Skip to main content

Biggest ever quantum chip announced, but scientists aren't buying it

Biggest ever quantum chip announced, but scientists aren't buying it

/

New chip from D-Wave boasts 2,000 qubits, or quantum bits

Share this story

If you buy something from a Verge link, Vox Media may earn a commission. See our ethics statement.

D-Wave

Quantum computing firm D-Wave has announced this month its largest ever quantum chip containing 2,000 qubits — double the capacity of its previous biggest system. The chip is scheduled to ship next year, and if it lives up to its promise, it would solidify D-Wave’s position at the forefront of quantum computing, a potentially revolutionary field that would change computing as we know it. But despite D-Wave’s confidence, scientists and academics say the company has never proved its advantages over normal computers. And, more damningly, that using the company's current methodologies, it never will.

D-Wave’s Colin Williams, the company’s director of business development and a former quantum computing scientist himself, is bullish. "[The new chip] isn’t just bigger," he told The Verge. "It’s improved in many other ways."

The Canadian firm’s quantum computing chips are based around a process known as quantum annealing. This renders a set problem (like, for example, trying to find the quickest route home passing through certain points) as a topographical map of peaks and troughs, with the optimum answer to the question defined as the lowest point on that map. While regular computers using static 1s and 0s would have to traverse the entire map to find that point, quantum computers — which use quantum bits, or qubits, that represent 1s, 0s, and both at the same time — can effectively tunnel through the landscape, find the lowest point much faster.

Williams says he’s certain that quantum annealing is the best way to make a quantum computer, and that other approaches are too theoretical. He points out that topological quantum computing (an approach that Microsoft has shown interest in) relies on creating exotic quasiparticles, which are difficult to produce and even trickier to work with. "We’re only at the very very beginning stages of being able to create these particles, let alone perform operations on them," says Williams. "[Quantum annealing] has tremendous advantages over other schemes."

no real proof of a quantum speedup

But researchers say the benefits of D-Wave’s method have never been proved. A study published in Science in 2014 found that tasks performed on the company’s machines were no faster than conventional computers. The scientists were looking for evidence of "quantum speedup" — the signature advantage of quantum computers which holds that the more calculations you throw at them, the greater a difference in speed they show when compared with classical machines. The paper in Science did not rule out the possibility of D-Wave creating quantum speedup, but certainly found no evidence for it.

"There was only ever a hope that a quantum annealer would be better," Matthias Troyer, who co-authored the 2014 Science paper, told The Verge. "It turns out that at least for the architecture implemented by D-Wave, [the computation] can be mimicked very efficiently on a classical computer." Troyer says that simply doubling the number of qubits in its chips will not help D-Wave overcome this problem. "We don’t have any evidence of quantum speedup in this architecture and building a bigger machine will not help that."

"Building a bigger machine will not help."

Other researchers agree with Troyer’s analysis. Scott Aaronson of the University of Texas and Greg Kuperberg of UC Davis tell The Verge that while there was theoretical hope that quantum annealing would produce results, the tests have not borne this theory out. The pair note that papers published by D-Wave and partners supposedly showing its quantum advantage are generally pitting its $15 million chips against the class of processor you’d find in your laptop. What’s more, they say, testers tend to pick computational challenges optimized for D-Wave's chips, giving the company’s tech a home-field advantage. This, they say, leads to impressive but misleading claims that D-Wave's technology has been proved to be "100,00,000 times faster" than classical computers.

Kuperberg adds that D-Wave’s qubits are also of low quality compared to those produced by other researchers. "Just because [their chips] are quantum, that doesn’t make them a quantum computer," says Kuperberg. "That's like saying that any invention that is influenced by air must be an airplane. Of course, it's not true; it might instead be bagpipes."

Jeremy Hilton, D-Wave's senior vice president of systems, defended the methodologies of these papers. In the case of the Google study claiming a 100 million times quantum speedup, he noted that the decision not to compare D-Wave's chips to the fastest algorithms available for classical machines was "intentional." These faster algorithms would not scale to "real-world problem sizes," says Hilton, and so would not represent the true potential of D-Wave's chips. "It’s worth noting that one of the inventors of this faster classical algorithm actually works at Google," said Hilton. "So it is safe to say they have a pretty good idea of how relevant it is for the problems they want to solve."

D-Wave says it's "a decade ahead" of rivals

Williams noted that while it's true you can't measure the quality of a chip in the number of qubits alone, D-Wave's new software functionalities would also deliver extra power. "We are at least a decade ahead in my opinion and if we can sustain our current pace of innovation we’ll remain a decade ahead, forever," he said.

Aaronson and Kuperberg would disagree, but say they’re still optimistic about the wider future of quantum computing. Troyer, too, mentions many other promising projects, including those at Microsoft, IBM, and the University of Oxford. Indeed, there have been rumors this year that a team at Google working under one John Martinis (separate to the group testing D-Wave’s chips) are getting near to a breakthrough, with results expected in the coming years.