Entangled Things
Entangled Things
Episode 154: Jonny Olson on the Cost of Proving Quantum Value
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
In Episode 154, Jonny Olson, Vice President of Strategy and Operations at Zapata Quantum, joins Patrick and Ciprian. Jonny is a physicist and a quantum algorithm IP attorney, and he opens with the story of Zapata’s early sales calls, where the first questions were always what quantum is, what it can do for you, and when it will matter. The team look at why the first wave of near term quantum solutions fell short against fast improving classical computing, and what DARPA’s Quantum Benchmarking program found when it tried to prove economic value. Ten use cases made it to full resource estimates, and none showed a clear economic win. From there the conversation covers how qubit count, runtime, and scaling shape which modalities fit which problems, why optimization claims are so hard to prove, and how enterprises are following fast while they wait. Jonny explains Zapata’s goal of cutting the cost of validating a quantum use case by a factor of a hundred, and how AI agents and formal verification could close the gaps between tools and experts.
In episode 154, Patrick and Cyprien speak with Johnny Olson, Vice President of Strategy and Operations at Zapata Quantum. The team discuss the DARPA Quantum Benchmarking Program, why optimization advantages are so hard to prove, and the case for a multimodal quantum future. Welcome to Entangled Things, your quantum computing podcast, hosted by Patrick and Cypri.
SPEAKER_01Hey Cyprian, how are you doing? Hey, Patrick, I'm doing good. Looking forward to another great episode of Entango Things.
SPEAKER_03Yes, today we're joined by Johnny, and I'm very interested to talk to him. Johnny, do you mind introducing yourself to our audience?
SPEAKER_00Patrick, Cyprian, great to meet you. Um I'm Johnny Olson. I'm the uh currently the VP of Strategy and Operations at Zapata Quantum. Um I'm uh, as far as I know, the only quantum algorithm uh IP attorney in the world. Uh wow. And uh and uh I have a background in in uh physics, a PhD from LSU. And uh yeah, I've been working on quantum computing for since 2012.
SPEAKER_03Talk about a rarefied space. I mean, you know, patent attorney and quantum physicist. That's amazing. So what does Zapata do exactly?
SPEAKER_00Well, you know, um let me tell you a story, Patrick. So I love stories. You know, we we started off a long time ago as uh as a quantum software company, and uh, you know, we tried to sell uh quantum computing services to people, uh, you know, to enterprise and government. And you know, what what do you think the first thing people ask when you jump on a sales call with them is?
SPEAKER_03What's quantum?
SPEAKER_00Yep. Uh then okay, what's the second question?
SPEAKER_03Why do I care?
SPEAKER_00Yeah, exactly. Or or maybe what can quantum do for me? Yeah, right? You can say. And then what's the third question?
SPEAKER_03How much does it cost?
SPEAKER_00Yeah, uh well, I might say, you know, when? When does it cost? Oh, yes, that there you go.
SPEAKER_03Yeah, when when when will this matter?
SPEAKER_00Exactly. And uh, and so we we um you know, for many years, right? It was all it was humorous. I mean, we would jump on a sales call with people and they would always say these same questions. And it was, I mean, you you we would always kind of internally laugh because those are the toughest questions to answer, right? You know, the the you know, whether quantum helps you is a deeply organizationally dependent question, right? Uh it depends on what's going on. And when does it occur is like the million-dollar question, right?
SPEAKER_03Yeah or trillion dollar question.
SPEAKER_00Exactly. Yeah. So I mean, you know, uh, and and that depends too on you know uh on the first question. And um and yet it's it's obviously the most two important questions you can ask, basically, right? It's like, well, you know, uh what's the relevance, right? And so um, you know, when we started off, we were selling uh, you know, we we were trying to sell solutions in NISC, right? Which is uh near-term intermediate scale quantum uh solutions, like primarily quantum chemistry, quantum machine learning, those kinds of things. And um uh so that changed a little bit. So, you know, uh I the the NISC solutions, I think we uh you know, back in in 2017-2018, when we started, the the quantum community was very optimistic about these solutions, us included. And uh and so we were trying to build software to support these uh these solutions and then and sell the solutions. Um but you know, more research needed to be done on whether these things really worked or not.
SPEAKER_03Right.
