Episode 111: Quantum Computing Is Here

Media Thumbnail
  • 0.5
  • 1
  • 1.25
  • 1.5
  • 1.75
  • 2
This is a podcast episode titled, Episode 111: Quantum Computing Is Here. The summary for this episode is: <p>In previous episodes of the Georgian Impact podcast, we’ve learned about the need to get ahead of the impact quantum computers can have on breaking encryption with Mike Brown of ISARA in episode 97. We also got a great overview of different quantum technologies from Vlad Gheorghiu in episode 104.&nbsp;</p><p><br></p><p>In this, our third podcast on Quantum Computing, Jon Prial welcomes Christian Weedbrook, CEO and founder of Xanadu. Xanadu is a photonic quantum computing company and a recent addition to the Georgian Partners portfolio. They discuss how quantum will improve financial analysis, help to make more efficient batteries, and optimize traffic congestion...and why the time for adoption is now.</p><p><br></p><p><strong>You’ll hear about:&nbsp;</strong></p><p><br></p><ul><li>The advantages of photonic quantum computing over other approaches</li><li>Use cases for quantum&nbsp;</li><li>Why Xanadu is building an open-source machine learning library to help the community adopt quantum</li><li>What you should be doing now</li></ul><p>&nbsp;</p>

Jon Prial: Quantum computing. In previous podcasts, we've learned about the need to get ahead of the impact that quantum computers can have on breaking encryption with Mike Brown of ISARA in our episode number 97. We've also got a great overview of differing quantum technologies from Vlad Georgiev you in episode 104. As there's so much new terminologies running quantum, if you haven't listened to these two episodes, I recommend starting there. Today for our third podcast on quantum computing, I'm delighted to welcome Christian Weedbrook, CEO and founder of Xanadu. Xanadu is a photonic quantum computing company, which came out of CDL, an incubator in Toronto. And they're a recent addition to the Georgian Partners Portfolio. We'll discuss the advantages of photonic quantum computing over other approaches and where, use cases for quantum. And why Xanadu is building an open source machine learning library, to help the community adopt quantum. It's a great episode, so stick around. I'm Jon Prial and welcome to the Georgian Impact Podcast. So, this is our third podcast on quantum computing, but I'd like to get from you, just a basic warmup, just why you see quantum as an important next step for computing?

Christian Weedbrook: Yeah, it's really a significant, next step for computing. We really haven't seen anything like quantum computing, I would say, since the beginning of computing, or since the mainframes really have come out and you saw the personal computer revolution of'70s and'80s. And people often go back to, there's the magic of it. When you want to understand, how is this machine powered? It's powered by the laws of quantum physics. And that's really exciting, there's some weird stuff happening there. Things that we don't normally see on our everyday scale. So, that's the fascinating side of it. But, what that leads to, is really the idea that you can have these incredibly powerful computers powered by the laws of quantum physics. And by powerful, I mean, there's usually two categories you can think of. Problems that may take a long time with traditional computers, and problems that have always been out of reach for any computer, even computers that you would run for thousands of years. And these, are typically known as intractable problems. So, the idea now is these things can actually be opened up, is quite incredible. And so, you can think of one quantum chip, potentially for certain problems, being equated to millions of normal chips.

Jon Prial: Wow. Now, we often talk about density as the key measure of quantum and how far along different companies get, across the different types of technologies. What's your sense of how well things with photonic, are progressing on density?

Christian Weedbrook: Yeah, one of the great things about us doing photonics, we believe is the idea that we didn't have to create the telecommunication industry and all the optical components to go with it. So, the optical telecommunication industry for say, internet fiber optics and so forth has been around for decades now. So, what that means is, we can actually stand on the shoulder of that history essentially, which billions of dollars has been put into it. So, what that means from a practical point of view, an everyday point of view, is we can call up companies that sell lasers, shell beam splitters, phase shifters, detectors, and all this optical equipment, and order it and get it sent to us. And that's something that thankfully, we didn't have to do ourselves. And also, we work with multiple foundries around the world because we have these photonic chips. Everything is miniaturized and put on chips and we can work with standard foundries. So, people label that as CMOS compatible-

Jon Prial: Oh, wow. Interesting-

Christian Weedbrook: Essentially, it means... Yeah, it's really fascinating, because some of these foundries cost billions of dollars, particularly when you want to scale up things, we often work with maybe foundries that are still very expensive, but we started off with prototyping, and then we go to the more advanced ones and then slowly, one day we'll be at the mass production side of things. So, the ability that we can call up a Foundry is very important too, and say," Well, here are our designs, if you can fabricate them for us and send them back." So, we can really leverage the history here. That's why photonics is one of the great points about it.

Jon Prial: Is there a bit of a Moore's law working? Obviously the chip density is something in the old- style chips. People talk about density of chips and we talk about number of cubics in a chip. How do you see the evolution and growth of cubits against, I guess, a timeline?

