Microsoft Research Podcast - 034 (rerun) - The future is quantum with Dr. Krysta Svore

Episode Date: July 25, 2018

First aired on January 17, 2018. If someone mentions quantum computing, and you find yourself outwardly nodding your head, but secretly shaking it, you’re in good company: some of the world’s smartest people admit they don’t really understand it either. Fortunately, some of the world’s other smartest people, like Dr. Krysta Svore, Principal Research Manager of the Microsoft Quantum – or QuArC – group at Microsoft Research in Redmond, actually DO understand quantum computing, and are working hard to make it a reality.Today, Dr. Svore shares her passion for quantum algorithms and their potential to solve some of the world’s biggest problems, explains why Microsoft’s topological quantum bit – or qubit – is a game changer for quantum computing, and assures us that, although qubits live in dilution refrigerators at temperatures near absolute zero, quantum researchers can still sit in the comfort of their offices and work with the computer programmer’s equivalent of Schroedinger’s Cat.

When I first got entangled with quantum computing researcher Christa Sfori back in January, I discovered that she herself is a rare embodiment of two qualities that are often mutually exclusive. She can both do the math and explain it to people like me. Whether you already know everything about the world of qubits, or you're learning about them for the first time, or both, you're in a super position to enjoy episode 8 of the Microsoft Research Podcast, The Future is Quantum. The problems we're looking at solving with a quantum computer are the problems that today

require age of the universe timescales. I'm not going to be around for that solution. Some of these problems literally require billions and billions and billions of years to solve. And on a quantum computer, what we've shown in some recent research is that you can solve some of these problems in a matter of, say, weeks, days, hours, seconds. I'll be around for those solutions. You're listening to the Microsoft Research Podcast, a show that brings you closer to the cutting edge of technology research and the scientists behind it.

I'm your host, Gretchen Huizenga. If someone mentions quantum computing and you find yourself outwardly nodding your head, but secretly shaking it, you're in good company. Some of the world's smartest people admit they don't really understand it either. Fortunately, some of the world's other smartest people, like Dr. Krista Svore, Principal Research Manager of the Microsoft Quantum or Quark Group at Microsoft Research in Redmond, actually do understand quantum computing

and are working hard to make it a reality. Today, Dr. Svore shares her passion for quantum algorithms and their potential to solve some of the world's biggest problems, explains why Microsoft's topological quantum bit, or qubit, is a game changer for quantum computing, and assures us that, although qubits live in dilution refrigerators at temperatures near absolute zero, quantum researchers can still sit in the comfort of their offices and work with the computer programmer's equivalent of Schrodinger's cat. That and much more on this episode of the Microsoft Research Podcast. So Krista, your research revolves around quantum algorithms. Tell us what's so cool about quantum algorithms and how you got interested in that line of research. Quantum algorithms, for me, they really serve the promise of being able to solve things that we can't solve today with digital computers.

Quantum, quantum computing more generally, really unlocks this amazing potential for computation. And so for me, quantum algorithms are the ability to actually do something with that computer, right? Algorithms being the recipes that you run on the machine. And so developing quantum algorithms that can actually solve real-world problems that are going to make a difference, that's a huge thing, a huge passion of mine to see how can we actually change the world, change, you know, our outlook for clean energy or food production, you know, things that really affect humanity. We have the potential to solve some or at least help combat some of those with quantum algorithms and with a quantum computer. When did you decide quantum was the algorithm of choice for you?

Yeah, that's a good question. So I, in college, I studied mathematics. During my junior year, I had a seminar from Andrew Wiles, who is a famous mathematician. He solved Fermat's last theorem. And in that seminar, he introduced the idea, it was a seminar on cryptography, and he introduced this idea that there was this model of computation that could break this cryptosystem that we use all over today, RSA, based on the problem of factoring. And he introduced that this model of computation, quantum computing, namely, could actually break this cryptosystem. And I became fascinated that there was this other model one could compute with. And I also had a keen interest in computer science. So I wanted to understand, well, how do you use that model? You know, how would you write an algorithm to do

