1
00:00:03,150 --> 00:00:15,350
Hi, I'm Leo, the founder and CEO over at Graphistry, and I'm going to be talking today about Project Domino, which is not even at Graphistry, but a community open citizen data science effort against COVID misinformation.

2
00:00:15,870 --> 00:00:26,190
And before we go into it, I do want to say thank you to the DEF CON staff, AI Village staff, and the Project Domino volunteers, and also AI Village in particular, Sven.

3
00:00:26,330 --> 00:00:32,890
This is a crazy time for all of us, and so to be able to put this together, a big thank you for that.

4
00:00:32,890 --> 00:00:35,410
I'm honored to be giving this as a keynote.

5
00:00:35,650 --> 00:00:49,070
I'm hoping folks here will benefit from hearing what we've been learning, and also if you are impacted somehow by COVID and want a way to give back, this might actually be an opportunity for you as well.

6
00:00:49,090 --> 00:00:53,430
With that, I want to start with an example.

7
00:00:53,570 --> 00:00:54,790
This is about three months ago.

8
00:00:54,790 --> 00:00:57,930
It might feel like three years for some of us.

9
00:00:58,330 --> 00:01:05,130
And around the time when the Seattle breakout was happening, our friends were kind of dealing with the hospital situations over there.

10
00:01:05,570 --> 00:01:10,030
I think it was already like people were actually dying already in Seattle.

11
00:01:10,870 --> 00:01:19,390
Governor Kevin Stitt from Oklahoma posted this strange tweet, eating with my kids and all my fellow Oklahomans at the collective OKC.

12
00:01:19,390 --> 00:01:22,950
It's packed tonight, support local, Oklahoma proud, basically.

13
00:01:22,950 --> 00:01:28,890
And then he clarified next day in a press release, we encourage Oklahomans to go out to restaurants.

14
00:01:29,130 --> 00:01:33,230
So for us, with ties to the medical system, our head kind of was exploding.

15
00:01:33,230 --> 00:01:35,850
Like, what's up?

16
00:01:36,150 --> 00:01:47,070
Fast forward to a couple weeks ago, and unfortunately for Governor Kevin Stitt and his family, he tested positive for COVID.

17
00:01:47,070 --> 00:01:57,070
And unfortunately for his constituents who are relying on him, they have about a thousand per day infections at this point, because kind of what we feared would happen is exactly what happens.

18
00:01:57,190 --> 00:01:59,550
And now Oklahomans are dying.

19
00:01:59,810 --> 00:02:06,250
So it's... I do want to do kind of some disclosures and stuff in a bit.

20
00:02:06,250 --> 00:02:08,170
But to us, it's not about a left or right thing.

21
00:02:08,170 --> 00:02:11,790
It's like we're, you know, this is a medical health emergency.

22
00:02:12,130 --> 00:02:13,410
We need to do stuff.

23
00:02:14,110 --> 00:02:18,730
And that's kind of a good note, kind of where this project came from.

24
00:02:19,210 --> 00:02:24,750
We're kind of looking at kind of what can we do.

25
00:02:24,930 --> 00:02:34,410
And there's this interesting statement by the... from the World Health Organization saying, we're not just fighting an epidemic, we're fighting an infodemic.

26
00:02:34,650 --> 00:02:50,230
And kind of the reasoning that we kind of believe in here is until there's a medical cure, a pill, a vaccine, been tested, it's safe, kind of the only thing we really have is behavior change.

27
00:02:50,230 --> 00:03:02,590
So for that, unfortunately, you either need to get top-down policy enforcement or bottom-up kind of citizen voluntary compliance.

28
00:03:02,590 --> 00:03:05,490
And unfortunately, we have folks standing in the way of that.

29
00:03:05,490 --> 00:03:14,450
And, you know, for Seattle, for Oklahoma, for all of us at home, we're seeing kind of the effects of failures to win that fight.

30
00:03:15,390 --> 00:03:17,390
Some quick disclosures here.

31
00:03:17,390 --> 00:03:19,410
We are going to talk about real people.

32
00:03:19,690 --> 00:03:22,350
We've published... actually, we haven't directly published.

33
00:03:22,350 --> 00:03:23,730
We publish through partners.

34
00:03:23,730 --> 00:03:26,730
So we empower researchers to do through our tech.

35
00:03:26,730 --> 00:03:29,150
So they've published on some of this stuff before.

36
00:03:29,150 --> 00:03:36,390
And based on some of the results today, I expect to see some accounts going out and hopefully some takedowns.

