Google's quantum supremacy is only a first taste of a computing revolution

Google's success at achieving quantum supremacy sounds like a momentous victory. But really, it's just the first step in making this radical new type of computing useful.

On Wednesday, Google published a scientific paper in the journal Nature detailing how its quantum computer vastly outpaced a conventional machine, an idea called quantum supremacy. Powered by a Google-designed quantum processor called Sycamore, it completed a task in 200 seconds that, by Google's estimate, would take 10,000 years on the world's fastest supercomputer. The importance of the achievement can be as hard to understand as quantum computing itself, a field made possible by the mind-bending behavior of atomic-scale physics. But if you want a takeaway, here it is: Quantum computing is only beginning to show some of the promise researchers have hyped for decades. We're still several breakthroughs away from seeing the true potential fulfilled.

Don't get me wrong. Google's achievement, documented by 77 authors in a prestigious peer-reviewed journal, is notable. Quantum computing skeptics should recalibrate their pessimism. Quantum computing ideas that Google has worked on for 13 years, and that famed physicist Richard Feynman described in 1981, are moving into reality.

Quantum computers work by embracing the strange nature of particles at the atomic scale. Where classical computers store data as bits that are either a one or a zero, the quantum computing equivalent, called a qubit, can store information that's part one and part zero. Next, a quantum computer gangs multiple qubits together, dramatically increasing the number of possible states they can record. Last, processing those qubits lets researchers explore countless possible solutions to a problem simultaneously instead of evaluating them one at a time. It's lousy for adding two and two, but potentially great for some problems classical computers just can't cope with.

Google's quantum researchers are already turning their attention to the next steps needed to make their machines more broadly useful, a step Intel calls quantum practicality.

"It will be a must-have resource at some point," Hartmut Neven, the researcher who began Google's quantum computing effort in 2006, said at a press event.

Google quantum researcher Marissa Giustina Marissa Giustina, a researcher with Google's quantum computer lab, draws a diagram showing "quantum supremacy" as only an early step on a path of quantum computer progress.

Stephen Shankland/CNET For those who want to try out Google's quantum computer, the company plans to make it available as a cloud computing service in 2020. That follows in the footsteps of IBM, which already has done so with its Q Experience.

What'll quantum computers be good for?

Google's quantum researchers are excited about the shift in their research from theory to experiment. "I started off ... bashing my head against the wall because all the algorithm development was for a machine that didn't exist," said Dave Bacon, who leads Google's quantum software work. Now comes the era when "I can just run it and see what happens."

Google has a lot of practical uses in mind:

Complicated optimization problems, such as calculating how to deliver packages in the shortest time while using the least energy. "Optimization problems occur everywhere at every company anywhere in the world," Bacon said. Addressing those challenges could both save money and help the environment. Improving encryption technology by generating random numbers. Google's quantum team is talking to its encryption key generation team about using a random-number generation tool it's already developed for today's Sycamore machine. Building machine learning systems better at tasks like distinguishing between real and fake items like bogus political videos. This was the original impetus for Neven's work, and Google researchers think it could be the first area to deliver on quantum computing's promise. Perhaps most interesting, simulating the real physics of molecular-scale materials. Revolutionary developments there could mean more efficient solar panels, a new way to produce nitrogen fertilizer without needing so much energy and better electric car batteries.

Google's quantum optimism Google believes it's on the right track, though, and that quantum progress will outstrip classical progress. It looks forward not merely to exponential performance improvements -- the kind that Moore's Law has charted for classical computers -- but double exponential improvements.

Google has a long to-do list, starting with improving how long qubits can run error-free. Errors mean a qubit flips to record bad information, stymying a calculation, and improving error rates is the top goal in the next year, said John Martinis, the University of California, Santa Barbara, researcher who now leads Google's quantum computing hardware team.

"The No. 1 thing we are trying to do is improve the errors of the device," Martinis said, standing in Google's lab with five hulking quantum computers suspended around him. "We've been kind of ignoring that trying to get to the supremacy result."

Later will come more fundamental changes, like quantum error correction techniques to sidestep qubit instabilities. Google researchers are unafraid to present plans stretching years into the future, when qubit counts rise from 54 to a million or more. And they're patient.

"We know that this decade-long march is going to require innovations across theory, engineering and actual physics," Bacon said.

Originally published Oct 25, 5 a.m. PT. Update, 9:19 a.m. PT: Adds comment on qubit longevity.

本站文章除注明转载/出处外，均为本站原创或翻译，转载前请务必署名,转载请标明出处
最后编辑时间为: 2019/10/27 00:07