that? How would you design a programming language and software stack to then implement that algorithm on the hardware? And that started many years of study and fascination with quantum computing. So we're assuming a fairly high level of technical sophistication in our listeners. So it's not really a quantum for dummies podcast. But I think there's room for a short elevator pitch on what quantum computing is and how it differs from classical computing. So two minutes, go. Yeah. So with quantum computing, you're really differs from classical computing. So two minutes, go. Yeah. So with quantum computing, you're really relying on the principles of quantum mechanics to compute. And these principles are just so vastly different than what we have classically. Things like superposition, entanglement, interference, you know, these terms,

if anyone's listened to Feynman, Richard Feynman and his lectures or, you know, studied Einstein's notes and papers, you would have come across these terms. So entanglement being this amazing ability to have two systems that can be correlated over, you know, universes apart, that when you do one to one thing, it automatically instantaneously does one to another. That's a pretty incredible property. And then things like superposition. So when I take a quantum computer, I rely on superposition to store the information. What that means is instead of just having a bit in my classical computer, you know, my classical computer, of course, just binary,

it's on or off, it's a switch. In a quantum computer, I like to think of it a little bit like a dimmer switch, right? Your information is actually stored in a quantum state, and that quantum state can take on both zero and one. It takes on basically a linear combination of zero and one. And this allows you to then scale up and achieve some of these amazing exponential improvements with quantum computing. So you lead the Microsoft Quark, which is quantum architecture and computation group at Microsoft Research and Redmond. Tell us about your team and your partnerships and your vision for quantum computing. Yeah, so we started Quark now almost, gosh, I think it's seven years ago. We started embarking on the idea that we wanted to determine what do we do with a quantum computer? It has all

this promise and all this potential. And then how are you going to make it do that, right? How are you going to program it? How are you going to get that algorithm into that hardware system? And it uses these vastly different principles, right? Vastly different systems. It requires classical computers unlike any we've built before, in addition to this quantum computer and quantum system we've never built before. And so our focus has been to really not only answer those questions, but then also make it accessible to a broader community.

So you've just done that. Talk about the kit that you just released. Yeah, we're pretty excited. We released what we call the Quantum Development Kit. And this Quantum Development Kit is really developed in order to make quantum computing more accessible to developers, quantum enthusiasts, students, academics, researchers, anyone who has a keen interest to learn how to take advantage of this amazing resource. And the kit includes, of course, a brand new quantum-focused domain-specific language for quantum computing. We call our language Qsharp. And Qsharp is embedded in Visual Studio. So not only do you get this brand new language, you know, then you also have in Visual Studio, you have all of these great things you can take advantage of, like debugging, code

colorization, you know, syntax highlighting, help information when you hover on functions and operations. And so it makes it that much more accessible and hopefully, you know, lowers the barrier of entry into quantum computing. Do you have to be an expert in quantum mechanics or quantum physics to understand this field and start developing for programming in it? You know, we really hope that you don't. You know, if we were limited to just looking at developers that understood quantum mechanics, it might be a small group. And we want to really open up quantum computing to a broad audience, make it accessible to everyone. And so the idea is that you need not know quantum mechanics. You know, after all, I'm in this field. I'm a computer scientist. I'm not a physicist.

I did not take quantum mechanics in college. You know, I came at this from the perspective of a computer scientist, someone who wanted to use this to change computation. And so here we're hoping, you know, more people want to come and learn about this space and develop algorithms and applications in this space. And you need not know quantum mechanics. Which is fascinating because for people that are wanting to get into quantum computing, if there's super high barriers to entry, you turn off a lot of people right off the bat. So how does Qsharp differ from other programming languages since it's a quantum tool? And does it work on both simulators

and actual quantum computers as we speak? Yeah, so Q-sharp has really been designed to target a quantum computer and the model of computation you have in a quantum computer. It actually operates hand-in-hand with a classical programming language as well. So what's important to remember about a quantum computer is that it's a hybrid device. It's like a coprocessor, an accelerator. So you would program much of your application and kind of the wrapper code in a standard language, and then you would call this Qsharp program that then is the program that's really specified to the quantum accelerator. Qsharp's designed with that in mind. And so the whole idea is that Q-sharp brings with it

the ability to easily set up things like entanglement, superposition, use interference to design your quantum algorithm and allow it to evolve. A functionality that's required in quantum algorithms is very, very different than the functionality we use in classical algorithms. And so in Q-sharp, we bring this right to the forefront so that it's readily accessible to the developer. One of the challenges in quantum computing is the instability of qubits that you think this concept of a topological qubit may address or is addressing. So what's a topological qubit and how will it help in quantum computing?