37
00:03:37,230 --> 00:03:40,430
Likewise, this is a volunteer organization.

38
00:03:40,430 --> 00:03:48,590
So the views expressed by us, me, this talk, they're not necessarily representative of the employers of the volunteers.

39
00:03:48,950 --> 00:03:53,010
Likewise, you should expect some bias here.

40
00:03:53,010 --> 00:03:59,010
So a lot of us, actually, some people on the team were even affected by COVID and didn't have their lives impacted by that.

41
00:03:59,010 --> 00:04:00,030
Family members do.

42
00:04:00,030 --> 00:04:02,290
Their work and work life has been impacted.

43
00:04:02,410 --> 00:04:04,850
Misinformation is a broader topic.

44
00:04:04,850 --> 00:04:11,610
I'm personally in a place some folks may remember called the Soviet Union, where propaganda was a thing and surveillance was a thing.

45
00:04:11,610 --> 00:04:13,650
So conflicting thoughts on that.

46
00:04:14,270 --> 00:04:20,690
But I do want to point out from the team, I do especially want to call out some of the kind of folks on the leadership team.

47
00:04:20,690 --> 00:04:24,010
Sean Griffin's been involved in pandemic response.

48
00:04:24,010 --> 00:04:28,730
Actually, he helped both Obama and the Trump administration with pandemic planning.

49
00:04:29,490 --> 00:04:35,230
Anita was over at the NSF and looking at a lot of digital crime, blockchain, stuff like that.

50
00:04:35,230 --> 00:04:36,870
She's at UIUC now.

51
00:04:36,870 --> 00:04:44,410
Cody Webb has been this research powerhouse for our team and researching misinformation forever.

52
00:04:44,610 --> 00:04:49,410
I shared the names of some of the engineers helping on our backbone.

53
00:04:50,650 --> 00:04:59,150
The first example today will be actually Jeff Goldberg over at Social Forensics and David Knickerbocker, an independent researcher, has been helping us a bunch on.

54
00:04:59,750 --> 00:05:08,050
And I think, oh, yeah, and then tomorrow, if this is interesting to you, kind of the same type of stuff we're doing for COVID.

55
00:05:08,830 --> 00:05:20,070
Andrea's going to be talking over at the Biohacking Village about imagine you're a cancer patient and you go to Facebook and then you see all the same misinformation giving you false hope or misguiding you.

56
00:05:20,330 --> 00:05:21,950
So it's important stuff.

57
00:05:23,150 --> 00:05:23,790
Okay.

58
00:05:23,790 --> 00:05:26,710
So today, I'm going to talk about basically three things.

59
00:05:26,850 --> 00:05:32,750
First, I just want to take you down the rabbit hole through an example from anti-vaccination misinformation.

60
00:05:33,190 --> 00:05:42,130
Second, I want to then give you more of a security data science perspective and how we're actually going about it, trying to figure out scale and autoresponse and all that stuff.

61
00:05:42,150 --> 00:05:53,130
And then briefly, but incredibly important, why we're doing this as Open Citizen Data Science and why I'm actually pretty hopeful about where this is going long term.

62
00:05:53,130 --> 00:05:58,890
And maybe this could kind of give you a chance or give you ways to help out the community.

63
00:06:00,390 --> 00:06:01,510
Okay.

64
00:06:01,610 --> 00:06:03,530
So let's dive in.

65
00:06:04,010 --> 00:06:13,810
So we were doing an investigation and this name popped up, Sherry Tenpenny, and we decided to see where we go for the purposes of this talk.

66
00:06:14,830 --> 00:06:19,670
And for some folks, if you're from Australia, this may be a familiar name.

67
00:06:20,530 --> 00:06:24,490
She's Wikipedia famous or infamous at this point.

68
00:06:24,490 --> 00:06:39,330
So her 2015 lecture was canceled in Australia because apparently she is involved in a lot of vaccine misinformation and private organizations in Australia wanted no part of that.

69
00:06:39,370 --> 00:06:43,210
And so they didn't support her.

70
00:06:43,510 --> 00:06:52,680
For our purpose and purposes, I do want to point out there's kind of an ecosystem with different incentives here.

71
00:06:52,680 --> 00:06:54,780
And so if we look at somebody like Dr.

72
00:06:54,780 --> 00:07:00,480
Sherry, we see that she does teach courses, like hence the book tour.

73
00:07:00,480 --> 00:07:01,420
She sells books.

74
00:07:01,420 --> 00:07:02,900
She trains folks.