SPEAKER_00And and over the the subsequent five years or so after we started, the dominoes kind of started falling over one after another, and it seemed like these NISC solutions just weren't quite performant enough. Um in part because classical computers are damn good. Right. Um and uh you know and still getting better as year as it sounds. Exactly. It's a moving target. I mean, during that time AI came out, right, which changed a lot on the classical side. Uh GPUs are just getting pretty crazy as well. I mean, in terms of their performance. Uh there's new technologies like you know, ASICs and well, I maybe not so new, but uh relatively speaking. And um and so you know, we really couldn't demonstrate a whole lot of quantum advantage. And so by I think it was 2022, um the government, so the DARPA, uh started a program called the Quantum Benchmarking Program, the QB program. Um and uh Zapata was was you know the only company that that was in sort of all three uh areas of this program. And and the the program was the the entire goal of it was to solve uh the question of will quantum computers be ever useful at anything? Uh you know, we can you prove economic value, right, for for a particular problem? Uh of course, you know, I I think uh Shore's algorithm, right, is is you know everybody's favorite example. Um but it's the boogeyman. Yeah, exactly.
SPEAKER_03Uh Bobby Yeager.
SPEAKER_00I mean, yeah, we probably wouldn't have modern quantum computing uh today, right, if it weren't for that.
SPEAKER_03But um certainly wouldn't have the funding in the non-winter um scenario.
SPEAKER_00Exactly, yeah. Um but you know, in in any case, um, you know, it shore's algorithm is maybe not the best example of economic value, uh, of course, right? Um maybe maybe there's value to the to uh nation states and things like that, but um you know, just just for uh uh you know companies, enterprises, like why would you be interested in quantum computing? There was no uh no proof anywhere that if you had a quantum computer, right, you could you could have a problem, uh, you could, you know, you understood the conditions under which it ran. So in other words, you know, if I'm if I'm running a computing task, let's say I'm like a quantum uh chemistry researcher, right? And I, you know, I have some workflow where I'm examining things, if I get an answer in an hour, right, and that's sort of what I'm used to, uh, then a quantum solution had better be able to answer it in an hour, right? Orabouts.
SPEAKER_04Yeah.
SPEAKER_00So you know, even if I get a hypothetically better solution in uh a week, right, uh, where oh you can you can argue it beats the classical solution, that's not a very useful one because my you know, my uh typical uh constraints for my job are I get the answer in an hour, not a week. Right. So so you know, this was and this is a really important part of the QB program. So we're really looking for you know tasks where uh it fit operationally within the constraints of the problem. Uh, and then there was some unambiguous calculation, right, where you could say the quantum computer is producing it, you know, producing a higher quality solution than the classical one. Yeah, that makes it a very reasonable question to ask.
SPEAKER_03Right. Right. If not a little rude.
SPEAKER_00Yeah, I uh you know, I think it's dare you. How dare you? It was very rude, yeah. Um uh, you know, somebody actually holding us to to all these uh statements we're making, you know. Um and it was I think it was a bit remarkable that by you know, even in 2022, we didn't have this answer, right? And yet, you know, all these companies are are building on the basis that there is going to be something. Um and well, long story short, this program lasted about three years. Um the program started with a hundred use cases, and uh and right the goal here uh obviously is sort of to find the most transformational ones, ideally. Right. Uh and the so the program started with 100 use cases, and by the end, there were 10 papers produced. Uh 10 papers where you know there was sort of end-to-end resource estimation done. And you have to appreciate how difficult that is, right? It's like not only do you have to figure out all these details about how valuable is the thing, how you know what's the uh how do you compare it to the best classical algorithm, right? Right, then you have to map the thing. So you have to build, of course, a circuit for a logical circuit, then you have to map it to real hardware, then you have to take into account error, quantum error correction, uh, and all the resources needed to produce the error correction. And then finally, you know, you have some examples of hardware that it could run on. And finally, at the end, right, you get some you know, qubit estimate uh and a runtime. Right. And so from there, you you can approximately assess right uh the value of this thing. And a c you know, of course, you have to sort of also take into account the cost of running the actual quantum computer, which is not zero, certainly. Um and so uh doing this through all ten applications, in fact, most of those were not a hundred percent complete. There were still a few missing details here and there. Um one of them came back as a definitive no. Okay, no value. This was uh computational fluid dynamics.
SPEAKER_04Okay.
SPEAKER_00Uh and in the other nine, which were almost all in the sort of material science uh or chemistry or um or differential equations areas.
SPEAKER_03Makes sense, yeah.
SPEAKER_00Um, they all came back as meh. Meaning maybe there's uh small value, right? Maybe a couple million dollars.
