#346 Get Quantum Ready with Yonatan Cohen, CTO at Quantum Machines

Generative AI is transforming industries at an unprecedented pace.

But as AI changes how you or your team work, one thing is clear.

Your skills also need to evolve. At datacamp, we offer everything you or your team need to adapt and thrive with AI. Whether it's business users looking to get the Most out of ChatGPT and Copilot, or developers and data scientists looking to fine tune models. You can learn the entire AI skills spectrum on Datacamp.

Power your AI transformation today.

Start learning@datacamp.com.

We may even see applications in a couple of years, right? Two, three, four years. We're hoping to see serious use cases, but these will still be very niche applications. In order to see something bigger, we need to build much larger scale quantum computers and also quantum computers where we deal with the biggest problem of quantum computers today, which is the noise or the error rate. So today we have working quantum computers that are somewhere between, you know, 100 to 1,000 physical qubits. We have billions of bits in a classical computer, right? But still, there are already things that the quantum computer can do that are very, very hard for a classical computer to do. Those things are not useful yet. But there are well defined mathematical problems that we can do with those hundred qubit devices that the classical computer cannot do. Even a supercomputer will take it, you know, many, many years.

Welcome to Data Framed. This is Richie. For decades we've known that quantum computing will be a revolutionary tech. That revolution is now close enough that you need to start preparing for it. A few years ago, self driving car experts were commanding crazy salaries. Today, experts in generative AI are earning megabucks. There's a good chance that quantum computing experts will be the next wave for tech rock stars. This is advanced notice for you to start learning today. We're going to get a primer on quantum computing, what the use cases are, who needs to care, what skills to learn and how to get started. Our guest is Jonathan Cohen. He's the CTO at the quantum computing control system company, Quantum Machines. Jonathan spent the last 15 years working on quantum computing problems. He went straight from a PhD to founding Quantum Machines. Let's find out how to get quantum ready.

Ty.

Nonsense. Welcome to the show.

Hey Richie. Good to be here.

Yeah, great to have you here. So just to begin with, I'm curious, what do you think the first big killer app for quantum computing is going to be?

Okay, that's the billion dollar question. So I guess it depends on how you define a killer app. You know, I think the first applications that we're Going to see for quantum computers are around, you know, quantum simulations, scientific computing being able to simulate all kinds of processes that we care about. But you know, then at some point we're looking for a killer app that's going to really make a huge impact on the world. I'm hoping. We cannot prove it yet that quantum computers could make a dent on how we do artificial intelligence. But this is something that is still kind of a question mark whether quantum computer would be good at or not. I'm hoping that this would be the killer app for quantum computers. Then of course there is also, you know, the kind of killer app in the sense that it's, you know, there is a, an application for quantum computers where we know that it could break pretty much all the codes, all the security for the Internet. So that's a, that's a real killer app. And so these are kind of the three main areas that we're thinking about.

Yeah, I think certainly using quantum computers to break encryption or to have quantum encryption, like that's the plot of many spy novels. So yeah, I think that's perhaps the most well known use case for quantum computing. But yeah, certainly the idea of having like scientific simulations doing science better and having quantum computing for AI, I mean these are huge use cases and I think quite beneficial to the world as well.

And I think that we pretty much know that scientific computing is within reach, hopefully in the, in the near term, like next few years. And that's very exciting. You know, if we can discover how certain materials work, how molecules interact with each other, et cetera. There's a lot that we can reveal first, like to understand how these things work and then you know, do material design better, save energy to do drug design. Those are kind of the low hanging fruits that we're hoping to get to. Yeah. And then AI, as I said, is still, we don't know yet, but we're hoping that this is going to be the case.

Yeah, no, it's very cool. We had an episode last year about doing like simulations for battery tech and the computational space is just absolutely crazy. And there were like, were partnering with Nvidia, just like running huge racks of GPUs. But it sounds like quantum computing is going to make that sort of thing a lot easier if you can do the smarter quantum computations. So you mentioned some of the use cases maybe a few years away. It seems like quantum computing has been this sort of coming soon technology for a while. So talk me through like what are the sort of expected timelines for doing useful things with quantum computing.

So I, as I said, I think that we think that for niche applications in the scientific computing area, then we may even see applications in a couple of years, right? Two, three, four years. We're hoping to see serious use cases, but these will still be very niche applications. And then in order to see something bigger, we need to build much larger scale quantum computers. And also quantum computers where we deal with the, with the biggest problem of quantum computers today, which is the noise or the error rate. And for that we need to implement quantum error correction. And that I think would take us a little bit longer. So maybe, you know, until we have large scale quantum computers with enough error correction so that we can solve large scale problems, I think it's more on the, you know, seven to 10 years or horizon.