75
00:07:02,900 --> 00:07:06,740
She has about 50, 60,000 followers on Twitter, not shown.

76
00:07:06,740 --> 00:07:10,860
She also has a big presence on Facebook and also on newsletters.

77
00:07:10,860 --> 00:07:12,880
So stuff that you don't even see.

78
00:07:13,100 --> 00:07:14,380
And she gets paid to do it.

79
00:07:14,380 --> 00:07:18,520
Now it's kind of like her beliefs and her livelihood are the same thing.

80
00:07:18,520 --> 00:07:20,240
So she's very invested in this.

81
00:07:22,660 --> 00:07:28,420
But once we start looking at it, there are kind of a few things that looks a little strange.

82
00:07:28,740 --> 00:07:37,700
So she talks about Nazis in her profile, like that's the first thing, that's the first word she shows you.

83
00:07:38,420 --> 00:07:42,520
And so that kind of is kind of like a whistle to one side of the house.

84
00:07:42,660 --> 00:07:44,540
And then talks about abuses by the state.

85
00:07:44,540 --> 00:07:47,160
So that's kind of like a different kind of thing.

86
00:07:47,160 --> 00:07:52,000
And so it's kind of interesting to ask, like, who is the community around her?

87
00:07:52,180 --> 00:07:53,620
So let's take a look.

88
00:07:54,660 --> 00:07:59,280
So what I did here is loaded in busy Dr.

89
00:07:59,280 --> 00:07:59,840
T.

90
00:08:00,500 --> 00:08:11,680
And so we're looking at her, and what we see in colors around her are basically, she's only following a couple hundred people, but she has 50,000 people following her.

91
00:08:12,020 --> 00:08:26,680
And so when we look at a sample of that, what we see up top is a bunch of folks, essentially public, basically anti-vax, natural medicine, a lot of kind of medically-minded folks that you'd sort of expect to be here.

92
00:08:26,680 --> 00:08:30,440
But then the surprise is once we start seeing the other colors.

93
00:08:30,440 --> 00:08:38,720
And so the tools here just kind of automatically colored and laid it out and everything, like it's kind of just traditional graph analytics.

94
00:08:38,920 --> 00:08:44,360
So we see Donald Trump, we see Michelle Malkin, we see Candace.

95
00:08:45,900 --> 00:08:51,800
Basically, like in the dark green, we're seeing these public figures, kind of national conservative public figures.

96
00:08:52,640 --> 00:09:05,180
On the right, we're seeing, but a bit on the right, and then actually, especially down here, we're seeing a mixture of QAnon, which is a conspiracy group we're hearing about a lot nowadays.

97
00:09:05,600 --> 00:09:09,700
We're seeing a lot of MAGA and Trump supporters.

98
00:09:09,700 --> 00:09:12,200
So not super surprising given we had Trump.

99
00:09:13,120 --> 00:09:17,940
And so it's kind of interesting that we have on the top, we have like, who do you expect with anti-vax?

100
00:09:17,940 --> 00:09:21,040
But on the bottom, we have like the whole MAGA thing.

101
00:09:21,040 --> 00:09:23,340
We have conspiracy theorists from QAnon.

102
00:09:23,340 --> 00:09:26,860
They're actually even like UFO stuff going on, on the bottom.

103
00:09:27,080 --> 00:09:30,060
It's kind of like interesting to see that mix.

104
00:09:30,660 --> 00:09:32,840
So I want to point out another phenomenon here.

105
00:09:33,260 --> 00:09:37,780
When we're looking at each community, like the question is like, how are they interacting?

106
00:09:37,780 --> 00:09:39,880
Is it like they all really love our Dr.

107
00:09:39,880 --> 00:09:40,480
Busy T?

108
00:09:40,480 --> 00:09:42,420
Like she just wrote this wonderful book.

109
00:09:42,880 --> 00:09:49,220
What we actually see is like, for example, from the QAnon zone, they're reaching into the medical zone.

110
00:09:49,220 --> 00:09:51,080
So they're following folks there.

111
00:09:51,140 --> 00:10:00,420
Conversely, if we pick somebody big in the kind of this top region, sometimes it's localized, but other times you see it's reaching into the QAnon zone.

112
00:10:00,820 --> 00:10:05,560
Kind of no matter which cluster I do, you're going to see those reach outs.

113
00:10:07,120 --> 00:10:07,900
Excuse me.

114
00:10:07,900 --> 00:10:17,100
So that's actually kind of strange, where there's no reason to believe one conspiracy has anything to do with another.

