 3388.72,
                    "word": " in",
                    "probability": 0.853515625
                },
                {
                    "start": 3388.72,
                    "end": 3389.14,
                    "word": " the",
                    "probability": 0.81103515625
                },
                {
                    "start": 3389.14,
                    "end": 3389.6,
                    "word": " snake",
                    "probability": 0.5908203125
                },
                {
                    "start": 3389.6,
                    "end": 3390.08,
                    "word": " agent",
                    "probability": 0.440673828125
                },
                {
                    "start": 3390.08,
                    "end": 3391.02,
                    "word": " is",
                    "probability": 0.76025390625
                },
                {
                    "start": 3391.02,
                    "end": 3391.8,
                    "word": " mostly",
                    "probability": 0.7939453125
                },
                {
                    "start": 3391.8,
                    "end": 3392.12,
                    "word": " due",
                    "probability": 0.89208984375
                },
                {
                    "start": 3392.12,
                    "end": 3392.5,
                    "word": " to",
                    "probability": 0.90380859375
                },
                {
                    "start": 3392.5,
                    "end": 3392.88,
                    "word": " the",
                    "probability": 0.81494140625
                },
                {
                    "start": 3392.88,
                    "end": 3393.32,
                    "word": " constrained",
                    "probability": 0.172119140625
                },
                {
                    "start": 3393.32,
                    "end": 3394.2,
                    "word": " discounting",
                    "probability": 0.852783203125
                },
                {
                    "start": 3394.2,
                    "end": 3394.6,
                    "word": " solution",
                    "probability": 0.84228515625
                },
                {
                    "start": 3394.6,
                    "end": 3394.7,
                    "word": ".",
                    "probability": 0.900390625
                },
                {
                    "start": 3394.7,
                    "end": 3394.84,
                    "word": " For",
                    "probability": 0.896484375
                }
            ]
        },
        {
            "id": 503,
            "seek": 341693,
            "start": 3394.84,
            "end": 3397.46,
            "text": " those of us who come from a reinforcement learning background, you're familiar with",
            "tokens": [
                50364,
                729,
                295,
                505,
                567,
                808,
                490,
                257,
                29280,
                2539,
                3678,
                11,
                291,
                434,
                4963,
                365,
                50504
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24272895390444463,
            "compression_ratio": 1.7032967032967032,
            "no_speech_prob": 0.0030269622802734375,
            "words": [
                {
                    "start": 3394.84,
                    "end": 3395.0,
                    "word": " those",
                    "probability": 0.168701171875
                },
                {
                    "start": 3395.0,
                    "end": 3395.12,
                    "word": " of",
                    "probability": 0.880859375
                },
                {
                    "start": 3395.12,
                    "end": 3395.26,
                    "word": " us",
                    "probability": 0.89599609375
                },
                {
                    "start": 3395.26,
                    "end": 3395.38,
                    "word": " who",
                    "probability": 0.80908203125
                },
                {
                    "start": 3395.38,
                    "end": 3395.58,
                    "word": " come",
                    "probability": 0.771484375
                },
                {
                    "start": 3395.58,
                    "end": 3395.76,
                    "word": " from",
                    "probability": 0.8173828125
                },
                {
                    "start": 3395.86,
                    "end": 3396.16,
                    "word": " a reinforcement",
                    "probability": 0.74462890625
                },
                {
                    "start": 3396.16,
                    "end": 3396.38,
                    "word": " learning",
                    "probability": 0.755859375
                },
                {
                    "start": 3396.38,
                    "end": 3396.62,
                    "word": " background",
                    "probability": 0.7744140625
                },
                {
                    "start": 3396.62,
                    "end": 3396.72,
                    "word": ",",
                    "probability": 0.2069091796875
                },
                {
                    "start": 3396.72,
                    "end": 3396.76,
                    "word": " you're",
                    "probability": 0.816650390625
                },
                {
                    "start": 3396.76,
                    "end": 3397.04,
                    "word": " familiar",
                    "probability": 0.79345703125
                },
                {
                    "start": 3397.04,
                    "end": 3397.46,
                    "word": " with",
                    "probability": 0.822265625
                }
            ]
        },
        {
            "id": 504,
            "seek": 341693,
            "start": 3397.46,
            "end": 3403.06,
            "text": " the basics of reinforcement learning. You probably know that the discount factor is typically chosen",
            "tokens": [
                50504,
                264,
                14688,
                295,
                29280,
                2539,
                13,
                509,
                1391,
                458,
                300,
                264,
                11635,
                5952,
                307,
                5850,
                8614,
                50772
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24272895390444463,
            "compression_ratio": 1.7032967032967032,
            "no_speech_prob": 0.0030269622802734375,
            "words": [
                {
                    "start": 3397.46,
                    "end": 3397.9,
                    "word": " the",
                    "probability": 0.6376953125
                },
                {
                    "start": 3397.9,
                    "end": 3398.12,
                    "word": " basics",
                    "probability": 0.87890625
                },
                {
                    "start": 3398.12,
                    "end": 3398.26,
                    "word": " of",
                    "probability": 0.89306640625
                },
                {
                    "start": 3398.26,
                    "end": 3398.5,
                    "word": " reinforcement",
                    "probability": 0.796875
                },
                {
                    "start": 3398.5,
                    "end": 3398.9,
                    "word": " learning",
                    "probability": 0.8291015625
                },
                {
                    "start": 3398.9,
                    "end": 3399.14,
                    "word": ".",
                    "probability": 0.0960693359375
                },
                {
                    "start": 3399.14,
                    "end": 3399.14,
                    "word": " You",
                    "probability": 0.88134765625
                },
                {
                    "start": 3399.14,
                    "end": 3399.54,
                    "word": " probably",
                    "probability": 0.9111328125
                },
                {
                    "start": 3399.54,
                    "end": 3399.92,
                    "word": " know",
                    "probability": 0.8232421875
                },
                {
                    "start": 3399.92,
                    "end": 3400.22,
                    "word": " that",
                    "probability": 0.87939453125
                },
                {
                    "start": 3400.22,
                    "end": 3401.2,
                    "word": " the",
                    "probability": 0.79150390625
                },
                {
                    "start": 3401.2,
                    "end": 3401.46,
                    "word": " discount",
                    "probability": 0.8720703125
                },
                {
                    "start": 3401.46,
                    "end": 3401.84,
                    "word": " factor",
                    "probability": 0.7490234375
                },
                {
                    "start": 3401.84,
                    "end": 3402.32,
                    "word": " is",
                    "probability": 0.90234375
                },
                {
                    "start": 3402.32,
                    "end": 3402.78,
                    "word": " typically",
                    "probability": 0.85546875
                },
                {
                    "start": 3402.78,
                    "end": 3403.06,
                    "word": " chosen",
                    "probability": 0.90869140625
                }
            ]
        },
        {
            "id": 505,
            "seek": 341693,
            "start": 3403.06,
            "end": 3410.04,
            "text": " to be 0.99 or something in the same ballpark and is left the same. It's treated as a constant",
            "tokens": [
                50772,
                281,
                312,
                1958,
                13,
                8494,
                420,
                746,
                294,
                264,
                912,
                2594,
                31239,
                293,
                307,
                1411,
                264,
                912,
                13,
                467,
                311,
                8668,
                382,
                257,
                5754,
                51124
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24272895390444463,
            "compression_ratio": 1.7032967032967032,
            "no_speech_prob": 0.0030269622802734375,
            "words": [
                {
                    "start": 3403.06,
                    "end": 3403.22,
                    "word": " to",
                    "probability": 0.90625
                },
                {
                    "start": 3403.22,
                    "end": 3403.34,
                    "word": " be",
                    "probability": 0.90185546875
                },
                {
                    "start": 3403.34,
                    "end": 3403.52,
                    "word": " 0",
                    "probability": 0.5859375
                },
                {
                    "start": 3403.52,
                    "end": 3403.94,
                    "word": ".99",
                    "probability": 0.988037109375
                },
                {
                    "start": 3403.94,
                    "end": 3404.18,
                    "word": " or",
                    "probability": 0.78076171875
                },
                {
                    "start": 3404.18,
                    "end": 3404.54,
                    "word": " something",
                    "probability": 0.73681640625
                },
                {
                    "start": 3404.54,
                    "end": 3405.62,
                    "word": " in",
                    "probability": 0.423095703125
                },
                {
                    "start": 3405.62,
                    "end": 3406.0,
                    "word": " the",
                    "probability": 0.8125
                },
                {
                    "start": 3406.0,
                    "end": 3406.16,
                    "word": " same",
                    "probability": 0.794921875
                },