SPEAKER_03Is this gonna end with us canceling this podcast?
SPEAKER_00It may.
SPEAKER_03It may.
SPEAKER_00Uh hopefully, hopefully, um when we get we'll we'll we'll get to the more positive okay part later, but uh right now, I mean, I can understand. I'm not filling you with a great deal of it's okay.
SPEAKER_01Surrealism is is always important. I'll be right back. I'm getting some popcorn.
SPEAKER_00Yeah. Well, yeah, the and and I mean you imagine what we're feeling at the time, you know, going through this thing, right? Uh and uh in so yeah, so three years later we produced the answer, and it there's nothing, you know. I mean, effectively three years was spent, probably 30 million plus dollars. I think there were eight different teams working on these different papers. Um and we basically came back with a nil, right? Um, maybe there were some so there are some examples where uh it looks like there's a uh you know a a new uh uh what would you say, an e economic advantage, but it's just outside of that window of usefulness, like I was talking about earlier, right? Yeah, it takes it's the runtime is a little too long, or you know, something like that. And so um the you know, this was not encouraging. Uh and we were like, oh my gosh, you know, even if we had a quantum computer today, right, yeah, we would not know what the heck to do with it. And that is in fact still the case today. Um, there are scientific examples. I mean, you know, uh scientific experiments and and scientific computing and stuff is maybe a different category, you know, because it's sort of uh there, you know, you're learning things and you're never gonna find economic advantage, you know, in a scientific calculation directly, right? Uh but uh but you know, in terms of useful things to run on something, we still don't have that today. Which is which is a little like frightening, right?
SPEAKER_03Um and and this is a this is a study that's been launched by the government.
SPEAKER_00That's right, yeah. So this was the DARPA, DARPA QB project.
SPEAKER_03Yeah.
SPEAKER_00And uh now there was enough, there were enough close calls in this case, and you know, there is enough, there has been enough progress that people are still encouraged that that there will be economic value here. So good, you know. Uh thankfully, yeah. And there's there's really been amazing progress the last couple years on uh on quantum error correction. That changes quite a bit.
SPEAKER_03And that and that's kind of the missing formula. Like if we had systems with 10,000 error-corrected bits, uh, you know, logical qubits, that changes the formulation, I assume.
SPEAKER_00Well, that certainly, you know, uh, I mean, there's really two numbers. Well, maybe let's say three numbers that matter, right? I mean, there's of course the number of qubits is an important number. Everybody focused on that in the beginning because the hardest thing to do was get a lot of them. But now, as we're getting to larger systems, people are going, well, okay, what's the uh, you know, number of qubits still matters, but the second is what is the runtime? Right. And there is a huge difference in a lot of these platforms on their runtime, right? Uh superconducting things run hundreds or maybe even thousands of time spanter trap systems.
SPEAKER_03Right.
SPEAKER_00And so, you know, uh for for time critical applications, right? That can be a massive factor.
SPEAKER_03Uh we've always thought that there's probably gonna be more than one modality in in usage based on that exactly what you're saying. That it's not gonna be. I mean, we we had vacuum tubes and then we had transistors. Transistors one. I don't haven't I can't point to a vacuum tube anywhere in my in my domain. Um, but I I don't think that's gonna happen with quantum. I think there'll be photonics and superconducting and trapped ions for different purposes. Is that does that track with what you found as well?
SPEAKER_00Uh there's good reason to believe that. I was uh I you know I've been I was doubtful of that that belief for quite some time, but it it is interesting to see the way things are shaking out because some systems are really, really hard to scale large. Um I mean, these superconducting systems, it's it's a it's difficult. And you don't get that grade of uh you know quantum error correction rates. That is the you know, the the number of uh physical qubits that you need to encode a logical qubit is often uh quite high on those systems. Yeah. And um, and so, you know, but they run incredibly fast, right? Uh whereas some of these other systems, which can scale up really, really well, uh, you know, neutral atoms are amazing. Uh and they have, you know, they're sort of the opposite in some sense. It's like very easy, uh, very efficient mappings of quantum error correction. Actually, their problem is decoding because they can do it so damn fast. Uh and um the uh you know, but the issue there, right, is um is speed. So uh so that yeah, there's a there's good reason to believe, you know, there's photonic platforms, right? Maybe you can get those, uh, you know, those have uh it looks like extremely low error rates, and so they can sort of keep those those systems maybe running for longer.