Okay, seven to 10 years. I mean, that still seems like fairly soon for such a big technology. So you mentioned the idea of error correction though, and trying to scale things. You just want to talk me through, like, what's the problem here? Why do quantum computers make errors? And why do you need to correct them?

Yeah, so basically, you know, we have errors even in classical computers, right? Every now and then you get a blue screen of death, right? And your computer turns blue and you don't know why. And actually, you know, for example, sometimes it happens because a cosmic ray hits your, your memory, the memory of your device, and literally flips a single bit of information, right? And that drives the computer crazy. And it's just that with classical computers this thing happens very, very rarely. And you can kind of think of an analogy of, you know, like a classical bit is like coin sitting on, on, on a table, right? So in order to flip the coin, you need to invest quite a bit of energy, right? There's sort of like an energy barrier to flip the coin, right? So it doesn't happen very often. Like you can't think about many processes that will just make the, make the coin flip, right? But a quantum bit is more like a coin floating in space, right? So it can sort of, you know, almost everything, it's analog. Every, almost everything will, we'll move it around a little bit and this will cause errors. But the nice thing is that we have a way to deal with it. We have to deal with a way to find these errors and correct them. That's called quantum error correction. And for that, what we need is we need to take several quantum bits, several qubits, several of these coins that are floating in space, and then we turn them into a single, what we call logical qubit and that qubit is actually going to be protected from noise. So it's going to actually be much more like that coin that sits on the table. And we sort of digitize, using error correction, we sort of digitize the quantum information, which is an amazing, amazing idea, a concept that we have, quantum error correction that can really help us protect quantum information in a way that we can scale up quantum computers and build quantum computers even though the qubits are very kind of fragile.

Okay. That seems like a very statistician approach. It's like the individual bits of data are very noisy. You know, once you take the average, it's a bit more stable. So I guess that's the idea with the logical qubits is they're. Yeah, they're less, a bit more robust in terms of the statistics you can pull out of them.

Yeah. And the problem with the problem is also. But the hard thing about it is it's, you need. Because how fragile the qubits are, you need to use quite, quite, quite a few, actually many qubits to make a logical qubit. And then the overhead that you need to pay in terms of both the quantum resources and also classical resources, because it's not enough to make, you know, one logical qubit from a hundred physical qubits. You also need then to do this quantum error correction process, which also involves getting some data from your quantum processor and processing it quite heavily on a classical processor. So there's this hybrid quantum classical thing going on here and the overhead both in terms of the quantum resources, just basically more qubits, and then the classical resources overhead is quite large. So it's going to take us a while to build those systems, but we're on the way there.

Okay, that's very cool. So I guess what's the current state of VR and how does it compare to classical computers? Like, where are we up to with this scaling process?

So today we have working quantum computers that are somewhere between 100 to 1,000 physical qubits. The leading quantum computers out there are somewhere between 100 and 1,000 physical qubits. And, and there are already things that those computers, you think about hundred, hundred qubits. Right. Like we have, you know, billions of bits in a classical computer. Right. But still, there are already things that the quantum computer can do that are very, very hard for a classical computer to do. Okay. And even impractical for, for classical computers, those things are not useful yet. But there are well defined mathematical problems that we can do with those hundred qubit devices that a Classical computer cannot do. Even a supercomputer will take it, you know, many, many years to do. Right. And we do quite quickly on these small scale quantum computers. That's very cool by itself. Right. Like you do something well defined, solve a mathematical problem with this 100 qubit device that you cannot do with the biggest supercomputer on Earth. It's. And that's where we are. Plus there is another very important milestone which happened several different groups in the last couple of years. And that error correction was demonstrated to be not just working conceptually, but also showed the right scaling so that the more we use, the more physical qubits we use, the lower and lower the error rates that we see on these logical qubits. And that's very important. So we see that the concept is working, the scaling is working. Now we need to scale up the system because, you know, if you have about 100 physical qubits for logical qubits, and practically right now you have one logical qubit or something like that. Right. So you can't do much with that. So once we do 100 or 1000 logical qubits with very low error rates, that's where we think we have like the first applications. Right.

