Podcast With Gustavo Ordoñez of Moody’s Analytics, Giorgios Korpas of HSBC, and Iordanis Kerenidis of QCWare

Transcripts

Yuval Boger: Hello, Gustavo. Hello, Iordanis. Hello, Giorgios. Thanks for joining me today.

Gustavo Ordoñez: Hello.

Iordanis Kerenidis: Hello.

Giorgios Korpas: Hello.

Yuval: Gustavo, let’s start with you. Who are you, and what do you do?

Gustavo: So Gustavo Ordoñez, I work for Moody’s Analytics. I’m a Senior Director Data Scientist, and I basically work in research. So I build new models, explore new ways of using the data to create new ways of looking at what can be done and how we can actually make better decisions, faster, more accurately. That then power our products, so that’s what I said. And part of what I do is to help the quantum ecosystem within modern analytics that is led by Sergio Gago, who I know has been in the podcast in the past. So I support Sergio in that journey of establishing a quantum computing group at Moody’s Analytics. And the reason for that is that I am very passionate about quantum computing and I was a research physicist in the past, many years ago, and that is where I discovered Quantum computing. And I’ve been keeping an eye on it, as a hobby, for many years. And then the last three or four years I’ve been actually looking at it on a more serious note because it’s becoming more and more likely that it’s actually going to have an impact on the industry in the medium term I would say. So that’s me.

Yuval: And Gustavo, what city are you calling from today?

Gustavo: Well, I’m calling from Scotland, actually, Stirling in Scotland. So it’s a small city, a bit in between Glasgow and Edinburgh, but I’m based in the Moody’s Analytics office in Scotland in Edinburgh, but I’m actually now working from home today, and I’m here in Sterling.

Yuval: Excellent. And Iordanis, I think you’re calling from Paris. Who are you, and what do you do?

Iordanis: Hi, Yuval. Yes, I’m calling from Paris. My name is Iordanis Kerenidis, I’m the head of Quantum Algorithms for QC Ware, which is the quantum software company. Before that I started working on quantum algorithms more than 20 years ago. I did my PhD in Berkeley and then I was at MIT. And for the last 15 years, I’m based in Paris, I was working with CNRS at the university. And for the last three years, I’m working for QC Ware where I’m working on optimization and machine learning for the finance domain, but also other domains. And we have worked quite a bit with different financial institutes, in particular on the Monte Carlo methods. So I’m very happy to be here.

Yuval: I’m glad you’re here. And Giorgios, where are you calling from and who are you and what do you do?

Giorgios: I’m Georgios Korpas, and I’m a research scientist at HSBC. I have a slightly different background than most people involved in quantum computing since I came from very theoretical aspects of physics. I have a background in string theory, topological Quantum Field Theories, and within HSBC I co-lead the programs in quantum optimization, Quantum replacements for Monte Carlo, and aspects of classical machine learning as well. And just like Gustavo and Iordanis, I’m extremely passionate about the topic, and I’m very happy to be here.

Yuval: Excellent. And I’m not an expert on Quantum Monte Carlo, certainly not to the level that you three are. And maybe, Iordanis, you can tell me what Monte Carlo is. And Gustavo, maybe you can tell what the difference is between Monte Carlo and Quantum Monte Carlo. So Iordanis, why don’t you go first, please?

Iordanis: Sure. I’ll try to say it in a simple way as I understand it, I’m also not a finance expert, but the idea is that we want to have some assets that want to price, for example, and we want to understand how to price them. And the way we do it, is by constructing some sort of model for the market, for example, which mathematically is some stochastic process. And what we do is, we try to run many different scenarios, according to this process. And for each one of these scenarios, we compute the right price for our asset. And then once we do it many, many times we aggregate all of these results and we compute on expectation what we say is the correct way to price this asset. So this is kind of the Monte Carlo way of pricing some asset and Gustavo can continue on the quantum, I guess, way of doing it.

Gustavo: Thanks, Iordanis. The idea behind Monte Carlo is actually very simple. Basically, you have a problem that is too difficult to solve analytically, so you can only tackle it numerically. And what you do is to draw from a random distribution that you can draw efficiently from, and basically the draws from that random distribution that end up in the solution space that you’re interested in, you keep and the rest you discard. And by doing that, you can infer the actual answer to the problem that you’re looking for. That is, in essence what the Monte Carlo technique is. And it’s extremely flexible, it works for a very large number of problems. It’s also very powerful in the sense that I can solve very high-dimensional problems efficiently. But at the same time, what I say efficiently actually is, in the sense that it’s very easy to code, very easy to get to run, et cetera, but it’s not very efficient in the sense that it actually takes a long time to run, because you have to draw many, many samples from these random distributions.