SPEAKER_03I keep seeing news of like they just had a photon that they kept uh going so that they believe they could have sent it 30 kilometers, and so they're trying to basically build the quantum internet. And that's why I think that multimodality is a no is is is gonna be a slam dunk because you need photonics for the communications. I I don't see us shipping packages of trapped ions across across wires. You don't?
SPEAKER_02I don't, maybe, maybe I'm wrong. You never know.
SPEAKER_03Um okay, so so okay, you said that the number of physical qubits, the number of logical qubits, the error rate is a big part of it. What you said there was another one. What is the other one?
SPEAKER_00Uh yeah, well, the other so so I guess there's uh the the size, size, okay, yeah, speed, uh it well, then there's also scaling, right? Uh and and so what what by that you know I mean it's in some sense the size question or the speed question, but it it's just right, how uh you know, how does that change in the future, right? So you may see some systems which are uh not winning right now, right? But they have amazing commercial potential, uh, right. And so the scaling, and maybe we could call scaling just cost too, right? But um uh right the you know, at some point, right, we may we may have a large need for a lot of quantum computers, right, running independently and so on, or deployed in places, and obviously the the cost of actually manufacturing these things is a significant factor, right? Um I mean you could throw in some other things like operating conditions, some of these things run at room temperature, you know, other than run it a hundred times colder than space. You know, yeah, the vacuum. Uh uh the quantum vacuum is is a hundred times hotter than uh you know than these systems. It's amazing. So I think you're right. I think you're right, Patrick. I mean, uh the there's also some belief that some of these systems will actually become uh they'll be like multi uh mode systems. In other words, like you may run, for instance, uh parts of the calculation on one system and you know, then use photons to hand it off to another system which you know calculates something faster. And yeah um uh you know of course that's you know uh very, very complicated, but you know, maybe we'll see that.
SPEAKER_03Yeah sometime. And and there's the concept of interconnecting. Like if I had if I had a hundred fifty logical qubit systems that I could connect and I could okay, you're gonna do this circuit over here to the you know, this depth, and you're gonna do this circuit, and we can we can bring that all together into one calculation and make one you know 5,000 logical qubit system in that way. I think there's a lot of ways to to to eat the apple, but we're still in early days.
SPEAKER_00That's right. And and you know, maybe and coming back to the question, I told you this long story, you know. I feel like Norman or something, you know. I just uh maybe there's a joke at the end. Um but uh the uh right, so the the you know the story was in in some sense to answer what we do here at Zapata. And um and so just to sort of reiterate, you know, this QB program took us years and many, many dollars, and we still don't have the answer for where the you know what the killer apps are gonna be. And so uh the question remains, uh we still have to we still have to find them. Um and if you really follow the literature and and see how many of these problems are getting the kind of treatment that these DARPA problems did, where they're you're you know, you're not just looking at is there a hypothetical advantage or an asymptotic advantage, but you're really looking at realistically, does this provide a user you know with something that they want to do uh better than a classical computer, and it's you know cheaper and and faster or whatnot. Um that process, I mean, if you you I said 10 use cases and 30 million dollars, okay, so that's three million dollars a pop. And it seems that most of the time the answer is no, quantum will not work. And so you sort of like do the math, right? If you're like, okay, let's say I'm an enterprise and I'm interested in you know quantum computing applications, am I gonna have to like how many applications am I have gonna have to uh validate in this way before I get something that's valuable?
SPEAKER_03Well do we know the answer to that?
SPEAKER_00Yeah, uh I I you know if the answer is a lot, then we we know it. Um but you know, the thing is that the the pace uh you know is it has been extremely slow. It's you know, you can count on one hand basically per year how many applications get this treatment.
unknownRight.
SPEAKER_00And that and the DARPA QB program's over. So, you know, even that.
SPEAKER_03Well, isn't the market deciding this and voting with their feet? Because we're seeing companies in the quantum space making revenue. D-Wave, and they're not all cash positive, but I I can't say the same thing about AI. Um we're we're seeing companies paying for things. Uh I you know, if you if you look at D Wave, they're they're advertising quite often that they have a webinar on how it's helping gas and oil exploration. Now they're adiabatic, and I know they're a different class, but they're still in the family. Is it how does that factor in? Is it is it that They're doing things in the hopes that they'll get an advantage down the line? Or is it they're it's arguable that you could do it all with classical? We we've had uh let me continue. We've had a um one one guest, Scott Gennon, who's been on a couple of times, and he's adamant that he doesn't care if quantum ever arrives, he gets a benefit from quantum inspired. So he does everything in simulation, and he doesn't care whether a quantum computer ever does anything for him, and he's doing like material science for OTI um or luminotic luminotics. So is it that the quantum inspired is the advantage and it doesn't matter whether there's a quantum computer behind it, or is it that the the we're still early enough days that the industry is starting to figure that out with their own experimentation?