Okay, so we're still, we're a few orders of magnitude away from having something that's really useful at the moment. But uh, but it tells, it's not like a real dramatic, like billions and billions and billions of times bigger before we get something useful.

Yeah. And the progress is fast today. The progress of a few groups is incredible. I, I can tell you I'm much more optimistic than I was three or four years ago about the timeline. We do still have very tough challenges, like engineering challenges to solve. But I think that this machine is, is a, is, can really be built and to work. And to be honest, I think that the bottleneck is starting to be more like, okay, we, it will take some time, but we will build this machine. And then the question, the bigger question is, you know, we need, we need more applications, we need more use cases, we need more algorithms. It start, it starts to show that we need, we don't need to shift our focus, of course, because we need to do the engineering and it's very tough and it's going to take some time, but we need to also have a lot of focus on, okay, what do we do with this computer once we have it and finding more algorithms, more applications and also connecting them to real world use cases of real companies and so on.

Okay, this is fascinating. So I guess talk me through what the current progress is at the moment and what the sort of the research on the algorithms needs to be.

So I think that first of all I have to say I'm less of an expert on the algorithm side, but I think that today we have at the end of the day very few quantum sort of routines. It's like you think about a subroutine, like something that could be done fast, much faster on a classic, on a quantum computer and we can use those. But then the question is, you know, can we map the real world problems that we have onto algorithms where these subroutines can really give a dramatic speed up for the quantum computer? Right. And so mapping those real world use cases, whether it's, you know, in again material design, whether it's in optimizing financial portfolios, optimization problems in general. Can we map some real world optimization problems through quantum algorithms and subroutines and really get the benefits from the machine in those use cases and applications?

Okay, that's interesting. There are so many optimization problems just around, I mean like almost everything. Like any business problem, it's like, oh, we're trying to optimize for some like maximizing some metric, but certainly like. Yeah, you mentioned financial portfolios, guess logistics as well.

Yeah, exactly. But you see, the hard part is that even if we even forgetting about quantum computers for a second, solving some of these optimization problems, if you just take the best kind of general algorithm that can solve these problems, for example, these traveling salesmen, we, we, it's an NP complete problem, we don't know how to solve it non exponentially, it's an exponentially hard problem. Right, but then for real world problems, we never do that algorithm. We actually use heuristics that would solve these optimization problems and we don't understand, you know, really stop. We have intuition why these heuristics work, but we don't, we cannot prove many of these heuristics. So in the same manner we cannot prove some of these quantum heuristic algorithms, okay? So the only way to know whether they will give us a speed up or not is just to build a large scale quantum computer and to try those quantum machine learning algorithms, for example, for these problems and to see whether it gives us any advantage over the classical heuristics or not. Because again, we cannot just do a mathematical analytical proof and just prove, hey, this is going to work, work much better than this. So for many of these optimization problems and AI and so on, it's just something that we, we would have to build the machine and try.

It does sound like a bit of a chicken and egg problem. You need to have the, the machines in place so you can test the algorithms, but you also need the algorithms to make sure you've got use cases for the machine as well. So there's lots of work to be done.

Yeah. So, you know, I could give you another like, interesting kind of thing that is happening to the, you know, someone has a Quantum computer with 100 cubits, as I mentioned, and they run, they run something on it. They say, hey, this is quantum advantage. Like, you know, we, we ran this thing and sometimes it's even connected to a real world problem. And they say, all right, that's great. Like I, I, I showed that quantum computers can do way better than a supercomputer here. But then someone takes looks at this algorithm and they go, okay, wait, maybe I can do it way better with a classical computer. I'll just do the algorithm this way or that way. And then takes them a few weeks and then they optimize it, and then they optimize it for whatever hardware they run, like the supercomputer or they take an FPGA and program it and then you see. Oh, actually, so you can do it classically and it's just as fast or faster or whatever. Right. So it's this also like a game where you come up with a quantum algorithm, you try it on a quantum computer, then someone comes with a classical algorithm. And unfortunately, in many of these things, I don't think it's going to be just things that you can prove. You just have to try new algorithms. They work, they give us an advantage. We use a quantum computer. Maybe someone can find a classical algorithm that runs with a classical computer and it's going to be a race like that.

Okay. I mean, I guess the competition is good. So as the quantum tech gets better, it means that all the sort of classic computationalists have then got a challenge to keep up. What do you need to do in order to like get started with this stuff? So suppose you're like trying to prepare your business for. Okay, quantum computer is coming soon. What do you do first?