What Quantum Monte Carlo does is leverage a very well-known set of algorithms in the quantum algorithm domain, namely the Grover algorithm, and the amplitude estimation algorithms, to do the same that you would do classically but quadratically faster. So that means, if the particular problem you want to solve classically takes one million simulations to reach the level of accuracy you’re interested, in principles, theoretically on a Quantum computer, you could do that with only 1000 samplings, which is a huge gain if you take into account that many of these problems are huge problems for which you need very large computer clusters in the cloud to run overnight to be able to get to the answer. That is why this is one of the key problems in quantum finance that have been explored by financial institutions. But as I said, it does, in essence, the same thing, is this sampling from distribution, in this case, because the quantum circuit runs on a machine that is already stochastic by nature, what you do is leverage these well-known algorithms that I mentioned to actually do the same quadratically faster.

Yuval: Giorgios, in your opinion, what makes Quantum Monte Carlo, at least theoretically, better than regular Monte Carlo? Is it that the random distribution for quantum could be real? Is it the ability to run multiple scenarios simultaneously on a quantum machine? Why the big interest in quantum?

Giorgios: I would say that it’s essentially what Gustavo mentioned, which is that it’s a sample complexity. So normally, if you have a very simple case where you want to run a classical Monte Carlo with a hundred samples, your confidence interval for this estimation would scale proportionally to one over the square root of hundred, that’s 0.1. But essentially quantumly, using these algorithms that Gustavo mentioned and Iordanis has been developing for a long time now, you can compute this confidence interval in a scaling that goes one over hundred. So what are your steps? So for hundred steps that would be 0.01, so that’s 1%. So it’s essentially this improvement in accuracy that you obtain, which can also be converted to speed. However, there are several limitations as to… And that we maybe discuss at a certain point.

Yuval: Iordanis, I think you work with customers in a variety of fields, maybe not just finance, is Quantum Monte Carlo applicable to additional areas, or is it primarily used in the finance field?

Iordanis: It is a very basic mathematical way of trying to calculate something. It does have applications in many other domains, but I think the finance one is one where it’s really used all the time and very much. And this is really where we’re trying to see exactly how quantum can help or when quantum will help because, as Giorgios was saying, there are some caveats in this algorithm. The first one is that, in order to get the speed up from a million samples to a thousand samples, which is amazing if we manage to do it right, we need to have these quantum circuits, which are very deep, let’s call it a thousand times deep. And this necessitates very good quality qubits, which are not the qubits that we have today, but we hope to get at some point.

One of the ways to maybe try to overcome this obstacle is also based on some work that we did with Goldman Sachs that says a very simple thing, which is that if you want to get this thousand times faster algorithm, then you need this very deep thousand-deep circuit, but you can get it ten times speed up if you have a 10-deep circuit.

So you don’t have to wait until you have these very good quality qubits to get the entire speed up. You can start getting smaller speed-ups, which hopefully will still be relevant from a business and finance point of view, with much simpler quantum circuits. So this is one of the two reasons why we now think that Quantum Monte Carlo may not be so far ahead in the future, but a little bit closer, let’s say, mid-term. And the second quick comment I want to make, as we said, up to now, what we were counting is how many samples we need to take from this process, from this distribution. One part that we don’t talk about so much in the quantum cases is how difficult it’s to actually take one sample from this distribution. Because to run this quantum algorithm, we actually need to do this sampling in a superposition in a quantum way to fit it into the quantum algorithm.

And this is not obvious how to do, like some things, like geometric Brownian motion, very efficiently in the quantum case. And this is also a very interesting part, how do we make this part also most efficient? So then, we may be able to not only 10 samples, but a hundred samples in order to get a hundred times speed-up. So putting these two things together, doing these drastic modeling very fast, and saving all the number of samples that we need can actually, we think, provide an end-to-end quantum solution to the Monte Carlo pricing problem in the near future.