SPEAKER_00Well, you know, quantum inspired at the end of the day is is a classical computing, right? And and if we decided tomorrow that God, quantum quantum itself is no good, it's just too expensive or whatnot, and quantum inspired is the way, well, then there would certainly be no need for these these computers, right? We can study all of this stuff independently. Uh, you know, uh may there might be some people that argue actually building the system will give you a little more insight or something. Um, or or people trying to develop quantum algorithms, i they'll accidentally invent sort of better classical algorithms like that.
SPEAKER_03And remember, the the the beginning of this whole thing was Richard Feynman saying, well, you know, in physics we probably need a quantum computer to help us with these quantum processes. He didn't he didn't anticipate the oil and gas companies or Shore's algorithm or any of that. So I I could see the the national labs building quantum computers because it's it's an interesting thing, and maybe it's a precursor to the biological computers we'll build someday.
SPEAKER_00Yeah, well, and let's let's give uh D-Wave credit where credit's due. You know, uh so they have you know, they have a platform that's annealing platform, and you know, the the struggle with annealing and optimization in general is typically that you can't, you know, it's it can be very hard to prove a universal advantage, i.e. or or or you know, uh prove something where you can say, this is unambiguously the best way, right, to calculate a uh, you know, to to optimize something or the most efficient way to do it.
SPEAKER_03Because every time you do, as Cyprian said, the the HPC people say, hold my beer, and they go all out to make sure that they show that they could have done it.
SPEAKER_00It's yeah, you know. That's right. Yeah, and and so it look, I mean, you can you can sit there and work on it, and uh, you know, and and the thing is optimization problems are so instance specific. So, you know, uh I can't, you know, in uh a lot of times what you see is uh in optimization, the actual physical problems that you deal with have a you know have a very particular property. I mean, I don't know, if you're I don't know, FedEx or something, you're trying to decide where to, you know, the most efficient way to deliver packages or something, right? That has a particular uh symmetry, right, or a particular um you know, physical uh reality in which you know the constraints are what they are. And so uh a lot of times it can you know heuristic algorithms end up doing ex extremely well. So, you know, I might say, oh, the best algorithm is you know, insert, I don't know, some you know, QAOA or something. But the the answer is usually somebody sits down and tweaks the parameters on these things you know until they're blue in the face, and then they figured out, oh, this very particular thing ends up being the best thing. Right? And there's you could not have proven that for at least the local minimum. Right, uh exactly. And so that's usually how those things go. And so I wouldn't be surprised if D-Wave was able to find some instances where uh their machine was doing better than the best, you know, classical thing that that could become, you know. But the problem is I can't necessarily generalize from that. I can't take D-Wave's solution for one company and then just apply it to the next. Um in some cases, maybe.
SPEAKER_03Uh that's good for the bottom line.
SPEAKER_00Right. And you know, the and then the other thing, you know, I and I'll of course there are many arguments that say, well, maybe uh you know, may maybe the classical, like if you spent as much time and effort doing that on, you know, improving your classical result as you spent on the quantum annealer, you'd end up with something better. So like the there's sort of these arguments that are just I mean, they just muddy the water, right, for those statements.
SPEAKER_03Well, it's it's almost the same as the argument about when Q Day will come. Is it is it right around the corner that will break you know elliptical curve for Bitcoin and and then eventually RSA, or is it decade away? And and the problem is nobody can nobody really knows, which is why the money is still pouring in.
SPEAKER_00Right. Uh you know, I think that's that's a reasonable. I mean, well, maybe just to differentiate that a little, I mean, uh of course Q Day is a uncertainty about the future, right? Uh and maybe this case of annealing is more of an uncertainty about uh alternatives, right? It's like, well, what if what if I had done X instead of Y? Right? And so well, there's no way to tell unless you actually do both. And you know, at any rate, I uh you know, I I don't want to, I certainly don't want to discredit uh uh you know what what D Wave may or may not not have been able to do. You know, it'd be nice to look at the numbers and actually see, but um in any case, uh let like we typically say, okay, like let's let's put that aside for a minute because that's that you that you might say is an uncertain advantage, right? You you maybe flip a coin and you get an advantage. Right. So um we are more concerned with is there something we can definitively prove if we had a quantum, you know, error-corrected quantum computer today, could it do something better, right? And such that it would be better for you to adopt that quantum computer than not. And uh still today, you know, it's uh what four years after the DARPA QB program, we still don't we still don't have an example. And uh and so our job, and I was I was telling you about the economics, right? It's like three million dollars. So we our job at Zapata is to lower the cost of that analysis by at least a factor of a hundred.
