Summary
Aurelie Helouis, founder and CEO of InfinityQ Technologies, developer of an analog quantum computer, is interviewed by Yuval Boger. Aurelie and Yuval talk about the types of problems that such a computer can solve, a unique application in the world of online gaming, and much more.
Key takeaways
- InfinityQ built an analog quantum computer aimed only at optimization problems, skipping the gate model that most quantum hardware companies pursue.
- Instead of encoding problems into qubits and gates, the machine represents optimization problems as continuous analog states that settle into a solution.
- One real customer use case is online gaming, where the system is applied to matchmaking type optimization problems in near real time.
- Aurelie's argument is that narrowing the hardware to optimization only lets it deliver practical results now, rather than waiting for general purpose gate based quantum computers to mature.
Read the full transcript on the Quantum Computing Report site
Transcript
Yuval Boger: Hello, Aurélie, and thank you very much for joining me today.
Aurélie Hélouis: Hi, Yuval. How are you? And thank you so much for the invitation. I'm super happy to be here today.
Yuval: My pleasure. So who are you and what do you do?
Aurélie: Oh, my God. So, I'm the CEO and co-founder of InfinityQ, a company in quantum-inspired computing. A little bit of my background: I was a naval aviation officer in the French Navy for 16 years, in France.
I had several responsibilities, especially on board the aircraft carrier, and I was in charge of the maintenance of the Rafale, the jet aircraft. So, a very interesting journey, and I had the opportunity to travel all around the world. And then I had a degree in IT, and at some point, after becoming the CIO of the naval air station, I moved to Montreal to do my MBA at McGill, and started working in a big company, Pratt & Whitney. And then at some point I discovered my passion for entrepreneurship.
And, you know, I quit my job and became an entrepreneur, and I've been an entrepreneur since.
Yuval: Excellent. So what is quantum-inspired computing? What is the product that your company is developing?
Aurélie: That's a very good question. So, to go back to the very beginning, I was working at Mila, the AI research center here in Montreal, the research center of Yoshua Bengio.
And I met a professor who wrote a new formulation of quantum computing, and we were wondering if there were ways to obtain the same type of results — maybe not that revolutionary, but at least enough to provide a significant speedup and the ability to solve new problems — without having the burden of, you know, building quantum computers, manipulating small particles, photons or atoms, building cryogenics, having all these complex materials and equipment.
So we looked at quantum-inspired technologies, and we found that developing an Ising machine was a paradigm shift in computing and a good way to solve new types of problems, these NP-hard problems. And there was a possibility to use hardware such as FPGAs, which we can configure ourselves, to solve the Ising model.
So that's where the idea comes from.
Yuval: Very good. So what kind of problems can the Ising machine solve? Could you give me perhaps a sense of that?
Aurélie: So, the beauty of an Ising machine is that it can solve all these combinatorial optimization problems, and any NP-hard problem can be translated into an Ising model.
Combinatorial optimization problems are everywhere, especially in logistics. Everybody knows the famous traveling salesperson problem, or, you know, pathfinding, delivery optimization, these types of problems. And right now we are also looking at gaming.
As I mentioned, pathfinding and the interaction between non-playing characters in online games is something that is very interesting. So for us, it's a very interesting and also challenging moment, because we have to use our technology to solve these problems and identify these problems in the real world.
So, yeah, that's the "techno push" problem, but I think it's very interesting because we have to open up, you know, to think a little bit out of the box, and to think a little bit beyond what exists.
Yuval: How does the performance of the Ising machine compare to classical computers, or maybe to a D-Wave machine, or basically any other approach to solving the same kind of problems?
Aurélie: That's a very, very good question. Actually, the performance is a bit better for these NP problems, because unless you have problems such as the TSP, the traveling salesperson problem, which is very well studied and has specific algorithms — for the rest, if you don't have a specific algorithm, classical computers are very, very bad at it.
And compared to the others, we do our benchmarks against, you know, the annealers. The performance is better as we increase the inter-complexity. We are still at the beginning, and we will publish a white paper by the end of the year to give more detail on that.
Yuval: Do you feel that this approach is a short-term solution to solving optimization problems? What I mean by that is, in a few years we expect quantum computers to be better, quieter, with more qubits and so on. Will there still be a need for an approach such as yours?
Aurélie: I think what is really interesting with our approach is the fact that it's easy to code.
There is no extra work to create, you know, quantum circuits. The simplicity is very important for customers because at the end of the day — and I think sometimes when you love technology you forget it — customers want something easy to integrate, easy to use, and they don't want too much disruption, otherwise adoption won't happen.
So what I think is, once customers adopt our technology, it will require a big revolution for them to change. So I think we have some time before that happens.
Yuval: And I assume you're selling a box, right? Is that a big box, a small box? Is that a cloud-hosted box?
Is that something that someone would put on premise? How would you deploy it?
Aurélie: So right now, we developed a local cloud, so it's more on premise. And we are working with Amazon and Azure to provide access to the machine via their cloud. And the idea is to have our own cloud totally ready in 2023.
So, yeah, we have three approaches.