                {
                    "start": 3406.16,
                    "end": 3406.5,
                    "word": " ballpark",
                    "probability": 0.92333984375
                },
                {
                    "start": 3406.5,
                    "end": 3406.6,
                    "word": " and",
                    "probability": 0.62744140625
                },
                {
                    "start": 3406.6,
                    "end": 3406.76,
                    "word": " is",
                    "probability": 0.73046875
                },
                {
                    "start": 3406.76,
                    "end": 3407.48,
                    "word": " left",
                    "probability": 0.86083984375
                },
                {
                    "start": 3407.48,
                    "end": 3407.92,
                    "word": " the",
                    "probability": 0.8076171875
                },
                {
                    "start": 3407.92,
                    "end": 3408.14,
                    "word": " same",
                    "probability": 0.8154296875
                },
                {
                    "start": 3408.14,
                    "end": 3408.28,
                    "word": ".",
                    "probability": 0.70263671875
                },
                {
                    "start": 3408.5,
                    "end": 3409.06,
                    "word": " It's",
                    "probability": 0.9287109375
                },
                {
                    "start": 3409.06,
                    "end": 3409.44,
                    "word": " treated",
                    "probability": 0.607421875
                },
                {
                    "start": 3409.44,
                    "end": 3409.62,
                    "word": " as",
                    "probability": 0.90478515625
                },
                {
                    "start": 3409.74,
                    "end": 3410.04,
                    "word": " a constant",
                    "probability": 0.849609375
                }
            ]
        },
        {
            "id": 506,
            "seek": 341693,
            "start": 3410.04,
            "end": 3416.46,
            "text": " throughout the training process. But this is very far from how our brain works and very far from",
            "tokens": [
                51124,
                3710,
                264,
                3097,
                1399,
                13,
                583,
                341,
                307,
                588,
                1400,
                490,
                577,
                527,
                3567,
                1985,
                293,
                588,
                1400,
                490,
                51440
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24272895390444463,
            "compression_ratio": 1.7032967032967032,
            "no_speech_prob": 0.0030269622802734375,
            "words": [
                {
                    "start": 3410.04,
                    "end": 3410.36,
                    "word": " throughout",
                    "probability": 0.76318359375
                },
                {
                    "start": 3410.36,
                    "end": 3410.54,
                    "word": " the",
                    "probability": 0.8076171875
                },
                {
                    "start": 3410.54,
                    "end": 3410.72,
                    "word": " training",
                    "probability": 0.7919921875
                },
                {
                    "start": 3410.72,
                    "end": 3411.18,
                    "word": " process",
                    "probability": 0.8525390625
                },
                {
                    "start": 3411.18,
                    "end": 3411.68,
                    "word": ".",
                    "probability": 0.82763671875
                },
                {
                    "start": 3411.68,
                    "end": 3411.9,
                    "word": " But",
                    "probability": 0.86474609375
                },
                {
                    "start": 3411.9,
                    "end": 3412.22,
                    "word": " this",
                    "probability": 0.87353515625
                },
                {
                    "start": 3412.22,
                    "end": 3412.56,
                    "word": " is",
                    "probability": 0.90625
                },
                {
                    "start": 3412.56,
                    "end": 3412.98,
                    "word": " very",
                    "probability": 0.74072265625
                },
                {
                    "start": 3412.98,
                    "end": 3413.32,
                    "word": " far",
                    "probability": 0.89404296875
                },
                {
                    "start": 3413.32,
                    "end": 3413.76,
                    "word": " from",
                    "probability": 0.81884765625
                },
                {
                    "start": 3413.76,
                    "end": 3414.74,
                    "word": " how",
                    "probability": 0.880859375
                },
                {
                    "start": 3414.74,
                    "end": 3415.12,
                    "word": " our",
                    "probability": 0.8193359375
                },
                {
                    "start": 3415.12,
                    "end": 3415.3,
                    "word": " brain",
                    "probability": 0.81640625
                },
                {
                    "start": 3415.3,
                    "end": 3415.52,
                    "word": " works",
                    "probability": 0.7509765625
                },
                {
                    "start": 3415.52,
                    "end": 3415.66,
                    "word": " and",
                    "probability": 0.779296875
                },
                {
                    "start": 3415.66,
                    "end": 3415.82,
                    "word": " very",
                    "probability": 0.732421875
                },
                {
                    "start": 3415.82,
                    "end": 3416.06,
                    "word": " far",
                    "probability": 0.89453125
                },
                {
                    "start": 3416.06,
                    "end": 3416.46,
                    "word": " from",
                    "probability": 0.822265625
                }
            ]
        },
        {
            "id": 507,
            "seek": 341693,
            "start": 3417.86,
            "end": 3423.06,
            "text": " the optimal approach or accurate approach to discounting. Our research group has recently",
            "tokens": [
                51532,
                264,
                16252,
                3109,
                420,
                8559,
                3109,
                281,
                11635,
                278,
                13,
                2621,
                2132,
                1594,
                575,
                3938,
                51776
            ],
            "temperature": 0.0,
            "avg_logprob": -0.24272895390444463,
            "compression_ratio": 1.7032967032967032,
            "no_speech_prob": 0.0030269622802734375,
            "words": [
                {
                    "start": 3417.86,
                    "end": 3418.42,
                    "word": " the",
                    "probability": 0.7939453125
                },
                {
                    "start": 3418.42,
                    "end": 3418.96,
                    "word": " optimal",
                    "probability": 0.69970703125
                },
                {
                    "start": 3418.96,
                    "end": 3419.88,
                    "word": " approach",
                    "probability": 0.71484375
                },
                {
                    "start": 3419.88,
                    "end": 3420.06,
                    "word": " or",
                    "probability": 0.8681640625
                },
                {
                    "start": 3420.06,
                    "end": 3420.44,
                    "word": " accurate",
                    "probability": 0.74072265625
                },
                {
                    "start": 3420.44,
                    "end": 3420.82,
                    "word": " approach",
                    "probability": 0.77734375
                },
                {
                    "start": 3420.82,
                    "end": 3420.98,
                    "word": " to",
                    "probability": 0.91064453125
                },
                {
                    "start": 3420.98,
                    "end": 3421.54,
                    "word": " discounting",
                    "probability": 0.885986328125
                },
                {
                    "start": 3421.54,
                    "end": 3421.88,
                    "word": ".",
                    "probability": 0.9736328125
                },
                {
                    "start": 3421.88,
                    "end": 3422.1,
                    "word": " Our",
                    "probability": 0.90673828125
                },
                {
                    "start": 3422.1,
                    "end": 3422.38,
                    "word": " research",
                    "probability": 0.75537109375
                },
                {
                    "start": 3422.38,
                    "end": 3422.52,
                    "word": " group",
                    "probability": 0.845703125
                },
                {
                    "start": 3422.52,
                    "end": 3422.74,
                    "word": " has",
                    "probability": 0.88427734375
                },
                {
                    "start": 3422.74,
                    "end": 3423.06,
                    "word": " recently",
                    "probability": 0.80126953125
                }
            ]
        },
        {
            "id": 508,
            "seek": 344007,
            "start": 3423.06,
            "end": 3428.14,
            "text": " started looking into this problem and is working on developing adaptive discounting",
            "tokens": [
                50364,
                1409,
                1237,
                666,
                341,
                1154,
                293,
                307,
                1364,
                322,
                6416,
                27912,
                11635,
                278,
                50612
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21551339839185987,
            "compression_ratio": 1.5022831050228311,
            "no_speech_prob": 0.0008559226989746094,
            "words": [
                {
                    "start": 3423.06,
                    "end": 3423.34,
                    "word": " started",
                    "probability": 0.28466796875
                },
                {
                    "start": 3423.34,
                    "end": 3423.62,
                    "word": " looking",
                    "probability": 0.79638671875
                },
                {
                    "start": 3423.62,
                    "end": 3423.84,
                    "word": " into",
                    "probability": 0.73046875
                },
                {
                    "start": 3423.84,
                    "end": 3424.0,
                    "word": " this",
                    "probability": 0.8837890625
                },
                {
                    "start": 3424.0,
                    "end": 3424.3,
                    "word": " problem",
                    "probability": 0.78173828125
                },
                {
                    "start": 3424.3,
                    "end": 3424.8,
                    "word": " and",
                    "probability": 0.79296875
                },
                {
                    "start": 3424.8,
                    "end": 3425.26,
                    "word": " is",
                    "probability": 0.8232421875
                },
                {
                    "start": 3425.26,
                    "end": 3426.12,
                    "word": " working",
                    "probability": 0.67822265625
                },
                {
                    "start": 3426.12,
                    "end": 3426.3,