SPEAKER_03Okay.
SPEAKER_00Because if if you can if you can do the analysis in you know for thirty thousand dollars instead of three million dollars, right? Now you can start saying, okay, uh, you know, I I could I could be an enterprise and I could have you know, I mean I could run uh you know, I I could do a hundred you know use cases for three million dollars. And you know, with a hundred done end to end, you would probably feel confident that there is at least a couple economies that might give you the advantage. And so and and you know uh these numbers have to fit what enterprises and governments like will actually pay to do this research, right? Nobody's gonna pay you know sixty million dollar uh coin flip, you know, uh based you know on their research funds on the off chance that there might be some quantum algorithms.
SPEAKER_03Well they will for the first few like they did here, but that would save them ten million dollars.
SPEAKER_00Right. And so the you know, I mean, some yeah, you know, some very very ambitious companies might take the risk. But effectively, you know, most will just say, well, why don't I just wait for that company to spend sixty million, you know. Yeah, yeah, I might, you know, it might cost uh, you know, I might only be second in line, but hey, if I follow fast, then I will uh you know, I'll I'll still be able to get the advantage. And uh and I think that sort of follow fast is what a lot of enterprises are doing right now, right? So they're like, I need to build up a quantum program, I need to get familiar with it, I need to, you know, get some team members, maybe play around with some algorithms, do some POCs, so that if somebody you know unveils something incredible, I can move quickly.
SPEAKER_03Yeah. Second mover status. I I'm okay with being Pepsi.
SPEAKER_01Yeah. It's right, like the the big shot use cases like the shores of the world, right? The regular company would not benefit from having the capability of breaking RSA. That's not gonna improve the economics of your average company, not even your large company, right? So, yeah, that's a use case for the government or whatever, um, or for freelancer agencies, but the economics of a specific specific company would not be helped by how about that capability.
SPEAKER_03But it definitely has kept the the winter at bay and it has put money into coffee.
SPEAKER_01Yeah. That's the super positive thing, but or or is it?
SPEAKER_03I don't know. Maybe it's positive for quantum, it isn't necessarily positive for the world.
SPEAKER_00I think I I think I saw a statistic saying this year uh this is the first time when private investment has exceeded public investment.
SPEAKER_04Yeah.
SPEAKER_00Uh and so that is I mean, that's a remarkable number, right? Which tells you that people are definitely interested. Yeah um, but you know, we are we're still worried about the damn thing. Uh, you know, we're worried about what happens if in five years, you know, we still don't have it, or maybe we have one or two use cases that are just you know, I mean, let's say you can't run it on an ion trap machine. You know, Ion Q's burning through a billion dollars a year. Uh you know, how long can they keep that up before you know finding something? I mean, uh at some point people will will lose their their interest, right?
SPEAKER_03And say, well, this is just you know I I think that's more likely to happen after Q Day. And I know I I I hate to belabor the point, but I think because the governments aren't going to slow down what they're spending. Now the companies like Ion Q might be in trouble. I don't think they will be. I think they'll find. I guess I'm a I'm a quantum optimist. I I believe uh before I met you, Johnny, at least. Um, this is a fantastic conversation, I must say. And I I'm I'm I'm not looking forward to it ending it because I would like to you know go on for a while, but we're getting close to the end. But um and and I asked, I'm I'm gonna play devil's advocate if that's okay. Uh I asked some of those rude questions. Uh it's kind of like we're in the infancy of quantum computing. You you got to admit, we you know, this is the uh we're bare we don't we don't even have the calculator, you know, if you go back to that. And we're competing against supercomputers. So is it is it fair to measure it now and say, well, you know, it's not it hasn't, it hasn't brought us to the moon yet? It's kind of like saying uh, you know, I've got the first car off the first, you know, bit built by Mercedes, but it can't get me uh cross country in two days. Are we asking too much by asking these questions? I think we should be asking these questions, but do you think that it just needs more time, or do you worry that there's no there there?
