The State of Quantum Computers

In this episode, Richard Moulds, General Manager of AWS Braket, discusses the current state of quantum computers. What developments we can expect in the coming years, what are the challenges, what are the applications, and how are organizations tapping into their capabilities now and in the future? Richard provides engaging insight into how he sees it playing out and how we might achieve the democratization of quantum computing making it available for all.

Transcript

Samantha Mabey: Welcome to Entrust Engage, an open forum for the most innovative leaders in security technology. I'm Samantha Mabey and I'm your host. Today's episode continues our conversation around post-quantum, but we'll be putting particular focus on the quantum computers themselves. And with that, I'm excited to introduce our guest today who's Richard Moulds, General Manager of Amazon Braket, which is the quantum computing service of AWS. Welcome, and thank you so much for taking the time to talk to us today.

Richard Moulds: Hey, Samantha, it's a great pleasure to be here. Fascinating subject.

Samantha Mabey: It is for sure. Lots of attention around it too. So we've been talking about the topic of post-quantum in this series so far. And, of course, for us having a digital security focus, a lot of what we've been discussing is the point in time where there will be a quantum computer that's powerful enough to break cryptography as we know it today. Of course, the general consensus is that's still roughly a decade or so away, but we also know that quantum computers do already exist. To get started, I'd love to hear from you, maybe do a bit of a level set on what the current state of quantum computers is, and maybe some insights into what the next few years might look like.

Richard Moulds: Sure, happy to. I mean, as you say, there's a lot of excitement around the technology. You read any technology publication these days, a lot of speculation about when quantum moving will be useful when it be real, and where it might impact. You talk about the threat to cryptography which is real. I mean, that's a known use case for quantum computers. Of course, we hope there are a lot more beneficial use cases for quantum computing and ultimately that it might benefit mankind on lots of different fronts. I mean, it has the potential to be a truly disruptive technology that could one day solve problems that are just intractable using a regular, we call them classical computers. But as you say, it's still some way away. The industry is growing really quickly but it's very much in discovery mode. It turns out it's really hard to build a quantum computer. These things work at the level of atomic physics.

To build them, you're fighting nature pretty much at every step of the way and it's hard. Still, this is not an engineering project, this is not a research project. But as you say, they do exist, they work, they give the right answer to problems when you submit questions to them, but they're still not at the point where they can deliver a commercial speed up relative to traditional computers. There's no production advantage if you like to running a quantum computer right now. There's lots of different ways you can build them whether you are using superconducting circuits or atoms or ions or photons so lots of different ways. And there are trade-offs between each of those different technologies. Some are fast, some are slow, some are more accurate, some less accurate, some do different types of calculations. And I think the industry, in general, is trying to figure out which of those performance characteristics are most important and which type of technology might be best suited to particular use cases. So a lot of discovery mode in the next few years.

Really the focus is on scaling these computers, making them able to handle bigger problems than this relatively small problems they can handle today. But really the fundamental stepping stone is about quality. These computers are extremely sensitive devices, they're extremely sensitive to noise, and that can create errors and those errors can be a severe limitation in what you can use these computers for. Much of the industry right now is focused on making better quality qubits. Qubit is the processing element of a quantum computer. And with that performance in quality, we can hopefully boost the scale of these systems, tackle bigger problems, and deliver more accurate results.

Samantha Mabey: Okay. Although the evolution of creating quantum computers is still sort of ongoing, what are some of the primary uses for quantum computers today?

Richard Moulds: To be honest, the applications for quantum computers as they exist right now is really to learn about quantum computing and to figure out how to build better quantum computers. As I say, we're still figuring out really thorny topics like scalability and error correction so we can use today's hardware to try and advance that work. Many of the people that use our service in AWS are researchers focused almost entirely on how to build better machines rather than necessarily how to use them today. And, obviously, we have to plan for the future so discovering what applications are most relevant, identifying what types of algorithms might be the most productive is clearly a very hot area of research.

But today, the use case is really about getting ready for the future and building better machines. We talked about cryptography, for example. We know there are specific use cases where quantum computers when they reach the appropriate level of performance, will offer a speed-up. For example, in the sphere of modeling molecular systems. It makes sense to use the quantum computer to model quantum systems. Figuring out how chemical reactions work better, figuring out how things like catalysts might be more efficient, figuring out better ways of improving the dynamics of materials, figuring out better ways of delivering against problems like drug discovery, for example.