115
00:10:17,640 --> 00:10:20,640
And so we're actually kind of almost like in this other world.

116
00:10:21,520 --> 00:10:25,700
And what we're seeing is the networks do cross over.

117
00:10:26,700 --> 00:10:30,500
And so the first thing is what we're seeing is the question is why?

118
00:10:30,500 --> 00:10:33,360
And that's basically to support influence.

119
00:10:34,460 --> 00:10:37,800
Sometimes that could be direct where you're just retweeting from each other's.

120
00:10:37,800 --> 00:10:40,140
Other times it might be to trick the Twitter algorithm.

121
00:10:40,140 --> 00:10:42,960
So for example, here is actually a Fox News reporter.

122
00:10:42,960 --> 00:10:44,460
So it's someone on TV.

123
00:10:45,560 --> 00:10:53,460
They have 40,000 followers, which is reasonable for a personality, but they're following 30,000 people back.

124
00:10:53,460 --> 00:10:58,780
So that probably suggests an inauthentic and automatic action of following back somehow.

125
00:10:59,040 --> 00:11:04,580
We don't know what order they happened in, but in general, that's prohibited by Twitter's terms of service.

126
00:11:04,580 --> 00:11:06,740
But they get clicks.

127
00:11:06,740 --> 00:11:11,460
And so even though it violates terms of service, there they are.

128
00:11:11,780 --> 00:11:15,940
And that's basically the welcome to the misinformation network.

129
00:11:17,360 --> 00:11:22,080
And I realized that I glossed over this, but the reason we care about Dr.

130
00:11:22,080 --> 00:11:34,440
T and that she's talking about this stuff is that she and her community are already pushing messages against a COVID vaccine, even though one hasn't even passed trials yet.

131
00:11:34,440 --> 00:11:37,800
She's against it before it's even been tested, just by principle.

132
00:11:38,140 --> 00:11:41,600
And so Twitter's been supporting her in doing that, essentially.

133
00:11:41,600 --> 00:11:43,060
And I'll get to that in a second.

134
00:11:44,380 --> 00:11:50,520
So I do want to take one more, I want to go a little bit deeper into our Dr.

135
00:11:50,520 --> 00:11:50,980
T.

136
00:11:51,460 --> 00:11:55,220
And so in particular, I want to understand the communities, like what they're doing.

137
00:11:55,280 --> 00:11:59,980
So here what we're going to do, and I want to see if there's kind of extra coordination structure.

138
00:11:59,980 --> 00:12:02,400
And so we're getting a bit more on the data science side.

139
00:12:02,460 --> 00:12:11,600
So what's going on here is we took the accounts, we actually took a look at about 100 million COVID-related tweets to see if there's kind of organized behavior within it.

140
00:12:11,780 --> 00:12:23,260
For each account, we just tried to look at for known sort of misinformation topics, what's sort of the signature, like, you know, how many times did it like activate on different topics.

141
00:12:23,260 --> 00:12:28,960
And for each account, we now have that sort of like misinformation fingerprint, essentially.

142
00:12:29,000 --> 00:12:31,800
And then we ran something called UMAP.

143
00:12:31,800 --> 00:12:36,400
I think it's a universal manifold approximation or something like that.

144
00:12:36,400 --> 00:12:40,480
But basically tries to smooth out and assume structure behind that.

145
00:12:40,760 --> 00:12:42,460
And then we took a look.

146
00:12:42,460 --> 00:12:44,840
So let's try it out.

147
00:12:45,620 --> 00:12:50,860
So yeah, so what's going on here is we're taking a look at those accounts.

148
00:12:50,860 --> 00:13:01,440
And we only are showing from all the accounts, we're looking at the 5000 or so that we actually saw active across multiple misinformation topics.

149
00:13:01,440 --> 00:13:08,960
And so these are kind of the really the ones that have kind of that behavior going across topics like are just like what's going on.

150
00:13:11,020 --> 00:13:13,080
And we ran UMAP.

151
00:13:13,080 --> 00:13:16,320
But then what we also did, you're seeing here, I'm drawing these little edges.

152
00:13:16,320 --> 00:13:20,760
We also ran a afterwards k-nearest neighbor and put them on screen.

153
00:13:20,940 --> 00:13:27,820
So that whenever we see accounts, kind of tweeting on similar topics and similar intensities.

154
00:13:28,060 --> 00:13:36,240
So it's really, for example, imagine two bots program to kind of talk about the same thing, we're controlled by the same person to kind of, you know, retweet someone else.

