Yeah, I'm Raf, everyone.
I worked at a quantum computing company for the last five years prior to that.
Academically, I was a physicist, but left that to do machine learning and actually make some real money for a while.
Got tired of that and came back to physics.
I work at a company called Xanadu.
But I feel like, as is tradition at DEF CON, everything I'm going to be talking about are more my personal viewpoints than data.
Anything about what I do, my day-to-day work, with Xanadu.
Also, all the information I'm covering here is publicly available.
So there isn't anything I'm going to be talking about that you probably couldn't find on your own.
But hopefully I'll provide a little bit of a unique perspective on where we're at with quantum computing today.
So, maybe before that I'll realize I should have probably introduced what NIST actually means.
You might have heard this term when applied to quantum technologies.
And it means Noisy Intermediate Scale Quantum Computers.
And really, we're talking about the quantum computers we have today.
And the noisy aspect is just from the fact that if you try to store information on a quantum computer, or you try to operate on that information, you're not going to have a perfect success rate.
By either reading that information back out, or being able to, let's say, entangle two qubits together, or perform some operation on them.
Mark mentioned his favorite machine right now is the IoT machine.
And that's because they have the highest probability of success for these operations of most available machines today.
And it makes work a lot easier when you're trying to do quantum computing.
But noisy really means for that fact that we are maybe at two nines right now in terms of what the success rate is for a given operation.
Different machines will vary.
Really bad ones are really bad.
Great ones don't go much past two nines.
So the real vision of the entire community right now, whether it's for-profit companies, whether it's tech giants, startups, academics, the open source community, is to really build quantum computers that are useful, and more specifically, valuable to people.
And I really don't want to take away from the fact that quantum computers drive a quantum-valuable brain.
But this is not the value I'm talking about.
Quantum computers today have changed how we actually teach quantum mechanics to people around the world.
The fact that there's this number of people sitting in a room right now that want to learn about quantum mechanics would have been unheard of 20 years ago.
And that's really the value that's been generated by quantum computing.
It has also really driven fundamental research forward.
Whether it be on materials, whether it be in quantum information science, whole new avenues of research have been opened up.
Of course, it's also allowed us to come up with future possibilities of what computing will look like after digital computing hits the end of the horse lock, if it ever does.
So it offers that promise moving forward.
And of course, I don't think Black Mirror episodes would be as fun if quantum computers didn't exist.
We wouldn't have the quantumputer or anything like that.
So a lot of value has been generated.
But that's not the value that we're talking about.
The value we're really talking about is stuff you would be willing to pay for, that companies would be willing to pay for.
And today you probably have about three or four publicly listed, pure-play quantum companies.
And they're trying actively to demonstrate that there is value today.
But a core part of this talk will be identifying that there is still a lot of work to do to move forward.
So the real question is, when are we going to hit a corner?
How are we going to hit it?
And I think that the dollars that have been invested in the quantum computer will actually be brought back through some sort of meaningful contribution to society.
I should mention, even though none of the views represented in this talk are from Xanadu, a lot of material is still in Xanadu.
But it's all publicly available on either YouTube as talks, from my colleagues, or directly on our website.
Your immediate kind of point might be, but Raph, we've heard over the last 20 years, starting with D-Wave, that quantum computers have a proven advantage over classical computers.
And I would say we probably see a headline like this now once a month, that a quantum computer has said something.
Most recently there was an exciting announcement of a quantum chemistry problem being solved by an IBM quantum computer.
Usually what happens to my inbox is everybody that I've always talked to and had a negative view on the state of the quantum today, emails me to say, I told you so.
And I usually wait about 48 hours to respond, because in 48 hours, usually there's a paper or an archive that says, but actually, you didn't consider the latest technique.
You didn't try this with a tensor network simulator.
This problem is completely, you know, complete or completely not meaningful for the industry that describes it.
So all of that actually ends up being a good point, and I'm often very satisfied in just sending back simply an archive link as a response to those emails.
There is one exception to this, and that has been the series of what I would call quantum supremacy or quantum computational advantage announcements that have happened over the last three or four years.