If we could find ways of simulating atomic systems, essentially creating digital twins for how molecules work, how polymers work, how reactions happen, then that could have an enormous benefit across, obviously, many, many industries. But we also think there are benefits in using quantum computers for decision-making in areas like machine learning and optimization. So there are a lot of mathematical problems out there that scale really badly. I'm sure many of the listeners to this podcast are familiar with problems like ... The set of problems called the traveling salesman problem where if you try and deliver lots of packages to lots of places there's a rapidly increasing number of combinations of how you might do that.

The same for routing taxis around a city or network packets around an optical part of the system. Decision-making, optimizing systems is a big challenge and we think that quantum computing will help in those areas. A good example. Just at the end of last year, BMW put out a public challenge, and they posted four use cases that they think most relevant to their industry so, obviously, building better cars. And those use cases sort of spanned that whole range. There were use cases that focused on materials development. Use cases that focused on using machine learning for better quality assurance in the manufacturing process, and using quantum computers for optimizing the use of resources in developing new vehicles. I think it starts with modeling molecular systems and materials development, chemistry, that type of stuff, and then evolving into decision-making, optimization, and machine learning.

Samantha Mabey: That's interesting. Saying BMW, that's an industry or a company I haven't heard of sort of leveraging or thinking about quantum computers that way, but, of course, it's just going to become more popular. You've talked a lot about how quantum computing right now it's being used primarily for research. Who is using quantum computers today? What does current adoption look like? Is that changing?? Becoming more popular? I assume there's a lot of academics or large companies.

Richard Moulds: Quantum computing has been in the research phase for a couple of decades now so not surprisingly that's a popular set of users on the service. But also educators. There's a problem in this industry that there's just not that many people that understand quantum computing, and how to program them, and how to build them. And as the machinery itself gets more powerful, clearly there's a general requirement to build up a broader community of quantum expert, both within corporations and within manufacturing community. So a lot of educators, universities, in particular, are starting up quantum computing courses at various different levels. There are a lot of specialist algorithm developers out there. It's easy to focus a little bit on the hardware, but there's a real gap to be filled in bridging the distance between a quantum computing hardware manufacturer and, for example, somebody building optimization tooling, or chemical engineering tooling, and material science tooling so a lot of algorithmic experts out there. So small quantum expert companies that are researching algorithms, and engaging customers, running proofs of concept so that community's very strong and growing quite quickly.

And then, of course, corporate users themselves that are trying to get ready. As we talked about in the context of BMW, getting ready for quantum build-up in-house expertise, establish small quantum teams, centers of excellence to try and understand the technology. And then, of course, just lots of people that are really interested. Everything from high school students to physics undergrads, anybody that just has an interest in the field. Traditionally it's been very difficult to get your hands on one of these machines and actually program a quantum computer by making it available on the cloud. Obviously, we've tried to make that easy.

Samantha Mabey: Absolutely, That would make it a lot more accessible for anybody who's interested in just sort of poking around or learning more.

Richard Moulds: Right. We use the phrase democratizing quantum computing. You don't need to necessarily be friends with a physicist to get access to one of these machines anymore. It's a cloud service, anybody can use it. That's the premise of the cloud. It's on-demand, you just sign up. If you're an AWS customer then you can just use the service and just you pay for whatever resources you use.

Samantha Mabey: That's awesome. Even just the conversation around quantum computing, right, it's opening up to a broader audience. People are just interested in so many different future opportunities.

Richard Moulds: That's right. But it's all about getting the right expectations, I think, and giving people an on-ramp, if you like, for technology that's as painless as possible and doesn't require a huge upfront commitment.

Samantha Mabey: Absolutely. A little bit more of a long-term prediction. You might have to get your crystal ball out for this one. But I'm just curious, where do you see the future of quantum computers going as far as commercialization and sort of mass adoption of the technology?

Richard Moulds: I mean, in the end, of course, we hope and expect that a quantum computer becomes just another compute resource in the cloud the customers can choose. If you look on AWS there's dozens and dozens of different types of CPU available, dozens of GPUs available, no reason why they couldn't be QPUs available in that mix as well. It's very important to think about how we'll consume quantum computers. In some ways, calling them a quantum computer is poor naming on behalf of the industry. These are not replacement devices for regular classical computers. Sometimes people joke about sort of quantum laptops, and quantum smartwatches, and quantum iPhones, and goodness knows what. None of that is real. Quantum computers are really quantum co-processors, they do certain sort of mathematics very well we think. In the same way that crypto code processors, something out of your industry, does a certain set of mathematical problems and calculations really well, much better than a CPU can.

