Richard Murray, CEO and co-founder of Orca Computing, discusses the company’s focus on building photonic quantum computers using existing telecom components and optical fibers. Orca differentiates itself by leveraging its origins from a quantum networking research group at the University of Oxford, emphasizing practical near-term applications in machine learning and optimization. Richard highlights the unique challenges of photonic quantum computing, such as manipulating single photons for qubit gates, and the company’s efforts in pioneering multiplexing techniques to scale their systems.
Full Transcript
Yuval Boger: Hello Richard, thank you for joining me today.
Richard Murray: Thank you, pleasure to be here.
Yuval: So who are you and what do you do?
Richard: Great question, my name is Richard Murray and I’m the chief executive and co-founder of Orca Computing.
Yuval: And what does Orca do?
Richard: So yeah, so Orca builds photonic quantum computers. So we’re a spin-off from the University of Oxford, although I’m not a spin-off, I joined from a different direction. And we build photonic quantum computers, so we assemble systems from a lot of optical fibre, a lot of existing telecoms components, a lot of optical switches and things that already exist, into a product, which I think is a little bit unlike your normal quantum computing company. And then we have a lot of activity on the software and application side. So we also spend a lot of our time looking at how those photonic quantum computers can be used to advance machine learning. And we also spend a lot of our time building error correction codes that are suited for photonic systems in particular. So we do a whole bunch of different things and my job is to try and keep it all together somehow.
Yuval: For the benefits to those that don’t know yet, what is the difference between a photonic quantum computer and a regular quantum computer, say a superconducting or a trapped ion one?
Richard: Yeah, good question. So I mean it relates to the thing, the object that carries the quantum information. And in the case of photonic quantum computers, or in fact our particular version of photonic quantum computers, we build them using single photons, so single units of light. So bearing in mind that a laser beam produces trillions upon trillions of photons at a time. If you think about what we do to isolate just one of those photons, one photon at a time, it’s quite remarkable. And in comparison, you have superconducting based on superconducting circuits and ions based on single ions. Photons are very different in many ways. For one good reason that they don’t stand still. So our version of quantum computing using photons is very different in that we have our qubits, our single photons flying through the system at all times at the speed of light. So that’s one major difference. The other quite significant difference is, well it’s a benefit but also a difference, single photons find it very hard to interact with, well, anything and also each other. So you take two single photons and you try and perform maybe a two-qubit gate and interaction between those photons is very very hard, they normally pass straight through each other. So we have to play several clever tricks to obviously get them to interfere and to create entanglement, and those are based on clever schemes that we can get into, I suppose on measurement based quantum computing and things like that, using measurement for your photons to create an interaction between them and a superposition state.
Yuval: I can think of several other photonic quantum computer companies. What’s unique about Orca in that subsegment?
Richard: Yeah, well I think one is something of a subtlety, but I think we push much more the reliance on optical fiber and telecoms. So I think other quantum computing platforms tend to start off looking at integrated circuits as their platform. By the way, I think many of those guys are now pivoting to include optical fiber, it’s really the only way you can have the ability to delay your photons for a very long time without losing them. But I think other platforms start off looking at photonic integrated circuits and adapt to include optical fiber. I think we started life very differently, starting by looking at optical fiber as the basis of our platform, and then maybe adapting to now include more and more on the integrated photonic side.
So that’s the very top level view, I suppose. So our use of optical fiber and telecoms I think is much more than the other variants. And what’s maybe interesting to your audience is that came about through the way we were spawned. So the research group that we spun out from at the University of Oxford, and they were building networks. They were a quantum networking group predominantly. And it was for that reason that they started life by looking at photonic quantum computing essentially as a network. I guess that leads me on to a slightly more subtle, more technical version of specifically what Orca does. So within telecoms, you have the ability to control photons in many ways. What is most intuitive is to imagine controlling single photons across different spatial degrees of freedom.
So essentially lots of switches that switch the photons one direction or another one or of course in superposition of those two. That’s pretty much the basis of a lot of the approaches taken by other photonic quantum computing companies. And it’s a good one. But we also recognize with photonics you have this ability to control photons across time and frequency as well. So we like to imagine that we can scale up our systems much better by allowing all the things that you would do to be done spatially, but also across the frequency space. So imagine across a number of maybe 10 or 100 different frequency modes and a similar number of time modes. And this is quite a critical technology to telecoms. It’s something called multiplexing, which I’m sure many people have heard of. And it’s sort of surprising that the idea of multiplexing with single photons with quantum light has always been a bit of a challenge. So we have a suite of components that allow us to perform multiplexing, the manipulation of single photons, without inducing excessive loss or noise in the system.
