If you’ve seen the video below, you know what it takes to run a quantum computer. The only version commercially available today, made by D-Wave Systems, takes a 10-square-foot appliance to keep a 1-square-centimeter chip at a temperature of near absolute zero. It’s not for the meek, or the budget-conscious, which is why the only known purchasers to date are organizations like Google, NASA and Lockheed Martin.
But don’t worry, D-Wave CEO Vern Brownell promises, the rest of the us will probably consume our qubits as a service. That’s right: D-Wave wants to deliver quantum computing via API.
Technology aside, it’s this vision why I’m so excited to have Brownell speaking at our Structure Data conference March 19 and 20 in New York. A big theme of the event is about taking advantage of advances in data processing and delivery models to do new things, and even if it’s a few years off, quantum computing delivered as a cloud service might be about as promising as it gets.
In the long term, Brownell said during an interview earlier this week, the company “absolutely” expects to be in the business of renting cloud cycles rather than selling entire quantum computing systems. “We have the capability already to do that,” he said, noting that Google actually tuned some of its quantum algorithms over the internet. “… It’s really kind of a business decision at this point that we don’t offer that.” (He even suggested it could be done via partnerships with Google and even Amazon Web Services, noting that Amazon CEO Jeff Bezos is an investor in D-Wave.)
The business factors involved are really three: support (as in D-Wave doesn’t have the resources to support potentially many thousands of users); abstractions (it’s in the process of developing interpreters that will make it easier for non-experts to program the computer); and algorithms. As it works with early customers and partners, Brownell said, the company is wrapping its head around more potential uses for the system and trying to develop the best algorithms for carrying them out.
The long and short of it is that D-Wave’s type of quantum computer (adiabatic as opposed to the more popular gate variety) is, theoretically, ideal for complex workloads. Google is using it to build try and better, faster and unsupervised classifiers for machine learning and artificial intelligence, Brownell said, while Lockheed is working on software validation and verification for flight systems. Another early tester is trying to optimize the pattern of radiation its cancer treatment system emits.
It could also be valuable in the financial services industry, where banks run untold number of Monte Carlo simulations in order to calculate the risk of their investments. “Goldman [Sachs, where Brownell used to work], they run hundreds of thousand of cores, 7 by 24, running nothing but these types of calculations,” Brownell said. Programmed correctly, a quantum computer could probably do this faster, more accurately and, because it doesn’t dissipate heat, much more efficiently.
“I think 10 years from now … you should have a level playing field for examples where it makes sense to implement an algorithm using our quantum computer [alongside traditional CPUs],” Brownell said. “… This is a good time to come up with a quantum computer because there is cloud … people are much more comfortable with that model than they were 10 years ago.”
Brownell also addressed recent criticism stemming from published benchmark tests in which the D-Wave system sometimes performed no better than an optimized GPU. He claimed the tests weren’t on the types of workloads for which someone would normally use a quantum computer, and noted that while researchers took advantage of 60 years worth of research in tuning their competitive processor, the D-Wave processor is only in its second generation (something even the researchers note). In fact, by the end of the year, the company expects to increase the qubit count on the processor from 512 to 1,000.
So while critics are questioning the validity of D-Wave’s claims about its computer, Brownell described the tests as “historic,” adding that, “It could be one of the most important innovations in the computer space in the last 60 years that something like this has happened.”