Christian Weedbrook: It's hard to say. So, one thing we can potentially say is that, once you reach 50 to a 100 cubits on chip, for certain problems, you're already bypassing the ability of whole data centers, in some sense, millions of chips. So, you're already with a small number of cubits, replacing something that was obeying Moore's law for decades. So, we leap- frog ahead of that, which is pretty cool.

Jon Prial: I see, yup.

Christian Weedbrook: But, having said that, you can think of going from hundreds of cubits, to thousands, and to millions and things of that nature. It's unclear how that will progress, whether there'll be an equivalent to a Moore's law for quantum. There perhaps can be, but we're in the initial stages now, where the next three to five years is really about building out the fundamental building blocks, rather than scaling up too fast. Still entering periods where we can actually solve important problems, but nothing like the doubling every 18 months or so.

Jon Prial: Got it. And it's probably not a fair parallel, and you look me in the eye, I'm a little bit of an older guy, it was before my time. Well, I started in computers in the'70s. Prior to my time people were wiring with little yellow wires.

Christian Weedbrook: Mm- hmm( affirmative).

Jon Prial: They weren't thinking yet about chip density. So, it's interesting that you're still building the scaffolding around all the pieces to make it operate before you can think about how the density really is the most important measure. There's a lot of other, I guess, scaffolding is a good word around that. Is that a fair categorization?

Christian Weedbrook: I think so. And one aspect about photonics, getting back to the photonics side of things, which is really interesting and very unique to photonics is, as you mentioned, you're thinking about scaling up with, say the number of transistors and squeezing as much computational power on one chip as possible. With photonics, what's very interesting is we have a couple of ways we could build a photonic quantum computer and we're investing in both approaches. The other one that's very interesting, that I haven't mentioned is this idea of using photons. They're not in any one place, meaning that they propagate and we call them flying cubits. And what's cool about that is you can scale up to in principle, millions. But, not actually needing a million quantum states existing at the same time. So, one way to think about, is you can create a small subset, let's say four to six cubits, and then you compute in time. So, instead of spatially laying out the chip, you have a small chip propagate through and then loop back on itself. And so, through that, you can actually show that's equivalent to having, say a million on one chip, where you've only got a very small subset. And that's because, normal electronics where it's classic or quantum, it's fixed on a chip, it's not moving anywhere, so to speak. But, with these flying cubits, you can actually propagate or compute in time.

Jon Prial: I think there's at least a t- shirt in there with a flying cubit on it.

Christian Weedbrook: Yeah, maybe something like,"I don't give a flying cubit."

Jon Prial: There you go. So, you had mentioned fiber and that's networking. So, let me do a little networking thing, 5G. So, 5G is either really important or it's really overhyped. More likely, it's the classic IT answer, and it's going to be all of the above or both of the above. Now I get that I can download a movie faster with 5G. Fine, fine, fine. But to me, 5G matters where load latency creates an opportunity for applications that never existed before. Remote telemedicine with surgeons working with robots through the network. If there's any latency there, the patient is going to die, perhaps. So, lag matters, speed really matters. Is there a parallel with quantum that you see and take me through some use cases as to where you see these applications of quantum coming first.

Christian Weedbrook: First off, this may sound obvious, but you want problems that have not already been solved by traditional computers. You just don't want to create something, that's going to take a lot of time and money, and it does just as well, or maybe even slightly better. Whenever you have a new technology, particularly if you want it to take off, it has to be a couple of orders of magnitude better in principle, for that to be a need for adoption. And the system, in addition to that, should be something that's very complex. So, as the number of aspects of the system increases, hopefully exponentially, it just becomes too much to manage. And so, putting those things together, the common answers to the applications in terms of verticals, would be finance. Going back to what I mentioned, then you can have a book or a portfolio of stocks, and as you add more and more, it doesn't scale very well. It doesn't scale linearly. So, it may scale quadratically or exponentially. And in that case, that's ideal for quantum. And also, you'd like to have a third characteristic where, if you move the needle a little bit in these verticals, it's very important. It's worth a lot of money in the case of finance.

Jon Prial: But, this is not for normal human beings, this is for large portfolios. Big companies managing large amounts of data, that are looking to apply some ML or AI to how they manage the portfolio.

Christian Weedbrook: Yeah, exactly, right? A more specific example, we just announced our work that we worked on with Scotiabank in Beaumont, and this is a good example. So, there's certain aspects of their computations in the bank, specifically capital markets, where they have to run it overnight. So, there's a problem. Someone pushes a button when they leave work and it starts churning and computing overnight, and they come back in the morning and they have the answer. And that's something that due to the complexity of the problem, it takes a long time to actually find the answer. In this case, six to eight hours, for instance. And so, a quantum computer can come in, and provided you have the right type of quantum computer, you can actually do this in minutes now. So, now that's just a huge game changer. So, you find these things, pharma and drug discovery is another one, where the simulations can take six months or more to do. So, if you can reduce the time it takes, particularly in a way that a quantum computer can, then this is going to be very important.

