We're kicking off 2021 with a new interview series: GOTO Unscripted, with our first round of interviews recorded back when we could still meet in person. GOTO Unscripted takes our conference speakers off the big stage and brings them behind the scenes for an intimate conversation on topics they know best.
Is quantum computing just a theory, or is it actually applied in practice? Join Murray Thom in an interview with Preben Thorø to discover the real-life implementations of this technology. You’ll learn how you can leverage D-Wave’s platform as a developer, which is open to all researchers to help solve health problems, and learn how quantum computing has evolved from a theoretical discipline to an applied science that helps researchers and developers solve complex problems 3 million times faster, without consuming any energy.
Murray Thom: My name is Murray Thom. I'm the vice president of software and cloud services at D-Wave. I've been with the company a little over 18 years, and I'm primarily involved in our software side, our lead cloud services and our quantum application environment.
Preben Thorø: So for 18 years, may I ask when D-Wave was founded?
Murray Thom: D-Wave was founded in 1999 by 2 graduate students at the University of British Columbia. I joined the company just a couple of years after that in 2002.
Opening D-Wave platform for COVID research
Preben Thorø: I'd like to dive right into this. When the whole COVID pandemic situation exploded one year ago, you launched an initiative, like you opened up your platform for research, COVID-related research. Could you mention a couple of examples of the outcome? What came out of that?
Murray Thom: Yes. Absolutely. Early on in the pandemic, we, like everyone else, were sort of learning about it and focusing on adapting. We'd actually been having some conversations with the Canadian government who had asked private industry to open up the resources that they have available to others who were developing solutions in response to the pandemic and we, like many others, decided to act quickly and do that. So we actually created an opportunity for people to sign up and get free unlimited access to our quantum computers and our hybrid solvers for them to be able to build and deploy applications that could be used on that with quantum hybrid compute technology. We did that. That was available to all of our users and open-source developers in 35 countries.
We also partnered with organizations like Denso, an auto parts manufacturer in Japan, Jülich Supercomputing Centre, NDR, Volkswagen, Menten and Sigma-i, to basically not only help them to be able to build responses to COVID-19 challenges, but also so that they could work with developers and forward-thinking businesses who were looking to adopt technology and see if we could help them. So we created that developer community, those who have experienced programming on quantum hybrid technology and those who had a need associated with the challenges in the pandemic.
Recommended talk: GOTO 2020 • A Quantum Computing Lab Tour at D-Wave • Murray Thom
Two interesting projects that rose off of that amongst a variety of different projects that people worked on, Sigma-i has developed a variety of formulations for taking challenges to do with logistics and optimization that arise from the pandemic and formulated them for a quantum hybrid resource, including ones which basically looked at the optimization challenge of assigning patients to hospitals, knowing sort of like how many resources the hospitals have available and what's the distance between the patients and the hospitals that they need to visit. They also developed other ones, especially with helping businesses to be able to adapt a new normal schedule for their employees who had to work partly remotely and partly in the office. Another company, Menten AI, has been developing protein therapeutics. These are proteins that have brand new designs for tackling challenges associated with medical therapeutics, and they do this not by taking existing proteins that are used in these contexts and modifying them, but rather by designing them amino acid by amino acid, and they were able to use our quantum hyper technology to design proteins. They got them computationally validated with protein folding, chemically synthesized, and there are now quantum hybrid-designed proteins that are out in live virus testing against COVID-19 in the lab.
Why is your platform is uniquely suited for that?
Preben Thorø: Why is it that your platform is uniquely suited for that?
Murray Thom: It’s sort of a question of what are the challenges with programming that so many applications are facing in the world today? A lot of developers and businesses are looking at an application and trying to figure out, how do you basically map the business problem that you're trying to solve to the computational resource that you have for tackling it? If you are a developer, you want to do that in the most natural way possible where you're taking the real representation of the world problem into how you're tackling it with your computer program. Some of the features like everyday life, features where you have to make decisions on optimization, I need to purchase this or not, I need to select this entity to participate in my solution, those events either happen or they don't happen. They're discrete. They don't happen. For example, if someone joins your team to play like a hockey game, you don't get 10% of them on your team. They're either playing on the ice or they're not playing on the ice. That discreteness actually makes it a challenging problem for a classical computer to optimize, and even more so, like many things in our life, they are interrelated. So these choices about where patients are gonna go to hospitals, these individual choices affect the subsequent choices that are gonna be made in that optimization. That feature of discrete choices that are interrelated makes these problems very challenging to solve with classical computers, and at D-Wave with our Leap platform, we've developed hybrid solvers which excel at optimization problems like this.
What is quantum annealing?
Preben Thorø: You're using a technique called quantum annealing, right?
Murray Thom: Yes.
Preben Thorø: What exactly is that?