SPEAKER_00Well, I am I am also an optimist.
SPEAKER_03Okay.
SPEAKER_00Um it may not sound like it.
SPEAKER_03Uh I I got a sense of it, but you were definitely an optimist who'd be beaten regularly.
SPEAKER_00I'm I'm uh, you know, I I certainly like to ground the the reality that we're in.
SPEAKER_03And and uh you don't want to be Pollyanna, I admit that.
SPEAKER_00If you follow as you know uh Zapata's history, you know, we we have a a a bit of a roller coaster ride, and uh and part of that roller coaster ride is maybe due to the fact that we were a little too optimistic, you know, uh early on. Uh but um Who could blame you? Uh who could blame you, indeed. Um But the the thing is that the technology, I believe, is is absolutely there to lower this by 100x and and get to the point where you know even a you know a mid-sized company can be interested in quantum computing and and do this analysis on their own, uh if they you know if they want to do that. And um the the thing that's really changed in the last few years is well you know, I hate to say it, but AI. And this, you know, the when you look back at DARPA and when you ask why was this thing so freaking expensive, there's a few answers. You know, there's a few reasons. One is um, you know, the the tools are rather fragmented, so it's it's sort of difficult. You know, there's no tool that just takes you from end to end, right? You have to use a half a dozen things. Um but the other the other problem is that the experts are fragmented in the process. Namely, I need I need probably four or five PhDs of differing uh expertise in order to go from end to end. So first I need a domain scientist, then I need a uh you know, a quantum algorithm scientist, then I need somebody uh you know who's like an optimization, you know, a compiler, right?
SPEAKER_03Sounds like sounds like five grokbots to me.
SPEAKER_00Um yeah, I mean it's it it it's funny, but it's uh it's a little bit true because some of these processes are uh I mean, so one aspect I said the tools. So one thing is you know, what Zapata is trying to do is is cover, you know, close the gap on the tools, at least so we can uh you know, some things still need to be built, some things just need to be glued together, and and agents are really great at doing that. Um the other thing is uh you know, we need to we do need to at some point sort of replace some of these scientists, like with the bots, you know. And some of that can be done through uh well, it has to be done very carefully, right? I mean, if you if you ask if you jump on chat, you know, GPT and uh and even on Astra and ask it to do an end-to-end calculation, then you're not gonna succeed, right?
SPEAKER_03I've we've seen failures in AI as well. Uh what you're more likely to see, I think, is is one of the scientists being able to do more products because they have a model that can they can delegate to. I don't think you'll replace the person in the middle.
SPEAKER_00That's right. Yeah, so to some degree, you can use you can use a lot, you know, you can use agents just to speed up the experts themselves. That's a major factor. But some of this, you know, we think can be can be removed altogether because really what um uh you know, especially with with uh what has been going on recently with something called formal verification, right? And in the right, this is the idea that you can basically prove the you know a mathematical statement in this computer language called lean, or and there's other formal verification languages, but that seems to be the most popular at the time, right? And so uh you know, the the biggest issue with agents is always hallucination, right? It's like, well, at what point did this thing sort of go off the rail? And so with these tools, you can really close the gap uh on some of these processes and say, well, I can actually prove that in fact the the thing produced by the agent is correct. And you and or you can say, I proved that it's incorrect, and so now I can tell my agent to go back and fix it. Right? And so that allows me to close these loops. And um, and that's why we think uh, you know, in a short amount of time, we can actually get to some you know a hundred X cheaper uh in in uh a relatively short amount of time. It's still gonna depend, you know, on uh uh you know a lot of good engineering.
SPEAKER_03But um so speaking of a relative short amount of time, I've missed the deadline of saying, hey, we're about halfway through. Uh so we're we're we're over time, but that's good. It's been a fantastic conversation. Is there anything I and I mean I we could go on for another hour, I I believe. Um, is there anything else, anything that we should be watching for from you guys?
SPEAKER_00Uh well, you know, we we will be making the next appearance on the show. We'll be making some exciting announcements in the next uh in the next few days, actually, or maybe weeks. Uh well let's just say month. Let's just say month and be safe.
SPEAKER_03By the time this comes out, probably.