Yuval: Gustavo, I’m curious about your perspective on how practical it is today. Obviously, there are algorithms, like Shor’s algorithm, that promise amazing things but are impractical using today’s computers. You work at Moody’s. Are you using Quantum Monte Carlo in a production way today? Or, if not, how far do you think they are?

Gustavo: No, we’re not using it, and I don’t think we’re going to be using it anytime soon. That’s why in my opening, I mentioned maybe the medium term. A medium term, I’m talking probably five years, maybe ten years, something like that. The limitation right now is the hardware, as Iordanis was alluding to. We don’t have hardware with enough fidelity to be able to really run these algorithms in a naive way. So we need to find ways to extract as much as we can with the noisy hardware that we’re going to have for now, for the foreseeable future. And therefore, I’ll go developers, what they need to do is basically find ways to play or find ways to deal with the fact that we are going to have noisy hardware, we’re going to be losing some of the advantages that are promised by the theoretical algorithms that we know today, to find ways around those limitations and still extract as much of event as you can.

But I don’t think that, as I said, I don’t think it’s going to happen anytime soon. However, another thing that I would say is that no one runs this in the classical term. When we are around classical Monte Carlos, we never use the more naive way of doing this, or very rarely do we do the more naive way of running these algorithms. There are many improvements that you can do on the classical side, like important sampling or other variance reduction techniques, there are many of them, that actually give you huge advantages already on the classical naive Monte Carlo algorithms. 

Yuval: Giorgios, maybe you want to add to that and Iordanis after that. You mentioned some of the work you did with Goldman Sachs. Maybe you could give us a lay of the land of the improvements that you’ve seen in recent years?

Giorgios: Maybe some understanding first for the people who may not be so much aware of what we mean as quantum alternatives from Monte Carlo. And that essentially it’s not one algorithm, it’s essentially an amalgam of three, four maybe quantum subroutines, quantum algorithms. And while, indeed, it’s a case that if you want to run the original ideas, that they were developed in the 2000’s, it’s very difficult, as Iordanis has mentioned. He also knows very well, he’s one of the co-developers of improvements to these algorithms. For example, you can take some of these subroutines and replace them with something else. This something else could be some classical part, for example, so there’s some classical post-processing in the way. So in the way, as your algorithm develops, you can do some classical post-processing that may not be the best thing you could do, but it can definitely bring it to this timeframe that Gustavo mentioned, the five-year timeframe.

There are other variations that Iordanis is much better to explain. For example, his coprime quantum amplitude estimation, which is also very promising. So I think that it’s not all negative by any means, but at the same time, I totally agree with Gustavo in the fact that there is some pressure, especially when you don’t work in a university, to focus only on the algorithmic side, but quantum computing is three steps. It’s data loading, it’s compute, and it is readout. And maybe the community should also pay more attention to, for example, the problem that Iordanis briefly mentioned, which is the data loading, to do it more efficiently in order to speed up your computations and not bury your quadratic advantage in these overhead costs.

And the same, of course, goes into having accurate readouts. So when we develop these algorithms that are so fundamental and that will be of such importance for financial applications, we need to take a step back and look at the bigger picture as well because improvements in these sites, in the beginning, and the end of the algorithm, actually can imply that the algorithm can come closer in the timeframe for actual application.

Iordanis: I totally agree with Giorgios, that if we really want to see any of these advantages in practice, then we really need to take care of every single part of the workflow. And he was mentioning the data loading part of, again, how do we load these Brownian paths or things like that. And we have been working lately on this, it’s not an easy thing to do, but we do think that we can provide very shallow circuits, like logarithmic side circuits, for doing this data loading part, which in turn would mean that the entire depth of the circuit and the quality of the qubits can now be much less. So I would still agree that we are five years out to see Quantum Monte Carlo starting to really bring something. I wouldn’t put it closer than that. And to make it happen in five years, we do need the hardware to continue to develop and to advance as the roadmaps are saying.

We believe that they will continue to advance the way they’re thinking they can. And on the algorithmic side, as Giogios was mentioning, we do need to keep working on it. We shouldn’t think of quantum applications as something where the algorithms are already there, and we’re just waiting for the hardware. We really need to push the algorithms as much as we can, because as Gustavo was saying, we do have to do with very noisy and not very big quantum machines and we have to make the best out of them, and we really need to understand everything about them if we can actually make a good use of them and really bring some sort of real world advantage.