Right. So with quantum computing, we rely on qubits, right? So qubits are how we store information in the quantum computer. It stands for a quantum bit. So we say qubit. And these qubits, what we have to realize is that they want to interact all the time with their environment. This is not a good thing for quantum computing. We want them to remain in vacuum, isolated from thermal noise and other noise processes. And so when they start to want to entangle with their environment, that means that they get noisy. That entanglement with the environment causes us to lose the

information in the qubit. So we have to prevent that. We have to really isolate the system. Now, there are various ways to isolate the system. And the approach Microsoft's taking is truly unique and fundamentally more scalable. So the idea is that we should seek a way to actually engineer into how we're going to store and compute on the information in the quantum computer. We should engineer that such that it's more robust, you know, from the foundation. So imagine you want to build a skyscraper. And imagine I have a brick. And all of us, you know, we look at bricks as extremely robust. But try to build a skyscraper out of bricks.

Right? It's very, very hard. It's probably not the way you want to go about it. A better way is to go employ steel. So steel, it's a different type of foundation from which to build this skyscraper. But it's going to allow you to scale quickly. It's going to be much more robust against changes, you know, natural disasters, whatever they may be for the skyscraper. And so our qubit is like the steel qubit. In our system with the steel style qubit, a topological qubit as we call it, we're going to be able to scale much more readily. And it's

going to require less overhead, less additional resources, less additional cost, less additional time, right, to get to higher numbers of compute capability, higher numbers of qubits. So, you know, one way to think about it is we can do more with less. We can do more computation with fewer qubits because they're like the steel. They're very, very strong. So it's a different kind of qubit altogether. And it's unique to Microsoft research. Yes, this qubit, the topological qubit. Actually, Michael Friedman, who joined MSR in the late 90s, he developed the ideas behind this qubit with Alexei Kataev, who was at

Microsoft at the time, and colleagues, collaborators at several universities. The ideas of this qubit have come out, indeed, out of Microsoft and Microsoft Research in particular. Conceptualized here. Yes, exactly. Let's move over to the system of quantum computing. A quantum computer doesn't operate on its own. It needs a bit of an ecosystem. Tell the listeners a little bit about what all has to happen to make a quantum computer work in our constrained world.

Here again, when we think about the quantum ecosystem, Microsoft has a very unique approach. And our approach is to say that we want to build a comprehensive, scalable quantum system. We don't want to just build the quantum computer, you know, that's going to enable 50 or 100 qubits. We want to enable the quantum computer that's going to scale to thousands, 10,000s, hundreds of thousands and beyond. That will unlock truly amazing solutions to the world's most challenging problems. So when we think about that ecosystem, what does it include? At the one end, we have to have a qubit to store and compute the information quantum mechanically. That qubit sits at millikelvin temperatures in our system. So to put that in perspective, that is very close to absolute zero. And so you have to

actually engineer first, you have this dilution refrigerator around the whole system that maintains this cold environment. And this is also to help protect the qubits against noise, against interacting and entangling with the environment. At the very top of this, you know, think of it as a system or stack, we want to ask the quantum computer to do something, right? We want to program it. So we're writing quantum algorithms. And I don't know about the rest of the listeners, but I write quantum algorithms in room temperature environments. And so I'm going to have to interact with this system that's sitting at almost absolute zero,

you know, close to minus 459 degrees Fahrenheit. And somehow we have to get that program, the control, the information, everything down to that chip. That means that part of the system requires an extensive control plane. And we also have to ask to run the quantum subroutines, if you will, from a classical program. So we still need a classical computer. And that computer sits at, say, 4 Kelvin. And so we have to engineer and design and develop not only the quantum chip, but this classical computer that is unlike any classical computer we've ever designed before. It's operating at temperatures we've never really tried to operate classical computers at