So I'm just going to briefly talk about what quantum supremacy is versus advantage.
So usually when people talk about quantum advantage, they talk about advantage on a meaningful problem that you would actually care about.
This is breaking codes, simulating molecules, being able to optimize subsystems, and being able to do it better and faster than in a classical computer.
When people talk about quantum supremacy, or more recently we coined quantum computational advantage, this is any task, any computational task, that you can certify to be able to perform faster in a quantum computer than a classical one.
And this was first done by Google, I believe in 2019, with sampling from random circuits.
So essentially picking some probability distribution that is incredibly hard for a classical computer to simulate, and pulling out random samples from that distribution.
This is a really specific but esoteric mathematical problem.
And even there, they originally claimed millions of years of, you know, simulation time required in order to be able to run it.
IBM came back and said, hey, if we had all our computational resources, we can do it in two days.
And then several years later, he from Google AI actually was able to demonstrate that they could scoop the whole experiment using AI methods and bypass kind of the whole thing altogether.
And since then, there's been a couple more announcements.
There's been two from Chinese groups from USTC, and there's been one from Xanadu as well.
So I'll just use that one to highlight how ridiculous the advantage is.
So we performed that same, slightly different experiment, but sampling from a random probability distribution.
And a task that took us about two minutes to run on our photonic quantum computer with 216 squeeze states, would take about 7.6 million years to execute on at the time, which was the largest supercomputer.
Sorry, 7 million years, 7.6 million years.
But the thing to highlight is, this is not a useful problem.
So with 260 qubits, we can do something that is really cool.
You know, yes, it's a nature publication, all the scientists are thrilled.
But it's still a really long way to now go to do that for a real world problem.
There are algorithms that we have.
VQE is one of these that we can use on today's quantum computers.
VQE allows you to measure the ground state energy of a certain electronic structure, so a molecule.
Essentially, if you're a quantum chemist trying to design new things, this is what you're really concerned about.
You want to be able to do this, we have methods classically that do this, but they're very computationally expensive.
Any paper that you pretty much read that says, you know, your favorite automobile company and your favorite quantum company have simulated batteries on a quantum computer, they run VQE.
The thing is, we can basically simulate hydrogen.
Anything bigger than that, we just don't have enough qubits in a quantum computer, or the quantum computer is not performant enough in order to be able to go deep enough to actually be able to simulate.
So these are things that you can actually effectively simulate using classical methods on your laptop today.
The other issue is an algorithm like VQE, or other ones you might have heard of, like QAOA, which is a algorithm for optimizing systems, need to be trained.
And the training is so long and prohibitive that most of these papers actually train on simulators, and then just do the prediction part on a quantum computer.
Because otherwise you could literally spend months waiting for an IBM device or an IRQ device to be available to go through enough iterations to actually do this.
Not all papers, but I would say the majority of papers you see out there do that.
And the real problem is that these NIST devices really have too much noise in them.
It's just way too loud.
And then the other issue is that fundamentally, classical algorithms are phenomenally good at solving many of these real-world problems.
So the other thing that we see a lot in industry is when we hear that a quantum computer has outperformed a classical computer, they cherry-pick the worst classical method that you can think of.
And this is realistically throughout optimization methods.
What happens the most often?
You know, a company has been stuck with an optimization approach from the 1970s.
A quantum company does something potentially quantum-inspired or quantum-adjacent, and they prove that, hey, there's this two-fold improvement on the solution.
But realistically, if they took the latest and greatest optimization solvers, they would get a much better answer.
So, it's not all bad news.
There is a way we can get around the noise in quantum computers, and that's using error correction and fault tolerance.
And the basic idea is the same as in communication channels.
If you know that you're only going to have some probability of transmitting your information through communication channels, you put in some additional tries in order to do that, and then you have some check at the end that uses all the information together,
whether it's a parity check, whether it's a vote, to find out what the actual ground truth is.
And we're lucky enough that we can actually do this in a quantum computer.
There's one downside.
It's that meme that we saw earlier, that the moment that you measure a quantum system, you kind of collapse it, so you can't have these measurements in the middle of your calculations.