                    "word": " on",
                    "probability": 0.91015625
                },
                {
                    "start": 3426.3,
                    "end": 3426.74,
                    "word": " developing",
                    "probability": 0.62109375
                },
                {
                    "start": 3426.74,
                    "end": 3427.38,
                    "word": " adaptive",
                    "probability": 0.8232421875
                },
                {
                    "start": 3427.38,
                    "end": 3428.14,
                    "word": " discounting",
                    "probability": 0.818115234375
                }
            ]
        },
        {
            "id": 509,
            "seek": 344007,
            "start": 3429.32,
            "end": 3436.94,
            "text": " solutions to enhance the resilience and robustness of RL agents, particularly",
            "tokens": [
                50680,
                6547,
                281,
                11985,
                264,
                19980,
                293,
                13956,
                1287,
                295,
                497,
                43,
                12554,
                11,
                4098,
                51080
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21551339839185987,
            "compression_ratio": 1.5022831050228311,
            "no_speech_prob": 0.0008559226989746094,
            "words": [
                {
                    "start": 3429.32,
                    "end": 3429.94,
                    "word": " solutions",
                    "probability": 0.82177734375
                },
                {
                    "start": 3429.94,
                    "end": 3430.56,
                    "word": " to",
                    "probability": 0.896484375
                },
                {
                    "start": 3430.56,
                    "end": 3431.36,
                    "word": " enhance",
                    "probability": 0.8125
                },
                {
                    "start": 3431.36,
                    "end": 3433.34,
                    "word": " the",
                    "probability": 0.79736328125
                },
                {
                    "start": 3433.34,
                    "end": 3433.7,
                    "word": " resilience",
                    "probability": 0.78955078125
                },
                {
                    "start": 3433.7,
                    "end": 3433.88,
                    "word": " and",
                    "probability": 0.87939453125
                },
                {
                    "start": 3433.88,
                    "end": 3434.64,
                    "word": " robustness",
                    "probability": 0.848876953125
                },
                {
                    "start": 3434.64,
                    "end": 3435.34,
                    "word": " of",
                    "probability": 0.908203125
                },
                {
                    "start": 3435.34,
                    "end": 3436.02,
                    "word": " RL",
                    "probability": 0.89208984375
                },
                {
                    "start": 3436.02,
                    "end": 3436.34,
                    "word": " agents",
                    "probability": 0.78564453125
                },
                {
                    "start": 3436.34,
                    "end": 3436.44,
                    "word": ",",
                    "probability": 0.7109375
                },
                {
                    "start": 3436.44,
                    "end": 3436.94,
                    "word": " particularly",
                    "probability": 0.76220703125
                }
            ]
        },
        {
            "id": 510,
            "seek": 344007,
            "start": 3436.94,
            "end": 3441.59,
            "text": " deep RL agents in complex environments for AI safety and security purposes.",
            "tokens": [
                51080,
                2452,
                497,
                43,
                12554,
                294,
                3997,
                12388,
                337,
                7318,
                4514,
                293,
                3825,
                9932,
                13,
                51276
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21551339839185987,
            "compression_ratio": 1.5022831050228311,
            "no_speech_prob": 0.0008559226989746094,
            "words": [
                {
                    "start": 3436.94,
                    "end": 3437.68,
                    "word": " deep",
                    "probability": 0.71533203125
                },
                {
                    "start": 3437.68,
                    "end": 3437.92,
                    "word": " RL",
                    "probability": 0.9521484375
                },
                {
                    "start": 3437.92,
                    "end": 3438.18,
                    "word": " agents",
                    "probability": 0.7998046875
                },
                {
                    "start": 3438.18,
                    "end": 3438.34,
                    "word": " in",
                    "probability": 0.830078125
                },
                {
                    "start": 3438.34,
                    "end": 3438.78,
                    "word": " complex",
                    "probability": 0.7421875
                },
                {
                    "start": 3438.78,
                    "end": 3439.24,
                    "word": " environments",
                    "probability": 0.6689453125
                },
                {
                    "start": 3439.24,
                    "end": 3439.76,
                    "word": " for",
                    "probability": 0.734375
                },
                {
                    "start": 3439.76,
                    "end": 3440.06,
                    "word": " AI",
                    "probability": 0.876953125
                },
                {
                    "start": 3440.06,
                    "end": 3440.34,
                    "word": " safety",
                    "probability": 0.9111328125
                },
                {
                    "start": 3440.34,
                    "end": 3440.52,
                    "word": " and",
                    "probability": 0.89697265625
                },
                {
                    "start": 3440.52,
                    "end": 3440.84,
                    "word": " security",
                    "probability": 0.9296875
                },
                {
                    "start": 3440.84,
                    "end": 3441.28,
                    "word": " purposes",
                    "probability": 0.8466796875
                },
                {
                    "start": 3441.28,
                    "end": 3441.59,
                    "word": ".",
                    "probability": 0.9814453125
                }
            ]
        },
        {
            "id": 511,
            "seek": 344007,
            "start": 3441.96,
            "end": 3446.2,
            "text": " Now, of course, there are naturally inspired approaches that can be looked at. For example,",
            "tokens": [
                51304,
                823,
                11,
                295,
                1164,
                11,
                456,
                366,
                8195,
                7547,
                11587,
                300,
                393,
                312,
                2956,
                412,
                13,
                1171,
                1365,
                11,
                51516
            ],
            "temperature": 0.0,
            "avg_logprob": -0.21551339839185987,
            "compression_ratio": 1.5022831050228311,
            "no_speech_prob": 0.0008559226989746094,
            "words": [
                {
                    "start": 3441.96,
                    "end": 3442.14,
                    "word": " Now",
                    "probability": 0.8701171875
                },
                {
                    "start": 3442.14,
                    "end": 3442.2,
                    "word": ",",
                    "probability": 0.39892578125
                },
                {
                    "start": 3442.2,
                    "end": 3442.28,
                    "word": " of",
                    "probability": 0.90771484375
                },
                {
                    "start": 3442.28,
                    "end": 3442.36,
                    "word": " course",
                    "probability": 0.87109375
                },
                {
                    "start": 3442.36,
                    "end": 3442.4,
                    "word": ",",
                    "probability": 0.90576171875
                },
                {
                    "start": 3442.4,
                    "end": 3442.48,
                    "word": " there",
                    "probability": 0.2274169921875
                },
                {
                    "start": 3442.48,
                    "end": 3442.52,
                    "word": " are",
                    "probability": 0.88916015625
                },
                {
                    "start": 3442.52,
                    "end": 3442.82,
                    "word": " naturally",
                    "probability": 0.81787109375
                },
                {
                    "start": 3442.82,
                    "end": 3443.22,
                    "word": " inspired",
                    "probability": 0.69580078125
                },
                {
                    "start": 3443.22,
                    "end": 3443.6,
                    "word": " approaches",
                    "probability": 0.673828125
                },
                {
                    "start": 3443.6,
                    "end": 3444.06,
                    "word": " that",
                    "probability": 0.8779296875
                },
                {
                    "start": 3444.06,
                    "end": 3445.1,
                    "word": " can",
                    "probability": 0.89892578125
                },
                {
                    "start": 3445.1,
                    "end": 3445.24,
                    "word": " be",
                    "probability": 0.90380859375
                },
                {
                    "start": 3445.24,
                    "end": 3445.4,
                    "word": " looked",
                    "probability": 0.72412109375
                },
                {
                    "start": 3445.4,
                    "end": 3445.52,
                    "word": " at",
                    "probability": 0.90234375
                },
                {
                    "start": 3445.52,
                    "end": 3445.56,
                    "word": ".",
                    "probability": 0.487548828125
                },
                {
                    "start": 3445.56,
                    "end": 3445.64,
                    "word": " For",
                    "probability": 0.89013671875
                },
                {
                    "start": 3445.64,
                    "end": 3445.96,
                    "word": " example",
                    "probability": 0.93359375
                },
                {
                    "start": 3445.96,
                    "end": 3446.2,
                    "word": ",",
                    "probability": 0.97998046875
                }
            ]
        },
        {
            "id": 512,
            "seek": 346547,
            "start": 3446.2,
            "end": 3453.4,
            "text": " approaches coming from, let's say, TD lambda models or dopamine models of psychopathological",
            "tokens": [
                50364,
                11587,
                1348,
                490,
                11,
                718,
                311,
                584,
                11,
                42606,
                13607,
                5245,
                420,
                37219,
                5245,
                295,
                47577,
                4383,
                50728
            ],
            "temperature": 0.0,
            "avg_logprob": -0.27648627576304646,
            "compression_ratio": 1.5879828326180256,
            "no_speech_prob": 0.0233306884765625,
            "words": [
                {
                    "start": 3446.2,