before. So we're designing that. We're designing the quantum chip itself and then all the software that connects all of this. And then from there, you have to compile that program to this system. And so we're developing not only a complete hardware stack, but also the complete software stack. So it sounds like you have a multidisciplinary expertise needed to pull this off. Exactly. So another key unique feature, I believe, of our quantum program here at Microsoft and Microsoft Research is really our team. So we have assembled this amazing global team of experts that have come together, coming from, you know, various fields. It's not just physicists, you know, you also need computer scientists, you need electrical engineers, you need cryogenic engineers, right? You need the developers, the software engineers, right?

And I go on and on. And we've built this amazing quantum dream team that's global, sitting in more than 10 countries around the world. And together, we're driving forward a comprehensive quantum system and looking at it from every angle. And the people that sit on this team are incredible. You know, Michael Friedman, who helped start the whole quantum program at Microsoft, he's a Fields Medalist in mathematics, right? We have someone on our team who founded Cray Computing. We have another person that's been the key architect at Intel. We have experimental physicists that have been professors at Harvard and Delft.

And we have people that have won awards at Harvard and Delft and we have people that have won awards in computational physics right you have all these amazing absolutely inspiring minds working on this problem together including yours well thank you talk about cool when we talked before you told me that quantum computers already actually exist on a small scale, real ones, not simulated environment quantum computers. How well are these working now and what are the challenges to scaling these systems? It's exciting to see that we have quantum computers in a very small prototype setting in existence today. Now, those quantum computers we have, they rely on this type of qubit that is similar to this brick

notion I mentioned, right? And so what happens is their error rates are not robust enough to scale without having a lot of overhead to achieve scale. And when I say a lot, it means, you know, for each qubit that I would express at the algorithm level, I may need somewhere between a thousand to ten thousand physical qubits in my quantum device to represent just one at the algorithmic level. We call that a logical qubit. Now with the topological qubit, the type that we're working on here at Microsoft, the idea is that you don't need that overhead. It's built into kind of the hardware level, if you will.

It's fundamental to how we design our qubit from the beginning. And so one topological qubit, we believe is going to be equivalent to roughly 1,000 to 10,000 of these other qubits. That's orders of magnitude. Yes, indeed. Is there even a hint that the topological qubit is ready for primetime anytime soon?

Well, we're working, of course, actively and as quickly as possible to develop a working prototype for the quantum computer based on topological qubits. That would be big. Right now, we really think it's imminent. This is just an exciting time to be in this field and in this place, isn't it? It's thrilling. It's like a total dream come true. Talk a little bit about the impact that quantum computing could have on things like AI and machine learning. Sure. So, you know, of all things, when I joined Microsoft Research a little over 11 years ago, one of the first things I worked on was machine learning here. And so I worked with the Bing team and wonderful people there on ranking algorithms for web search and other

things as well. And it's pretty remarkable that now, you know, at that time, I had done a PhD in quantum computing, and then gone on to learn some machine learning at that time, you know, I would not have expected that we would then be merging those two fields in this short, you know, roughly a decade period of time. And what do you know, I think six, maybe six or so years ago now, we started working on our first quantum machine learning algorithms. And I really think we've pushed the boundary there in the last few years. So we're really excited to be able to show speedups in quantum machine learning. So by using a quantum algorithm to do training of, say, a deep neural network or a Boltzmann machine or a perceptron, you can get speedups in terms of those training

times. I think what is perhaps even more fascinating is that in quantum machine learning, you have this device, you have this quantum computer at your fingertips that can model nature, which goes back to things that Feynman referred to, you know, this amazing entrancing capability of nature to do things. With a quantum computer, you can be able to actually compute closer to how nature computes. And when we think about that, you know, what does that mean for unlocking patterns or, you know, models in speech and language and vision? I think it's pretty remarkable. And what we see is that when you add some of these quantum properties or quantum terms to models, this can greatly enhance the model of the data or the model of the information at hand. You actually get fundamentally different things out.