It's, it's a great question depending on the level of depth that you want to understand the technology. Okay, so, you know, at the lower layers or the lower levels of the stack, quantum computers are very different. They're programmed completely different. Algorithms are different. You know, the basic operations are different. So, you know, the programming languages, the algorithm, the way we express things is just different. And if you Want to get to that level, you just need to really learn the language of quantum computing. You need to learn a little bit of the language of quantum physics, to be honest. Like the world behaves in a different manner, so our intuition is slightly different. And so you have to do that. On the other hand, I think, I believe that a lot of the end users at the end of the day may be masked from this. Like we will have the experts would build the applications in a way that users can use them for their applications and program directly in the application language. Right. Especially for the first use cases. I think that that's going to be the case. So given that I think that many of the, for example, companies that think that quantum could be relevant to them, I think that maybe, you know, they need to be less focused on understanding all of these low level details and maybe more focused on what problems do I have that perhaps a quantum computer could solve faster? And that's again, as I said, not a trivial question. But if they find those use cases, then they need to really care about those use cases because that's, there might be just a huge difference in the economics of using quantum computers to solve these problems. And so if you have large scale computational problems, you really need to be focused on do I have problems that a quantum computer is going to be transformational or disruptive when it comes and work with the researchers, frankly, to maybe. And that goes back to what I said about the community needing to also put more effort on looking for these, these use cases. So I think the end users could also help in that respect.

Okay, so that's interesting that if you, I guess on a corporate level, if you're going to like really push for quantum computing, you need to think about what are the use cases. And that's going to be what, where am I spending vast amounts on computation and I guess probably on optimization problems in order to make it worthwhile. Does that sound about right?

Yeah, exactly. And I think it's a very healthy thing because, you know, we need the end users to come and say, here is, here is my hardest computational problems. Can a quantum computer make a dent here or not? So because that will give us a lot of focus on first, it will give the algorithm, the quantum algorithm guys and the quantum application guys focus on where do they need to look. But it will go all the way through the stack. Like even we do something much more lower level in the stack. We're enabling the people who build the quantum computing computers. But even that, you know, there's certain optimization that we could do. Once we understand where the applications, or at least the first applications are going to, what are they going to look like? Right.

Okay. So it does sound like there's a big need for trying to figure out what the big applications are going to be. So it sounds like you as, like a quantum computing builder, like, you're quite receptive. Like, well, what should the applications be? Give us your hard problems and we'll try and solve them.

Yes. Yes. Yeah. Okay. All right.

And so you mentioned that your company works slow down in the tech stack. I'm really not entirely sure what the tech stack consists of. Can you talk me through, like, what does a quantum computer even consist of?

Yeah. So quantum computer, first of all consists of quantum hardware. We call it the quantum processor. This is where we have the qubits. That's where the magic happens. You know, there's superpositions and entanglement, and that's where the magic computation happens. But there is another very important system which is called the control system. That's actually what we do at quantum machines, and that's a classical system. It's a classical. Classical. Classical hardware, which we call it drive. The quantum processor basically sends the signals that make the qubits, do the basic operations that qubits can do. So if, you know, if you want the gates of the quantum computer, the basic operation, mathematical operations, for example, flipping a qubit, so you want to flip a qubit from 0 to 1, or you want to put it in a combination of 0 and 1, you need to send a little pulse to the qubit. And that pulse, electromagnetic pulse, for example, in some implementations need to come from this classical control system. So that's the second part. And that basically constitutes the hardware of a quantum computer. And then you have several software layers. So you have the lowest level API, which is basically the interface of the quantum controller. But then you have a compiler and quantum computing language, programming language, where you basically program this gate operation. You say, I want to flip the qubit, I want to flip this qubit, I want to do this operation on these two qubits, et cetera. And that needs to get compiled to the instruction set of the quantum controller that then drives the quantum processor and at the end gets the results and returns them to the user.

Okay, wow. So lots of layers of tech there. I mean, I guess I should have expected, like, you start off with like some very low level with flipping qubits. And then it goes up to like, I want to be writing some code to control this thing or actually, so you mentioned there's a quantum, like a programming language. What's the programming language?