Jon Prial: In this case, the quantum chip really is, if it's fair to call it, a sideboard or a co- processor. So, you know what needs to be done, you're Scotiabank, or you're a pharma company and you somewhat kick off this process and the system knows to spin it off to this chip, so that everything is still stays in place. And this is an add- on for particular application and usage. Is that fair?

Christian Weedbrook: Yeah, exactly right. We don't see quantum replacing everything, in terms of computing. You see it as a co- processor, as you mentioned, or an accelerator, particularly in the early days. There's certain things like loading email, loading Facebook and things like that, where it's not believed a quantum computer is suited for. Use your normal chips for that.

Jon Prial: You don't know how many friends I might have, no.

Christian Weedbrook: I look at you and I think you have quite a few.

Jon Prial: So, we've got finance, which is pretty cool. Absolutely, the drug simulation, these are obviously high computing workloads. What else might be something that you guys are thinking about?

Christian Weedbrook: Yeah. So those two, along with material design, for instance, that would fall under the area also of quantum chemistry. Designing new batteries, more efficient batteries, that's a usual suspect in use cases. Another area would be logistics. So, traveling salesman is a great problem. Working with, for instance, autonomous vehicle companies. Uber, for instance, last year, I think did a billion trips. So, if you can somehow optimize the routes that Uber drivers or Lyft drivers, whoever it is takes, then first off you're reducing the gas consumption. But, also it's a great thing for the environment. This has a characteristic also, as the number of stops increases, it scales very badly as well. So, these are things that do take a lot of computational power and effort.

Jon Prial: And just to make that clear, this is really important. Today, if I use my GPS and there's a traffic issue and they want to take you off the highway, seven zillion cars go on this little tiny road with three traffic lights. You're talking about optimizing all the cars and they all go to different places. So, they can all get off the highway, but they all don't go to the same little tiny road with three traffic lights-

Christian Weedbrook: For instance, that's right. That's right.

Jon Prial: Okay.

Christian Weedbrook: Optimizing traffic congestion, is definitely one of the aspects of this. And again, it's an important problem, but it also, as the number of cars or situations increase, it becomes a very tough problem, a very complex problem. And it's outside of the usual use case of traditional computers.

Jon Prial: Very cool. So, we talk about these different industries and different applications. Let me build up to a question for you and how we build that. So, we talked, you mentioned main mainframes and the PC explosion. So, we know about chips and they are Intel chips and there were risks chips, or I guess there are risk chips that they original running Unix and there were mainframe chips. And I think that chips became a little less relevant and operating systems began to cross over. Linux ran on different platforms and the like, but obviously what happened to the world of IT, we began to generalize. We began to build some interfaces, make systems more consumable. And so, therefore middleware showed up. We have some interface tools. So, I guess it's a two- part question, what's Xanadu doing about it? And what is it with you in the Beatles?

Christian Weedbrook: I see the connection there. Well, I think there was a time where software was King and hardware, obviously it still existed, but it wasn't as crucial or as important. And really in the last, five or so years, hardware's come back in a big way. If you just look at the number of startups and companies, and I don't even mean a quantum hardware companies. You have a lot of special purpose or AI accelerator chip companies. Actually, some of them using photonics, others electronics, nothing to do with quantum. But, you see with a Nvidia, for instance. People just found use cases in terms of Bitcoin mining and also graphics cards and things like that. And then, you throw a quantum on top of things as well. Where, it's our belief. You really need to the full stack, you need to do hardware and you need to do software in order to create the best product really for customers. Now, the second, which is to me more important is about the Beatles. It's definitely my favorite band. And I have been since I was very young. And they've been inspiring for me and it was cool. Let's bring out some of the famous songs like Penny Lane and Strawberry Fields and label them with products. Xanadu itself, is named after the song from the movie, Xanadu. So, there's a lot of, I guess, pop cultural references there.

Jon Prial: So, Strawberry Fields is an open source project. So, obviously you mentioned you want to control the stack, but nowadays control is quite different than what it was back in the day is a proprietary operating systems, through proprietary interfaces. So, you're doing some open source work with Strawberry Fields. Tell me more about how that's going to work? And what you're driving forward with that?