Meaning that the overhead, the way that you actually check whether computation has been successful or not, requires many, many qubits.
To give you an idea of the best estimates right now, we'll need somewhere between 1,000 to 100,000 physical qubits in order to get one good qubit that we call a logical qubit.
So now when we're talking about an algorithm that needs 50 qubits, we're really talking about one that requires half a million qubits in order to work.
So today's quantum computers are sitting around a couple hundred qubit mark.
We need to get to millions of qubits, plus all the other development that has to happen around error correction in order to unlock that beautiful space of all these applications.
I've also been disappointed.
I have like two cliche slides that I use in most talks.
One of them shows like quantum computing can cure cancer, you know, solve climate change, build long-lasting batteries, and the other one is that Moore's Law is anything.
But all those things are, again, predicated on the fact that we actually build volatile quantum computers.
But then there's another question.
If we do build a volatile quantum computer, and let's be generous, let's say it's one with a thousand logical qubits.
So you should be thinking now, you know, this is a data center the size of Las Vegas, consuming huge amounts of power.
What can you actually do with these thousand logical qubits?
And I'm going to make the statement that today we know three things to do.
We know essentially the code breaking.
We know a search algorithm that potentially will never work in reality.
And we know some things that are related to quantum chemistry.
And everything else is a big question mark of whether or not the latch is going to work.
And then the bigger question becomes, even if we do figure these things out, would you pay for it?
Code breaking, we probably have maybe some folks here from three-letter agencies that would be willing to pay for it.
More likely they would kidnap some of us in order to make sure that it happened.
But, you know, the market isn't actually that huge outside of a few state actors for being able to break RSA.
And in fact, once you do, everybody shifts over that value and that drops by quite a bit.
The real problem is, of course, that fault-tolerant quantum computers don't exist.
And the other issue is that classical algorithms are still too good.
For the majority of problems, we actually know really phenomenal solutions for classical devices that make it really hard to come up with real quantum devices.
So this is a very difficult question of, what would we actually do for quantum computing?
And the real, now, issue comes down to, we need to work in a very collaborative, multidisciplinary way in order to develop what these algorithms are.
And I was not a believer in this statement for a very long time.
For a very long time, I believed, like, you know, as long as quantum information people do what they do, eventually we'll get to a good enough space that people will just use this technology like they did in digital computing.
The difference is that we do have a very good computing framework right now.
And it is really hard to identify, in the near term, where quantum computing's passion is.
So, you know, we work a lot in the quantum chemistry space.
This means we're working incredibly close with quantum chemists with decades of experience.
That, you know, when we point out, hey, we can accelerate this part of your workload, they tell us, listen, we don't care, that represents less than a fraction of a percent of the time you spend.
How about you fix this massive problem?
Of course, we don't know how off the bat, but that feedback is incredible.
So, the issues with building new applications of quantum computing are underestimating what classical solutions are out there.
And really, if one thing you take away, anytime you read a press release on any quantum computing related thing, just know that there's definitely a better classical solution.
Until somebody's really just exploding and becoming a huge company based on your quantum devices, realistically, you know that there is no inherent value to this.
The other one is, of course, overestimating the quantum technology and capacity of computing that exists today.
And on the algorithmic side as well.
This is not easy, it's incredibly difficult.
And then the biggest one is a misunderstanding of where we actually need to push computational technology forward.
You know, it's easy to say large language models need tons of computational power.
That is so far for quantum computing right now to be able to make an impact there.
But there's definitely going to be smaller wins that we'll be able to get to within this decade.
That will actually have a major impact, as long as we can find where those corners are.
So, most people, what they end up doing is actually still running the same variational algorithms on this hardware.
Because, unfortunately, this is what executives love to see.
Executives love to have that press release saying, you know, a major corporation is working with somebody doing quantum tech.
They've run on a quantum computer.
My argument is we should be focused on algorithms that will have proven exponential speedups.
These really forward-looking projects.
And that's an incredibly difficult thing to do.
Many of those publications and press releases that you would have seen probably would have covered so many different topics for quantum computing.
You'd hear of pipelines in automobile factories being optimized.
