WE NOTE THAT THE FOLLOWING TRANSCRIPT WAS CREATED BY A ROBOT SO PLEASE FORGIVE ANY TYPOS.
[Scott] Welcome, everyone. Thank you for joining us. I’m Scott Dybwad. I’m a lawyer and patent agent with Borden Ladner Gervais. We’re a full service law firm with offices throughout Canada. And I’m based with the intellectual property group out of our Ottawa office. Pardon me. I also sit on the editorial board for Dead Cat Live Cat, a quantum computing focus collaboration of IP lawyers, with super Lawler in the US. Marks and clerk in the UK Schindlers in South Africa, and of course, BLG in Canada, and I’m very excited to have joining me today, Wojtek Burkot. He’s the co-founder and Chief Physics Officer of BEIT, a quantum computing startup based out of Poland. And on that note, welcome Wojtek. Great to have you here today and joining me here at this conversation.
[Wojtek] Thank you very much for having me. Just a small correction. We are formally incorporated in Delaware like every other startup, right, we do have an R&D office in Poland, Krakow, Poland, but also one in Toronto, Canada. And we have just hired our first employee, and luckily the second physicist, that was the hardest time for us to get enough physicists there. So indeed, I’m Wojtek Burkot. I’m co-founder of BEIT together with my business partner for over 20 years, Paulina Mazurek, our CEO. We met a long time ago at Motorola then continued through Google, Allegra, and we co-founded BEIT when we thought it’s a right time for the quantum computing hardware to actually deliver on the promise within my lifetime. I’m pretty old. So that’s, that’s important.
[Scott] Well, that’s I mean, you’ve had a really interesting journey, Wojtek. I mean, you start as a physicist, pardon me, you went to Motorola, then to Google, then to Allegro. And now you found that it was the right time to sort of jumped into the quantum space. So tell us a bit about that. What inspired that move? What made you think it was the right time?
[Wojtek] Well, I don’t think I made that to a real move. Since my physicist times, I was always interested actually, when at the university, I lectured on this. What are the physics paradigms for computation? What’s compatible with SNOP? Why? Why do we use computers to model mathematical abstractions. And that was my core interest forever. Throughout my career, regardless of who was paying for bread and butter, and especially the tipping point was my strange visit, Poland was still a communist country. And I got invited by people from MIT. The names famous in the quantum space, Professor Tommaso Toffoli gate and Professor Norman Margolus gate. And they were coming from the LCS at MIT. And we were working on something some other physics motivated paradigm for computation. Yet, they were already starting on the Professor Richard Feynman, with quantum computing, and missed Feynman, but the portfolio and Margolus made a lasting impression on me. And ever since that visit, whoever was paying for bread and butter, I was doing quantum computing right in my free time, on and off along the same lines, and why the switch to the startup, because we thought that the hardware is about to be ready. And also, because of that past of, like, 30 years focusing on this, I didn’t feel comfortable sharing everything I had with the with Google, by American job contract, everything I was working on in my private time belonged to Google. So I didn’t think it was fair. And that’s why I quit Google. And then we that’s why we started BEIT with the really noble adversary that third co-founder mathematician, who was just the basic who was just the designated future of the crazy physicist ideas, right, with his formal approach to problems and ability to formally prove them in the language I could understand. So some of those ideas were just shut down, initially, but once he was having difficulties shutting down some of those and coming up with his own, we decided it’s time because the hardware might be ready in our lifetimes.
[Scott] And, yep, go ahead.
[Wojtek] One more thing, you know, the, there’s more about me now. I missed the mobile revolution while working at Motorola on it, right. We weren’t delivering 3G networks globally. And I didn’t think it was anything big Then I essentially underestimated the Internet revolution, despite being exposed to the early stages of the web browser at CERN when I was still a physicist, and I thought he was like, come on, I just fixed an important problem in the CERN library. And those guys are doing just the documentation system. So when they bring the head of the CERN data division at that moment, he introduced me to his team, I was like, Nah, it’s an unimportant, right. So that’s how I missed two revolutions of my lifetime. And I was adamant that they will not let the QC revolution the pass by. So that was the real motivation was starting this now.
[Scott] The third time’s the charm, as they say.
[Wojtek] Third time’s the charm.
[Scott] I understand that BEIT is software focused and hardware agnostic. But you’ve mentioned that the hardware was finally kind of getting to the right place where you felt it was sort of time to jump in. So what about what was happening at that time with hardware that’s now that, you know, allowed you to see sort of a maturation in the hardware technology that you felt made it the right time to kind of jump in?