Right. So on the same note, tell me what the potential impact is on things like cryptography, security, privacy, because on that side of things, it seems in my mind anyway, that there's great promise and also great risk. Right. So one of the earliest quantum algorithms we have, dating back to 1994, is this algorithm by Peter Shore that shows that you can perform what's called factoring on a quantum computer efficiently. Now, why do we care about factoring? Well, this is what underlies

pretty much every digital encryption scheme we use, pretty much every crypto system on the internet. And of course, one thinks, oh, wow, right, a quantum computer is going to attack and break this crypto system, right? That feels like a threat. However, the great news is that there are cryptosystems that aren't able to be attacked by a quantum computer. And this is something Microsoft Research is also doing. We have a whole team here looking at what are those protocols that are going to be robust against quantum attacks, against quantum computers breaking them. And these are classical systems, classical systems that can easily and readily replace things like RSA. These protocols have been put forth. Brian Lamacchia's team in Microsoft Research, for example, has been working on this. And they have protocols that they believe are robust. And of course, it will take, you know, maybe a decade to switch over the

software, all the calls to RSA in our present software. But if we start today, we will take, you know, maybe a decade to switch over the software, all the calls to RSA in our present software. But if we start today, we will be, you know, fundamentally ready. So when a quantum computer of the size required to break RSA exists, we won't be using it to break RSA. We're going to be using it to solve, you know, these amazing problems we have in the world, you know, clean energy, food production, so forth, carbon capture, because we will have robust security systems. So you're looking at both sides of the issue, both making it and preventing it from being malicious. Yeah, we know that there's a way to prevent this, and that's by bringing forth what we call post-quantum crypto methods.

These are the classical methods that will be robust against quantum attacks. And they exist and they're being well studied. NIST, for example, in the United States is looking over these protocols and is going to put forth which one, you know, one should switch to. Interesting. That's all happening in real time. That's cool. You said one of the good things about quantum research is that it's revealing issues both about quantum computing and classical computing at the same time. So what kinds of things are you discovering? Yeah, what's great is when you start studying another field, often it lends these

amazing insights back to some other field. And in this case, when we look at quantum computing, it's pretty amazing. We've learned a lot about classical computing. Actually, when we started working on quantum machine learning, we thought, oh, my gosh, we found an exponential speed up over the state of the art for quantum computing, right? A quantum algorithm to do machine learning. And when we dug under the hood, we realized, oh my gosh, these ideas apply to classical machine learning. And so we sped up the classical algorithm. That's a great outcome of studying quantum computing. And this is now happening, you know, that's just one example, but this is happening repeatedly in machine

learning, optimization, theoretical computer science. We're learning more and more about classical functions, about how to do classical optimization, and we're improving those methods greatly, drawing from ideas that we're learning because we're looking at quantum computing and quantum algorithm design and so forth. What was your path to Microsoft Research? How'd you end up here? So I did my PhD at Columbia University in computer science, and my advisor there was a huge fan, or is, I should say, is a huge fan of Microsoft Research and had suggested I look at Microsoft Research as a potential place to come. And at the time, I was doing a PhD in quantum computing. My dissertation was on fault-tolerant scalable quantum computing, which is what I'm still

doing today. And so I became quite interested in looking at Microsoft. I had heard it was this incredible environment and it employed the most PhDs in computer science of anywhere on the planet. In school, you often think an academic university position is the only place that you can go forward and continue research and publication and collaboration and so forth. And yet here was this amazing, amazing institution that had more PhDs and more people studying computer science and how to use it to really change different aspects of our technology landscape. And so I thought, wow, that looks like a pretty amazing place to look at. So I

looked at it and what do you know, I ended up here. Yeah. I forget who said it. I think it might have been Michael Friedman, that Microsoft is the largest curiosity-driven research enterprise in the world. Oh, I love that. I think that curiosity and creativity, those are two things you see so much of here. And such an enthusiasm to learn something new and to work with others that aren't quite in the same, you know, vein of research you're in. There are so many amazing people. They're all here right at your fingertips. And the more you reach out, the more you learn.