Yuval: You spoke about the hardware and the need to have deeper circuits. Would more qubits help? If you had a choice between a computer that can run a ten times deeper circuit or has ten times more qubits, would one be much better for quantum Monte Carling than the other?

Iordanis: It depends on the application. I think, on the Monte Carlo side, the quality of the qubits is more important than how many qubits. We do need more qubits than we have today, but I think the bottleneck is really that we need to do these deep computations, so we need more depth. Of course, you can get more depth by error-correcting your physical qubits. So if I had many physical qubits, I can use some of them to make fewer, better qubits let’s say. But we need a couple of orders of magnitudes better fidelity than we have today. It is in the roadmaps of the hardware providers for the next five years. So we are not expecting something out of the ordinary to happen. And we are continuing to reduce, as you said, how many qubits do we need? Or how good these qubits should be and still get a good result.

Gustavo: If I may add a little bit to Iordanis answer. I think it also depends on the particular problem in finance we’re talking about. Because Iordanis mentioned already that one of the uses of Quantum Monte Carlo would be in option pricing. And there, typically, you have a few sources of uncertainty in the problem. So we’ll have the price of the stock, maybe the volatility, you can also assume that follows some kind of stochastic process. You can add jumps to the process as well, so you might have a third source of uncertainty or stochasticity. And basically, for each of these sources of uncertainty, you would need more qubits to deal with. So now, if you move into a basket of options where you have 10, or 15 assets, then the problem multiplies by 15. And if you move into the realm of, for example, create portfolio, where Monte Carlos now deal with potentially millions of assets, then at the minimum you would need, even if you assume that it’s only one source of uncertainty per asset, you would need at least one million qubits, just to deal with uncertainty from the assets.

So it really depends on which application we’re thinking of that we can get away with just having fewer, better quality qubits than the more complex applications in finance that would actually need millions of good quality qubits.

Yuval: Now, I think that the original Monte Carlo was invented in Los Alamos as part of the Manhattan Project, and obviously, there were a lot of physicists and mathematicians there. When you guys work on these algorithms, even with that five-year horizon, what kind of people do you need on your team to improve Monte Carlo? Are they more finance people? Are they biologists? quantum scientists? Is there a particular type of person that you’re looking for to add to your team?

Gustavo: If I start, we’re typically looking for people with a technical education. So we’re looking for people with a Masters or PhDs in a quantitative subject. And it can actually vary from people from math, engineering, physics, and economics, but you can also go to the more specialist, financial mathematics type of expert. The advantage of having someone with more financial mathematics type of thing is that they know already the type of problems that we’re solving in finance. But at the end of the day, these quantitative skills are very easy to transfer from one domain to another. So if you have good mathematical skills, good analytical skills, combined with a decent amount of IT programming skills, then you’re in a good place. If you tick these boxes, it doesn’t matter so much if you’re coming from physics, mathematics, or financial mathematics, you will get there. Although, I would say that in the beginning, it’s always a steep learning curve to actually get to the level of knowledge about the financial industry that you would need to really be able to use those mathematical analytical skills to solve real-world problems in finance.

Iordanis: On our side, I think one important thing is the experience in building and developing new quantum algorithms. And this is something which is quite rare because we’re not a very big community, unfortunately. So we need people who are very confident with classical algorithms and quantum information so they can start developing new quantum algorithms. But I agree with Gustavo that the most impactful research happens when the expertise on the quantum side comes together with expertise from the finance side because we do need to understand what is really the bottleneck and what do we need to care about or care less about, and what is really the problem that we’re trying to solve. So understanding the financial part of this is really the difference between theoretical publications on quantum algorithms and real financial solutions with a quantum computer. So I think we need both.

Yuval: And Giorgios, do you have something to add?

Giorgios: I totally agree with both Iordanis and Gustavo. I would just like to add a small comment that, in my experience, I found particularly useful. I found particularly useful that interaction with people who understand stochastic mathematics, so stochastic differential equations, stochastic optimal control and real analysis. And the reason is that often when we’re at the initial stage of quantum algorithm development, sometimes we omit certain technical stuff about the speed-ups we give and such. However, it is the case that you have to take into account small technicalities. For example, what type of function, what distribution are you working with? There’s a difference between Hoelder continuity and Lifschitz continuity and this affects a lot, the bounds you give on your algorithms. There’s no single answer. So I believe that a lot of disciplines can contribute at different levels in algorithm development, but they all have a difficult learning curve.