Christian Weedbrook: We decided nearly, from day one to keep our software open source. And really the reason is, where are we offering the most uniqueness? And it's in our photonic chips. So, everything is really built around, how do we drive customers back to using our chips? And what better way? Or one of the better ways, is to really have software that is really amazing. And people love that and naturally feeds them back to our chips. So, from a very high level, as you mentioned, the two products, Strawberry Fields is actually a photonic simulator. It's very unique, there's none like that anywhere in the world. It's very much tailored to our photonic quantum computer. At the very top, you actually have an interactive device where you can drag and drop gates onto a circuit and press run. And it'll output the circuit or the app for the circuit. And IBM has one like that, and we're the only two in the world that has that. And Strawberry Fields will be used as a way to connect to our cloud platform, when you want to connect to the cloud, which is the backbone, which has that hardware. So, that's really key. On the other side, Penny Lane. And by the way, both Strawberry Fields and Penny Lane were released on the same 45 disc and'66. So, there's also another connection there, that's key. And what's really great about Penny Lane. The first thing, when I think about Penny Lane, I think agnostic. So, it actually runs not only on our hardware, but every other hardware that's out there. So the

Jon Prial: And that's the machine learning platform, correct?

Christian Weedbrook: Yeah, exactly. So, we think of it as the TensorFlow or PyTorch for quantum computers.

Jon Prial: Great.

Christian Weedbrook: It's a way to abstract away all the potential challenges or difficulties in understanding the hardware for developers.

Jon Prial: So, if a CEO thinks about this, they actually should think about Penny Lane as a layer of machine learning, and then they could begin to see what it might do on a photonic chip with Strawberry Fields as a simulator.

Christian Weedbrook: That's correct, that's right.

Jon Prial: And a CEO who would think about this really, as I listened to your use cases is really, although they're related, I categorize the two places. One is just raw compute power, because you'll get more compute power of it. But, obviously the number one use case against that might well be optimization cases.

Christian Weedbrook: Mm-hmm(affirmative).

Jon Prial: So, if you're a CEO, that's beginning to get a huge AWS bill, or whatever cloud service they're using, to do some work. If they see a lot of number- crunching, is it time to think about how I might not optimize that and move over to quantum? Is it time? Should they call you? Should they start to play around? What's the timeframe for a CEO who is recognizing that there is something coming down the pipe, that could impact his or her business.

Christian Weedbrook: I think the time is now. And well, I should say, it depends on what industry you're in. So, if you think of the industries I mentioned before, banking, finance, hedge funds, for instance, and pharma and drug discoveries. These are industries that always try to stay on top of the latest technological developments. So, for sure in those industries, it's very important to start understanding this now. Our works with the banks here in Canada, have so far shown that it actually takes a while to really connect the dots. That's really a big thing because this shows you, when you have such a paradigm shifting technology, it's not just the same as Intel releasing or Nvidia releasing the latest chip and people know what to do with it. It's very different. Even if we had the most powerful type of quantum computer today, and we released it, people wouldn't really know what to do with it. There's still a lot of help that needs professional services and things like that. So, what we've done at Xanadu, we've actually started a Quantum Readiness Program. And essentially, it's for CEOs, as you mentioned, big companies in those two industries that are like," Well, what is quantum computing? How do I get into it? Is now the right time? What's this roadmap look like, and we can really educate the customer on those dances.

Jon Prial: When I had a product business way back when, we always had a handful of Skunkworks projects going on, that we were just really working on... So, we knew what was going to happen. It sounds like it is time, people shouldn't assume," I'll wait two years down the line to pick up the phone." If they think they're getting close, they're not going to be on the quantum bandwagon tomorrow, but they should start thinking about it, maybe do a small Skunkworks project, start to understand better. I like the fact that you could consult them on how to fit the pieces together and recognize that 12, 18 months down the line. This could be real and you don't want to be caught behind the eight ball.

Christian Weedbrook: For sure, and you can imagine saying those two industries, finance and farmer, imagine the competitive advantage you can have by starting these things early. I mean, it's potentially huge. If quantum lives up to its potential, then these will be some of the most dramatic, competitive advantages that we've ever seen.

Jon Prial: If that doesn't get someone's attention, I don't know what will, what a great interview. Thank you so much. Thanks for taking the time to be with us.

Christian Weedbrook: No problem. Thank you for inviting me.


In previous episodes of the Georgian Impact podcast, we’ve learned about the need to get ahead of the impact quantum computers can have on breaking encryption with Mike Brown of ISARA in episode 97. We also got a great overview of different quantum technologies from Vlad Gheorghiu in episode 104. 

In this, our third podcast on Quantum Computing, Jon Prial welcomes Christian Weedbrook, CEO and founder of Xanadu. Xanadu is a photonic quantum computing company and a recent addition to the Georgian Partners portfolio. They discuss how quantum will improve financial analysis, help to make more efficient batteries, and optimize traffic congestion...and why the time for adoption is now.

You’ll hear about: 

  • The advantages of photonic quantum computing over other approaches
  • Use cases for quantum 
  • Why Xanadu is building an open-source machine learning library to help the community adopt quantum
  • What you should be doing now