D-Wave is leading the market in quantum computer hardware and software. CEO Vern Brownell tells MarketBrains why finance in particular is a strong market for this type of technology.
MarketBrains: What led you to quantum computing?
Vern Brownell: I ran technology for Goldman Sachs globally back in the 90s, I left in 2000 when finance was very different than it is today. Over the years, I’ve seen many exciting new technologies, but I hadn’t seen anything more interesting than D-Wave and I felt I had to be part of it. I joined D-Wave eight years ago. It is a hard project building a quantum computer, but I think it’s the most important technology going on in the world today.
I am not a scientist, so I am at a disadvantage sometimes here. I am just a plain old engineer and IT guy.
MB: But I hear it’s an engineering problem now?
VB: It’s becoming an engineering problem, there’s still deep science involved for sure. The engineering problems are becoming increasingly complex. Our systems have to run at these amazingly low temperatures – 150x colder than outer space. They need to run in these pure magnetic vacuums and develop these superconducting chips that are unlike anything else. The hardware aspect of it is off the charts complicated.
“It will increase in power exponentially and become almost necessary to implement once other competitors get this kind of technology”
MB: What are the applications right now, particularly in finance?
VB: I think there is enormous opportunity in finance for everyone from the big bulge bracket firms to the big hedge funds, and ultimately everybody in this space. The leaders will be the ones that are the most advanced computationally today and are looking toward using this kind of technology.
It’s an exciting time because we are really at the dawn of this industry and when finance picks up a new technology, it implements it very quickly. We hope that will be the case as well for quantum computing.
MB: I actually hear that sometimes big corporate mindsets are not always so great at adopting innovation…
VB: It will increase in power exponentially and become almost necessary to implement once other competitors get this kind of technology. There are probably still a few years before quantum computing becomes something that’s used by a wide variety of people in finance.
MB: So, like low latency technology, the way it’s now a must have?
VB: Exactly, and the same thing will happen with quantum computing I believe, but it’s going to take some time.
“We ain’t seen nothing yet on the AI side”
MB: I hear that D-Wave’s technology, the quantum annealer, is specific to solving optimization problems?
VB: That’s partially true. For classical computers, you would say everything has to be put in terms of a multiplication, division or addition and subtraction. Classical computers basically work at these very low level operations, or “assembly language.”
The quantum computer uses a much more powerful low level assembly language, if you will. That low level primitive is an optimization problem. It’s really using nature to do that optimization problem because quantum mechanics are the most fundamental laws of how the universe operates.
Much like we don’t think about assembly language and those low level primitive likes multiplication and division anymore when we use our computers, it will be the same with quantum computing.
You won’t think about optimization problems when using a quantum computer, you will think about solving AI problems or value at risk calculations using Monte Carlo simulations – things that are already being done and we’re familiar with. It’s just a different, more powerful, approach to those problems.
Sometimes folks like IBM say, “Oh, D-Wave can only do optimization problems,” but that’s absolutely not true. We have solved many interesting problems that aren’t just optimization problems, some of the most notable are in the AI space.
If you think about machine learning, or deep learning in particular, a lot of what’s buried in the training that goes on is optimization problems.
You are trying to find and build the best model, and the quantum computing capability can be integrated into algorithms to provide very effective, even better model building, than you can with today’s AI. We ain’t seen nothing yet on the AI side.
“There’s a lot of exploration of early applications of AI in finance, but frankly I think they are a bit behind the other industries”
MB: Can you give me an example of what you mean?
VB: That’s a tough question because a couple of the banks that we work with are being completely proprietary about what they’re doing. They think they have developed an algorithm that is going to provide market advantage, so they’re not talking freely about that.
But I can tell you that we have seen people explore portfolio optimization and different forms of Monte Carlo simulation. In fact, Monte Carlo is the largest computing workload in finance today. It certainly was when I was at Goldman. We spent millions of dollars a year just running Monte Carlo simulations, and now the regulatory agencies are mandating more simulation.
There’s a lot of exploration of early applications of AI in finance, but frankly I think they are a bit behind the other industries. They are certainly behind what the giants like Google, Amazon and Facebook are doing in AI.
But it’s really starting to ramp up dramatically now, you see every bank trying to explore AI applications ranging from fraud detection, to things like exploring whether you can use historical data to find trends, or even intelligent customer interaction for the big retail banks and so on.
There’s a pretty wide range of AI activity, although I think it’s still pretty early in finance.
“A lot of quants are physicists, or certainly mathematicians or computer scientists, and what we’ve seen is they instantly get the value of quantum computing and particularly the D-Wave quantum annealer”
MB: And where does quantum computing fit in?
VB: We believe quantum computing will add to the way that AI is done today. The particular kinds of algorithms that we’ve developed for our platform do something called generative machine learning. You can train with less data, and less labeled data for instance, which is a problem in finance, the cumbersome process of data clean up — it’s both a problem and an opportunity.
The other piece of it is understanding how the model is built and what the model is actually doing. We believe that this kind of AI that we are working with in conjunction with the quantum computer will be better than the methods that are used today.
That’s quantum machine learning (QML), which is the fusion of quantum computing and machine learning.
MB: You’ve created a group called Quantum for Quants, how is that going? Do you have traction?
VB: A lot of quants are physicists, or certainly mathematicians or computer scientists, and what we’ve seen is they instantly get the value of quantum computing and particularly the D-Wave quantum annealer.
Going back a few years now, we have worked with a number of quants and thought that it would be a good idea to share algorithms, or insights into how to use this kind of capability. That was the original genesis of the idea.
Not as many algorithms have been published as I would have liked to have seen on the website, but the idea of getting quants together across the banks, and then hedge funds, to look at ways to do this is very important.
Another aspect of this is we try to present, or do training, at some of the global quantitative finance conferences. It’s partially outreach and education, but ultimately I hope it’s an opportunity to share algorithms and ideas across banks.
“You can use higher level abstractions and interfaces so that programmers who don’t have a background in quantum computing can understand”
MB: How many people can work with these algorithms, what kind of population are we discussing here?
VB: Today, we’re talking about really just the highest end quants in the banks and hedge funds. What we are working on, particularly in the AI area, is providing algorithms and libraries that are usable by anyone who is a data scientist, machine learning engineer, or anyone that has that kind of background.
Most of the work on Quantum for Quants has been more fundamental algorithm research. For example, how you use it directly for a portfolio optimization. It’s been less about the intersection of quantum and machine learning, but we will add that as well.
Machine learning is just another tool, and quantum machine learning is as well, that quants, engineers, and software developers can use. We want to make that as ubiquitous as possible, but it’s not quite at the point yet where everybody has access to it.
We have to create a developer ecosystem for technology like this. You want the technology to be ubiquitous so people can use quantum computing anywhere, through cloud access and so on, and we’re working on that. We already have that capability but we want to broaden it out.
But maybe more importantly the tools and the software stack need to be there because that makes it easy to use, so you don’t have to be a quantum computing expert to use this at all, or even an AI expert to use this technology. You can use higher level abstractions and interfaces so that programmers who don’t have a background in quantum computing can understand.
This interview is edited and condensed