Yuval: Very good. So as we get close to the end of our conversation, I want to ask you one of my favorite questions. If you could have dinner with one of the quantum greats, dead or alive (don’t say both dead and alive, no superposition here) whom would that person be? Maybe Iordanis, you’re first, and then Gustavo and then Giorgios, please.

Iordanis: Sure. I think I had dinner with many of the people who are alive, so I’m very happy with that, and I’m sure I’ll get some opportunity to meet with the rest of them. So I would go for someone who is dead for a long time, and I would say Niels Bohr. And the main reason is because Bohr actually had the right idea and the right intuition about quantum mechanics and there was a big fight between him and Heisenberg on what does quantum mechanics really mean. And even though we talk more about Heisenberg and Heisenberg’s uncertainty principles, it’s actually Bohr’s idea which is the one that we believe today. So I would really want to have dinner with Niels Bohr and understand how he came up with all these counterintuitive ideas that actually are the ones that we believe today.

Gustavo: I think I would want to have dinner with Paul Dirac. And the reason is that I have to say I admire the passion that he had, and the self-belief that he had found the right solution when solving for his famous equation that, in essence, predicted the antimatter and the positron was found shortly, or I think it was a couple of years after he actually published his findings. But the fact that the entire community was thinking that it cannot be right because his solutions predicted negative energy, and that didn’t make any sense. And he found the solutions that he had found so elegant that it had to be correct, and it was just a matter of finding the right interpretation. I think that was incredible. And I would love to hear from his own mouth how he actually coped with potential pressure around him to basically move away from that theory, and he managed to keep to his guns and really, basically dig up, and dig, and dig until he found a more plausible explanation. It turned out to be the right one and a massive new discovery for science.

Giorgios: I would go with Alan Turing because I think he’s a giant and he’s one of the people responsible for the age of information we live in right now, including quantum information. And the ideas he sparked from philosophy to pure math, to applied math, to engineering, and his contributions are enormous. The notion of computable functions, what is computable and what is not computable. And I would be very curious to ask him a lot of questions and get intuition and inspiration out of him. I think that would be my choice. Maybe I would like to invite John von Neumann as well, if I were allowed.

Yuval: Excellent. And Gustavo, I had a previous guest that said that Dirac was not a very good conversationalist. He spoke very little, but when he spoke, it had tremendous meaning. So I think my previous guest said, “I’d love to Dirac, but maybe we also have Feynman or someone just to keep everyone engaged.”

Gustavo: Apparently, once Paul said to Feynman was, “Do you have an equation named after yourself?” That Feynman didn’t, so that was the extent of the conversation. I think it could be quite awkward, the dinner with Paul Dirac, you’re absolutely right, but it’s still one of my absolute idols. So I would love to spend the time, even if it was a very one-way conversation in the end.

Yuval: Perfect. So gentlemen, thank you so much for joining me today. What’s the best way for people to get in touch with you to learn more about your work? Maybe Gustavo, and then Iordanis, and then Giorgios?

Gustavo: Anyone that is interested in quantum opinion and how we are thinking about utilizing quantum computing at Moody’s can get in contact with me, and I can leave my contact details with you, and if anyone wants to reach out, can do. I’m also on LinkedIn, so anyone that is interested in knowing more about what we do, can send me a contact request I’d be happy to accept, and we can have a conversation.

Iordanis: The same, through LinkedIn or through the QC Ware website. We are collaborating with many people in finance, so we are always happy to talk more about these topics and work together to make it happen.

Giorgios: For me, it’s the same. You can find me on LinkedIn. And if you want to know anything about how we use quantum in HSBC, and what we’re interested in, what we’re doing, I’ll be happy to provide answers and engage in a nice discussion. And also, of course, through arXiv papers and such. So there are ways. Absolutely.

Yuval: Excellent. Well, thank you so much for joining me today.

Giorgios: Thank you.

Gustavo: Thank you.

Yuval Boger is an executive working at the intersection of quantum technology and business. Known as the “Superposition Guy” as well as the original “Qubit Guy,” he can be reached on LinkedIn or at this email.

January 3, 2023