Yeah, so there are several programming languages. And again when we talk about programming languages, there's like several layers. But typically people talk about what we call gate level programming language. This is this level that when I mentioned like you talk in terms of what operations do I do to qubit? So I do this to this qubit and that to that qubit. I want to do this operation of this to qubits. It's very low level when you think about it in the classical computer. You never actually talk in this level. So there are many companies just wanted to clear that, you know, there are many companies actually above that line that takes it from this level to a much higher level where you know, you program like similar, more similar to how people program in classical computing. But still today when people talk in the, in the community about languages, they talk about these gate level languages. And there are several languages like Qiskit by IBM and CUDA Q from, from, from Nvidia and Circ from Google and, and, and, and many others. So there is still not a single programming language that is kind of controlling the field. And I'm not sure there will be, but there would probably be several other words to sort of take the lead.

Okay, all right. So yeah, sounds like lots of opportunities to learn different things. I guess those all sound like quite low level. Like is there like a Python interface? Can you just be like, okay, I'm running this with the language I know.

Yeah, yeah, I should say like most of them are embedded in some other programming languages. So for example, qiskit is essentially, it's embedded in Python. So you work in a Python environment and in, there's a library in Python, there's a package in Python you install and then you connect to a quantum computer and then using this Python package, you create these quantum programs that then you send to the quantum backend. Right. In similar manner, CUDA Q is in Cuda in the CUDA platform and so on. So in that respect it's very similar to typical software development environments. It's just that those basic operations that the language itself, the, the syntax of the language is, you know, these qubits and operations on qubits and you need to build your quantum program, your quantum algorithm from this new basic operations and mount on the device.

Okay, yeah, so you got a very like different math at the sort of low level. So I guess. Yeah, yeah.

Other than that it's the same, it's this low and that's why I'm saying, like, at the end of the day, I think people will wrap those, you know, especially the. Because, you know, at the end we don't have that many quantum routines that are going to give us split ups. They might be taking different shapes in different applications, but I think they will wrap those in, you know, certain things that are going to be much higher level and you will just use them as a user. You may not even know you're programming on a quantum computer or a classical computer. You just say, okay, like, I want to try to run this on a quantum computer and see if it runs faster or something like that in the future.

Okay, that all sounds like there's potential for it being quite convenient eventually. I'm curious if I want to run my first quantum computation, how do I go about doing that?

So today there's actually a few cloud platforms you can use. You can go to Amazon, Braket, for example. This is where Amazon is, has their quantum cloud service. And they actually give you access to different machines, different backends from several different quantum computing vendors from the industry. And you can log, basically you can connect to a quantum backend and then again, they give you a language that you can write the quantum algorithms. It's all in Python, as I mentioned. And you, you write it. You write your quantum circuit or quantum program and you run it and you get the results. And you can play around with that, you know, today, tonight.

Okay, that sounds very straightforward. I love that. And in terms of like, getting stuff like is, what's the sort of quantum equivalent of hello world? Like, what's a good simple first thing to try?

So the first thing to try is, yeah, just take even a single qubit and do some operations on it and measure it. Because, you know, we have this thing in quantum mechanics where, you know, this qubit, you know, it's like a bit in the sense that it has. It can be 0 or 1, but it's not 0 or 1. It's always in some kind of a combination. We call it a superposition. And so, you know, you can start with the qubit in zero, completely in zero. And then you do some operation to it and then it's, you know, half zero, half one. And then you can try to measure it and first see that, you know, half of the times when you do this, you get zero and half of the times you get a one. It's a way to truly create a random number generator, for example, right? You truly, you may do this operation on your qubit. It's now Half and half. And whenever you. And if you repeat this experiment a thousand times. About half of the times you'll see it in zero. About half of the time you see it once. So that's the very basic behavior of a quantum bit. Right. And then you play with two qubits. And see how you can make them talk to each other. And what does that mean on the 1st. On the fact that you measure one? Does that mean, do you know the state of the other one? Do you not know the state of the other one? Does it matter what operations you did before to your qubits? So that's kind of just to play a lot, I think. To be honest. I think that quantum physics is not so complicated. But it was never taught using these basic quantum bits and quantum operations. It was always taught in terms of, you know, the structure of the atom and, you know. And all these, like, complicated things where you need to know complicated math. Here we have very isolated at quantum bits. And if you just play around with some basic operations. You actually get a pretty good understanding of how these qubits behave pretty quickly. And that's pretty cool to just learn quantum mechanics this way. Right. And get some intuition of how these things work.