SPEAKER_00Yeah, well, that's a good point. That's a good point. Uh it'll probably have already come out. So uh look in the past and uh and uh you know we have uh a new product coming out, and that this product is precisely this question of how do you take enterprise problems uh and verify and verify them, right? And um and and not only that, we also uh update them. So you know, tracking is an issue. So once you can do once you can do verification cheaply, it also allows you to track them cheaply. Because now let's say a new quantum error correction paper comes out, right? I can rerun that estimation or even a part of that estimation efficiently on this new technique. And so I can almost in real time keep track of these use cases, which is something it's kind of like watching your tokens. Right. I mean it's some it's something that's been just uh uh vexing, I think, uh a lot of companies because they'll do a report sometimes by I mean we uh uh back in the day, uh I remember doing this report for uh BP, you know, British Petroleum. And I remember by the end of the by the time we generated the report, it was already out of date. Like there were new developments while we were writing it that were just like, ah, well, I wish we could go back and do it again, right? Uh and update this. And then but but that's the that is a really vexing problem, too.
SPEAKER_03So when what once that comes out, we'll put it on your page because every every guest gets a page on the website for you know entangled things. So by the time if if it is in fact released, by the time this airs, people should be able to see it on your page. And if not, we definitely encourage them to come back and check it out.
SPEAKER_01And and Patrick, just before we we close, I just want to make sure that we point out an important thing because I I believe the work that you folks do is extremely important. I just want to highlight it, make sure our audience understands it clearly, right? Is your work allows companies to look at their potential use cases way before the actual stable quantum computers are there, right? So it's not a matter of we have to wait to see, right? You help them basically understand if a certain use case would hold water or not uh today, or maybe it will change tomorrow, right? Even before the computers are are here. Would that be correct?
SPEAKER_00Thank you, Cyprian. I you know, uh, we are hiring for business development. Would you uh would you be open to uh nice try.
SPEAKER_03Nice try, Johnny.
SPEAKER_01Uh I I really think this is important. Patrick and I discussed, like, I think many, many times in the show how difficult it is like building the computers and doing the engineering and everything, but at the same time, it's perhaps even more difficult to translate and map the problems that you have in real life, right, into something that could be used on a quantum computer. And and this is why I believe like your work is really, really important. Yeah.
SPEAKER_00That's right. And that's our belief. You know, in some sense, you know, when when Zapata went through this restructuring, we had we had a little bit of time to sit down and really look ourselves in the mirror. And that is really precisely what we decided was look, the you know, the quantum ecosystem is very bottom-up built. I mean, uh, by that I just mean hardware, right? Sort of starts at the hardware and it and it comes up. And there are very few people who meet, you know, meet the customer where they're at and then say, how do we make this useful? I mean, those that typically do are like the consulting companies, right? And so, in some sense, what we're doing is maybe competitive with the consulting companies that are offering, but but you know, and we think it's it's much better because it allows you to really come up with a data-driven strategy.
SPEAKER_03You're like the Gartner of quantum.
SPEAKER_00Right. Right. You know, I mean, can you uh, you know, because it that's that's what people are wondering, especially, you know, Patrick, you mentioned like Q Day, right? As we get closer and closer, it's like, how you know, how am I gonna be, how do I know when I'm gonna be ready to take advantage of these things? And so, you know, by by looking at these use cases, figuring out which ones are actually useful to you, right? And then encoding them into the platform that can be updated over time, you will know the time uh when the time comes, right? And and sometimes and part of that is knowing when the time comes to start preparing for exactly because there's there's always that lead time.
SPEAKER_03Yeah. So I think we need to see you uh regularly in the fall to winter to to see like, is Johnny getting more or less optimistic? Let's see what's going on with that. Like the ground our groundhog day. Uh it's been a wonderful conversation. Um, thank you for for taking the time. I I wish I didn't have to end it here, but we're coming up on the limits of what we normally publish. So uh, but it's been great. Uh I appreciate your time.
SPEAKER_00Thank you so much, Patrick, and thank you, Cyprian.
SPEAKER_03Thank you so much. It's been a pleasure. Thanks, everybody. We'll see you next time. Bye. Bye, everyone.
SPEAKER_02Cybercrime is one of the biggest threats to businesses of all sizes and industries. With almost half a million open cyber positions, the problem is compounded by the lack of available talent in the marketplace. At Pulsar Security, our elite team of highly credentialed experts collaborate with you to assess your current defenses and develop solutions tailored to your specific needs. With services ranging from cybersecurity education to advanced penetration testing and red teaming, you can start reducing your risks today. Visit pulsarsecurity.com and let's secure your digital future together.