So I’d say it’s for those two reasons we’re different. And then the third is more on the application side. I mean, I will say slightly controversially, we are all in on sort of short-term quantum systems being useful. Not that we think that’s the end, I mean, we also, as I said, have an error correction team, but we really spend a lot more time looking for applications of small-scale quantum systems as quickly as possible in areas like machine learning and optimization. Whereas I think, by and large, a lot of the other photonic quantum computing companies have decided or come up with some evidence, which maybe sort of has convinced them for some reason that they’re going to bypass that loop and go just for the route for building fully error corrected large scale systems.
Yuval: Ignoring gate model technologies, if I were to compare two different gate model quantum computers, we could say, OK, how many qubits are there? What’s the connectivity between them? What’s the single qubit and two-qubit gate fidelity? What’s the depth of the circuit that you could reasonably execute? And so ignoring the software, how would I compare different optical computers?
Richard: Yeah. So ultimately, those questions can be asked and are comparable. So, you know, our version of a universal quantum computer has all of those metrics the same as everyone else. I think that it’s reasonable in time to ask those questions. I think the thing which makes it a difficult question to ask and is an important point to stress before a photonics platform answers them is, in some ways, the nature of photonics makes, for example, a two-qubit gate is very challenging.
And it involves a lot of measurement based quantum computing. It also involves a lot of extra activities, measurement, feed forward and those types of things, which you wouldn’t require in your normal two-qubit gate interaction between, say, two superconducting qubits. So the reason it’s sort of difficult and shouldn’t be a direct comparison today is some things that you would find natural to do with a superconducting qubit are naturally hard for a photonic qubit. So I believe that’s why very few companies have announced their two-qubit gate fidelity in the world of photonic quantum computing.
But that said, even though that’s hard, some other things are very, very easy for photonics. So once you’ve got that building block, your ability to carry out a two-qubit gate operation, for example, there’s much less crosstalk involved because your photons don’t talk to each other. So the thing to bear in mind is, I always describe it that the first hurdle for photonics is so much harder to meet and is, by the way, performed quite differently from the other platforms underneath the hood. But beyond that point, the point about scaling, I think, is there and maybe much more easy for photonics than the other platforms, if that makes sense.
Yuval: Listening to your focus on near-term utility, and also I saw some of your press releases, I’m getting the sense that the machine works today. Does it work and what can you do with it today?
Richard: Yeah, so it works. So it’s a four-photon machine across eight qumodes. So you send four photons into the machine and it has eight different available modes, so time segments that those four photons can pop out in. So it does that, and it’s a good machine to do that. Again, a qumode isn’t the same as a qubit, so already we’re different. Very, very approximately, many of your audience will know that this isn’t a very precise comparison, but very, very roughly speaking, a qumode can be thought of just more or less as a qubit. Again, some people will really object to that, they’re very different, but so roughly a sort of 8-qubit equivalent type size system.
And that allows you to perform some very rudimentary but interesting proof of concepts on how you apply quantum systems to machine learning. So of course, that’s not a system that will deliver quantum advantage or anything that a classical computer can’t do. Like others, we’re on the road map to increase the number of photons and to get to that point. But what it does do is to start to say, okay, if I have this quantum thing, well firstly, yes, I can plug it into a much bigger high-performance computing system or NVIDIA or other GPU cluster type system. And I can start looking at how quantum intersects and can be used within a much larger classical model.
And what we find is that there are interesting things that you can find. Of course, you can plug the quantum system into a machine learning model. And we find that some of those quantum statistics that are generated can be useful and it can be put to use within a machine learning model. So already we’re helping customers identify the first steps that they might take, the general application areas. You can do benchmarking, which is of course very interesting to look at how far you are away from running a real-life large-scale use case. And many other important things that are critical to our customers. Of course, it’s not up to us to decide whether they buy a system or not. And some of those customers are deciding to purchase those systems in order to carry out those exercises.
And ultimately, maybe as well, it’s important to say training of people, so training up people and getting them educated and hands-on using those systems is also a big part of the activity that’s needed before anyone will use any type of quantum computer. So facilitating them doing that is, you know, it can be sniffed at. It’s a very, in my view, very worthwhile activity.
Yuval: When customers purchase your computer, do they need any special infrastructure to install it?