                    "end": 3446.52,
                    "word": " approaches",
                    "probability": 0.689453125
                },
                {
                    "start": 3446.52,
                    "end": 3446.74,
                    "word": " coming",
                    "probability": 0.78564453125
                },
                {
                    "start": 3446.74,
                    "end": 3447.38,
                    "word": " from",
                    "probability": 0.814453125
                },
                {
                    "start": 3447.38,
                    "end": 3447.44,
                    "word": ",",
                    "probability": 0.568359375
                },
                {
                    "start": 3447.44,
                    "end": 3448.5,
                    "word": " let's",
                    "probability": 0.929931640625
                },
                {
                    "start": 3448.5,
                    "end": 3448.74,
                    "word": " say",
                    "probability": 0.88623046875
                },
                {
                    "start": 3448.74,
                    "end": 3449.28,
                    "word": ",",
                    "probability": 0.77099609375
                },
                {
                    "start": 3449.28,
                    "end": 3449.44,
                    "word": " TD",
                    "probability": 0.83984375
                },
                {
                    "start": 3449.44,
                    "end": 3449.74,
                    "word": " lambda",
                    "probability": 0.452880859375
                },
                {
                    "start": 3449.74,
                    "end": 3450.24,
                    "word": " models",
                    "probability": 0.771484375
                },
                {
                    "start": 3450.24,
                    "end": 3450.66,
                    "word": " or",
                    "probability": 0.72607421875
                },
                {
                    "start": 3450.66,
                    "end": 3451.1,
                    "word": " dopamine",
                    "probability": 0.494140625
                },
                {
                    "start": 3451.1,
                    "end": 3451.68,
                    "word": " models",
                    "probability": 0.7939453125
                },
                {
                    "start": 3451.68,
                    "end": 3452.4,
                    "word": " of",
                    "probability": 0.888671875
                },
                {
                    "start": 3452.4,
                    "end": 3453.4,
                    "word": " psychopathological",
                    "probability": 0.923095703125
                }
            ]
        },
        {
            "id": 513,
            "seek": 346547,
            "start": 3453.4,
            "end": 3458.2,
            "text": " problems or neurological problems and the solutions prescribed to those, as well as",
            "tokens": [
                50728,
                2740,
                420,
                48185,
                2740,
                293,
                264,
                6547,
                29099,
                281,
                729,
                11,
                382,
                731,
                382,
                50960
            ],
            "temperature": 0.0,
            "avg_logprob": -0.27648627576304646,
            "compression_ratio": 1.5879828326180256,
            "no_speech_prob": 0.0233306884765625,
            "words": [
                {
                    "start": 3453.4,
                    "end": 3453.86,
                    "word": " problems",
                    "probability": 0.68798828125
                },
                {
                    "start": 3453.86,
                    "end": 3453.98,
                    "word": " or",
                    "probability": 0.83349609375
                },
                {
                    "start": 3453.98,
                    "end": 3454.34,
                    "word": " neurological",
                    "probability": 0.50537109375
                },
                {
                    "start": 3454.34,
                    "end": 3455.02,
                    "word": " problems",
                    "probability": 0.70751953125
                },
                {
                    "start": 3455.02,
                    "end": 3455.16,
                    "word": " and",
                    "probability": 0.7763671875
                },
                {
                    "start": 3455.16,
                    "end": 3455.3,
                    "word": " the",
                    "probability": 0.798828125
                },
                {
                    "start": 3455.3,
                    "end": 3455.8,
                    "word": " solutions",
                    "probability": 0.84375
                },
                {
                    "start": 3455.8,
                    "end": 3457.02,
                    "word": " prescribed",
                    "probability": 0.71337890625
                },
                {
                    "start": 3457.02,
                    "end": 3457.22,
                    "word": " to",
                    "probability": 0.90478515625
                },
                {
                    "start": 3457.22,
                    "end": 3457.5,
                    "word": " those",
                    "probability": 0.833984375
                },
                {
                    "start": 3457.5,
                    "end": 3457.62,
                    "word": ",",
                    "probability": 0.578125
                },
                {
                    "start": 3457.62,
                    "end": 3457.84,
                    "word": " as",
                    "probability": 0.90625
                },
                {
                    "start": 3457.84,
                    "end": 3457.98,
                    "word": " well",
                    "probability": 0.8759765625
                },
                {
                    "start": 3457.98,
                    "end": 3458.2,
                    "word": " as",
                    "probability": 0.9140625
                }
            ]
        },
        {
            "id": 514,
            "seek": 346547,
            "start": 3458.2,
            "end": 3463.58,
            "text": " approaches in social sciences, which may help with the security problems arising in multi-agent",
            "tokens": [
                50960,
                11587,
                294,
                2093,
                17677,
                11,
                597,
                815,
                854,
                365,
                264,
                3825,
                2740,
                44900,
                294,
                4825,
                12,
                559,
                317,
                51240
            ],
            "temperature": 0.0,
            "avg_logprob": -0.27648627576304646,
            "compression_ratio": 1.5879828326180256,
            "no_speech_prob": 0.0233306884765625,
            "words": [
                {
                    "start": 3458.2,
                    "end": 3458.48,
                    "word": " approaches",
                    "probability": 0.7021484375
                },
                {
                    "start": 3458.48,
                    "end": 3458.64,
                    "word": " in",
                    "probability": 0.53466796875
                },
                {
                    "start": 3458.64,
                    "end": 3458.88,
                    "word": " social",
                    "probability": 0.80712890625
                },
                {
                    "start": 3458.88,
                    "end": 3459.26,
                    "word": " sciences",
                    "probability": 0.763671875
                },
                {
                    "start": 3459.26,
                    "end": 3459.26,
                    "word": ",",
                    "probability": 0.73486328125
                },
                {
                    "start": 3459.26,
                    "end": 3459.5,
                    "word": " which",
                    "probability": 0.91552734375
                },
                {
                    "start": 3459.5,
                    "end": 3459.86,
                    "word": " may",
                    "probability": 0.904296875
                },
                {
                    "start": 3459.86,
                    "end": 3460.22,
                    "word": " help",
                    "probability": 0.8193359375
                },
                {
                    "start": 3460.22,
                    "end": 3460.7,
                    "word": " with",
                    "probability": 0.82080078125
                },
                {
                    "start": 3460.7,
                    "end": 3461.14,
                    "word": " the",
                    "probability": 0.80126953125
                },
                {
                    "start": 3461.14,
                    "end": 3461.5,
                    "word": " security",
                    "probability": 0.923828125
                },
                {
                    "start": 3461.5,
                    "end": 3462.16,
                    "word": " problems",
                    "probability": 0.69873046875
                },
                {
                    "start": 3462.16,
                    "end": 3462.58,
                    "word": " arising",
                    "probability": 0.8720703125
                },
                {
                    "start": 3462.58,
                    "end": 3462.9,
                    "word": " in",
                    "probability": 0.88134765625
                },
                {
                    "start": 3462.9,
                    "end": 3463.2,
                    "word": " multi",
                    "probability": 0.78369140625
                },
                {
                    "start": 3463.2,
                    "end": 3463.58,
                    "word": "-agent",
                    "probability": 0.81591796875
                }
            ]
        },
        {
            "id": 515,
            "seek": 346547,
            "start": 3463.58,
            "end": 3471.6,
            "text": " RL settings. Very well, thank you very much. And I believe at this time I should be available for",
            "tokens": [
                51240,
                497,
                43,
                6257,
                13,
                4372,
                731,
                11,
                1309,
                291,
                588,
                709,
                13,
                400,
                286,
                1697,
                412,
                341,
                565,
                286,
                820,
                312,
                2435,
                337,
                51648
            ],
            "temperature": 0.0,
            "avg_logprob": -0.27648627576304646,
            "compression_ratio": 1.5879828326180256,
            "no_speech_prob": 0.0233306884765625,
            "words": [
                {
                    "start": 3463.58,
                    "end": 3464.1,
                    "word": " RL",
                    "probability": 0.7353515625
                },
                {
                    "start": 3464.1,
                    "end": 3464.46,
                    "word": " settings",
                    "probability": 0.84130859375
                },
                {
                    "start": 3464.46,
                    "end": 3465.1,
                    "word": ".",
                    "probability": 0.9853515625
                },
                {
                    "start": 3465.54,
                    "end": 3465.68,
                    "word": " Very",
                    "probability": 0.6083984375
                },
                {
                    "start": 3465.68,
                    "end": 3465.8,
                    "word": " well",
                    "probability": 0.86767578125
                },
                {
                    "start": 3465.8,
                    "end": 3465.86,
                    "word": ",",
                    "probability": 0.76123046875
                },
                {
                    "start": 3465.86,
                    "end": 3466.0,
                    "word": " thank",