Murray Thom: Quantum annealing is an approach to use quantum mechanical properties to help you solve a problem. I think most commonly when people are talking about what is a quantum computer, they will tell you, "Okay, we've got this idea. We want to leverage quantum mechanics with computation. We're going to take the fundamental building block of a regular computer and just modify it slightly so that now it can have quantum mechanical properties. Instead of just being at a zero state or at one state, it can be in both states simultaneously." A lot of developers are sort of wondering, "Okay, well, how am I going to make use of a variable that's in two states at the same time when I'm trying to solve a problem?" The notion of quantum annealing is to effectively use that superposition to allow you to search a solution space.
So a developer will say "Here's a problem I'm trying to solve. Here's an optimization challenge that's stated from our application space, and we need to look through a variety of different solutions in order to be able to find one that is optimal for the problem we're trying to solve." In that scenario, that quantum annealer is effectively doing this process where it turns up the quantum properties of the computer, and that actually causes it to search a solution space by spreading itself out very broadly in the solution space.
That solution space can be larger than the estimated number of particles in the universe. That's the ability it will have. Then slowly turning down that quantum annealing and sort of really localizing towards the best possible solutions. Very much as if you were thinking about this in a classical system, it would be like you move to high temperature and then you gradually turn the temperature down as we have with thermal annealing. In quantum annealing, you turn up the quantum search, and then you gradually tune that down, allowing it to localize to the highest quality solutions. It is being able to be in multiple places at once so that it can move between those solution spaces quickly, all the quantum properties that it has there basically reinforced the task of finding those high-quality solutions.
The history of quantum computing
Preben Thorø: To me, this sounds like pure science fiction. But you have been around for a little bit more than 20 years. Could you tell me about the history of quantum computing?
Murray Thom: Yes. Absolutely. Quantum computing is something that I think we all want to get a bit of an idea around. Imagine taking ourselves back to the 1980s where people were using computers to try to solve an expanding variety of complex tasks facing society at that time. There were quantum physicists back then who were trying to simulate quantum properties and quantum materials with computers and they found there's no way we're going to be able to do this. Not only could they not do it in the '80s, they knew they were never going to be able to do it because quantum properties require such a huge number, an exponential amount of classical resources that it was always going to be out of their reach. And they said, "You know, if we actually built computers to harness those quantum properties, this would all of a sudden become possible."
Richard Feynman really popularized that idea in 1982. In 1994, Peter Shor, building on some of those ideas said, it's possible to write algorithms with computers that have access to these quantum properties to solve some really important problems like factoring numbers, which is actually the basis of all of our sort of encryption mechanisms that we use this for transferring information around over the internet. That protocol really got a lot of people interested in quantum computing. In 1998, some researchers at Tokyo Tech proposed an idea for taking optimization problems which were being approached by a technique called thermal annealing, where you would move through a solution space with high energy accepting lower quality solutions and then gradually reduce that temperature so you kind of settle down to high-quality solutions and they said: "If that concept of thermal annealing is helpful for solving problems, we can actually do that with quantum annealing and give it quantum properties and help that search for solutions."
Recommended talk: GOTO 2019 • Fueling the Quantum Application Era with the Cloud • Murray Thom
One year later in the United States, a group of researchers actually demonstrated with quantum materials, they could do both thermal annealing and quantum annealing, and they were able to demonstrate at large scales with no error correction that quantum annealing did, in fact, accelerate the obtaining of high-quality solutions and low energy configurations of these sort of optimization challenges. So since that time, D-Wave has been building effectively like a programmable version of that, a version which basically allows us to use traditional fabrication techniques to build devices that we can control so that we can create sort of connections from common optimization problems like the ones associated with allocating resources or optimizing and designing proteins, and putting them into a programmable quantum system where its quantum properties allow it to very quickly work its way to high-quality solutions. That's quantum computing over the last 40 years and sort of a quick snapshot. That approach is key for practical quantum computing because, although some of the original lead came from quantum physicists who wanted to discover sort of quantum material properties, the larger, broader impact is our ability to leverage those quantum effects for optimization challenges facing business and finance and manufacturing logistics as well as in the pharmaceutical industries.
Is quantum computing energy efficient?
Preben Thorø: You have mentioned big numbers, complexity, the huge magnitudes, but still quantum computing is very, very energy efficient. Isn't that right? How can that be?
Murray Thom: In fact, I have some quantum computers actually sitting next to me right here and you can see the computer chip up in the monitor here next to me. So one of the fantastic features about the quantum computers that we're building is that the actual metals that are used to print the circuitry on the chip are made so that when you cool them down to a really low temperature, they can superconduct, which means they can conduct electricity with no loss whatsoever. When you think about the computer chips that are in our phones or in our laptops, their operating temperature is room temperature and when you turn them on, there's resistance and then they heat up. They warm up and then you have to use a lot of energy that's going into that heat and then use a lot of energy to pull that heat away so they don't overheat.