You don't see the crypto code processors as a crypto computer that might one day replace a regular computer, and the same is true for quantum computers. They really code processors. They'll always exist alongside large chunks of classical compute resource. Quantum computers will always be relatively expensive, I think, and relatively scarce compared to classical chip-based computers so customers should always see it in that context. Never think of quantum computing as being something that exists in isolation, it's always parts of a broader compute fabric. And so you think about the future, yes, it'll be a compute option, if you like, as you think about assembling the infrastructure to go solve hard problems. So I think the communities that we're focused on, the folks that are pushing the limits of classical computers today.

We talked about some of the problems earlier that scale really badly as the problems get larger. For example, simulating molecular systems or optimization problems. So the customers that face those problems day in and day out, customers like BMW, are in the cloud right now and they like the cloud because it's elastic. They can leverage tons of classical compute resource pretty much on demand to solve these big computational problems. Those are the sorts of customers that one day we think quantum computers will enable them to stretch their thinking, solve harder problems, solve hard problems more quickly, and hopefully more cheaply. I think cloud is the right place to think about quantum computing because that's where these hard problems are being solved right now just because of the on-demand elastic nature of compute resources in the cloud. And we see quantum computers one day complimenting that range of compute resources.

Samantha Mabey: Well, there's already a lot of benefits today with classical computers, and IT infrastructures, and moving and migrating to the cloud so it just makes sense that would be a convenient place to access this technology.

Richard Moulds: The challenge is delivering it in a way that doesn't require everybody to be a quantum physicist to actually use it. All right. I mean, today if you want to program a quantum computer you sort of need to know how it works. That's okay for now because we're in discovery mode. But in the future when we think about customers using a quantum process unit, a QPU, in the same way they might use a GPU, they don't need to understand how it works, they shouldn't really need to understand anything about physics. There's a real opportunity, I think, to build that tooling and to try and make a bridge between the physical hardware and the use cases that these types of customers are likely to have.

Samantha Mabey: And it's good to know that we won't be doing away with classical computers as we know it. Rather this will be, it sounds, like just more of a compliment and more specialized for different organizations as opposed to the consumer level.

Richard Moulds: Right. You're not going to be streaming video through your quantum computer, you're not going to be running spreadsheets on your quantum computer, rendering animations, that's not what they're for. They solve certain sort of mathematical problems we think very well. And people will use them for what they're best at frankly, and they'll use cheaper resources, more available resources for the bulk of the rest of their work.

Samantha Mabey: So I won't hold my breath for the quantum I watch.

Richard Moulds: I wouldn't, no, no. No. Some of these quantum computers run at tiny fractions of one degree Kelvin, so they're colder than outer space, so I'm not sure you want to be carrying that around.

Samantha Mabey: Oh. That doesn't sound great. I mean, I'm Canadian but probably still not a good idea. All right. So going back to the primary uses of quantum computers today. Again, we know that quantum computers will unlock some opportunities for things like medical research. You gave your BMW example. Are there any interesting or notable achievements that you've seen today? I know Google talked about their computer being able to achieve calculations that were more than three million times faster than what could be done on the world's fastest computer. Are there more examples similar to that?

Richard Moulds: So you have to be really careful. So yes, there have been a couple of claims of what's called quantum supremacy, which is an unfortunate term I think but nonetheless, it's out there over the last few years as you mentioned. But that doesn't mean a quantum computer is useful. These experiments, these demonstrations of where a quantum computer has done something that is either impossible or just very expensive on a regular classical computer, even as classic as supercomputer. But these aren't useful problems. These are essentially synthetic problems that have been created that are as solved as easily as possible on a quantum computer and solvable as hard as possible on a classical computer.

It proves the point in a sense that quantum computing ... Quantum computers can be special, that they can do things that are intractable or impractical at least on a classical computer. So we've crossed that milestone. That was an important milestone, but it's only the first step in the road because the problems that were part of those supremacy calculations are not useful problems, essentially they were randomly generated problems. They don't play into any use case that any real customer might actually have.

The next big step down the road is what we call quantum advantage, which is where we demonstrate a real speed up in terms of performance for a problem that somebody actually cares about, something that's actually useful and ties into a real use case. We are not yet at that point. There have not been any demonstrations of quantum advantage only quantum supremacy. Quantum supremacy is synthetic. Interesting, but synthetic and not very useful. Quantum advantage is when these machines become useful, that's something that somebody cares about. So we are not there yet. So whether that's a couple of years away, whether it's five years away, we really don't know. That's what the industry's working towards now.