Richard: No, no. Well, by and large, no. I mean, one of the huge advantages of photonics as we see it is they’re very easy to work with. They’re very, you know, largely they work at room temperature with maybe the exception of the detectors you might use. Largely, the thing can work using a lot of existing and quite mature componentry, optical fiber, switches and things like this.
So I think one of the reasons we’ve, one of the ways we’ve differentiated ourselves is by really trying to push a system which is quite mature and a system that can be delivered off the back of a shipping lorry or something and sort of rolled into a data center and plugged in and obviously sort of, you know, set up and things, but otherwise just set up pretty quickly and ready to go. And, you know, that’s very important.
The ability to not have an army of PhD students maintaining the system is quite important to several customers. If you are inside of a data center, by the way, you know, a proper data center shouldn’t have access. You know, you should have restricted access. So there’s like an ability to go in and say, calibrate components every day and things like that, which if for a proper data center is not available to you.
And yeah, because of the ability of Orca to build this thing from, you know, a lot of mature componentry, we made life a lot easier in that installation process. And I mean, hopefully touch wood, that persists. Obviously another important quality of any computing system is the uptime and lifetime. So you don’t want it to go wrong very often, especially if you’ve perhaps signed up to a service contract to guarantee that it’s available for a certain amount of time.
And not breaking very quickly, having a certain lifetime is quite important as well. So all these features might sound strange. They’re very product orientated, which is probably hopefully an impression you’re getting from me. I do believe that technical performance matters a lot. You know, don’t get me wrong. I mean, we know that delivering a system with a lot of qubits or in our case, a lot of photons, that matters a lot to deliver larger and more performant algorithms.
But I guess I like to stress all this stuff because it’s often not said that if you’re launching a product or if you’re thinking about meeting the demands of a future market, especially if you fall into the sort of way of looking at the future quantum market as being billions or trillions of dollars, that stuff just does not exist or cannot exist unless there’s this product maturity side to it as well. And, call me shallow if you like, but I do believe that if products aren’t mature and workable within the customer’s environment, that they’re a non-starter. They don’t work in the same way that a system not having technical performance wouldn’t work either. If that makes sense.
Yuval: Tell me a little bit about the company. Where are you based? How many people? How much are you looking to grow in the near term?
Richard: Yeah, so we’re based, well, we were set up in London in the UK, and that’s where most of our team still works. So we’ve probably got two thirds of the company based in London. And if anyone ever visits us, we’ve got this amazing office that’s in Zone 1. It’s about as central of London as you can get. If you come in, we’re in Paddington, which is the station you get to when you fly into London and get the train in.
So we’ve got an office and a lab and an assembly facility right in central London, which is great. I mean, great from a talent perspective, being able to attract people from all over the world and so they can get home. But we’ve also been growing our offices elsewhere. So for a long time now, we’ve recognized the talent base in Toronto and Ontario in Canada. So we have a lot of our theory team based out there, which is great, you know, close proximity to world leading universities.
And just very recently, you know, you might have seen over the new year, we acquired a company based in Austin in Texas who do all of our integrated circuit work, all of our photonic integrated circuit stuff. So all in all, we’re 60 people, which is a good number. I like it because, you know, you don’t want to grow too quickly. You do want to preserve quite a high quality of talent. And we all know talent’s hard to come by in quantum. So, you know, by being about 60 people, it’s that, you know, for right now, it’s the right size for us to do some really, really cool stuff with some really great people.
So that’s all good. And then beyond the next few years, like others, we are currently fundraising which will bring the resources to grow even more, to deliver and grow the company in all aspects. Take any area of the things that we’re doing, where we see a need to grow more, more researchers, more engineers, more commercial people, more operational people. So we’re just growing out on all fronts in those areas to meet all the good stuff that we want to do.
Yuval: We are recording this conversation in the middle of the year. Do you think since the beginning of the year, what have you learned or what has most surprised you in the quantum world?
Richard: Oh, that is a good question. It is, I mean, this just reflects, it is amazing what’s happening across the quantum industry. If you don’t mind, I won’t call out specific things because I think people know what they are. And if I do start talking about specifics, I think I’ll offend people from forgetting their thing. But, everyone sees every day advancements across every qubit modality and hardware and software.
So what surprises me constantly is you think how hard this technology is. You know, it really is hard to get really good qubits and lots of gates operating on those qubits and scaling. But yet, the optimists are winning. The pessimists are saying, oh, there’s a quantum winter coming and blah, blah, blah. You know, the end is nigh. By the way, most of those people tend to be academics, in my view, who aren’t in the thick of it doing all this hard work.