Air traffic routes.
You'd hear people talking about building new batteries, photovoltaics.
You'd talk about people trying to predict weather.
I'm not going to name the company, but you can look it up.
There is even one talking about simulating human consciousness on a quantum computer.
These are all, again, I would say, potential future applications.
But none of this is reality today.
And if anything, these are incredibly small, tall models that you can do on your laptop, even on a quantum simulator, much more efficiently than on a quantum computer.
Out of all of these, realistically, there is a few that we would focus on.
One would be, of course, breaking public encryption.
Mainly, this is the one that we know incredibly well.
It's been very well studied.
We know how many qubits we need, of what quality.
And in fact, a lot of work has been done to push that algorithm closer and closer to reality.
When it was first released, it would have never actually run on a real quantum computer.
And over the last two decades, it's actually three decades now, people have really focused in on, how do you make this run on real hardware?
Not of today, but real hardware that we might have in a decade, let's say.
The other ones are, and it's kind of listed in three points, it's really one, but batteries, catalysts, photovoltaics.
But really, hard quantum chemistry problems that require you to have an understanding of the quantum nature of the actual molecules that you're dealing with.
So, some people include drug discovery, but drug discovery is still a very, I would say, physical macroscopic event, like how drugs bind to proteins, for example, don't really require you to understand on a quantum mechanical level, it's much more on a physical structural level.
But in batteries, when you're talking about small molecules, like electrolytes, cathodes, atoms, and catalysts, these are small transition metal catalysts, or photovoltaics, that quantum nature is very difficult for us to simulate.
But, if you break those problems down using the best approaches in quantum today, using quantum algorithms, you can actually start seeing, maybe we can run these with a thousand or two thousand logical units.
So, you can kind of plot this on potential for industrial impact, things that will actually have a huge, you know, impact on society, versus what is actually amenable to a quantum computer today.
But, I will say, this, of course, can change at any point in time with new innovation.
This is given where we're at today.
And that simulation of quantum systems using a quantum computer is probably going to be the first killer application, but we still don't have all the details working.
So, at Xanadu, the team that works on this, they are specifically focused on batteries, and it's because it's a nice position between everyone cares about batteries right now, it has a great societal impact, there's a lot of work to be done there, and,
of course, it's something that we can actually see running on a quantum computer in the next decade.
But, there really is, I would say, an important thing to note, and it's that the devil is in the details.
A lot of papers that you see, even in this space, in quantum chemistry, are solving problems that nobody actually has to solve.
So, this is from a great paper that actually identifies, it's clearly saying, although quantum chemistry is a strong candidate for a killer application of quantum computing, the lack of details in how quantum computers can be used for specific applications makes it difficult to assess whether they will be able to deliver on their promises.
And I'd say 95% of the papers out there, definitely 100% of the press releases, definitely fall into the category of, they did it because they could do it on quantum computers, but it actually had no impact on the underlying problem.
The other thing I really want to identify is, quantum computers are not fast.
They are incredibly slow.
Mark's favorite quantum computer can probably do about a couple thousand operations per second.
And I really just want to highlight this plot, which gives you kind of the gate count that you need right here.
The Pauli gates, these are specifically some of the difficult gates, but I really want you to take a look at the number of gates that you need here.
If you're implementing a thousand per second, it's going to take you an awful long time in order to get through an algorithm today.
So not only do we need to make quantum computers a reality, we also need to make quantum computers fast as well.
And this actually becomes a really big issue for some core technologies.
You might know the superconducting approach to quantum computing, but there's also trapped ion, there's photonics, neutral atom, there's people looking at silicon spin qubits, and people often focus on how good is their qubit, and how can it scale to building out a million qubit or greater device,
but I'll add one other thing people should be focused on.
Is there a fundamental limit to how fast your quantum computer can go?
Because if there is, you might be dead in the water anyways, while we talk about generating real value, because you do need these quantum computers to be operating at, let's say eventually megahertz, in order to do the type of things that we want to do,
code breaking and simulation of quantum systems.
And then on the other side, of course, is the qubit count.
So how many qubits do you need?