155
00:13:36,940 --> 00:13:38,900
That's when we're going to draw that edge.

156
00:13:39,180 --> 00:13:45,880
And so again, they're not necessarily following, they're actually just on certain topics, they're actually pushing those topics in the same way.

157
00:13:45,880 --> 00:13:47,860
So it's entirely content based.

158
00:13:48,400 --> 00:13:54,420
And what we're seeing here is some interesting structure pops out for across COVID Twitter.

159
00:13:55,080 --> 00:14:03,520
And then the second thing I did here is coloring, you're seeing some of these are a bit pinkish.

160
00:14:04,340 --> 00:14:06,620
So if we jump, let's take a look at this one.

161
00:14:07,000 --> 00:14:12,000
So for this kind of group of activity, we're seeing a lot of this q&a and stuff.

162
00:14:12,000 --> 00:14:18,440
And so what I'm doing here is if in those tweets talking about COVID, they also talk about q&a.

163
00:14:18,720 --> 00:14:22,480
And we have a couple words on that, we call that out too.

164
00:14:22,960 --> 00:14:28,340
And so what we're seeing is like, you know, for this general topic, a bunch of accounts never explicitly mentioned q&a.

165
00:14:28,340 --> 00:14:34,400
But it's definitely maybe it's like two coordinators, or who knows, or like something like that.

166
00:14:34,400 --> 00:14:40,500
But we're actually seeing a lot of q&a influence across these different organized groups.

167
00:14:40,940 --> 00:14:44,460
And actually, in particular, almost all of these have a bit of q&a.

168
00:14:44,720 --> 00:14:47,980
So you're seeing it's not all white, it's all a little off white.

169
00:14:48,000 --> 00:14:52,220
And then what's more, you're seeing some of them are you're seeing these pockets of heaviness.

170
00:14:52,300 --> 00:14:57,960
And we calculated out about 15% of these accounts spread throughout our q&a.

171
00:14:57,960 --> 00:15:02,020
And so at least for COVID Twitter, that was actually kind of interesting.

172
00:15:02,020 --> 00:15:03,520
And for us, that was a surprise.

173
00:15:03,800 --> 00:15:08,980
So that kind of put q&a on in terms of medical misinformation on our radar.

174
00:15:09,120 --> 00:15:10,600
So that's not good.

175
00:15:12,920 --> 00:15:15,600
Okay, so jumping ahead.

176
00:15:17,260 --> 00:15:20,460
So okay, so we're looking at our busy Dr.

177
00:15:20,460 --> 00:15:21,160
T.

178
00:15:22,100 --> 00:15:26,500
And clearly blatant activity going on here.

179
00:15:26,500 --> 00:15:38,780
The last thing I want to point out was, then we look on the right here, and Twitter is saying, hey, you should probably now that you're here, you should also check out this Judy Mikovits.

180
00:15:38,980 --> 00:15:40,820
This PhD sounds authentic, right?

181
00:15:40,820 --> 00:15:41,800
Sounds smart.

182
00:15:42,460 --> 00:15:53,120
You may have also known about this Judy as a video going around called Plandemic, which, you know, got pulled down as pretty bad misinformation.

183
00:15:53,860 --> 00:15:58,040
And so that was pretty shocking to see, like, I was like, oh, that name sounds familiar.

184
00:15:58,040 --> 00:15:59,360
So that was a little crazy.

185
00:15:59,360 --> 00:16:04,540
That was like the number one recommendation of Twitter as soon as you go to this web page is to learn about Plandemic.

186
00:16:04,880 --> 00:16:07,960
So and this is like, you know, way after the fact.

187
00:16:08,440 --> 00:16:25,880
And so kind of in summary, what we're seeing in this, in this basic, pretty basic, honestly, analysis, very blatant anti-vaccine misinformation, a lot of it about COVID in particular, we're seeing it's not just anti-vaxxers, but we're actually seeing a lot of mis,

188
00:16:25,880 --> 00:16:29,220
cross-community support by misinformers.

189
00:16:29,520 --> 00:16:31,080
It's pretty blatant.

190
00:16:31,080 --> 00:16:37,500
It's also it's pretty, pretty obvious how they abuse, like, the very few terms of service Twitter does have.

191
00:16:37,500 --> 00:16:39,580
So that's a little strange.

192
00:16:39,980 --> 00:16:52,900
It is a little disappointing to see Fox, the Republican Party, MAGA, and a lot of them kind of involved in this anti-medical misinformation.

193
00:16:52,900 --> 00:16:54,700
And again, it's so blatant.