When you see examples and news stories of content machine being used to route taxis or direct buses around a city or plan some operation in a production line, for example. You see these periodically at conferences. They're not arguing that the quantum computer is faster, they're just arguing that the quantum computer came to the right answer which is important and sort of table stakes for computation. We're demonstrating that these machines work, which is most important, and that they can be applied to regular everyday problems. What we've not seen yet is a demonstration that they actually do deliver any sort of operational advantage. That's the next step, and that's still I think some years away.

Samantha Mabey: I think that's great. Quantum advantage. I think that's important to make that distinction so I appreciate your sort of outlining that. So linking back to the beginning of the episode sort of talking about how, at least from that digital security lens, we discuss that point in time where there will be a quantum computer powerful enough to break cryptography as we know it. Those headlines and concerns around the threats of that quantum computing will bring. Obviously, you get a lot of researchers, a lot of people poking around in interest in quantum computers. Is this something you see as well, customers just having these concerns? And if so, what sort of recommendations do you make? What can people do today to prepare for that?

Richard Moulds: Of course, it is a huge concern and it's a real concern. We know that quantum computers, there are algorithms that have been established already that will help in the factorization problem. It's a matter of scale. To crack, for example, an 2000-bit RSA, key needs tens of millions of qubits that are at quality levels. Many orders of magnitude higher than the current quality of qubits in current quantum computers. And remember, the current quantum computers are typically anywhere between 10 and 100 qubits so we are five or six orders of magnitude away in terms of scale, and five or six orders of magnitude away in terms of quality. Both of those problems are hard, and solving them both at the same time is even harder. We're a long way away from machines that can routinely break today's crypto.

As we know, it takes a long time to introduce new crypto algorithms. It takes a long time just to prove that they're safe. And they might actually physically deploy them into routers, and software and goodness knows where else, cryptography is used, isn't it? It's sort of everywhere. Thinking about how to address the arrival of sufficiently powerful quantum computers is something we should certainly do right now. And, obviously, there are a lot of initiatives underway to come up with either software algorithms that we think are immune from attack by quantum computers so sort of post-quantum software. There are competitions and standardization processes that happen right now to try and find those algorithms. AWS is heavily involved in that process. We've deployed example algorithms on our own externally facing networks to try to get some experience about how these algorithms actually work.

That standardization process is interesting. Obviously, there are also techniques that actually use quantum mechanics to try to deliver quantum-safe systems, they're being explored in various different industries as well. It's a real threat. And given the time scale to roll out new cryptosystems, we should be very focused on that now and gladly, thankfully rather lots of industries really are and there's some really promising approaches. It's going to be a long battle to prove and deploy true quantum-safe systems. We have time but we need to crack on.

Samantha Mabey: Absolutely. We've experienced similar things. I mean, even SHA-1 to SHA-2 migration, right, that took a lot longer than people expected. We've brought that up a few times. And this is just sort of worlds apart from that as far as how long it's going to take. That's definitely something we hone in on too, just sort of seeing the time to start thinking about this and looking at it is now.

Richard Moulds: I mean, just think of the time it took to approve even just AES was a decade or so, and that was solving classical problems that we could test with classical systems. Building a quantum-safe algorithm, we are trying to test it, obviously, against a classical quantum computer that doesn't exist yet, at least at that level of scale. So building something that's robust against something we actually can't test against is pretty tricky. And, of course, any quantum system also has to be classically safe as well. It's a pretty hard problem solve but there's real progress I think in terms of standardization and real-world implementations that are being tested right now so I think we're off to a decent start but we just got to keep focused and take it seriously. Obviously, customers should be tracking the evolution of these standards and stay tuned. Pay attention, this is real and we need to make sure that we're moving forward as a community.

Samantha Mabey: Absolutely. I mean, I'm sure you're seeing it as well, there's just an uptick, in general, of people who are interested in this and definitely starting to look at it now so that's promising for sure.

Richard Moulds: Again, which is the reason why we launched the service. One of the ways you can try and understand the security threat is to understand the trajectory that current systems are actually on. The only way to do that in many cases is to roll up your sleeves and go test it so we try to make it easy to do that. And, obviously, at the same time to plan forward for how we as a service provider, and the applications that all of our customers build on our platform, be protected against it which is why it was so critical that AWS is part of the development process for quantum-safe systems.

Samantha Mabey: And it's actually a pretty perfect little takeaway and note to leave our listeners with today. This was great. Thank you again so much for taking the time to have this conversation with me I really appreciate it.

Richard Moulds: No, you're most welcome, it was a great pleasure.

Samantha Mabey: That's it for today's podcast. Please keep up with new episodes by following us on LinkedIn, Twitter, and using the links in the episode description. Thanks for listening to Entrust Engage.