                    "probability": 0.818359375
                },
                {
                    "start": 3466.0,
                    "end": 3466.1,
                    "word": " you",
                    "probability": 0.89404296875
                },
                {
                    "start": 3466.1,
                    "end": 3466.3,
                    "word": " very",
                    "probability": 0.7275390625
                },
                {
                    "start": 3466.3,
                    "end": 3466.64,
                    "word": " much",
                    "probability": 0.814453125
                },
                {
                    "start": 3466.64,
                    "end": 3467.28,
                    "word": ".",
                    "probability": 0.80517578125
                },
                {
                    "start": 3468.42,
                    "end": 3469.06,
                    "word": " And",
                    "probability": 0.85302734375
                },
                {
                    "start": 3469.06,
                    "end": 3469.5,
                    "word": " I",
                    "probability": 0.9833984375
                },
                {
                    "start": 3469.5,
                    "end": 3469.84,
                    "word": " believe",
                    "probability": 0.71240234375
                },
                {
                    "start": 3469.84,
                    "end": 3470.02,
                    "word": " at",
                    "probability": 0.689453125
                },
                {
                    "start": 3470.02,
                    "end": 3470.22,
                    "word": " this",
                    "probability": 0.87890625
                },
                {
                    "start": 3470.22,
                    "end": 3470.34,
                    "word": " time",
                    "probability": 0.8271484375
                },
                {
                    "start": 3470.34,
                    "end": 3470.44,
                    "word": " I",
                    "probability": 0.437744140625
                },
                {
                    "start": 3470.44,
                    "end": 3470.6,
                    "word": " should",
                    "probability": 0.92626953125
                },
                {
                    "start": 3470.6,
                    "end": 3470.72,
                    "word": " be",
                    "probability": 0.9072265625
                },
                {
                    "start": 3470.72,
                    "end": 3471.04,
                    "word": " available",
                    "probability": 0.7265625
                },
                {
                    "start": 3471.04,
                    "end": 3471.6,
                    "word": " for",
                    "probability": 0.90185546875
                }
            ]
        },
        {
            "id": 516,
            "seek": 346888,
            "start": 3471.6,
            "end": 3473.14,
            "text": " your questions.",
            "tokens": [
                50376,
                428,
                1651,
                13,
                50492
            ],
            "temperature": 0.0,
            "avg_logprob": -0.537109375,
            "compression_ratio": 0.6521739130434783,
            "no_speech_prob": 0.09246826171875,
            "words": [
                {
                    "start": 3471.6,
                    "end": 3472.04,
                    "word": " your",
                    "probability": 0.228759765625
                },
                {
                    "start": 3472.04,
                    "end": 3472.5,
                    "word": " questions",
                    "probability": 0.939453125
                },
                {
                    "start": 3472.5,
                    "end": 3473.14,
                    "word": ".",
                    "probability": 0.958984375
                }
            ]
        }
    ],
    "text": " Welcome to the last talk of the AI village from Fahid. I'll just let him take it away. Hello and thank you for joining this talk on the security issues and challenges in deep reinforcement learning. I'm Fahid Behzadan, I'm an assistant professor of computer science and data science at the University of New Haven and I also direct the Secure and Assured Intelligent Learning Lab, or SAIL for short, working on AI safety, security and applications of machine learning to cybersecurity and safety of complex systems. So here's the outline of my talk. I'm going to quickly go over the basics of deep reinforcement learning and reinforcement learning. There will be some math but I'll keep it to a minimum. Then we'll talk about vulnerabilities of deep reinforcement learning and whether deep RL is susceptible to classical adversarial machine learning attacks like adversarial examples and such. We'll talk about and develop a threat model for deep reinforcement learning. We'll identify different attack models, attack surfaces, we'll identify different types of vulnerabilities, and we'll introduce a number of attack mechanisms and corresponding defenses that have been developed in recent years. And we'll talk about the frontiers and areas of future research and work in this area. So just a quick overview. I assume most of the audience here is already familiar with the terminology used here. We can classify machine learning algorithms as supervised, unsupervised, and reinforcement learning algorithms. Supervised learning algorithms are those where the training data set includes labeled data, meaning that each data point in that data set comes with the correct label or the correct output expected from a model trained on that data set. Then there is unsupervised learning where clustering and anomaly detection and some other algorithms fall under where there are no labels available for data points. There's no feedback available on the data. However, the goal is to find some underlying structure in the data. And then there is reinforcement learning, which is concerned with the problem of sequential decision making. There's a reward system included in reinforcement learning, but there are fundamental differences between the settings of reinforcement learning and supervised learning, and that's why it merits its own category. We'll talk about those differences in the next few slides. So you can see the general settings of reinforcement learning, or in general, sensory motor agent problems and settings. In RL, we have an agent which is to be trained by a reinforcement learning algorithm. This agent interacts with the environment by performing some action. This action causes the state of the environment to change. Then the new state is observed by the agent. There's also some sort of reward associated with this change, provided or inferred by the agent based on this change in state. Well, you can think of this in terms of playing a game. The environment can be the game environment where different actions may result in scoring or loss of score, and the actions are, well, essentially the actions that the player can take. The state of the environment can be the state of the game. For example, in Breakout, the configuration of these breaks can be one of those where the agent is, where the ball is. If you recall the game of Breakout, of course, if you are old enough to remember Breakout, you're probably familiar with the dynamics. But you can think of any other game, and this setting still applies. What's the goal here? The goal is to learn how to take actions in order to maximize cumulative rewards, not instantaneous rewards, but cumulative rewards. So an RL agent is not just concerned with maximizing the current score. It wants to learn how to act so that at the end of the game, its total sum of rewards is maximized. What are the different applications of RL? Of course, game playing is well publicized because it's a very good testbed, and it has become one of the common testbeds and experimental settings for RL research. However, there are some major real-world applications for reinforcement learning. In essence, RL, or reinforcement learning, is the machine learning response to the need for data-driven control problems. Where do we encounter such problems? Well, in robotics, like autonomous navigation, object manipulation, and such. Algorithmic trading, this is one of the areas that my research group has recently become active in. Critical infrastructure controlling smart cities, resource allocation in smart cities, traffic management, intelligent traffic systems, smart grid management, and control. Healthcare, such as clinical decision making, this is actually one of the better known applications of RL in the real world. And other types of resource management and applications in operations research and such. So, RL is either envisioned or is already heavily adopted in various industries, many of which are critical and may become targets of malicious actions. Let's formalize the RL problem a little more. I'm just going to quickly go over this. There will be some math, but this is just to introduce the basic settings. The underlying framework to formulate the RL problem and provide a framework to think about and reason about the RL problem is the Markov decision process. Why do we call it the Markov decision process? Because it's based on the Markovian assumption, or the Markov property, which says the current state completely characterizes the state of the world. So, if you have the current state and you perform an action, the next state is only going to depend or be a function of the previous state and nothing before it. You don't need to know the history of the environment. All you need to know is what the previous state has been to infer what the coming state is going to be. Now, in Markov decision processes, a problem or setting is formulated by a tuple of five main parameters. One is the set of possible states, S. Set of possible actions, A. Distribution of reward given state action pair, or R. This distribution can also be a function. It doesn't necessarily have to be probabilistic. It can be just a function that tells you if at a particular state, S, I, an action AI is taken, then the reward is going to be, let's say, RI. Then there's transition probability, or transition dynamics, P, which represents the dynamics of the environment. Which state are we going to end up in if we are in a particular state and we perform a particular action. And there's finally a discount factor which defines how myopic the agent is, how much values, rewards that may occur further in the future, further down the line. Now, in general, the RL setting is based on the following process. At time step T equals zero, in the beginning, the environment samples initial state S0. Then, for T equals zero, this is a loop, until done, the agent selects action AT based on some criteria. It can be completely random in the beginning and then slowly becomes more targeted and more policy-driven. It performs some action, the environment produces a reward signal, and the next state, and the agent receives those signals. It may be partial, it may be incomplete, or it may be noisy, but the agent receives some signal resulting from the transition to the new state ST plus one, and a reward that has emerged from that transition and has resulted from that transition. Now, we define a new entity here called pi. A policy pi is a function, or a distribution in the probabilistic case, that maps from state to action. So, it tells the agent what action to perform given any state. And the objective is to find the optimal policy pi star that maximizes the cumulative discounted reward. We also call this return. So, cumulative, meaning the sum of rewards in the entire duration of interaction. It can be one episode of a game, it can be throughout the training period or training horizon, and it's discounted. This gamma is the discount factor that defines how myopic the agent is, or how much it values events that occur further down the line. It's typically a constant value between zero and one, and as you can see, as t increases, this gamma to power t decreases. So, the value, the preference, or the observed value of something that happens further down the line, which means with greater t, is going to decrease as t increases. So, again, the objective is to find the optimal policy pi star that maximizes this sum here. All right, we are going to quickly define two other definitions here. If we want to evaluate a particular state in these settings, one approach is to measure the value of that state using the value function. The value function of state s is the