[Wojtek] Right. So you know, I have big doubts about some of the hardware solutions touted as the quantum computers, and I will not mention the names out of kindness, actually, they get to that idea too. And switching tax. Let me talk about the NMR computers right, there is no entanglement. This is pseudo pure state, what is the so it cannot provide the advantage. So it’s useless as a better computer, we need a better mousetrap for some of the problems which we are facing. And if something does an offer that is of no interest, even if it is driven, in a sense by the quantum mechanics, right? If you think about like the floating gate field effect transistor works, the quantum device, right? The current is so small that you are struggling to overcome quantum effects in such a transistor. And why don’t we use them rather than fight against them to get some speed up, which was luckily proven possible early on theoretical works by Shore and Grover, essentially showed us that there is a way and this was initially this was a crazy idea. And you know, the rather stupid device might say this, if you draw the line provided by the most low, and then derivatives, I know, it’s about something completely different, the optimal size of the chip on the die of a given size, and that the givens rate of failures. But essentially, people draw many conclusions and say how the technology is progressing. So all those lines get to an interesting point sometime in 2023. So we are switching by a single electron, you have a size of the of the transistor, comparable to the size of an atom or the bigger ion if you just blow it up by reducing of some of its electrons. And then essentially building one of those costs about the GDP of the global GDP of the whole earth. Right. So this is, obviously they are pointing to a singularity sometime in 2023. And that was the initial motivation. But then we realized that the was a breakthrough. I think when I learned about the global approach to quantum computing, Harvard never asked me to go to the execs I was an old hand. So I could go to Alan Eustace, and actually ask for the funding for this thing. But then I realized it might be getting serious, Google is really serious about technology. So once they started it, with all the tax they did, initially, with the really bad ideas, but some extremely good, especially one with incorporating the group from Santa Barbara, into Google to develop the machine. So that was a great idea. But then I realized that it’s happening, right. And that motivated the timing as much as the this funny thing about the singularity in 2023. So that’s more or less what, motivated us to start at this moment. And then there were pivots. And we’ll get to that why it matters because you know, demonstrating something like two five cubits is completely different thing of providing this advantage which I mentioned which is extremely important for me there is no use for the quantum computer, which is weaker than the HPC’s of today. So that’s something which made us pivot several times along this journey.
[Scott] Well, let’s maybe talk a bit more about those pivots that so you know, you have this initial vision, and then you’re coming to certain realizations along the way. So how has BEIT changed course? Over the last? I guess it’s been about four or five years now, since its founding? Like, can you maybe tell us a bit about how you’ve changed your identity?
[Wojtek] Yeah, so the changes, right, first thing, you know, the being a small startup, essentially, from Poland, as you mentioned, required us to prove our worth, right, the in that space, people were expecting physics PhDs, we had one physicist with PhD and one mathematician with PhD. And then we thought that we’ll be doing good with just a bunch of the really talented individuals, regardless of their academic credentials. This is not very convincing for people who are bound on hiring physics PhDs, even if there’s cannot really program and kind of really think in terms of the complex systems being solved numerically on any hardware. So we did the change here. And then we wanted to establish the our name in what we are doing, which was kind of tricky, until you are able to show that you have something other people don’t, right. So people that the something which was very closely linked to my ideas from the past related to the Grover search. And you know, there was a nice moment and doing the fast forward. Now, when we got a review of our paper in physical FA, one of the reviewers said, Well, this is for the first time since 2009, when anybody was able to prove the scaling and the optimality of the new quantum algorithm. And it’s like, wow, indeed. And that’s, something when you are not running a department at the University, but a small start that with youngsters who are just working on their masters, that’s really something getting this approved, and actually making sure that those ideas were not only theoretical, theoretically interesting, but we can make them run on the actual hardware, that was really something because then you do something which in math would be two steps of the induction proof. So here is a very small version of that stuff. And it runs for real, like you know, people say that to do six qubit Grover’s search, which is useless, it’s like searching for the spent match in the match book full of matches, right, because of the sizes involved. Nevertheless, that’s the limit of the of the hardware capability, provided you use our software, our method of solving that problem, nobody else could do this, because it takes something like 1002 qubit gates, which is out of bounds for any of the existing nice hardware, right? If you do our stuff, then you can reduce that and bring essentially the point where those things will be useful forward, but something like two to three