Skip to main content

Google’s quantum computer just flunked its first big test

Google’s quantum computer just flunked its first big test

Share this story

A processor from the D Wave 2
A processor from the D Wave 2

When the D-Wave 2 was first released last year, it was accompanied by a tidal wave of hype. The machine was a self-proclaimed quantum computer, commercially available to anyone with $15 million to spend, and attracting the attention of everyone from NASA to the NSA. One of the computer’s buyers was Google,which launched a new lab to test the device's powers more rigorously than they’d ever been tested before. This October, the lab announced a major discovery, providing stronger evidence for quantum effects within the D-Wave 2 than anyone had previously found. As D-Wave had claimed, its device really was quantum-powered — and Google’s big research bet seemed to be paying off.

The team found no clear advantage to the quantum computer

But today, the D-Wave 2 is facing its first big stumble. A study in Science found that the quantum device is no faster than conventional computing, calling into question the entire premise of Google's lab and D-Wave's machines. Led by scientists at ETH Zurich, the research team matched up the quantum machine against conventional computers on a set of problems intended to suit the quantum machine's strengths. (The project also got a crucial assist from Microsoft Research, which tested out the classical simulations on its high-powered computer clusters.) Once the results were in, the team found no clear advantage on either side. The D-Wave machine might be quantum-powered, but it didn't run any faster because of it.

"We haven't seen any fundamental limits to performance."

The research was hoping to find what’s called "quantum speedup," the property that lets a quantum computer perform tasks faster than a conventional one, with more and more advantage as the problems grow more complex. The speedup is quantum computing's big advantage: with it, a quantum computer could tackle mind-bogglingly complex problems that would take a conventional computer years or even centuries. Without it, you're left with a lot of expensive hardware and no clear advantage. "It's hard to show that the speedup isn't happening," says Matthias Troyer, who co-authored the Science paper. "But if you don't find the speedup, you can't show that the machine is better than a classical device."

The results don't rule out the speedup entirely, but they're not encouraging — and there are plenty of other reasons for researchers like Troyer to be pessimistic. D-Wave's computers are based on a structure called quantum annealing, which many quantum computing experts think is doomed from the start. Since D-Wave's work outpaced much of the academic literature on the topic, there's still no theory for how the quantum speedup might emerge from an annealing-based machine.

"Most scientists would not set out to build a machine without knowing what it can do."

It's still possible that testing other problems might turn up the missing speedup. In fact, D-Wave representatives point to recent work by USC's Itay Hen as evidence that this is already happening. And even without the speedup, D-Wave is still going toe-to-toe with some of the most powerful conventional computing clusters in the world. Jeremy Hilton, D-Wave's VP of processor development, says the company's computers could surpass state-of-the-art classical computing within just a few years. "We haven't yet seen any fundamental limits to performance that cannot be improved with design changes," Hilton says.

And then there's Google's lab. The Quantum AI Lab is still testing D-Wave's limits too, although the company declined to comment on when the lab might publish its own results. As with most research, it's a slow, gradual process — a process that’s only now testing out the ambitious promises of the D-Wave 2. "Most scientists would not set out to build a machine without knowing what it can do," Troyer says. "Most scientists would first think a bit more about it."