expected cumulative reward from following the policy pi from this state. So, let's assume we already have a policy. What will be the expected cumulative reward if you start in state s and keep following policy pi? Now, we also define a Q value, or a Q function, which tells us how good is a particular action A if it's performed in state s. In other words, if we are in state s, if the agent is in state s and performs action A, the Q value of this setting is the expected cumulative reward from taking action A in state s and then following the policy. Remember, the policy is a function that tells the agent what action to perform given any state. When the agent performs action A in state s, it goes into a state s prime, or the next state, and then we can use this policy function here to see what the action should be, what action should the agent take at that state, which takes it to another state, and then the policy is used to figure out the action to take in that state, and this goes on until the termination of the episode or the horizon. Now that we are familiar with the value function and the Q function, let's go a bit deeper. The policy pi, the value function, or the reward model are all functions. We want to learn at least one of these from experience. This is the essence of RL. If there are too many states, however, we cannot just tabulate everything and try to experience every state and the corresponding result from that state or that state and any action performed, because as the size of the problem, as the state space and the action space increase in settings like, let's say, playing GTA 5 or a self-driving navigation policy in the real world, these state spaces just explode. Dimensionality is too high and it's just not feasible to store every possible state. In those cases, we need to approximate. In general and traditionally, this is called RL with function approximation. If function approximation is done using deep neural networks, then we call the training setting or the RL setting deep reinforcement learning. The term is relatively new. It was, I believe, introduced in late 2013, early 2014 by me and David Silver and others at DeepMind in their DeepQ learning paper. However, the concept is not very new. That said, there have been fantastic and even mind-blowing advances in this area in the past three or four years. We'll talk about some of those as we move forward. A quick overview of the taxonomy of different RL approaches and RL agents. Remember, we have value, function, policy, and model. There are different approaches to solving the RL problem. Sometimes we just want to find a policy directly. The approaches that respond or satisfy that need are called policy-based RL. Sometimes we want to find a Q function or a V function first and then derive the policy from those functions. These approaches are called value-based. Sometimes we want to first learn the model of the environment and then solve the problem. These are called model-based approaches. Sometimes we don't have the model and we don't want to learn the model explicitly, and those are called model-free. When we're dealing with both value, function, and policy at the same time, in other words, we have one agent learning the value function, another learning the policy, and then contrasting those with each other in a kind of zero-sum setting. We call those actor-critic RL approaches. As you can see, there are different approaches to different settings and different problems. However, all of those can still be grounded on top of the Markov decision process framework and the general solution approach we looked at before. One of the better-known approaches to RL, which falls under the value-based approaches, is Q-learning. The objective in Q- learning is to derive the optimal policy pi star based on optimal Q. So the optimal Q is one that maximizes the value of each state and action. And from that point onward, with an iterative formulation based on Bellman equations or dynamic programming, it becomes possible to find this through bootstrapping and iterative re-estimation of the Q value. There are different approaches to Q-learning in large state spaces or large action spaces where function approximation is required. One such approach is to parametrize Q as s, a, and an estimation parameter theta. If we solve this parametrized Q function by neural networks, where theta corresponds to the neural network, the approach is called Q-networks. This solution approach was proposed very early in 2000s, even earlier, but was perfected in some sense in 2013-2014 by DeepMind, by David Silver and his team, which resulted in the proposal of deep Q-networks or QNs. So we're talking about deep networks. What does that mean? It means that these networks use deep neural networks, like CNNs, which help with both function approximation and also end-to-end feature learning. One of the advantages of deep learning is its superior performance in learning feature representations, especially from images. Now, for those of you who are more statistically oriented, you may have seen the term IID. It means that the data are independent and identically distributed. It means that one data point does not depend on another data point, and also each data point is equally likely to occur. These do not hold for the RL settings. A particular state, for example, is highly correlated with its previous state and action. Why does this matter? Well, a lot of our supervised and deep learning approaches are based on the assumption that the training data is IID. When this is not possible, what happens is, in response to this problem, what happened was the DQN approach introduced experience replay. It's like a bag of all data which is randomly sampled in each training iteration to reduce the correlation, the sequential correlation, the temporal correlation of data points, and also make it more likely for data to be evenly distributed. Also, to reduce the effect of oscillation during training, DQN uses fixed parameters for a target network. The target of optimization is fixed and is updated every few thousand iterations, so that reduces the oscillation problem. And, of course, the rewards are normalized to minus one to one to make sure the performance, the reward signals are bounded. Now, now that we have a preliminary understanding of deep reinforcement learning and one of its implementations, one of the parallel approaches called DQN, let's take a quick look at adversarial machine learning. And I assume that by now, those of you who are not familiar with adversarial examples have become, have been introduced to this concept. The idea here in adversarial examples, we are now talking in the realm of, we are speaking in the context of image classifier supervised learning. Let's say we have an image classifier trained on a set of images of pandas, cats, and others to identify the object in those images. In the beginning, we pass the image of a panda to the classifier, and it correctly classifies it as panda. Now, it's been demonstrated, it's been actually established by now, that it is possible to make the classifier, to induce incorrect classifications in deep learning models or machine learning models in general by adding minute, minimal perturbations to the original image. As you can see, it's almost impossible to detect or see any changes in this final image. These perturbations, these pixel perturbations are very small. Now, this is one example of different attacks or different vulnerabilities in the classical realm of adversarial machine learning, which is mostly concerned with supervised learning and sometimes unsupervised learning. In general, the adversarial objectives in AML or adversarial machine learning can be classified under the traditional CIA triad, confidentiality, integrity, and availability. So, with respect to confidentiality and privacy, an adversary may wish to target the confidentiality of the model parameters, architecture, no intellectual property theft is an issue in larger models, or it may target the privacy, the adversary may target the privacy of the training and test data. If medical, for example, if medical records were used in the training data, there have been proof of concept attacks showing that it is possible to infer whether a particular patient or the records pertaining to a particular patient were used in the training data or not, or sometimes it's possible to reconstruct the training data set by just having access to the model itself. And that is a major HIPAA violation. In essence, it's a privacy violation. Also, with regards to integrity and availability, the attacker may target the integrity of predictions or the outcome of the model, the performance of the model. For example, adversarial examples are an attack on the integrity of image classifiers or supervised machine learning models. And also an adversary may target the availability of the system that is deploying machine learning, for example, a facial recognition system or an autonomous navigation system in a driverless car. Now, now that we know so much about adversarial machine learning and security vulnerabilities of classical machine learning, supervised learning and unsupervised learning, there is a major question. And it's whether deep RL is immune to those attacks. Back in 2016, late 2016, when the research community was just beginning to start, was just beginning to pay attention to both deep reinforcement learning and the issue of adversarial examples. I came up with this question and decided to experiment with it a little to find out whether deep RL can also be vulnerable to such attacks. I started from a simple observation. The deep neural networks in DQN models and classifiers are both function approximators, are both at training time, there are function approximators at test times or just functions. And I came up with this hypothesis. If classifiers are vulnerable to adversarial examples, then action value approximators of DQNs may also be vulnerable. All right. So I started to set up an experiment where the aim was to perform adversarial attacks. The adversary's goal was twofold. Test time attack to perturb the performance of target's learned policy. So the target at this point is fully trained and is deployed in the environment and the adversary wants to somehow cause the agent to perform incorrectly, to manipulate its policy. What does the adversary know about the target? It knows the type of input to the target. For example, it knows whether the target policy is looking at image data, text, audio, and such. Why? Because it helps with estimating the architecture. For example, if it's images, the adversary can come up with a good guess that the target architecture includes CNNs or convolutional neural networks. I also assume that the adversary knows the reward function. So the adversary may have access to the environment and if it's, for example, a game environment, it knows what the scores are, how the scores are generated. What is not known? The knowledge of target's neural network architecture is not known. So it's a black box attack. And also the initial parameters, the initialization of the target neural network is also not known. What is available to adversary in terms of actions? The adversary may perturb the environment where the target performs. For example, it can change pixel values in a game environment through a man-in-the-middle attack. I consider in this work two techniques for perturbing that environment. One is the classical fast gradient sign method for generating adversarial examples and the other is the Jacobian-based saliency map attack or the JSMA approach introduced by Paperno in 2015, I believe. Now, in that experiment, I used the classical DQN approach, the classical DQN architecture introduced by Nie in an Atari game, the game of Pong. And this was all implemented in OpenAI Gym with TensorFlow. Back then, PyTorch wasn't really and trained the agent against heuristic AI. And here's the initial proof of concept result. This is for the white box attack. Of course, later on, I will show you the results for the black box attack. And you can see that for FGSM and JSMA, regardless of how far along the training the agent is, both policies, the policy for the Pong agent is highly vulnerable to adversarial perturbations through simple techniques like FGSM and JSMA. You can see that for JSMA, the success rate was 100% for all of the cases. For FGSM, it was slightly lower and it was mostly because of the termination criteria and the perturbation threshold that I had defined. But you can see, it can be observed that the policy can be very easily manipulated through adversarial example attacks. So, at 100 random observations, perturbed with FGSM and JSMA, the results were fed, the perturbed images were fed to