But the optimists are winning. You know, the field is moving forward at a dramatic pace. Yeah, we are making progress on all fronts. And I do think it’s only a matter of time before we’re going to start getting beyond the proof of concepts into, hey, this stuff really matters. It’s really valuable. We’re using it for real use cases. So, yeah, that’s not quite specific enough to be interesting, but I think it’s worth noting that it’s just a really exciting time and a field to be in at the moment.
What surprises me? I still think it’s surprising sometimes how little investment is happening, to be honest, that we are making progress and it’s such an important technology, and yet, you know, maybe I’m asking for too much, but you don’t have to think very hard to think of the true significance of what we’re doing in the quantum computing industry.
And yet, investment rounds are still smallish, I guess, with one exception. I mean, let’s admit the great success that PsiQuantum had with the Australian government. I mean, that’s surprising. I don’t think anyone would admit that it wasn’t surprising. So, by and large, I think it’s exciting thinking about where it’s going. And I think hopefully the amount of investment will change and we’ll start seeing many, many more companies grow and much more, very high valuations to reflect the significance of everything we’re doing.
Yuval: As we get closer to the end of our conversation, we spoke earlier about the differences between photonics and other modalities. And you mentioned error correction a couple of times. I’m curious, is error correction for photonics different than error correction for other gate-based modalities?
Richard: So two answers to that. It doesn’t have to be different. I think that the correct way of saying it is the way you implement error correction, because of what I said about it being reliant upon measurement-based quantum computing. So the way you implement it, the way you develop resources to perform error correction is very different.
And you, yourself and your listeners probably know you grow these massive entangled states and you perform operations on those entangled states, which is a world away from, how you perform normal gate-based operations. But that’s just an aside. The way you perform error correction, once you can do that, once you can perform the same gates as everyone else is the same.
I think the twist on that, you know, and when I mentioned that we look at novel ways to perform error correction is, well, firstly to also look at novel ways to create the resources that you need. There’s a lot of opportunity to explore new ways to do things. But the way that photonic quantum computers can be connected is very different. I mean, you’re not talking about nearest neighbour connectivity with a photonic system. No way near.
And so there is, there is only, we’re only just at the start of people looking at novel error correction codes that are not surface codes. They’re not nearest-neighbour codes. I can list all of the new exotic codes and things like that, which are interesting and exciting because of the possibilities to perform error correction better with fewer resources. So driving down the scary point at which you will be able to perform an error correction code with these new ways of doing it.
So photonics just opens up a different way of looking at error correction, but I’ll repeat what I said at the start, not vastly different. It still can be the same. There’s just an opportunity to do it differently based on this high connectivity that we have.
Yuval: A last hypothetical, if you could have dinner with one of the quantum greats dead or alive, who would that person be?
Richard: I knew this, I should have known this question was coming because you ask it of everyone. That is a good question. I, well, I have two completely conflicting answers, to be brutally honest. And it probably shows two sides of my personality. I love the stories of two characters in particular. I love the stories of, of course, Paul Dirac, just his focus and his bluntness and his ability to get right to the heart of the problem. And of course, a clear genius. I really don’t use that. I shouldn’t use that word, but I don’t use it very often.
But there is this clear, huge intellect. But also I like his personality. Like these great stories about people asking or making points in his lecture and he didn’t answer. And then, you know, the story, there’s this great story where someone in his lecture theatre sort of called out and said something like, I think you’ve made a mistake, Professor Dirac. And he carried on writing. And then someone else put their hand up in the lecture theatre and said, well, aren’t you going to answer his question? And he just pointed out, well, it wasn’t it wasn’t actually a question, so I’m not going to answer it. Which is just great. I don’t know. It’s a silly story, but a great one, I think.
So on the one hand, you know, Paul Dirac would be awesome. And of course, a lot of my understanding of quantum physics comes from all of his work. For a completely different reason. I love the story behind Rutherford, who came up with the first models for what the atom looks like and things like this. And I like his character because he was just so apparently so boisterous in the lab. You know, he used to get into the lab and just be like, come on, let’s do this. So, yeah, for two very different reasons, I think having both of those two people around a dinner party would be quite fun, quite interesting.
Yuval: Wonderful. So hopefully you’re right and the optimists are winning. But in any case, thank you so much for joining me today.
Richard: Thanks very much. It’s been a real pleasure. Really enjoyed it.