Absolutely. Yeah. I remember, like, doing little bits of studying this back in university sort of 25 years ago. I was like. Yeah, it was actually. It was like all sort of complex linear algebra. It was quite tricky math. So I like the idea that you can just play around, write a bit of code. You can just see the results of what's going on with the computations. And get a feel for it. You mentioned superposition is an important sort of thing to know about. Are there any other quantum physics principles that you think are important to know if you want to try some of these, like, just playing around with these computations?

Yeah. At the end, there are two quantum phenomena. Which are at the heart of quantum mechanics and then quantum computing. And they're at the heart of the difference between quantum and classical. And that's superposition and entanglement. So superposition is this phenomena that basically says that, hey, we have a system, it has states. It could be 0, 1. Or if we have four states, it could be 0, 1, 2, 3, 4. And in classical physics, you just need to choose one of them at every point in time. And you move between one and the other. Right. So a car, it moves with time on the road. Right. And. But at every point in time, it has a single position out of all of the possible positions. Same way bit will be either 0 or 1. It can move from 0 to 1 in time. In quantum mechanics, we say the system can be basically, at least mathematically. In all of the states at the same time, right? With different sort of weights. So you can be a little bit. The car can be a little bit here And a little bit here and a little bit there, right? And this is the superposition, right? And the other phenomena is entanglement. That's when we say, you know, when we start combining system. If we take, you know, two systems, two bits, for example. Each one can be 0, 1. Now, there are four states to the combined system, right? It can be 0, 0, 0, 1, 1, 0 and 1, 1. So we have four states. And now the entire system. The combined system can be again. In a superposition of all of its states. So a superposition of 0, 0, 0, 1, 1, 0, and 1, 1. And the number of states that such a system will have. Is exponentially with the number of systems, right? So if I have, you know, only a hundred bits, 100 qubits. I already have two to the hundred states. That I can be in all of them at the same time with different ways. So I have two to the hundred ways. Which means that nature and 100 bits is nothing. I can have a system with a million bits. So two to the million. That's way, way bigger than the number of atoms in the universe. So somehow nature sits there With a list of numbers which is larger. The length of the list is larger than the number of atoms in the universe. So God has all of these numbers that it needs to keep track of. And they're changing with time, right? And that happens because, you know, we have so many degrees of freedom. And it happens because this combined system has, you know, the combined states. And you can be in all of the different states. And that also means that you cannot anymore separate the system to its subcomponents. There are situations where you cannot tell me. That this part of the system. Like this qubit. Is in 0 or 1. Or even a combination of 0 or 1. Without telling me what the other one is, right? They're only together in this superposition of their own. Like the total kind of combination of states, right? And that's called entanglement. And we really need to be able to create this entanglement. By having these qubits interact with each other. In order to explore the entire space of this enormous number of states that we have. And that, together with superposition. Are the two phenomena that we use in quantum computing. So to complete this at the End of the day, the only advantage that we get from a quantum computer is when we can take advantage of what we call interference. And that's the fact that these superpositions, these waste that we talked about for these superpositions, they, throughout the computation, they can either combine because the system goes in kind of two different paths or exponential number of paths. And those paths have those weights that can either combine and give us a higher probability for the right solution, hopefully, and a lower probability for the. For the. For the wrong solution. So maybe I should say there are three phenomena. Superposition, entanglement, and then interference. Interference is the way that we eventually get the quantum spin up. Because we use the fact that those quantum amplitudes, these weights that I talked about interfere, kind of like waves in the ocean and the other amplify one another or kind of cancel one another to get some advantage from the fact that the computer goes in this crazy parallel universes and connect with faster.

Okay. I mean, you make it sound very straightforward. Just three concepts to understand. And then, yeah, you've got computational magic. So I love it. I mean, so the superposition idea, this is. This is like Schrodinger's cat analogy, right? Where you've got a cat in a bag and until you open the bag of the cat can simultaneously be alive or dead. And once you open the bag, it's like it resolves one state or another. I'm not sure if there's an equivalent analogy for the. For entanglement and for the interference. But, yeah, it sounds like once you get the hang of that, you got the. I guess you're moving into a probabilistic world from like rather than just trues and falses everywhere. And that seems like you've got most of the concepts. So in order to sort of take advantage of this. And you've got a few quantum physics skills, you've got a bit of programming skills, I guess, using these cloud services. Is there anything else you need to know to sort of take advantage of quantum computing at the moment?

No, I think today that's pretty much it. Again, today it's mostly about learning those things, understanding how they behave, and starting to identify the use cases where quantum is going to be important from this knowledge of how they operate. Right. So yeah.