194
00:16:55,040 --> 00:17:03,680
And then on the platform side, it's when we saw that overlap and like, oh, wait a minute, Twitter makes money off of some of this stuff.

195
00:17:04,740 --> 00:17:08,820
They're clearly looking the other way for the terms of service violation.

196
00:17:09,000 --> 00:17:12,260
And then like, what was the the the Plandemic promotion?

197
00:17:12,260 --> 00:17:14,280
Just that was just that blew my mind.

198
00:17:14,280 --> 00:17:18,920
So this is not, this is not a healthy situation.

199
00:17:19,580 --> 00:17:27,480
So from a data perspective, or actually security analytics perspective, I do want to step back.

200
00:17:27,960 --> 00:17:36,720
If we think about attackers, defenders, ultimately, the targets are individuals and organizations going on a radicalization journey.

201
00:17:36,720 --> 00:17:44,220
And so every time Twitter recommends you do another crazy thing that just takes you deep, deeper.

202
00:17:44,520 --> 00:17:53,340
Interesting extra level on the threat model here is that folks are essentially influenced based on their life conditions.

203
00:17:53,340 --> 00:17:57,280
So maybe they're going through a hard time, maybe they're elderly and don't really know what's going on.

204
00:17:57,280 --> 00:18:04,640
And also by their community, you know, if your if your pastor or somebody or your family friends tells you something, you're more inclined to believe it.

205
00:18:04,700 --> 00:18:12,060
And so being able to infect communities with this stuff is having mastery over that and platform level support, it's not good.

206
00:18:12,680 --> 00:18:17,060
For actors, it's a lot of what you'd expect.

207
00:18:17,560 --> 00:18:28,040
I think that the surprising thing for for threat actors here is we actually consider the, at this point, the, the platforms to be part of the threat.

208
00:18:28,320 --> 00:18:34,280
In particular, we we break down the platforms as the social network system.

209
00:18:34,280 --> 00:18:41,760
So Facebook owns a bunch of social network, Google owns a bunch of social networks, a few are just big on their own, right, like Reddit, Reddit and Twitter.

210
00:18:42,280 --> 00:18:45,040
And, you know, Twitter is a multi billion dollar profitable company.

211
00:18:45,380 --> 00:18:47,280
But here we are.

212
00:18:47,460 --> 00:18:49,760
A little less obvious to folks is messaging.

213
00:18:49,940 --> 00:18:59,220
And again, actually, a lot of the same companies like Google, Microsoft, Yahoo, own essentially the messaging channels, those are completely unmonitored.

214
00:18:59,220 --> 00:19:03,540
And so a lot of times the radicalization actually hops over to newsletters.

215
00:19:03,540 --> 00:19:05,800
So that's stuff basically dark data.

216
00:19:06,680 --> 00:19:14,060
And also the local news, I think like Fox and folks like that.

217
00:19:14,060 --> 00:19:22,620
A little surprising is there's kind of a monopoly effect where because you do have regional different ways of having monopolies on people's attention.

218
00:19:22,620 --> 00:19:24,960
That's all the reason these companies got so big.

219
00:19:25,380 --> 00:19:29,440
Same reason we're basically at the whimsy of these companies.

220
00:19:29,440 --> 00:19:35,580
And so when you kind of calculate it out, the 80% essentially is just about 13 of these groups.

221
00:19:36,140 --> 00:19:50,760
And so if we have a fickle Silicon Valley CEO, if we have employees who just go along for the ride, if we have advertisers keep feeding them money, you know, so that's kind of from the threat model perspective.

222
00:19:53,020 --> 00:19:58,800
Shifting gears a bit to more like kind of data like security, data engineering, data science.

223
00:19:59,080 --> 00:20:02,780
In order to do the data science, I did want to talk a little bit about this.

224
00:20:03,380 --> 00:20:08,620
We end up having to build a lot to work at kind of the speeds and scales we're targeting.

225
00:20:08,620 --> 00:20:15,340
So if you want to have an instant response pipeline, that means you have to kind of have streaming, you know, can you take hundreds or thousands of things a second.

226
00:20:15,340 --> 00:20:23,960
So our V1, you're kind of seeing the performance of that where you're architecting to a V2 so that we can actually go to arbitrary scale, just have essentially budget limited.

227
00:20:25,060 --> 00:20:27,540
You are noticing we're doing a lot of graph technology.

228
00:20:27,920 --> 00:20:39,180
And so basically that's a nice way to deal with a lot of heterogeneous data and a lot of like linkage of data, even actually in a machine learning world where instead of just direct links, you might have vectors, feature vectors.