the trained neural network representing the agent's policy as test input, and then the success rate was measured. Now, is this type of attack practical? Is this really something that we should be worried about? So, a few years later, in 2018, Clark and his co-authors published the report of a similar attack on an autonomous robot based on DQN policy using ultrasonic collision, ultrasonic sensory input for collision avoidance. And they had shown that they can use adversarial perturbations to manipulate the trajectory of the robot and make it follow a path defined or desired by the adversary, not one that the robot itself wants to follow. There are more recent examples of how this sort of attack on deep RL, or in general, attacks on deep RL, can be of concern. One of the recent works by my graduate students is on attacks on automated trading algorithms based on reinforcement learning. The paper is coming out in a couple months, so I can't go into more details, but this is one of the more severe and urgent cases for security researchers to consider. Deep RL is already being used by many major financial players and stock traders, and it can be easily manipulated in the real world. There are other cases, of course, but this is one of the examples that demonstrates the practicality and the applicability of this attack to real world scenarios. Now, before we go further into different types of attacks, let's develop a threat model for deep reinforcement learning. Again, the adversary's objectives can follow those of the CIA triad. The adversary may wish to access the internal configurations, like model parameter, reward function, policy, and such, to still the model for intellectual property theft and such. There can be attack on integrity, which means compromising the desired learning or enactment of the policy. There can be attacks on availability, which are essentially compromises of the ability of the agent to perform training or actions when needed. Now, let's look at the attack surface of the RL. This is the general diagram. This is the general block diagram of a DRL agent or DRL system. Now, we have the agent. The agent typically has some memory where it stores its experiences during training, and then those experiences help with function approximation, data-driven policy learning, and such. And then there is an exploration controller, which controls how the agent explores the environment during training. There's an experience selector, how to select experiences from the bank of data or observations stored. In this data set, there is an actuator, which then enacts the actions of the agent inside an environment. The environment is connected back to this agent block through an observation channel, observation of the state, and reward channel. And it's of no surprise to the more seasoned security researcher and professional that all of these components can be a subject or target of attack. As we go forward, we'll cover some examples of attacks that can occur on each of these components. We've already seen an attack on the observation channel. There are attacks on the reward channel, which we'll hopefully touch on. And there are attacks on the agents during training and the actuator. Let me give you a quick example of the actuator attack. Let's assume we have a robot, an actual robot, learning to navigate in an environment while avoiding obstacles. If the robot commands or decides to, let's say, move the left wheel forward, but there is some sort of obstacle in front of the left wheel, and the left wheel doesn't actually move, then the actuator, the resulting data is going to be, the resulting observation is going to be skewed, because the agent is going to assume that the actuation has happened and then look at the changes in the observation and use that to retrain its policy, to optimize its policy based on faulty data. What are the adversarial capabilities? Well, we first look at different attack modes. The attacker can perform a passive attack where it's only observing the target, it's not changing anything, or it can perform an active manipulation. In passive attacks, the attacker can perform inverse reinforcement learning to learn about the reward function of the agent, or later on we'll see it can perform imitation learning to steal the policy. Active measures include attacks on the actuation, observation, or the reward channel. And attacks on observation can be targeting their representation model, how the agent sees the environment, or perturbing the transition dynamics, how the agent sees the changes in the environment. Okay, going back to our initial proof of concept, remember our original goal was to perform a black box attack. So, to achieve this objective, we introduce an approach based on the transferability of adversarial examples. So, we create a second DQN, the adversary creates a second DQN with similar architecture, but different initial parameters. When I say similar, it doesn't necessarily have to be a match, it just needs to be a convolutional neural network with some functional approximation techniques, but it doesn't need to have the same parameters or the exact same architecture. And it trains that agent, that model, on the same environment. The assumption is the adversary has access to that environment, and then uses the knowledge of that architecture and the trained replica policy. It crafts adversarial examples the same way it did for the white box case, and we know that many of those adversarial examples can transfer to similar models, train on similar data, which also you can see applies to DQN policies as well. And this is how we implemented a black box attack against deep RL policies. Now, what about training time attacks? In the same paper we introduced the policy induction attack, where the attack is of the adversarial example type against a DQN agent during training. Now, what are the different steps in this attack? First, the adversary derives an adversarial policy from the adversarial goal by training on the same environment where the target is going to perform or be trained in. So, if the adversary wants to minimize the reward gained in the game by the agent, then the goal, the optimization goal, is going to be the exact opposite of that of the target policy. The target wants to maximize the reward and the adversary wants to minimize it. Of course, there are different ways of formulating this adversarial goal. Then, the adversary creates a replica of target's DQN and initializes it randomly. And then comes the exploitation phase, where the attacker observes the current state and transitions in the environment, then estimates best action according to the adversarial policy derived in step one. Then, the attacker crafts perturbations to induce adversarial action based on the replica of target's DQN. This is exactly the same as our black box test time attack. The attacker applies the perturbation as a man in the middle in the observation channel. The perturbed input is revealed to the target and the attacker waits for target's actions. And this is a loop. This loop can go on until the training process of the target either converges to a suboptimal policy or up to a certain number of iterations. This is a very rough plot. This is not smooth yet, but you can see that the unperturbed agent moves towards... this is in the game of Pong still... moves towards convergence to an optimal total sum of rewards, while the attack agent moves towards a convergence to the minimum possible return of zero. It's getting closer and closer to zero, which indicates that the training process of DQN and DeepRL in general can also be targeted through adversarial attacks. Now, I'm going to introduce another type of training time attack. Again, this aims to induce some form of misbehavior. We call this misbehavior addiction, and this is a follow-up to a work I did with my colleague Roman Niempolski on psychopathological modeling of AI safety problems. This is a proof of concept. We consider the game of Snake. Many of you probably remember Snake from older Nokia phones. The DQN agent is a snake and is learning to play in this environment. The attacker adds a drug seed with more instantaneous reward than the typical seed, but it also results in more increase in the length of the tail. You can see that this can end up... well, if the increase in the tail length is more than a certain amount, then a longer-tailed snake is bound to eat its own tail sooner rather than later. So, we show that we actually derive some theoretical closed-form solutions for what the additional reward and increase in the tail length should be for addiction to emerge, meaning that the agent learns a more myopic policy instead of the optimal policy, and you can see that it's actually possible to make the agent addicted to the drug seed, as we call it, and this results in learning a sub-optimal policy. And, of course, due to time limitations, I'm going to introduce only one more type of attack, and it's that of targeting the confidentiality of a deep RL policy. The problem here is, or the question is, is it possible to extract a deep RL policy from observations of its actions? Why does this matter? Well, the security challenge posed by this sort of action is, of course, model theft. A company like Google or, let's say, Uber or Waymo may have spent billions or millions of dollars on coming up with a very accurate deep RL policy for autonomous navigation, and if it can be stolen by an adversary, then the intellectual property becomes worthless. And also, a stolen policy, an extracted policy, can be leveraged in integrity attacks in the same way that we mounted black box attacks on DQN policies. So, let's see. As it happens, a branch of reinforcement learning, or in general, one solution to the sequential decision-making problem, is not in the RL domain, but in the supervised learning domain, and it's called imitation learning. Imitation learning is the supervised learning of policies from observed behavior of an expert, and by behavior, I mean state action behavior. What is the policy of an expert? Based on this concept, in 2018, Hester et al. proposed DQFD, or DQ Learning from Demonstrations, which is DQN, where the initial training is done based on observed data using deep learning. So, they have data from human players playing a certain game, or human performance doing a certain task that they want the agent to learn. The initial step of training for a DQFD agent is supervised learning on observed data, and then it starts building on top of it through reinforcement learning approaches, and it was shown that it can result in faster convergence, better sample complexity, and sometimes more interesting and robust policies. As security researchers, you can probably see where this is going. This wonderful algorithm, DQFD, can also be used to replicate policies. Instead of applying it on observed data collected from human performance, it can be applied on observed data from a target policy. So, here's a proof of concept attack the procedure. The attacker observes and records n interactions of the SARSA type, state action, the next state, and the reward based on this transition of the target agent in a particular environment, and then the attacker applies DQFD to learn an imitation of the target policy and Q function. Now, at this point, the attacker may either just go away and sell the extracted policy, or it may decide to target it using different adversarial perturbation attacks, some of which we've covered so far in this talk. So, as a proof of concept, we consider a slightly less complex environment data cart poll, where the objective is to stabilize this poll on this cart by moving the cart to right and left. The reason for choosing a simple environment is merely economical, because we didn't want the experiments to take days or weeks. We start with a simple case, and we consider different types of policies. DQN with prioritized replay, an enhanced version of the classical DQN proximal, policy optimization, and asynchronous actor-critic. And, of course, we also train an adversarial RL agent, a DQN agent, whose objective is to incur maximum loss of reward, or in other terms, in more technical terms, maximize the regret of its target. And here are the results. First, with regards to replication progress, based on only 5,000 demonstrations, 5,000 state action, next state reward observations, we see that all three policies are almost exactly replicated. We can see convergence to the optimal performance of those policies in the environment. And then we perform adversarial training. We train an adversarial RL agent to attack and maximize the regret of those policies, and you can see that this can also be easily achieved within very few iterations of training in CardPol. I believe we