Elham Kashefi, Chief Scientist of the UK National Quantum Computing Center, is interviewed by Yuval Boger. Elham and Yuval discuss the history and future of this national strategy, the new Quantum Solutions Lab, quantum-inspired algorithms, what blind quantum computing is, and much more.
Full Transcript
Yuval Boger: Hello, Elham, and thank you for joining me today.
Elham Kashefi: Hi, thanks for having me, Yuval.
Yuval: So who are you and what do you do?
Elham: I am Elham Kashefi, professor of quantum computing at the University of Edinburgh, superposed with being the director of research of CNRS at Sorbonne University. And I look at quantum computing from a computer science background, that a bit of math background and I like to make sense out of it in any sense, in any way I can.
Yuval: And I think you’re now involved deeply, I would say, with the UK National Quantum Program, is that correct?
Elham: That is correct. I’ve been there since the birth.
Yuval: What do you do for them?
Elham: So the history goes is like the national program. My education started in UK, maybe I start saying that, you know, so I did my PhD back in Prior College, then Oxford University, then University of Edinburgh, but meanwhile I have been always, you know, have this French part of my life that we come and talk about that as well at CNRS.
But particularly about the national program, when it was announced almost a decade ago, I was involved first in the first phase of the hub, in KIT, UK Quantum Technology Hub. I was the Associate Director in charge of applications. Then came the second phase, I was a member of the senior science team, again in charge of software and applications.
So you can see all the time I have been trying to be the connector between our colleagues from physics, engineering, computer science, math and trying to focus on application.
And very recently in November, I was elected as chief scientist of NQCC, National Quantum Computing Center. And again, trying to push algorithm and software activity and trying to make sure we have a co-developed program between hardware and software.
Yuval: When people think about algorithms and software, they usually think about computing. Is the quantum, is the national quantum program broader than that? Does it also cover sensing or communications or other things?
Elham: Yeah, definitely. So the, when the national quantum technology program was announced, prior to many other national programs, it actually set the agenda that we have quantum sensing, so that was one of the hubs led by Birmingham, many many other universities, then there was quantum managing and metrology led by Glasgow, and then quantum computing and simulation, how led by the folks were that we were just talking about, and quantum communication led by York.
So indeed, the UK from the very beginning had this sort of four pillars that we saw it repeated almost in many many other programs, but the national program is covering it all, but the NCCC, which was only recently, in 2019, was announced, is a dedicated center for the quantum computing aspect. But the national program and future program, which is just again announced, 2.5 billion by the UK government, it goes for all the technology, the full spectrum.
Yuval: You mentioned other national quantum programs, and there seem to be many of them. Is there something that’s unique about the UK program from your perspective in terms of the goals or other aspects?
Elham: want to do the proper verification benchmarking and then I need to first sit down, do my homework and study all the other national programs which I haven’t done. So I’m a bit familiar with the EU flagship. I’ve been a part of it so I can keep that in mind. And I hear here and there from the other programs. I would not be able to really say so unique. I would say in UK started very early on. So that’s something unique in terms that, you know, they started, because it was 10 years ago, so it’s a decade ago. And many other places, they also have this sort of dedicated quantum program, DARPA, we heard about it and all these things.
I think the UK was very early on trying to put this perspective, okay, quantum science has been going down and they have been supported for a long time and they wanted to start this program of the translation, this transition to making sure the science that has been expanded in all the academic domains will be step by step going towards the technology deployment.
And I would say probably because they started early on, they have done it, it’s hard to
say rigorously, but they have done it over time. So it has been a very patient program, which I think I appreciated a lot that, for example, it was very clear quantum imaging, quantum sensing, it can come out early than quantum communication, but it was very clear and thanks to all the policy makers and all the lobbyists and all the influences of the national program that they knew quantum computing is going to be a long program and another 10 years, another 20 years.
I think this combination of what politicians want to see versus what is realistic in terms of timeline, it was something that they have done in practice. As a result, they managed to attract a lot of talent. managed to be early on to say this is the place.
And I would say something else, again, I don’t have the numbers and all the facts in my hand, but I feel because it started 10 years ago, and not just me, many of my other academic colleagues who were part of the hub and part of this ecosystem, we have been trained through national programs to become entrepreneurs. I co-founded my own company VeriCloud, which happened to be in France, But many of the colleagues that we started 10 years ago, are now co-founders of various quantum startups in the UK or elsewhere.