In the quantum computer case, the operating temperature is very low, but when you turn them on, they don't heat up because they can flow these currents with no resistance. So while you have, let's say a refrigeration system that just uses a constant amount of heat, no matter which quantum computer you're using, the quantum processor itself, the one that I have in the picture above me, uses 1/10th of a millionth of a watt.
It has actually been that level as we've been progressing through different processor generations. So as we are growing the scale and our ability to handle complex problems with quantum annealers and quantum computers, we're not actually having to scale up the power requirements of the system, and that's a transformative change in terms of high performance compute because typically when we're using computers that have larger expanding scales, our society is using computers much more broadly in an automated way, it has huge power consumption requirements. In fact, there are power plants set up to supply power to data centers. These new technologies could fundamentally change that where we could offset big portions of the computational tasks to a technology that consumes virtually no power.
Preben Thorø: So what you're saying is that once we have cooled down that environment, I could add another chip so I would not even double, but multiply my computing power with a huge factor without spending more energy?
Murray Thom: Yes. Think about it from this perspective: over the last seven years, the levels of integration on our classical computing chips have been increasing exponentially each year. With quantum computing, we're actually able to leverage many of those same fabrication techniques. At D-Wave, over the last 15 years, we've been doubling the size of our quantum processors every sort of 18 months to 2 years. That's an exponential growth in the size of the quantum processor. But what's interesting is that the problem itself is growing exponentially with its size. So the performance opportunity here is sort of like super-exponential growth. This is something that's quite dramatic. If you have that scenario where you're taking one processor and then you double it in size, the computation for a classical computer can become square as complex. However many steps of data, you have to multiply that number of steps by itself to then tackle that new, larger task with a classical computer. So that can mean a huge growth in terms of the classical computer you need to tackle that problem and the corresponding growth in terms of the power consumption needs. Whereas for the quantum system that the quantum computer chip is in, you've taken, one chip and then you've just doubled it in size, it becomes much more powerful in terms of the problem complexity you can deal with, but its power consumption has effectively remained flat.
Where is quantum computing heading in the next 10 years?
Preben Thorø: So where do you see this going over the next 10 years?
Murray Thom: Well, that's a great question. I mean, what do they say? Those who gaze into the crystal balls are destined to eat broken glass. But quantum computing technology is gonna allow us to take problems, which are beyond our reach. There's sort of the reason, beyond our imagination with classical computers, and bring them into the realm of the possible so that, although we've seen so much scaling in terms of computational resources in the last 10 years that have allowed us to open up into so many new areas, many of us are still familiar with some challenges that have remained inaccessible, challenges associated with really effective pattern recognition or the ability to do really high quality, like personalized medicine. Also, basic logistics problems, scheduling teacher schedules in a school or scheduling nurses in a hospital, allocating resources in response to the pandemic, even at small scales, remain very challenging for our society and that's because the nature of the problem themselves requires optimization over discrete variables that are interacting with one another.
So in the next 10 years, our ability to solve problems in that space becomes much stronger, we're not going to need to sort of approximate these problems anymore. We're not necessarily going to need to reformulate them into something that's better suited to our computational resources. Our new computational resources are now going to be well-adapted to the more natural problem. So I'm hoping that things associated with really high efficiency, like business and logistics operations, things where airline schedules become simpler and easier to run, and the cost associated with logistics of a business in terms of delivery schedules, vehicle routing, those types of things become really easy. So in some sense, it will become easier to use these technologies in everyday tasks, even the ones that are extraordinarily complex. I don't know if that's the right way of thinking, like what might become possible, but those are some thoughts on that space.
How can a developer benefit from the platform today?
Preben Thorø: But all of this is, in fact, possible today. So how can I as a developer benefit from your platform? How can I join your community today?
Murray Thom: Well, I'm really glad you asked me that. So for developers and forward-thinking businesses who are excited about the prospect of quantum computing and they're looking at how do I begin writing quantum hybrid applications, you can actually just search D-Wave Leap or D-Wave Cloud, or D-Wave Applications and find a link to our Leap Cloud platform. Anyone can sign up right now. We have plans as low as $0. You can get in and have a month of free access, run some of the interactive demos, and go through some of the open-source examples to see how applications can be mapped to quantum hybrid resources to build applications. We have a community portal with users who are experienced with programming the systems. They're actually posting questions and responses to one another to help people as they're building applications and a help center which actually does a search over all of our documentation and information online about how to use quantum hybrid resources for building applications.