years, because it’s a factor of 10. And that’s something which happens in that industry in about two to three years. So that was, that was the first moment when we realized that we can do something better than anybody else. And this still holds right there is a benchmark on the Grover search, you’ll see that it requires about 1000 CXC gates, we did that in 23. Right, not exactly the same thing, but still better than the classical solution, which is important. And then, besides this prove on the existing hardware, you are formally proving the scaling, saying that if it works for six, it will work for 6000. And that’s what we did. And that’s the power of what we are doing protecting the IP. And I think we’ll get to that, because it was also very important for us. So that was the first thing right, making sure. And you know, this also touches on our hiring strategy bill. It wasn’t accidental, haphazard, it was just designed that way. So having those guys and being able to come up with those solutions and being able to partner on the basis of those solutions with the best hardware manufacturers, you know, those conversations are strikingly similar with those guys. So it’s like, we know better. Alright, can you do this? No, nobody can well run our stuff. Oh, it works. Can we partner right and there are many things around this but that was how we established our position in that industry. Some other methods We’re also employed, you know about our results in the quantum challenges, we essentially stopped participating in them, we did what we wanted there. And that that’s enough. We have a pipeline of candidates which we are satisfied with. And we’re extremely selective. This just works as a design process with the expected outcome and measurable along the way. So that was our first pivot, we are not expecting for the reason that our unstructured search will be something extremely important in the far future, it’s not giving enough of the power advantage over the HPC for that matter, but that was letting us work with other guys who needed the that kind of proof to establish that the hardware actually works. And us, it helped us to establish our position in the industry. So that was the first step. And we are equally comfortable generating the ideas pushing the manufacturers like a provide the mid circuit measurement, this is something which we believe can benefit the computation, despite what the textbooks say. But by the way, anybody not doing this is kind of missing the boat, you cannot do the fault tolerance, you cannot do the error correction, you cannot measure and feed forward the measurement result to the subsequent computation. So why bother, it will be just a toy. So that was something which we were stressing from the first moment. And we got the partner understanding this right they between first and second set of our runs, Honeywell provided mid circuit measurement, and we show that it’s actually beneficial and we protect, before that we protected part of that with the with the appropriate patterns. So we are essentially saving resources by doing it our way. So that was a second tack. And the third one, I’m kind of again, jumping forward a bit was realization. And again, it was serendipitous. So we met people who are at completely different stage now use what they say. And then you might know, who am I talking about? But you know, essentially, we are for the quantum computing, we are in for the moonshot, right. And getting to the moon, like all the nice machine, regardless of the underlying physics, and the brand of the provider, all those essentially are trying to build a ladder to the moon, one step at a time. Well, everybody knows you need a rocket engine. And there are people. Luckily, there are people who are building those rocket engines, right? Meaningful tolerance. And that’s our most recent pivot, we did re-examine all our previous patent applications, about half of them pertains to the fault tolerant machines, because it saves the critical resources there. We were optimizing for the two qubit gates, because that’s the limitation on the existing NISC machines. And then we moved to optimizing for smaller rotations, like the T gates, half of our patents for NISC I usable in in the in the fault tolerant model. And obviously, I think I thought about businesses in answer to some of your other questions. So we might get there. But essentially, that’s also the difference. And that’s the pivot, yet another pivot we made. So now we are like committed to fault tolerance. And we treat the remaining stuff as you know, the ability to talk to people show what we can do, and fight some hype, you know if we’ll get to that, I guess, as well. But that’s something which, which we’re really focusing on right now, on the way we started working with D wave initially as a part of the CDL accelerator. And as such we generated several patents for the quantum inspired technologies. So massively parallel, a kneelers like D wave like Fujitsu Tao, and then trying other architectures, whether they can run this you can go to our website and you will see the emulator of the way the completely classical one and we try to employ the techniques in there for some of the parallel architectures like the cerebral city too we are talking with those guys we are after first test. We were talking to Fujitsu and their Tao machines and our own stuff runs either on the CPU or on the GPUs as we want. So but this is straight and so are some of our patterns for the NISC specifics, the specific machines, because in the in reality, if there is a quantum advantage in the future, it will not come from the NISC machines, it will come from the fault tolerances.
[Scott] Tell us a bit about how patents are perhaps helping draw investment or growing the BEIT and positioning yourself for your, for your moonshot.