I think that’s also a credit to this sort of long procedure that naturally assisted us to go through this step and create a very promising ecosystem. I think we are second to probably US in the number of quantum startup for a nation, which is impressive. So that’s another aspect of this transition that I was talking about earlier.
Yuval: Having been involved both in the UK side, in CNRS, on the French side, in dealing quite some time with quantum software and applications, what do you know today? What have you realized today that you didn’t know 12 months ago about quantum software?
Elham: 12 months ago? So I kind of knew it already then, but now I let’s say that I truly believe in it. That is some sort of a motto that we use a lot about our recent activity that I’ll talk about in a second. It really takes a whole village to build a quantum application. I always said I take a village to raise a child and I take the entire ecosystem to build a useful quantum application.
So going back to your question, looking at what’s happened in UK, what’s happening in France and what’s happening everywhere, and you always have this academic industry, this pull and push. Today I’m even more confident than ever that if you’re really serious about useful quantum application.
A) It’s not just hardware development. We need to actually influence hardware when we’re talking about application. B) and it’s really we need to start talking all together in the sense that who has the pain point, who has the solution, who has the… It’s become even more than ever that we need to have this interconnection rather than somebody figures it out that has the killer app and off you go, they’re going to build the quantum computer and let’s protect the IP.
This sort of interconnection, it’s becoming a new thing in my own way of perception of what’s happening to the ecosystem.
Yuval: Speaking of new things, and I think you mentioned it in passing, there’s this new quantum software lab that you’re heading, right? Could you tell me about that, please?
Elham: Yes, I already spoiled it. It was like the recent baby, so we spent the last few months really launching it. So after my appointment was announced as an NQCC chip scientist, we have been developing the first physical node of NQCC, which is based in south and now in north in Edinburgh. We have the first physical node of NQCC and we launched the quantum software lab.
The history of it goes back in Edinburgh. We have had already one of the largest quantum software teams in the UK with six faculty soon with two new to be announced. So it was this momentum of various people with different backgrounds coming together. We have been building our village for the last decade and then the momentum with my appointment with the support of NQCC is now we have created the Quantum Software Lab two weeks ago.
So we signed a memorial agreement between NQCC and the University of Edinburgh. And what we’re trying to do, what I’m trying to do with the NQCC software lab, and we have an initial pilot project which is called Quantum Advantage Pathfinder, is really again contributing to this interconnection space I was telling you.
So in the UK, if I take one step back in the UK, we have technology hubs, quantum hubs, which have been pioneering the science. All the top scientists in the UK, they are involved in various help including quantum computing. They are discovering the resources. Then we have this very rich program of Innovate UK, which has already brought industry-led projects with the academic and focuses a little bit on the use case discovery and application. And then we have NQCC and other national labs, which have their own position like NPL, the position in the national lab for testing, NQCC, now a national lab for quantum computing.
So what we’re trying to do with the NQC Quantum Software Lab and a bit of what is my role as NQC Chief Scientist is to connect all these three pillars and try to really bring the science and technology and use case deployment together. So now how are we going to do that? I wait till you ask, but essentially it is an in-between place that I have been dreaming for a long time that is not just pure academic and it’s not just pure industry. We call it like We’re trying to create a startup culture but it’s still within the academic because we still need to do blue sky research but it’s much more focused to the use case delivery.
Yuval: How would that work? Let’s say I’m a commercial company, not in the quantum space, I’ve got an interesting machine learning classification problem, I’m not happy with the classical solutions or not happy enough, I want to investigate whether quantum is the place for me. do I come to the QSL? Is there some other path to see if the QSL can help me solve my problem?
Elham: Right, so you’re already a very good, an ideal customer for us because we actually set up to even help with the people who don’t even know anything about quantum. So we have this sort of this pathfinding methodology that we have in mind is that, okay, industry end users come with their pain point, they might not even know that this is actually machine learning classification.
So in your case, you’re ahead of the curve, but we even want to help with the hand-holding of translating this pain point that I want to have better production, I want to have cheaper data loading, etc. Translating to the mathematical.
So we had this use case discovery work packages that we were organizing as well as NQCC had lots of programs like Spark and Aim Day, etc. So we’re dragging those use cases, we’re doing the translation to understand what is the mathematical problem related to this use case, and then we go through the user journey, as you were mentioning.
Okay, now a user journey means that I need to find what’s the best algorithm, whether it’s simulation, optimization, machine learning, etc., matching to that particular mathematical problem. Then comes the question, okay, how do I benchmark? And that’s the expertise of the lab that we have a large classical approach of benchmarking, whether it’s error mitigation or verification and testing versus HPC versus classical.