We even have an integrated developer environment right in that cloud platform to make it really easy to program the system. If you are building an application that you are interested in and willing to open-source, we'll actually renew that access every month afterward. For those businesses who are not interested in open sourcing their projects but they're actually looking to build businesses around what is now becoming possible on quantum computing technology, we've got plans that allow you to get subscription access that are on those business-to-business terms.
Programming languages supported
Preben Thorø: Which programming languages do you support with your platform?
Murray Thom: We have an open-source software development kit called Ocean. It's written in Python. It's actually a combination of Python and C++ in order to be able to make the programming really easy and accessible for Python programmers, but it also has components that are written efficiently for some of the more computationally intensive tasks.
Developers can find that we have our tools, open-source. We've got a set of D-Wave examples open-sourced with documentation around them. Those are things that you can then load into our cloud services quantum application environment and begin programming an IDE right in your browser to help you get started so that you can find examples that are the closest to the applications that you have in mind, load them, review them, look at their references, and then start modifying them, and import in your data so that you can get a quantum application built quite quickly.
Preben Thorø: And do we have debug facilities like in any other IDE? How rich is the environment here?
Murray Thom: Yes. So it's a fully-featured integrated developer environment. We've actually added extensions that integrate it into our cloud service so that you can easily move back and forth between the resources that you have available in our cloud services and the developer environment itself. It's based on Thea. You can enter into debug mode and actually enter into the code. We've also got extensions to our libraries that allow you to dig in, and you can actually see the quantum machine instructions that are being submitted, inspect the sort of instruction that's being sent as well as the data format of the answer coming back. But also, I think a lot of developers are really interested in what can we build quickly with this system? So we have hybrid solver services. These are services that will actually deploy computers in the cloud and start working across the problem that you're solving with the best of classical and quantum computing technologies.
So it will actually break the pieces out of the problem and submit them to the quantum computer for you, and that allows developers to work with problems up to like a 1 million variable scale, really getting up to full scale for quantum application development. I think though, we don't wanna divorce developers from the actual excitement of the underlying technology. So we also had features that will allow you in the debug in the developer environment, where you can look at those calls in the quantum computer and actually visualize them, and you can see the connection between the problem that you submitted and the actual instruction that went to the quantum processor graphically in front of you and explore them as well as the set of solutions that you got back from the quantum processor. So I think it's a combination between the utility, the function of getting something done, and also the form of seeing an exciting interface, learning about it, and really getting a chance to interact with quantum computing technology and programmable quantum systems very early on in the development process.
Quantum computing more efficient than classical techniques
Preben Thorø: I know that you have recently published a research paper in "Nature Communications". Could you tell me, what was that about?
Murray Thom: That work was done in collaboration between D-Wave and researchers at Google. We were looking at simulating a quantum material where its properties were occurring because of quantum mechanics. So it actually entered into a new phase because it had quantum mechanical properties that actually affected the way this quantum magnet was working. In the study, we both did the experiments using the quantum annealer as well as some really optimized classical code which is sort of the best alternative way for solving that problem. In that research study, they were able to demonstrate that it was 3 million times faster to do that with a quantum computer than it was with the classical techniques. And what's really significant about that is that that is an advantage that has been demonstrated on a problem which is an example of a practical application problem. Material simulation is one of the applications that people are really excited about with quantum computing. Along the path of developing quantum computing, that is one of the milestone markers of the development of the systems and their capabilities today.
Preben Thorø: So that ties into the original thoughts from the early '80s, right?
Murray Thom: Yes. Absolutely. In fact, many of the origins associated with quantum computing were: can we build computers that will allow us to simulate the properties of quantum systems? Because it was known that with classical computing technology, the resources you would need would grow exponentially as those quantum systems become larger in size, the ones that you wanna study. So this creates a connection to some early lectures that Richard Feynman gave, popularizing the idea of quantum computing through to a set of researchers at Tokyo Tech who proposed ways of using these optimization heuristics and modifying them to incorporate quantum effects that were then demonstrated in the United States in 1999.
So we have basically gone from those early demonstrations at large-scale quantum systems using quantum annealing that can accelerate calculations to a programmable quantum annealing system which we can now use in those applications and demonstrate that advantage.
Preben Thorø: Thank you. This has been amazing. Thanks for joining us today.
Murray Thom: Thank you for having us. It's great to talk to you.
About the interviewee
Murray Thom is Vice President of Product Management at D-Wave, responsible for the Leap quantum application environment, Ocean tools, system software, and documentation. Previously Murray led a team engaged in customer projects related to algorithms, applications, and performance testing. Since joining the company in 2002 Murray has been involved in all aspects of systems engineering and processor development for D-Wave's quantum computers. Some of these project areas include cryogenic refrigeration systems, superconducting electrical filters, cryogenic chip packaging, magnetic screening and shielding, QPU signals, and automated test systems.