Okay.

All right.

So it seems like from an individual point of view, it's fairly straightforward to get started. We started chatting briefly before about like, what. What organizations need to do or businesses need to do. And do you have a sense of like, are there any organizations that are kind of getting started with quantum computing, trying to take advantage of it. And what are they doing?

Yeah, I think many organizations are, are starting to gear up towards quantum computers. So of course there are the companies who are building quantum computers like Google and Amazon and IBM, and they are also exploring, of course, use cases, potential use cases, both for themselves and for others. Then there are the like, sort of like pure end users. So, you know, banks like HSBC and Goldman Sachs and companies like Volkswagen or Loched Martin have all done research on various use cases and how quantum can help solve some of their. They're, you know, heavy computational problems, and they work typically with the quantum computing vendors together to, again, as I said, and with the quantum algorithm guys. So there's a lot of that going on. And I think it's, it's really the pace of, of companies that are doing, you know, more and more in this space, especially on the sort of research level, that they really look into what, what can we do with this thing that is relevant to us? I think that that is growing, and I hope that it will grow further because we really need as much of that as possible. We need as much of that as possible also because we don't want to wait until we have millions and millions of qubits. So the more people work on it, besides finding new use cases, many times they find that for this use case that we thought that we need 20 million qubits, all of a sudden now, oh, we found an algorithm. They could do it with half a million qubits. And that's critically important to where we are in the progress with the hardware and software, the development of the machine. Then there is other types of organizations who are even putting more resources on them. That's, of course, governments. So the DOE labs, for example, in the US are putting a lot of emphasis and we're seeing an increasing interest in quantum computers, how they can solve for certain use cases and how these computers can also be integrated into their data centers and their software environments. How can they be programmed in a way that also works together with, for example, how you would program a supercomputer? Because at the end, the supercomputer and quantum computer will have to sort of work together to solve some of these difficult computational problems. So the quantum computer may just solve a problem, part of the problem, not the entire problem. And so there is a lot of interest from these supercomputing centers, national labs, et cetera. And I believe actually these will be the early adopters, the first adopters of quantum computers.

Okay. Yeah. So it seems like finance is. Well, I mean, finance is usually the cutting edge of technology. So finance use cases for quantum computing seem to be leading the way. And then scientific use cases where particularly the big supercomputer sort of organizations. Okay, all right. Before you mentioned that progress has been very rapid in the last three or four years. Can you talk me through, like, what the big breakthroughs have been and what might be coming soon?

Yeah, I think there are two very important elements that are making very fast progress these days. One of them is quant error correction, as I mentioned. I mean, I think that, you know, maybe up until two or three years ago there were demonstrations, but we didn't see the full scaling and how it works, like really well. And now I think this has changed. It's clear that quantum error correction is. I mean, we're at the point that the performance of the individual elements is sort of good enough so that we can really see how it scales, how we could build a machine if you just scale it. It's not that simple, of course, but you can see how you scale it and the error will go down. You will build more logical qubits. And so the work that we've seen from Google in this respect, I think was perhaps the most important and interesting, but also works from Quantinum and Creera. And that's another. The other big thing that's happening is sort of the. So, you know, in quantum computing, we still have. We don't have the silicon transistor yet, so we have different types of qubits that are still competing. You know, what would be the sort of silicon transistor? So we have Superman qubits, we have trapped ions, we have neutral atoms and spin qubits and neutral atoms. Based quantum computing, I think has been steadily moving at a very fast pace, especially towards scaling up these systems to at least tens of thousands of qubits in the next few years. And it's so you're sort of like this kid on the block that is making everyone also work harder because the other platforms need to match superconducting qubits, I think was and maybe is still leading in many respects. But neutral atoms are just kind of moving very fast. So everybody has to work harder. And I think that that's the second thing that's happening, which is really shaping how the field is going to look like.

Okay, so error correction and scaling with the neutral latam qubits. Okay, that's cool. And what are you most excited about, like, with the world of quantum computing? What's Exciting you?