[Wojtek] Alright, so actually, I think that’s easy. But this is less of the focus with the VCs, they do realize that the revenue from patents comes late, right? The timeline for this is pretty long is extremely valuable, especially the European patents are more valuable than the American ones, German ones, for example. But the timeline is long. And you know, luckily, the timeline for the quantum computing hardware is similarly long. So we made that connection. And we think that we have that kind of time. So it’s less for the investment community, if somebody understands that buys this, that’s perfect, right, that’s makes the conversation easier. But we are really driving for the real impact. If you need our patented software, to bring the quantum revolution forward by if we are going the estimation is by the NISC trajectory. So that will be 10 years right now, we know that the hardware will be sooner, but the fault tolerant hardware will be sooner. But judging by the improvements, that trajectory of the NISC machines, with our 100 fault improvements, we are bringing the quantum revolution 10 years earlier, because it’s like they 10 fold every year, it’s like you know, hand-waving kind of the argument we have proven. Well, we did our structured search in 2019. Nobody could, nobody have beaten our results today. So it’s like those two, three years seems reasonable. Nobody was able to match what we did with our patented software. So we are aiming this at the real benefit rather than just for the investment needs, which again, makes it harder. But in the long run, I do think that the eventually, people selling snake oil will get routed out of business for the reason it might take some time and you will be there, we will definitely be there with investment or without it will prefer waste because then we’ll be bigger and more important. And we want to be a certain layers of the QC stack, because it brings the revolution forward. And also it lets you handle larger problems. And again, and jumping ahead. A lot of the industry is working at sub optimal solutions, because there is no computing power are able to find the optimal solution for the problems they are facing. So that’s where we want to make money. And we are patient, we can do that in couple of years. It’d be great to survive till then, preferably on the investment.
[Scott] So Wojtek if we are now two to three years into the future from now and having this conversation again. Or what do you think the likelihood is that BEIT will be making or have made another pivot?
[Wojtek] Well, I think the there are no more pivots, which I can imagine needed forward. Right? If the fault tolerant fault tolerant systems are what we need right to it, you know, when I was talking to those guys previously, and they were asking whether I need fault tolerant system, I was like, you know, no, I just need to be able to run a circuit of 200 qubits for something like 2002 qubit gates. And if you provide me with that, I’m happy. But then if you do the calculation, it shows that Well, essentially, you are in fault tolerant regime, if you provide those parameters, right. So I was avoiding the answer. But those guys are providing the answer in the sense that they might be hardware which does exactly what we need there. And then that would be the focus. So we think that we selected the layers in the quantum software stack we want to operate for the reason we want to stay there. And I don’t think that we need to pick up much from out of this that was actually the same layers we were working on when we were thinking NISC. We claim that we do hardware agnostic algorithms and their hardware aware implementation. So we are just changing the implementation also optimizing our circuits for something else. It used to be two qubit gates, because that’s what limiting risk. And now it is the number of key gates or small rotations, because that’s what’s limiting the fault tolerant architecture. So not much of the table that the ideological level. But yes, it is a pivot if you think that you are now optimizing for something else. But you know, for the people we have here, right? The cost function is different. That’s all we change. So it’s not really a pivot at that level.
[Scott] And, and finally, Wojtek, we’ll just kind of pivot to the end here. It’s becoming a bit of a tradition and in our interviews here at that cat life got to finish on a on a, on a softer tone. And we were wondering if you could maybe tell us if you could invite any three people from history to a dinner party to discuss quantum computing? Who would you invite, and why?
[Wojtek] Well, I miss missing Feynman at MIT. So obviously, he’ll be the one of those. And I missed him by a couple of years, but his students made a lasting impression on me and actually, that they directed what I’ve been doing since 2000. Sorry, since 1993. So finally, for sure, and two other geniuses, Turing, very clearly, he had everything there was already quantum mechanics when he was having his thoughts, right. So knowing that it can be done, I would love to pick his brains on the on the potential of the quantum computing compared to the Turing Machine, which is essentially an abstract model for the classical computing. And the third one would be, I have a difficulty here, like the tie for the third position, it will be either Von Neumann or his close collaborators, Stanislav Olam, the Polish guy pretty shy, so you will not see his name, especially that some of their works are still confidential. But those guys knew a thing or two about deficient computing. And they found Norman automaton. And they this stuff, you can read out whether this was for Neumann, or alum, they were both working on that. From Other Sources, it is clear that for no man is a genius was a genius, right? So one of those guys would be the third one to have that dinner with, and top quantum computing.
[Scott] Or that would certainly be a very interesting, interesting discussion with that collection of people. And obviously a lot of impact on the kind of work and research and developments you’re doing today. So Wojtek, I just thank you again for your time today. We’ve really enjoyed the conversation here and learning more about Bates and getting your deep insights into what’s going on at Bates and in the quantum computing community at large. So thank you again for your time. It was our pleasure to host this interview and have this conversation.
[Wojtek] Thank you very much again for having me and listening to my ramblings on the quantum computing and the technology in general. Thanks a lot.