So we do all of this benchmarking, having access to the hardware, we have announced a partnership with AWS, but we have other partnerships through NQCC. So we will benchmark. And the idea is that after the cycle of going back and forth, we will be able to answer this question of whether indeed there is an advantage in quantum or maybe there isn’t, in fact, you’re stuck with your classical solution, and because we are academic and we are actually at the end of the day, we want to publish a paper and feasibility report, we might for many of these use case, happily provide the negative results, but also being able to provide the real feasibility that when and how you might achieve that advantage.
And maybe, who knows, but some of the particular applications, for example, the security, etc., might be readily made and we can even point it out. So what I’m trying to say is that we like people to come to QSL because we are not a company, we’re not having any product, we want to close that gap. So if it’s too expensive or too soon for somebody to already talk to the particular member of this ecosystem, we would be able to help. But that again is my role, to be able to matchmaker when the problem is actually an academic problem, when the problem is already made, here’s the code, and when the problem is actually required domain expert that we don’t have in QSL lab, so we pass it around to the other startup or national lab we are working with.
So it is a bit of helping if you wish, we are like a first step. Have a chat with us for free, and get access to the experts. We are 30 plus researchers in the lab. And then this would allow us to identify this pathfinding journey, hopefully.
Yuval: How are you thinking today about quantum-inspired algorithms? Because some companies have started with quantum computing, then they said computers are not good enough, we’ll go back to quantum-inspired, which are essentially running on classical computers. Do you see that as just an improvement in classical computer science, or do you see that as part of the QSL or something that can truly advance quantum science?
Elham: Well, I would say this is like, you know, we are scientists and we are remaining scientists. So that’s another mission of QSL. You know, when you’re a researcher, you know, bring it out. You know, we usually even don’t even know what is the problem. We just come up with random solution and we want to do it the other way around. So for me, the reason that I gained, the only novel thing in my own thinking for the lab is that, okay, let’s start from use case.
And we don’t know what is this use case. Let’s start at, you know, I want to have a better synthetic data. I want to have a more secure differential privacy. I want to have more optimized bed allocation, etc. And then when you start so far away from the pain point, it’s very natural. At some point, you don’t have a quantum solution that fits it. It’s very natural that in this path of going forward, sometimes a classical problem, an optimization problem might be useful.
And I’m pretty sure everybody in different companies already sees that you know, most of the time, maybe some of these para-end teams are not even aware of the best classical solution. So if our journey will show that from these 100 use cases that we play around, maybe 90 of them have an ideal ready-made classical solution, why not?
But there are also two ways to interpret the classical-inspired or quantum-inspired algorithm that naturally, like some of the work we did recently with our team, we didn’t intend to do, but we showed that a large family of the VQE variational circuit actually can be classically simulated. I mean, I’m happy about that because it’s again, it’s like because we’re trying to understand where is the advantage of where is not the advantage, which is a very key question for us. We come up with a new method of simulation of this quantum circuit and many people are doing that.
So in short, I mean, I don’t have anything against it. My mission is not to come up with a new classical algorithm, I’m actually very much excited to see where the quantum advantage is coming from and how we can capture it. But on the way to capturing this thing, we’re clarifying the pathway, you know, we understand it’s not here, it’s not here. And if for those we don’t get the advantage on top of this, we come up with a novel solution of classical, even better. But this is for me, it’s like a clarification. It’s not the target, is the byproduct of this exploration.
Yuval: Given your software background and now doing it in quantum, what do you think is the missing piece or the biggest pain in software development today? Is it an abstraction layer? Is it actually better control of the hardware? How do you do a circuit with a large number of qubits? What do you see as the next pain point that’s coming to software developers?
Elham: Just to clarify, I’m not really the world expert on software. I’m a computer scientist in the sense that if you go to the traditional study of computer science, we have algorithm complexity, format, method, programming language, compiler, optimization, etc., etc., etc. That we take all of these courses, so we have a bit of familiarity. I think for me, overall, the biggest pain point for our field in quantum is that we are trying to reinvent a lot of wheels. So, it’s that, you know, we know there is a large class of classical ICT, as we say it in UK, is already very much relevant when you look at the control, the calibration, optimization, programming language.
