Tuesday’s first panel session at Global Derivatives 2016 gave us some fascinating insights into the world of quantum solutions.
First up was Vern Brownell, CEO of D-Wave systems, who took us on a whistle-stop tour of the current picture of quantum computing. D-Wave shipped their first product in 2010 and their latest system, the D-Wave 2X, has 1,000 quantum qubits.
First off, what is a quant computer?
“It’s a device which directly exploits the laws of quantum mechanics to do a quantum computation,” explained Vern. “Built around the concept of the qubit, which is analogous to the digital bit, except that it can be 0 and 1 or 0 and 1 at the same time. When you start to string together qubits you can represent very large numbers and you can also do calculations that are quite powerful.”
One of the challenges of the quantum computer, added Vern, is that it needs to operate in a quiet environment:
“IT HAS TO RUN AT LOW TEMPERATURES, AND IT NEEDS TO BE SHIELDED DRAMATICALLY FROM THE EARTH’S MAGNETIC FIELD. THE TRICK IS HOW YOU SEPARATE THE CALCULATION FROM THAT NATURAL ENVIRONMENT,” HE EXPLAINED.
While there are different ways to build a quant computer, the most popular being the Gate model, Vern and his company focus on quantum annealing. “It’s more focused on optimisation, sampling and machine learning,” he said.
“Ultimately there will be different types of quantum computers in the field, but we felt that quantum annealing was the fastest way to get a quantum computer to market. We’re probably at least five or ten years ahead of anyone else because of that choice,” explained Vern.
What is quantum annealing?
“It’s a method for finding the lowest energy in a two-dimensional landscape,” explained Vern. “You can map your problems in terms of a landscape and then the system searches through this enormous answer space in one clock cycle. There are a huge number of possibilities that the machine can go through in one clock cycle. Ultimately, this is in the service of trying to find the lowest energy state or the best answer, sorted by optimality.
Two particular advantages of quantum computing that Vern pointed out were the fact that the machine uses no energy, and its affect on Moore’s law:
“WE’RE USING MATURE IN A VERY EFFICIENT WAY. THIS MACHINE USES NO ENERGY, BECAUSE THE CHIP RUNS AT SUPERCONDUCTING TEMPERATURES, IT DISSIPATES NO HEAT AT ALL.
“The other advantage is that we are seeing Moore’s law flatten out. That slope that we’ve enjoyed for a long time is definitely flattening. This is a new technique that will allow us to extend that computation advantage we have seen over the last decade.
“Ultimately, this is really a device that is meant to solve hard problems faster and hopefully solve problems that are beyond what one can do with classical computing today.”
Why does quantum matter?
“It matters because of the new capability that researchers, scientists, users and developers have never been able to tap into before, because they would just take too long in the classical computing environment, even with GPUs and all the exotic computing architecture that we have today,” stated Vern.
“WE HAVE REAL DEVICES DOING REAL COMPUTATION AND IT’S A REALLY EXCITING TIME. WE ARE AT THE DAWN OF THE COMPUTING AGE,” HE SAID.
Davide Venturelli from NASA Ames Research Center, whose turn it was to venture onto the stage, agreed:
“I really share the same vision on the impact that this technology can have. We’re really at the beginning,” he said.
Why have people got excited about the idea?
“Imagine if today you want to factor a large number that is used for a cryptosystem, it’s impossible. But if you had a quantum computer, a back of the envelope estimation would say that it would take seconds.
“That’s what exponential speed-up looks like – it’s game changing,” said Davide.
He went on to explain how he and his colleagues at NASA are investigating the near term applications of quantum computing in optimization using the D-Wave device and others, “because we have optimization problems, which are related probably to the optimization problems in the financial field.
“We also do a lot of fundamental research, where we craft problems, which are useless from a practical standpoint, but they teach us something so we can feed back what we learn to the manufacturers. Empirical research is really important, we try to understand the limits of the approach.”