229
00:20:39,700 --> 00:20:43,100
Two other interesting things about the stack I did want to point out here.

230
00:20:43,640 --> 00:20:53,260
One is we do do end-to-end GPU computing just to be able to, you know, you can actually, if you want to have a billion node graph in memory, you actually could do that stuff with, for example, a Rapids AI stack.

231
00:20:53,480 --> 00:20:58,300
Likewise, neural nets in this space, graph neural nets are getting interesting.

232
00:20:58,540 --> 00:21:02,800
We actually do more today on the traditional ones, but we're looking at those.

233
00:21:02,940 --> 00:21:06,160
And the last part, and I think most important, is kind of division of team.

234
00:21:06,580 --> 00:21:09,080
We have basically heroes for data engineers.

235
00:21:09,080 --> 00:21:15,980
And then a big part of this is figuring out ways to get out of the way, just given it is an open volunteer effort.

236
00:21:15,980 --> 00:21:18,060
And so we actually do a lot on the self-service.

237
00:21:18,060 --> 00:21:21,040
And that's a big theme for look about powering.

238
00:21:21,060 --> 00:21:29,120
So think notebooks, streamlet dashboards, graphistry, do visual analytics and automatic ML.

239
00:21:29,120 --> 00:21:34,680
Just increasing the ability of people just to go and handle the stuff.

240
00:21:37,040 --> 00:21:41,360
From the data science side, I kind of wanted a fun example here.

241
00:21:41,900 --> 00:21:47,820
For example, we had a news organization ask us about Obamagate.

242
00:21:47,820 --> 00:21:57,720
And so one of our analysts, Cody, did a good job of using hashtag analysis to connect the dots to a month back to some QAnon stuff.

243
00:21:57,840 --> 00:22:03,720
And it was kind of impressive because the Trump tweet made no sense at all, even when I read it today.

244
00:22:03,880 --> 00:22:05,720
All these kind of code words.

245
00:22:06,060 --> 00:22:14,940
But once we moved to more modern NLP techniques and go away from exact match of terms.

246
00:22:15,220 --> 00:22:19,360
So Cody had to look for Obamagate.

247
00:22:19,480 --> 00:22:24,460
But with NLP, we can do things like look for the word Obama and the word conspiracy.

248
00:22:24,460 --> 00:22:27,280
But not the word, but just the fuzzy topic.

249
00:22:27,840 --> 00:22:34,000
And when you have those things, then we can kind of more loosely kind of walk back.

250
00:22:35,080 --> 00:22:38,400
And so that means we can do kind of traditional analysis more easily.

251
00:22:38,400 --> 00:22:40,340
And so we've been building out a lot around that.

252
00:22:42,800 --> 00:22:47,580
I want to modeling from basically what everyone does now.

253
00:22:48,640 --> 00:22:56,000
And then the other thing I want to point out here is this is also opening up types of analysis that just were straight up, like, you couldn't even think about.

254
00:22:56,500 --> 00:23:07,420
So maybe one example could be for looking at masks in a community where you might ask, like, which messages are against mask use?

255
00:23:07,420 --> 00:23:11,960
That's kind of a strange... that's kind of a gets into semantics and meaning a bit.

256
00:23:11,960 --> 00:23:14,040
And that's hard to do just by keywords.

257
00:23:14,660 --> 00:23:16,760
And then could you summarize it?

258
00:23:16,760 --> 00:23:18,420
Could we actually have a demographic breakdown?

259
00:23:18,420 --> 00:23:21,800
There are just certain things that with traditional NLP, you couldn't have done.

260
00:23:21,800 --> 00:23:28,640
But over the years, over the last, like, five years, and actually even two years, it's just been really nuts what's been going on.

261
00:23:33,160 --> 00:23:37,780
Then kind of the last piece I'm going to share here is kind of just where we are, where we're going.

262
00:23:38,280 --> 00:23:48,200
You're kind of getting a sense that we can now bring in a researcher on a topic, and if we have the data, or we help them pull in the data, and kind of do this kind of analysis.

263
00:23:48,520 --> 00:23:57,100
Next level of empowerment we're looking at is, like, as we're building this out, is can we actually partner with less technical organizations, and if we have to kind of do this to protect their community?

264
00:23:57,100 --> 00:24:03,900
Imagine, like, a local civic leader talking about the cancer community with Andrea, with kind of at-risk patients.