can see that for PPO2, which is a somewhat robust deep RL algorithm or approach, it's possible to incur maximum damage or find a policy that incurs maximum regret on the target within 60,000 iterations, which is a relatively low amount. Now, what about defenses? So, of course, similar to adversarial examples, one approach or one technique for reducing the impact of adversarial example attacks is through regularization. And one common type of regularization in supervised learning for mitigating adversarial example attacks is adversarial training, training the model on adversarially perturbed samples to make sure that it sees different perturbations of the same image and knows that all of those results in the same label, in the correct label. This is called essentially data augmentation in general as a regularization technique. So in 2017, when I was just starting to look into this problem or this domain, I had gone through the adversarial training literature and thought that the same may also hold true for deep RL. I made a hypothesis, actually two. One is with regards to recovery. If training time attacks are not contiguous, if not all of the observations are perturbed, then DRL adapts to the environment and adjusts the policy to overcome the attacks. This is training time. And with regards to robustness, I made another hypothesis. Such policies, policies trained under attack are more robust to test time attacks. And this particular investigation is published in a paper titled whatever does not kill deep reinforcement learning makes it stronger. So similar as before, we are looking at DQN in Atari games, Breakout, Enduro, and Pong. Now, the way I designed the experiment was based on a probability of attack. So as an attacker, I assign a certain probability for each state, for each observation during training time to be perturbed. I perform experiments, different experiments with different values of this P attack, 20%, 40%, 80%, and 1, which means contiguous attack. And it's interesting to see that for values of P less than 50%, the agent actually recovers. But for values greater than 50%, the training process plummets and either does not converge or converges to a very, very low mean return value or mean total reward value. I later on publish a theoretical analysis of why this happens. It's available in my PhD dissertation, which I'll reference in the final slide. Also, it was interesting to see that the robustness is also, the robustness hypothesis is also true. You can see here that after training, if we attack the test time policy with probability of 1, the plane or vanilla policy, a policy that was not trained adversarially, performs very poorly. However, policies trained under adversarial attacks with probability 0.2 or 20% and 0.4 perform really well. For 80%, it's even for policies trained at 80%, as you can see, the policy itself is already performing very poorly. It's surprising to see that at P equals 1, the performance gets slightly better. It's comparable with 40%. To this day, I'm not entirely sure why this happened. I've repeated the experiments a number of times and still get the same result. I still don't know why this happened. This is one of the interesting problems that we are looking at right now as a research group. Another defense that we introduce is based on parameter space noise. This is very much like dropout. The idea of parameter space noise was introduced in 2017, I believe, independently by Plamper et al and Fortunato et al. And the idea here is, again, similar to dropout, to introduce zero mean random noise to the learnable parameters of neural network in deep RL to enhance exploration and convergence in deep RL benchmarks. Now, in another paper in 2018, we investigate whether this approach can be used to mitigate the impact or severity of policy manipulation attacks on DQN, and it was shown that it actually performs very well compared to vanilla or classical DQN. And these are the training time results. You can see that if noisy net is used, if parameter space noise is used, the performance degradation for all environments is at a much lower slope, at a much lower rate than the vanilla architecture. Finally, to have proposed a solution for the policy extraction problem, I, along with William Hsu at K-State, came up with the idea of watermarking DRL policies. Watermarking has already been introduced in deep learning in general. The idea is to come up with a unique signature that is both difficult to remove and does not impact the performance of the policy itself, or the model itself, but it still provides a unique signature, proof that a model is the same as another model, or is a replica of the suspected model. So we introduce an interesting watermarking procedure. I know it's arrogant of me to call my own work interesting, but I still get excited when I think about the moment I came up with this idea. The idea is to create a second environment whose state space is this drawing from the main environment. So, create an environment. If you're training an agent to play a game, let's create another environment which has no states. None of these states are the same as the original training environment, or deployment environment of the agent. But the dimensionality of the states are the same. So, if each state in the original environment is represented by, let's say, three values, three features, then each state in the second environment is also represented by three features. It doesn't really matter what the second environment looks like. It's just some other environment that the agent may interact with. And then, we craft the transition dynamics and reward procedure for the second environment such that the optimal policy follows a looping trajectory. So, an optimal policy for an agent trained in the second environment is going to be one that follows a loop, goes through, let's say, state one, then state two, state three, and then goes back to state one. During training, what happens is, we periodically alternate between the two environments. So, let's say, at every N iterations of the training process, we take our RL agent from the original environment, we drop it in the second environment, we train it for a few iterations, and then bring it back to the original environment. Now, once trained, if we want to examine the authenticity or whether a policy is copied or not, we apply the policy in the second environment and measure the total reward. Here's the experimental setup. Again, we are working with CardPool. So, the watermarking environment is defined with five states, states one to four, plus a terminal state, which should never be reached if the policy is optimal in this environment. And none of these states, as represented here, can be found in the original, can occur in the original CardPool environment. These are all highly impossible, not highly, definitely and absolutely impossible to occur in the original environment. As for the transition dynamics, this is how we've defined it. Let a particular state be A0 and A1. If we are, I'm sorry, actions. Actions A0 and A1. If the agent is in state I and performs action I modulo two, if I is even, this is going to be action zero. If I is odd, this is going to be action one. Then the next state is going to be state I star or I modulo four plus one. So, and it receives a reward of one. If the agent performs any other actions, instead of going from one to two, it performs any other action, it will immediately go to the terminal state and receives a reward of zero. So, the optimal trajectory is state one to two to three to four back to state one and so on. All right. Let's see how it works. Let's look at the test time performance comparison of watermarked and nominal, non-watermarked policies. The watermarked policy performs exactly as well as unwatermarked policies. It reaches the optimal or best performance of 500. And when it's applied to the watermark environment, the second environment, it also performs optimally. It gets the maximum reward possible. However, you can see that if we try to apply the non-watermarked policies to the watermark environment, we'll see very, very small values of score, the total reward. So, you can see that it's possible to determine whether a policy is authentic or not or whether a policy is an exact copy of another policy by just applying it to a second environment and see whether it performs optimally or not. Now, there's there are many other things that I really wanted to touch upon in this talk, but unfortunately we are a little short in time. For practitioners, it may be of interest to have some way of benchmarking or evaluating the resilience and robustness of policies and compare different policies, different approaches with regards to the resilience and robustness. Some of my work already introduces or proposes an RL-based approach to perform this evaluation and benchmarking. I've also done some work on investigating the impact of hyperparameter choices on resilience and robustness of DQNs in particular, but also other model-free and actor-critic approaches. This can be very helpful to those who want to engineer and design the new RL agents to be deployed in critical environments. Also, something that I wanted to mention, but unfortunately I don't have time to do so, is that adversarial training is not a silver bullet. It's not an answer to all of the problems in DQN agents. There are certain limitations for robustness and resilience obtained from adversarial training of DQN agents. And also adversarial training is very costly in general, especially when it comes to real-world scenarios, real-world environments and actions. And some of my recent work is focused on improving the sample efficiency and computational cost of adversarial training via a new exploration mechanism called adversarially-guided exploration, or AGE. All of this work can be found in my Ph.D. dissertation, which bears the same name as this talk, False Narcotic Star Security and Deep Reinforcement Learning. You can find it if you search my name on Google Scholar. If you're interested, of course, all of these are published in separate papers in slightly more details under the same titles. And finally, some of the open areas of research in this domain. With regards to training time resilience and robustness, not much has been done with regards to policy search and actor-critic methods, as well as model-based and hybrid methods. Of course, when we talk about model-based, there are some approaches from optimal control theory and approximate dynamic programming that may be applied here, but very few have looked at this problem from a security point of view. So if you're interested, this is one of the areas that is in dire need of a security-oriented investigation. As for mitigation of policy replication, one of the ideas that my research group is currently working on is constrained randomization of policy. So randomize the policy such that the replication through techniques like imitation learning becomes more costly, more samples, more observations will be required, while preserving the performance of the policy. There's almost no work done in multi-agent settings. Of course, adversarial reinforcement learning has been investigated in settings where there are zero-sum agents, but not really where there is an external adversary or adversaries trying to exploit the inner workings of the agents, the RL components of the agent. One more thing that is of note is the importance of discounting. The addiction problem that I demonstrated earlier in the snake agent is mostly due to the constrained discounting solution. For those of us who come from a reinforcement learning background, you're familiar with the basics of reinforcement learning. You probably know that the discount factor is typically chosen to be 0.99 or something in the same ballpark and is left the same. It's treated as a constant throughout the training process. But this is very far from how our brain works and very far from the optimal approach or accurate approach to discounting. Our research group has recently started looking into this problem and is working on developing adaptive discounting solutions to enhance the resilience and robustness of RL agents, particularly deep RL agents in complex environments for AI safety and security purposes. Now, of course, there are naturally inspired approaches that can be looked at. For example, approaches coming from, let's say, TD lambda models or dopamine models of psychopathological problems or neurological problems and the solutions prescribed to those, as well as approaches in social sciences, which may help with the security problems arising in multi-agent RL settings. Very well, thank you very much. And I believe at this time I should be available for your questions."
}