There are two things. One is just on a personal level, I think that quantum comp, like I'm, you know, I'm a physicist. So the thing that drives us crazy as physicists is this, you know, this outstanding problem in physics of, you know, how you unify quantum mechanics and general relativity. And I personally think that quantum computers may help us solve or at least understand if we're going in the right direction in solving this. This issue, which is in my mind at least the biggest issue in physics today. It's like unifying our basic understanding of the universe. These two theories are very fundamental to how we understand reality on a philosophical level. And we have to find a way to unify them and a quantum computer can help us. So I think that's on a personal level and on the more like how do we make an impact on the world kind of level, I think that I'm excited to understand how quantum computers may be impactful for AI. Again, it's an outstanding problem. It's not something I think we can prove and there are lots of ways to dismiss it. There are also lots of ways to why, to believe that it might be helpful for AI. And I'm just excited to first of all see whether it's going to accelerate AI and second, understand why because again, I think it's a very deep kind of question.

Absolutely. The idea of combine the quantum physics and general relativity is a very, very long standing problem in physics. It's a very different problem to I'm trying to optimize my financial portfolio or optimize my logistics. But yeah, I love it. It's, it's a fundamental thing about understanding the world. So I like that idea.

Yeah. And you know, sometimes these, you know, optimizing like an optimization problem, like fine, maybe you know, we're using it in order to optimize our financial portfolio. But there's something deep about computational problems and whether something is computable and how fast you can compute something given certain rules on the physical system that computes. So there is something very fundamental about those things on a scientific level that I think is interesting. And perhaps we discover some of these more fundamental philosophical questions because we're, we found a way to optimize the financial portfolio, so there's some connection.

Excellent. And because you mentioned it's a very unclear thing whether quantum computing is going to have an effect on AI. Do we have any sense of what the impacts are going to be?

No, I think we don't. And I'm not the person who would say it will have a huge impact. I frankly don't know. There were very. Before the AI explosion, so basically up until 2012, there were neural networks, was a very small community of people who believe that this is going to be crazy impactful. But it was mainly a niche, right? It was pretty much a niche and was not so much mainstream to think that this is going to work so well. And then someone made it work for a specific problem, 2012, and it was, you know, and since then it exploded. And the reason again is just that it worked specifically on the GPU and, you know, and it just worked. Right. And before. So it's not like our understanding the day before and the day after changed. We had the same algorithms, we had the same kind of the basic understanding of what it, what those neural networks do and why we maybe have an intuition that they will work was the same. And it's pretty much, you know, we understand maybe a bit better today, but we cannot prove exactly why this works. Right. So the key element was to try those things on a real machine and see whether it works or not. And I hope that with quantum computers it will be similar, but there is no reason to necessarily believe it could or it wouldn't.

So, okay, so maybe new algorithms coming, maybe some, I guess, form of better intelligence that just magically works. Not magically, but works because we have access to a different computational approach. Okay. It's exciting times. We'll see what happens. So just to wrap up, I always want you people to follow. So whose work are you most excited about at the moment?

A couple of companies that one I mentioned, neutral atoms. So there are several players in neutral atoms. Quera is perhaps the leading kind of player now in neutron atoms. And we're seeing, you know, really great progress by cura and I'm very excited to see, and I'm following their progress and seeing how, how, how, how fast they're moving, which is really fascinating to see. And I think this, as I said, it's, it's shaping the field. So neutral atoms in general, but Carrera is also the leading player there, I believe today. And then, and then I think that in, in trapped ions, we have a new player here in Israel, actually quantum art. So always the problem with trapped ions is that they're actually working incredibly well. The error rate is very, very low. But scaling up is hard. And Quantinum and IonQ are doing great work, but in many cases it has to do with moving ions, like individual ions around. It's very difficult to do. And quantum ort, which is this new company here in Israel. They have a different way way of doing the scale up and I think that's a very interesting way. Maybe trapped ions all of a sudden becomes easily scalable and that's going to change the picture. And then of course Google and Amazon and IBM, they're, you know, these are the sort of biggest companies that have invested tons of resources on quantum and are making, you know, huge progress with superconducting qubits. And that's still, I believe, you know, the, perhaps the leading pair still today. And they're doing now a lot of work on error correction. And I want to see how fast we're making progress in building those logical qubits and scaling those systems up because those systems are also the fastest system. So if we can do scale up of these systems and do the error correction where Google sort of demonstrated already on a single logical qubit, but now scale the systems up, which is difficult engineering problem, then those systems have a huge, huge advantage because they're just very, very fast. So, so these are the different players.

Okay. I do love that there are lots of different competing technologies. The competition is kind of rife. So I think that's going to benefit everyone else with all these different ideas being thrown around. So, yeah, exciting times and lots of interesting things to watch out for. All right, brilliant. Thank you so much for your time, Jonathan.

Thank you so much. It's been great.