265
00:24:03,900 --> 00:24:13,220
And there we're looking at, are there folks who are not even maliciously misinforming in their community, but people are just confused?

266
00:24:13,220 --> 00:24:14,900
Or are there people kind of targeting them?

267
00:24:14,900 --> 00:24:18,220
Can we start mapping out and building tools and stuff around that?

268
00:24:18,820 --> 00:24:23,760
So that's kind of the phase shift we're starting to go into, so that's pretty exciting.

269
00:24:23,820 --> 00:24:28,220
You're kind of seeing, as you build out that state of capability, all this stuff unlocks.

270
00:24:29,120 --> 00:24:34,480
And then, finally, I just kind of want to close out and tie together some of the stuff here.

271
00:24:34,800 --> 00:24:37,400
So the beginning, I was kind of sharing that investigation.

272
00:24:38,300 --> 00:24:44,100
Hopefully, what you got across is, like, yeah, the analysis is cool, and we're able to, like, do these giant analyses.

273
00:24:44,100 --> 00:24:53,900
But, like, ultimately, a lot of the important stuff is so in-your-face and so huge numbers that, like, again, that analogy of fishing with dynamite.

274
00:24:54,420 --> 00:25:07,480
And it's, you know, those 13 companies, they shouldn't be gifting, essentially, millions or billions of dollars in free targeted ads to these groups and targeted recommendations.

275
00:25:07,480 --> 00:25:10,620
It's kind of that stuff that I think is more systemic.

276
00:25:11,260 --> 00:25:18,900
I am disappointed to see kind of these otherwise legitimate groups appear to be corrupted by misinformation.

277
00:25:18,900 --> 00:25:23,400
So call out, essentially, we're seeing MAGA is just so ever-present here.

278
00:25:23,400 --> 00:25:26,280
We're seeing social media companies, regular media companies.

279
00:25:26,280 --> 00:25:33,000
So there are things you could do if you're an employee, if you are anywhere near the advertising spend of your company.

280
00:25:33,000 --> 00:25:34,420
All that stuff's important.

281
00:25:35,860 --> 00:25:41,420
And then kind of turning it around, that's all kind of social policy.

282
00:25:41,640 --> 00:25:45,760
A lot of folks tuning in are technologists of different stripe.

283
00:25:46,940 --> 00:25:55,340
And the bad news is we're seeing these monopolistic media companies are not able to self-regulate.

284
00:25:55,340 --> 00:25:59,740
Like, the left hand and the right hand are always at each other, and it's kind of brought us to where we are.

285
00:25:59,800 --> 00:26:09,200
Likewise, they've kind of gutted investigative and journalist teams just because their ad companies sucked all the money.

286
00:26:09,220 --> 00:26:14,740
And likewise, we're seeing private companies are sort of compromised as they're kind of trying to deal with revenue.

287
00:26:15,020 --> 00:26:17,020
So that's bad.

288
00:26:17,160 --> 00:26:21,560
What's good is we actually have an open source community.

289
00:26:21,920 --> 00:26:25,940
We, I mean, just like the technology community, that is huge.

290
00:26:26,040 --> 00:26:30,700
And now at the same time, data science is, you know, our Python that's almost like built into the DNA.

291
00:26:31,140 --> 00:26:35,800
So I see strong potential and gathering potential for that.

292
00:26:36,300 --> 00:26:38,480
I encourage that.

293
00:26:38,880 --> 00:26:42,400
You're seeing some of that happen here with Project Domino.

294
00:26:42,400 --> 00:26:50,180
And so on that point, if you are into this kind of stuff and you could help out with coding, head into our Slack.

295
00:26:50,560 --> 00:26:59,800
Likewise, as we launch in the coming months, these kind of more self-serve and tiers for folks sitting on top, reach out for partnering.

296
00:26:59,800 --> 00:27:17,500
And likewise, while we are looking at doing sustainable approaches to our growth, especially as sort of a bootstrap phase that we're in for the next year or so, we're definitely, you know, grants, whatever, that stuff does help with the core team do what they need to do.

297
00:27:17,500 --> 00:27:20,500
Likewise, you know, serverless compute time, anything like that.

298
00:27:21,380 --> 00:27:23,900
With that, thank you for your time.

299
00:27:23,900 --> 00:27:32,760
And hopefully this changed a little bit about either how misinformation is working in practice in the COVID era.

300
00:27:32,760 --> 00:27:39,900
And also, even more importantly, that we actually could do things and we actually detect it and then can actually build tools for it.

301
00:27:40,280 --> 00:27:41,280
Thank you.