And again, it goes back to contributions going all directions, right? That helps us advance not only quantum computing, but it feeds back ideas to their line of research. And it's just this incredible growth of ideas and output and impact. So you can't go down and get coffee without bumping into three geniuses. Right, which is a great thing. It's not the coolest thing. It's a treat to go get coffee around here, right? Because you ultimately end up in some incredible conversation and you learn something new every single time.

I was going to say, I bet the water cooler conversations around here are sort of on another level. Right. Sometimes. And I would have to say it's encouraging for me to note on the blog that announced the kit, the dev kit that just came out, there's a quote that even some of the smartest people in the world confess they don't understand quantum computing. And then when I clicked on that link, because it was in blue, it took me to Bill Gates and Satya Nadella. It was like, what? I don't feel so bad. As we close, I have to ask, is there any work going on in the quantum snack food lines of research? Because I'm always looking for something yummy that's both more filling and less filling at the same time. Exactly, right? Something that's good for you and bad for you. Right? More filling tastes great. Less filling tastes great. Actually, I do want to circle back to another question, because I think every time we talk about something that has the big potential for good, we also have to look at the big potential for bad. And quantum seems the most intimidating because how hard it is to understand, but also what the promise is. Talk to me about the length of time needed to solve, say, an NP-hard problem versus what we can ask today of a computer. Yeah. So when we think about quantum computing, it's important, I think, first to realize that

quantum computers aren't going to replace all computers we have today. It's really going to solve a special class of problems, a certain class, kind of similar to how we use supercomputers, right? We're not using supercomputers to do our email or Word documents or, right? You're not? Yeah, I wish. I mean, maybe I'd get more email done. But, you know, it's really a special purpose device. We don't send every subroutine to a GPU when we're programming, right? So, you know, you have to look at what problems you want to send to a quantum computer. And not every problem will be sped up by a quantum computer. But a pretty incredible set of problems are sped up by a quantum computer. And that happens to be

problems in the areas of, you know, as we talked about machine learning and artificial intelligence, but also the simulation of physical systems. So studying quantum chemistry, studying problems in the space of like catalysis, potentially medicine, health, optimization problems. These problems are great candidates for getting a quantum speed up, as we say, or, you know, for getting a fundamentally different solution. And the unlock there, really, the problems we're looking at solving with a quantum computer are the problems that today require age of the universe timescales, right? Wait, wait. Yeah, age of the universe. I'm not going to be around for that solution.

I don't know about you, but... I might. But age of the universe timescale is a very very long time to wait for a solution now you know problems as i just mentioned in catalysis like when we look at trying to understand how a catalyst will improve a reaction rate some of these problems literally require billions and billions and billions of years to solve using not just you, today's supercomputers, but post-exascale supercomputers also will see the same type of time requirements. And on a quantum computer, what we can show and what we've shown in some recent research is that you can solve some of these problems in a matter of, say, weeks, days, hours, seconds. All of those timeframes seem very doable. I will definitely,

you know, I'll be around for those solutions. So going from billions of years to a week or a month, you know, we train machine learning models for months all the time. So we're looking at really bringing down, you know, really unlocking solutions for some of these problems with quantum computing. So is there anything in what you've just said that keeps you up at night about that information? Well, what keeps me up at night is I want us to get there faster. You know, I think about, gosh, you know, how can we unlock even more the ability to program this device? Because the potential is so huge. And I think that's so exciting that we could potentially combat global warming with a solution coming from a quantum computer in conjunction with

solutions from classical computers, right? That's pretty amazing that this device could help us do that. It could help us better produce artificial fertilizer and in turn help food production. It could help us better model machine learning. And the unlocks are so huge that what keeps me up at night is just, I want to, there's not enough time in the day to get all of this done. You know, you want to work so fast at this because it's so promising and so exciting that it's hard to want to go to sleep. That's awesome. That's a closer. It's hard to want to go to sleep.

Yeah. Oh, Krista Svori, thanks for coming in today. Well, thank you so much. It's inspiring to hear your passion and cool stuff you're doing. Yeah, well, everyone go download the Quantum Development Kit, and we'd love to see a quantum community really, really develop here. No pun intended. Yeah, exactly.

To learn more about Dr. Krista Svori and Microsoft's distinct approach to quantum computing, visit Microsoft.com slash research.