And I really want to identify that this plot shows you logical things.
So you can multiply this by 10,000 to understand where we need for those qubit numbers that are actually out there.
And you know, you have your favorite RSA algorithms right here, but you're taking a look that you need, you know, huge numbers of qubits in order to do that, and also huge numbers of gates.
So for example, just to break a 4k key, we need about, in this case, would be 10,000 qubits, and we need what would be 10 to the 11 gates.
Today, when Mark jumps up to his IoT device, he can maybe run 16, maybe 30 gates, before everything kind of turns to mush, and the noise breaks into the system.
So that error correction fault-tolerant algorithm is really required, but more importantly, it needs to be performant enough to actually get you to algorithms that can run for that amount of time.
So there's one exception, maybe, and that's quantum machine learning.
So quantum machine learning is a pretty weird thing that actually was a side note in a paper from a Japanese group that identified that when you look at a quantum circuit, that we normally kind of represent like this, with each little horizontal line being a qubit,
and those boxes being gates that either operate on a qubit or allow you to operate on multiple qubits, you can actually forget all that, and you can treat it as a black box that can have some trainable parameters.
And the really, really cool thing, that even on hardware, you can calculate the gradient of a given circuit analytically.
Again, not through some sort of difference approach, but actually an analytical approach to getting that gradient.
And this means that your favorite gradient descent methods that you've learned from your work in ML will actually work on a quantum computer as well.
The basic idea is that you can actually train a quantum computer knowing absolutely nothing even about the circuiting side, the same way you would a neural network for any other ML, or differential, specifically.
And it doesn't work just for one circuit with one free parameter.
You can actually build incredibly complex circuits with many free parameters, and you will see.
If you're interested, this will be my only pitch for anything, Xanadu, and it's an open source project called Penny Lane that was built out on taking advantage of this idea, and how it has become a general purpose software development kit for quantum.
It allows you to do this very easily.
If you can use TensorFlow, you can use Penny Lane or many other quantum computing or quantum machine learning languages in order to start playing on real hardware today.
I'll just highlight one thing you could do, and it's completely pointless, but I encourage you to do it because it's super fun.
You can actually do quantum transfer learning, and you can do this in about 20 lines of code.
What you would do is you take your favorite neural network, this is ImageNet, which is an image classifier.
You feed it your favorite image, and it comes out with tags of what's in there.
It was originally built out by Google.
You can actually blob off the last few layers of it.
What you put in there instead is a combination of a classical embedding network along with a quantum circuit.
In this case, we put in a 4-qubit quantum circuit that runs an IADF machine.
Then you can retrain this entire thing, training the embedding, the output embedding, which will give you two categories.
In this case, we trained on bees and ants.
You can get performance that is equal to that if you didn't include this at all.
At least the quantum part is not making it any worse.
It's not going to make it better, but it's not making it worse.
There's in fact no reason why you would believe that in this specific case, including the quantum piece, would make it better.
But there is research now suggesting two directions where quantum machine learning can provide some advantage.
It won't be computational advantage in terms of schema, but it might do one of two things.
One, if you're dealing with quantum data, quantum machine learning might actually have a big impact.
I think over the course of the two days, you might be exposed to some quantum sensors and the idea there is if you're actually getting quantum information from a quantum sensor, maybe you could use some QML methods in order to improve how you process that data.
The other one is there's been some recent research from Lawrence Berkeley National Lab and Los Alamos National Lab that suggests that potentially the amount of data you need in order to train quantum machine learning models is lower than what you need for classical models.
The amount of pain you have to suffer to get that advantage is huge.
Again, you're not going to train on real hardware, it's going to take way too long, so you're going to train on simulators.
Right now, the barren plateau issue, which the machine learning world faced several decades ago is a real big issue right now in quantum machine learning where you find yourself in areas where there is just no brain in where you can sit there.
And many other issues that people are facing and working towards.
It's a really exciting field and the best part about it is it's super straightforward to get into there.
The other thing I'll just mention about TinyLane is that you can embed this with broader machine learning frameworks.
So in fact, you can work with TensorFlow and treat the quantum part just as a node in the broader graph.