Davide concluded his talk by inviting the audience to submit their own proposals to play with the machine at NASA: “But maybe try not to make it too much about finance,” he warned, because NASA and USRA wants brilliant mind to think of problems with applications beyond finance.
Landon Downs, from leading software company for quantum computing applications 1QBit, was next on the podium and told the audience how to redefine their problems using quantum computers.
“What I really want you to think about,” he urged, “is tackling your problems from a different perspective.
“The types of tools that are available to you through quantum mechanics are very different to what you have with classical systems, and what that means is that you want to try and think about your problem from a slightly different angle. Before you try to stay away from NP hard components, you want to stay away from discreet components, non-convex, non-linear components, all those things are a great places to think about when you’re trying to solve a problem with the quantum computer, and those are the areas where you might actually see some benefit.”
Landon then ran through five different applications that are still at the toy problem stage at the moment: applying an NP hard technique to traditional regression analysis; looking at optimal trading trajectories; tax-loss harvesting; AI machine learning; and graph comparison and graph analytics.
“What I’d like to leave you with,” he concluded, “Is that we’re really at a stage now where we can move from theory to application. The tools are available: you can use our stuff for free online. Microsoft, IBM and other groups like that have put up great simulators and they have put cloud access out to their hardware, you can go and sign up. I would encourage you to go try it; go and experiment and see for yourself what you find.”
Finally, the session’s moderator, Marcos López de Prado, Senior MD at Guggenheim Partners, presented some of the applications that he was using for quantum computing. Marcos manages large international funds, and outlined the reason for his interest in quantum computing:
“We think we will be able to monetise the applications that you have heard about so far,” he said. In particular, he went on to explain, four use cases: Dynamic portfolio optimisation; clustering; scenario analysis and option pricing.
“There is this thought from Richard Feynman on why quantum would be needed in physics research. His idea was that, since quantum physics involves probabilistic variables, not deterministic phenomena, it would be interesting to come up with computers that are non-deterministic, so that rather than having to simulate random processes through deterministic machines, we would operate random processes through probabilistic machines. The same idea can be translated into finance. Very often we deal with random variables, and rather than having to use deterministic machines to simulate those random variables, it would be interesting to use the probabilistic machines to solve these probabilistic problems.”
Like Landon, he ended his talk with the call for everyone to get involved, with the announcement of the launch of a new online community, quantum for quants.
“It’s a community and a community needs people,” he said. “We would welcome everybody in this room to begin contributing. We are very excited about getting people involved in the discussion of these topics. We think the more people think about the applications and what are relevant problems for the finance community the more this will help the researchers in the field to lead the discussion and research into helping us.”
The dark side of quantum computing
During the initial talk, Davide had mentioned cryptography, and a member of the audience was keen to know: if we can factorise the number that supports all the encryption security, what happens then?
“It’s a very important question,” answered Vern. “My view is that it’s a not a focus area for application, at least not for D-wave. Having said that there will be quantum computers of this type and other types that will eventually be able to break encryption. There’s a lot of research being done on post-quantum cryptography, so what needs to happen over the next two decades is a movement towards cryptography that assumes that quantum computers exist, but it will take decades.”
“At least one decade,” interjected Davide.
ANOTHER QUESTION OF THE DAY TOUCHED ON HIGH FREQUENCY TRADING, WOULD QUANTUM COMPUTING MAKE THIS WORSE?
“It will make it better,” stated Marcos. “The reason that high frequency algorithms thrive is that many people make short term decisions that need to be corrected later on. Someone optimises a portfolio, of course the optimal solution is not robust, it needs to be fine tuned, and as a result there are many transactions involved. Now we have a computer that can give us optimal trading trajectories, that will require minimum rebalance over long periods of time, that will reduce your need to go to the market and demand liquidity. Ideally the solution to HF trading is not to regulate but to eliminate the incentive that HF traders have to go after liquidity demand.”
Finally, the panel were asked, how would a user know if they logged onto use it in the cloud, that it was a quantum computer at work?
“Why would you care,” declared Marcos, “if it solves the problem?”