Yuval: What stage is the company in? How many people do you have, roughly? When will a product be released? What can you tell me about that?
Aurélie: So we incorporated in January 2020, so almost two and a half years ago. We are around 10 people, so still a small team.
But we already have a working prototype, and we're working on POCs with different companies. So, still early stage, but more in a growth phase.
Yuval: When you listen to quantum computing vendors, the TSP, the traveling salesperson problem, comes up very often. But I must admit that I haven't heard too many people speak about gaming.
How did that application come about for your company?
Aurélie: So first, we were looking at different industries where our technology could be a good fit. We did an analysis, a kind of SWOT analysis with opportunities, and we discovered that gaming could be a good fit, especially, as I mentioned, with online games.
And it's becoming more and more interesting, and it's, I would say, the future. Especially after COVID, a lot of people became gamers. And also with the metaverse, it opens a lot of opportunities. And I think what is great with gaming is that they really need solutions. They won't buy a consulting agreement to explain what can be done in five years, for example.
They really want a working solution to help them solve their problem today. So you can't really cheat with them. It's like, either you provide something, or you come back with a solution. And another good thing: Montreal is a great city for gaming. We have a lot of studios, a lot of big companies in gaming.
So for us, it's also a good way to find talent and potential customers.
Yuval: Let's see if I understood the application. So if you and I are gamers, then we have a character, and that character travels through, you know, a virtual world. But then there are other non-playing characters, right? Crowds or, you know, villains, and they also need to move through the space.
And today it sounds like their movements are not very realistic, and so you're saying that your machine would help figure out the best path for a non-playing character. Is that about right?
Aurélie: Exactly this. And not only optimization of the path, but also optimization of their interactions — making interactions smarter.
So that's what we are working on right now.
Yuval: And that sounds like almost a real-time activity, right? I'm not gonna wait five minutes to figure out how a character moves through a virtual universe. So is the machine built to provide very fast answers?
Aurélie: Yeah, that's our goal. And what we see is also for the future, you know — with the metaverse, we will need more and more real-time environments.
So that's kind of the mindset we have right now: improving interaction, and at some point we will be able to apply that to the metaverse.
Yuval: You spoke earlier about cloud and on-premise deployment. Are there any specific requirements for the machine? Does it need any special cooling, or lots of electricity, anything in particular?
Aurélie: Oh, not at all. It's very easy. Right now we developed a solution on FPGAs. The beauty of FPGAs is that the hardware is off the shelf, but you configure everything yourself, so you customize your own hardware. And for us, it's a great opportunity. It's cheaper than building your own chip.
But at some point we aim to build our own chip. So right now we develop FPGA solutions, and then at some point we will be able to have an ASIC. That's our timeline and our roadmap.
Yuval: And going back to TSP, roughly how many cities can you solve today, and what do you expect to be able to solve in the near future?
Aurélie: So that's a good question. We have a new approach, a digital approach. We shifted to a digital approach at the beginning of the year to ensure better scalability. So right now, I think we can solve around 100 cities, something like that. And, as I mentioned, we are working on bigger FPGAs to be able to scale more rapidly.
Yuval: Tell me a little bit about the composition of your team. You said about 10 people. I assume they don't all need to be quantum scientists, or maybe none of them are quantum scientists. How is the team made up?
Aurélie: So we have a professor from Polytechnique, which is a university here in Montreal.
He's a professor in computer science. We have a couple of interns, actually, with a background in software development, a cloud expert, a hardware expert with FPGA experience, the software developers I already mentioned, and mathematicians.
So actually, yeah, no quantum experts per se. We have a consultant who is a quantum expert, but the team itself is really more computer scientists.
Yuval: And when you go to customers, do you basically say, "Look, you have these problems. Classical computers are gonna have a hard time solving them.
You don't want to wait five years for quantum computers. Here's a solution that you could deploy in the near future"? Would that roughly be the sales pitch?
Aurélie: Exactly. But as I mentioned, the good thing is we can explain what we can do, and then we listen to their problems, and then we fill the gap. We explain that the solution can solve part of their problem, or the whole problem, and then we work together to create the solution.
So that's the approach we use right now to work on POCs. And it also helps to educate people, because they are not very familiar with these combinatorial optimization problems. They know there are problems in the real world, but they usually don't know what the mathematical model underneath is.
So we are there to help them.
Yuval: Are you able to share anything about pricing? Is it per machine? Is it per shot or per thousands of shots? How does pricing work?
Aurélie: So right now, for pricing, we work with services, so it's more project-based pricing. But we aim to sell computing time, and at some point, once we develop the infinityQube, to sell licensing for the software.
So those are kind of the three approaches we have in mind.
Yuval: Excellent. So how can people get in touch with you to learn more about your work?
Aurélie: So please contact me — send me an email at aurelie@infinityq.tech, or contact me on LinkedIn, and I would be happy to get back to you, and we can discuss with the rest of the team.
Yuval: Excellent. Aurélie, thank you so much for joining me today.
Aurélie: Thank you, Yuval, for the conversation. It was very interesting, and I hope you enjoyed the chat.