So, like I said, if you're used to working with Jax or TensorFlow, PyTorch, whatever your favorite way of doing gradient descent is, you can do it right now.
So, overall, the point of the story is we need better quantum algorithms.
But it's not all bad.
There's a lot of fantastic people working on this.
But the thing that they're actually lacking is tooling.
So, the pitch I will make to everybody here is get involved.
The expertise that you have in everything, whether it be from networking, whether it be cyber security, compilation, classical programming, it's desperately needed in this field.
You'll hear some horror stories about what quantum hardware looks like from a cyber security perspective.
That's online today.
Tons of room for improvement.
But I would also encourage you to get involved in that software development pit in the compilation stack.
This is something that's really limiting researchers being able to develop better algorithms.
With that, I'll leave it there, but happy to stick around for the next little bit and answer any questions you may have about anything quantum or quantum machine related.
Yeah, so, I mean, Mark will answer this from his perspective.
It's what he does for a living.
But maybe what I'll say is, a lot of this is focused around the fact that quantum is believed to be good for a number of optimization problems.
And whether that's true or not is a point of debate between many people.
But that's when things like portfolio optimization come into mind.
Arbitrage opportunities, so you can imagine having a big network and depending on how you traverse that network, one path might be shorter than the other.
And this can be important for if you're looking at an equity market and you realize there are interest rate markets or exchange markets, you could potentially get some arbitrage opportunities.
I would say, a lot of people are in finance because finance pays well.
And they're trying to find out anywhere where it might apply.
But my general position is it's not been a fruitful experience.
There's one maybe exception.
Quantum Monte Carlo is a thing .
And Monte Carlo is used pretty heavily as a method to look at evaluating risk and pricing in a lot of financial products.
Probably Monte Carlo underpins far too much of the financial system.
And Quantum has the potential to provide a quadratic scheme.
But you pay in a lot of ways, and probably the worst way is that you can't peek inside the underlying paths.
You just get the answer out in the end.
And it's really hard to actually take a look at all the underlying information.
Which today, people really leverage in order to understand the models, in order to evaluate risk.
So, a lot of effort is happening.
I would say, we're like a decade probably out before we see quantum finance really making an impact.
I'm leaving out Quantum Inspired, which is a whole thing, where there have been a lot of cool optimization methods that are applicable to finance, that have come out from how we solve quantum physics problems, or how we simulate quantum mechanics.
But writing on an actual quantum computer, again, is going to do well.
I mean, I just say, actually, that's a really good response, which I'm not going to add anything to.
Only because there's actually some of the largest amounts of bullshit I've seen in quantum, that have been from claims made around finance.
Forgetting that, when you analyze financial data retrospectively, it always looks great.
You know where the fit is.
But does that fit working for the future, is the real question.
And that's where there's, I think, a lot of effort in looking into it.
Quantum finance is a thing.
Google it, that's the phrase, quantum finance.
You will find all sorts of...
about how money markets move, and how quantum physics work with it.
There's really interesting ideas.
Applying different algorithms to different areas, like graph-frenshing, like graph-absorption, or max-learning-cut, problems are really applicable.
But the big one that we're looking at more and more is quantum-inspired intermediate.
And there's a talk at 3pm today on quantum.
So, any other questions for Matt?
Yes?
You mentioned that you need a point, you said for one qubit per day, it's going to be like 100,000?
More like 1,000 to 10,000.
Is that because, like, the operation going through a gate collapses the wave function?
Is that a problem?
So, I would say, the exact underlying physical mechanism depends on how you build your quantum computer.
So, in a photonics approach, the number one source of noise is loss.
So, your qubits are photons, or a-photons, depending on the method you take, that are traveling through a fiber-optic cable, or a chip, or a waveguide.
And every so often, they have the potential to scatter away or be absorbed.
And if they don't all make it to the end, that introduces some sort of noise into your system.
And superconducting qubits, it's coupling to the environment, they're coupling to other qubits that they shouldn't be coupling to.
So, think of it as thermal noise.
It's why they're sitting in supercooled environments.
