By Tom Simonite, senior writer of WIRED on 24/10/2019
Google shows off its Sycamore chip that it says reached “quantum supremacy,” but it will still be years before quantum computing is useful.
Wednesday was warm in Santa Barbara, California, but Google’s quantum computing labs there were pleasingly cool—colder than outer space in some spots. Three fridge-sized silver cylinders hung below crowns of ducts and cables. Inside, superconducting quantum processors were at operating temperature, a fraction of a degree above absolute zero.
Googlers gazed at one of the gleaming cylinders, sporting a ring of green LEDs, with special fondness. In the early hours of the morning, the journal Nature published a peer-reviewed paper describing how the chip inside, named Sycamore, had completed in minutes a math problem estimated to require a supercomputer to work flat out for 10 millennia. Although IBM has questioned Google’s results, many researchers say they appear to represent a scientific milestone termed quantum supremacy: When a quantum computer gives the first taste of the technology’s potential by doing something beyond any conventional computer, even if on a contrived problem.
On Wednesday, Google took a victory lap. “We compare it to a Sputnik moment,” Hartmut Neven, the executive who started Google’s quantum program in 2006, told reporters, wearing a silvery jacket and boots apt for the space analogy. In an interview and blog post, Google CEO Sundar Pichai giddily likened the moment to the Wright brothers’ first flight, the first rocket to escape Earth’s gravity, and chess champion Garry Kasparov’s historic loss to a computer.
Despite the mondo metaphors, Google’s quantum team didn’t have a big party planned that night. Neven said he was thinking of perhaps a small dinner. “We have a lot going on right now,” noted a more junior member of the team.
The superposition of celebration and sobriety on display at Google’s quantum HQ was fitting. The company’s supremacy experiment is a signal moment in a recent burst of progress in quantum computing. Tech giants, startups, investors, and governments are plowing more money into quantum computing. Nature estimates that private firms working in quantum-related technology raised $400 million in 2017 and 2018, more than double the amount in the previous five years. At the same time, no one quite knows when today’s quantum technology, even Google’s champ Sycamore, will mature into something useful—or profitable.
“This is not a technology milestone, it’s a scientific milestone,” says David Poulin, codirector of the quantum information program at the Canadian Institute for Advanced Research, and a professor at Université de Sherbrooke in Quebec.
Quantum computing’s promise of vast computing power springs from devices called qubits. They encode data into the quantum mechanical properties that invisibly underpin reality. Like the components of conventional computers, qubits can perform calculations by manipulating digital bits of data, 1s and 0s. But qubits can also attain a third state that is a superposition of both 1 and 0 at the same time, a quantum mechanical phenomenon without analog in everyday human experience.
“This is not a technology milestone, it’s a scientific milestone.”
DAVID POULIN, CANADIAN INSTITUTE FOR ADVANCED RESEARCH
Researchers proved decades ago that a large and well-controlled collection of qubits should be capable of solving problems impractical for any conventional computer—even breaking encryption. But making qubits that function reliably enough, in large enough numbers, has been difficult.
Error-correcting codes have been invented that could allow a large number of qubits to clean up their own messes, and perform like a much smaller number of near-perfect qubits, probably on the order of thousands fewer. But no one is close to fully implementing them.
Google’s Sycamore chip has 54 qubits; Intel and IBM have shown off processors of similar scale, also built on superconducting chips. A rough road map shown by Google Wednesday suggested that it could build quantum processors with full error checking once it has processors with a million or more physical qubits. But to run error correction, those qubits would need to be much more reliable than those of today. Google’s quantum hardware chief John Martinis said Wednesday that qubits needed an error rate of 1 in 1,000 to run full error correction. Right now, the error rate on Google’s Sycamore chip is more than five times as high, he added.
Even as it works to meet that threshold, Google must also reinvent the way quantum processors are built and controlled. Martinis admitted Wednesday to a fixation with redesigning the wires that snake between the supercold qubits on a quantum processor and the racks of more conventional electronics that control them. Google’s supremacy experiment used only 53 of the Sycamore chip’s 54 qubits because one of the delicate control lines broke.
“I feel we know how to scale up to hundreds and maybe thousands of qubits,” Martinis said. “Scaling up to 1 million, that’s really hard.” On Wednesday, several members of Google’s quantum team estimated that it will be about a decade before there’s an error-corrected quantum computer.
While the world waits for Google or its competitors to make the necessary breakthroughs, quantum computing remains stuck in what researchers call the NISQ era—one of Noisy Intermediate Scale Quantum technology. “The million-dollar question is: Can we use these devices in the near term without error correction?” says Poulin, of the Canadian Institute for Advanced Research.
Answering that in the affirmative could turn quantum computing into a money-spinner much more quickly. But Poulin says discovering what NISQ devices can usefully do will largely be a question of trial and error, not something that can be worked out in theory.
The WIRED Guide to Quantum Computing
Google’s quantum researchers said Wednesday they are talking with the company’s security experts about adapting the Sycamore quantum supremacy experiment to create random numbers for encryption keys. Another project is exploring how Sycamore-like chips might help machine-learning algorithms that can generate realistic images.
The company and others with quantum hardware hope early customers will help figure out what the technology is good for. Google said Wednesday that it’s working to allow access to its quantum hardware via the cloud—similar to quantum cloud access offered by IBM, whose partners include J.P. Morgan, and by startup Rigetti, working with companies including drugmaker Merck.
Many companies exploring quantum computing depend on fundamental science, requiring more patience than internet technology. Daimler, which is using IBM’s quantum cloud and also working with Google, is interested in battery chemistry. Airbus, an investor in quantum hardware startup IonQ, hopes to reduce the burden of simulating the physics of new aircraft, according to IonQ CEO Peter Chapman. “The computers themselves aren’t ready, but early adopters are going to have first-mover advantage,” he says. “If you start development after the machine plops onto your desktop, it’s probably too late.”
IonQ builds qubits using a fundamentally different technology than Google or IBM, known as ion traps. The company announced $55 million in new funding Tuesday from investors including Amazon and Samsung—as well as Google parent Alphabet. It’s a reminder that with so much still to do to turn quantum supremacy into quantum practicality, no one is sure which tech will win out.
Asked how sure he feels that Google is on the right track, Neven says he’s not sentimental about the technology. “We live by the Silicon Valley motto of ‘Only the paranoid survive,’” he points out. “We would shamelessly pivot if there’s a better technology that comes along for qubits.” Then he strode inside to brief Alphabet’s board on the day’s news, space-age jacket gleaming, metaphors at the ready.
Updated, 10-25-19, 12:40pm ET: This article has been updated to reflect that Google parent Alphabet, and not Google, had participated in the funding of IonQ.