So, I think for us, we are reaching into that era because these devices are emerging because the application is arriving. So finally we should be able to knock on the door of our classical colleagues and say that, okay, as we did 20 years ago, 30 years ago, the first computer scientists who were very much involved with quantum was algorithm complexity. We immediately find a common language and we could compare complexity classes in quantum or algorithm design.
But the rest of the field of computer science, taking anything from verification testing, informal method, optimization, and we are not really tapping in that space. So I think maybe the pain point is again, the language barrier, somehow the subclass of track A of CS of the people who managed to overcome that barrier, but it’s been a long time in the University of Edinburgh. School of Informatics is one of the largest in Europe with 1,000 plus staff. And they are all excited. The more they hear about quantum, they become more excited. And I feel like I wish all of my colleagues would understand quantum just like this, or at least be involved, because they have the solution to a lot of problems that we’re trying to solve them.
So we don’t need to reinvent all of them from scratch. So it’s a translation of existing knowledge.
Yuval: As we get closer to the end of our conversation today, I wanted to ask you, I was looking at your background preparing for this call and I saw that you did a lot of work on blind quantum computing. What is blind quantum computing and why is that important?
Elham: Thank you for asking that question. So this is one of the protocols that I co-invented with two of my colleagues, Joe Fitzsimons and Brodwin back in 2009. And as an example of that pathfinding that, you know, we were working a lot on measure-based quantum computing, which is a different model of quantum computing that the classical quantum separated. And we will be all exploring, how could we exploit this thing? And before we realized that actually we invented this model for secure cloud computing.
So, in the future, right now, that’s when you look at all the quantum computing services are provided via the cloud, we will ask the question about the security of this thing. You’re very likely that when we seriously get involved with trying to connect to any of this cloud provider, we don’t want our algorithm to be known, we don’t want our data to be known, and et cetera, et cetera.
This is a problem that exists classically and a homomorphic encryption is trying to solve it. So we come up with a solution for the quantum, which provides you information security for delegating your computation to the remote server while keeping all of your information and data intact. I think for me it’s very important because it’s almost like a tool for using quantum because we believe for currently the model of the quantum, it will be this sort of giant beast that is sitting somewhere like in the data center and the privacy of the data in whatever application domain is become very crucial.
So, BlindQC, which is also the target of my startup, VeriCloud, is really providing that next level of trust so we can scale up. And on top of this BlindQC, we are able to provide integrity or verification of quantum computing. So to me, it’s like, if we’re serious to make quantum computing as a service, then the question of integrity and privacy can be provided with this framework of BlindQC, which also connecting quantum communication and quantum computing. So they are no longer separate fields. You need to have quantum communication on the client side to be able to exploit quantum computing on the server side. So that’s blind quantum computing. Thank you for that.
Yuval: QSL is new and I’m wondering, professionally speaking, what keeps you up at night? What are you worried about the QSL, if anything?
Elham: Oh, it’s like it has already started worrying, you know, because we’re changing the direction. See, as a researcher, we sit down, we discover new things, and then we push it to the outside role. And that’s something that I have been doing for the entire half of my career, you know, like, and leading all my team to let’s explore whatever that it comes and then pushing it. And now I’m trying to change the gear. I’m trying to, I can say that, okay, we know a lot of tools, We know verification, we know testing, we know algorithm design, we know abstraction, we know all of these things and I brought a lot of people together. But we want to do the game the other way around, bringing outside question and translating it.
So I think till I see few examples of this full process, it’s really working. And I figured the machinery that it’s not at heart, but really there is a common journey that whether the problem is on accuracy or privacy or different sector, whether it’s a fintech or I don’t know, quantum chemistry, at the end of the day, this user journey is working and I can bring this different expertise as I was mentioning earlier, that they can talk together and we work together rather than just coming up with the next big breakthrough. That’s just, that’s keeping me, I was excited and worried because it’s a new way, it’s a new way of dealing with the problem, but it’s also scary. So we’ll see. That’s me in six months’ time and we’ll see about where we made it. I’d be happy to.
Yuval: And so my last question, a hypothetical. If you could have dinner with one of the quantum greats, dead or alive, who would that person be?
Elham: Oh, okay. a dead or alive quantum person. It’s going to be John Bell, you know, it’s like the quality is just so amazing. Although we think we know everything about it, but the fact that as a mathematician, he went and really set up that framework, it’s going to be an amazing dinner, just to chit chat. And there are plenty of other people alive, but I know I can have dinner with them anyways.
Yuval: Excellent. Well, Elham, thank you so much for joining me today.
Elham: Thank you very much for having me. That was a great pleasure.