And there's many different sources of noise for each modality, but usually one dominates the other ones.
But what it fundamentally means is that, unlike digital computers, which are pretty much perfect, you know, you do have a huge source of air.
There's work being done on what I would call intrinsically air-corrected qubits.
So, we can imagine ways of building qubits physically that already have some air correction or fog tolerance built into it.
It's a very hard thing to do.
There's been several paperwork tractions from early scientists' qubits that have already happened.
There's an in-between that we leverage at Xanadu, which is you can use kind of, instead of encoding your qubit on a two -level, a 0 and a 1 type of quantum system , you can use a quantum system that has many levels, but you need to do a much fancier encoding in order to get it in there,
but then it gives you some air resistance.
It's still not air correction, it's not fog tolerance, but at least it gives you a little bit more resistance.
And then the last thing is air-correcting codes.
So the codes that check when an operation has gone bad and then correct it require way more research.
So there's a really big opportunity that there'll be some transformative impact coming out of that academic community that would reduce that number down much lower.
But with today's codes, the way we understand how to use today, realistically we're talking about 1,000 to 10,000 depending on how good your qubit is and what codes you're using.
So quantum communication is a different thing, but yeah, I mean, I'm sure it'll be discussed on the fact that...
The next talk, in fact.
In fact, yeah.
The answer is yes.
I'll give my negative pitch for QKD.
We've had it for 20 years?
39 years since BB84.
39?
Yeah.
Okay.
It's 2023.
It has not gotten mass adoption, and also if you ever want to see terrifying side-channel attacks, look at existing QKD hardware.
Because it is just like obviously getting a bunch of businesses to build a cyber security device, and then being like, oh wait, we have to protect this from a classical side?
Oh no, so...
On the side, the original side-channel attack on the original paper for BB84 if you can find it, the original side-channel attack was they used motors to pull the polarisers in and out, and when they landed, they made different noises.
So, QKD's had a plethora of problems.
There's also...
There's one that they did in Bristol where they had four separate lasers to do the four separate states that we've been doing before.
And the four separate lasers had different powers, so you could...
switch lasers on and off.
Just by looking at electricity.
So yeah, so we've already had the whole power analysis for fun.
And now we can probably also look at...
By the way, on this topic, if anyone finds a vulnerability in a QKD system, do get a CVE for it.
Because there are at least a dozen papers on archive that tell you how to hack QKD.
None of them are documented in Cybersecurity.
None of them.
And we use these things, you might have heard of the CVEs, but you know the CVEs.
There's no CVEs for quantum computation tables.
Even the stuff that crazy Russian hackers have been doing, or teens from universities and all that, there's no CVEs.
So if you're CSO looking at QKD, there's literally no way of telling that it's absolutely screwed.
I will add, if you're CSO is looking at QKD, it's time to find a new job.
That's it.
I'm going to put that out there.
Any other questions?
So mine kind of actually leads into what you were just saying, which is, kind of in this call to action or escape, we need to even better talk about genvolves and what channel force it is.
Yeah, that's great.
So there is actually pretty fantastic resource communities right now.
Again, the one I know the best, because the project was started with Xanadu's paneling, that there's active discord, Slack, user conference, developer conferences that happens.
Qiskit has a really active community.
Potentially, based on some signs we see maybe that's slowing down a little bit, but there is a number of very open conferences right now on the quantum software side.
I don't have my QHack shirt, but QHack would be one of them that I would suggest getting involved in.
I would say, just trying to think of where other places would be good.
The National Labs, also specifically Oak Ridge National Lab, is a leadership coaching facility.
They put on an annual conference as well that would be quite good to get involved with.
But I would also say feel free to reach out to myself, anyone here with kind of in the quantum industry, we can definitely point you in the right direction, depending on what your interest is.
And I would say, probably, I don't want to knock on wood, but it's also an industry that is very short on talent and growing at a scary scale right now.
Don't know whether that's the same or not, but your favourite large multinational now has a few hundred people focus on quantum software and quantum networking.
So it's also a pretty solid career to get involved in this space right now.
Any other questions?
Then let us thank Raf.
Thank you!