Ten thousand years versus 200 seconds. That was the gap Google claimed in 2019 between the world’s fastest supercomputer and its own quantum chip, and it pushed a phrase coined by a Caltech physicist into every newspaper. John Preskill, the physicist in question, spent the aftermath explaining, with some patience, what the phrase does and does not mean.
What the term was meant to say
Preskill proposed “quantum supremacy” to name the point where quantum computers can do things classical ones cannot, whether or not those tasks are useful. Wired dates his first use to a 2011 talk on how researchers could prove quantum hardware offered real benefits; Preskill himself points to a 2012 paper. He wanted to stress that humanity had entered a privileged moment, when information technology built on quantum physics was on the rise.
The word drew fire for two reasons, he wrote in Quanta Magazine. Its echo of white supremacy struck some as repugnant, and he admits he failed to foresee that. The other complaint was that it feeds the overhyped reporting on quantum technology, which he did expect. He weighed alternatives, “quantum advantage” among them, but felt “advantage” lacks punch. A horse that wins by a nose has an advantage. A quantum machine, for certain tasks, is far faster than that.
The Google experiment, under the hood
The test came from Google’s Sycamore processor, 54 qubits by Wired’s count, though Preskill refers to a 53-qubit device. It was asked to sample the output of a quantum random number generator. Google’s paper said a top machine like the Summit supercomputer at Oak Ridge National Lab would have needed approximately 10,000 years to match what Sycamore did in 200 seconds. The paper also admitted that technical leaps were still needed.
Preskill was quick to point out the catch. The task was picked to suit the hardware, and the computation has little structure, so the answer is not very informative. “The problem their machine solves with astounding speed has been very carefully chosen just for the purpose of demonstrating the quantum computer’s superiority,” he told Wired. Still, he called it a “truly impressive achievement in experimental physics.” Matching the output against a classical supercomputer, where the classical machine can manage it, showed the team understood its device.
Noisy machines and a long road
He also coined a second label for the period now beginning: NISQ, for “noisy intermediate-scale quantum,” which he says rhymes with risk. These are machines large enough to do certain specialised jobs beyond today’s supercomputers, but with imperfect control over the qubits, so errors build up and a long computation is unlikely to return the right answer.
Others tempered the excitement. Jonathan Dowling of Louisiana State University said breaking encryption was still “many years out”, and he and colleagues estimated it would take millions of high-quality qubit devices to threaten it. Dario Gil, IBM’s director of research, worried that the word “supremacy” could inflate expectations. Scott Aaronson of the University of Texas offered the opposite comparison: “You could have said similar things about the flight at Kitty Hawk, [that] it’s just an irrelevant proof of concept.”
Preskill’s own forecast was the most measured. He expects quantum computers to transform society, but this may still be decades away. In 2012 he had asked whether controlling large quantum systems was “really, really hard” or “ridiculously hard”, the difference between decades and centuries of work. Google’s result, he wrote, bolsters confidence that it is the former.
Sources
- Why I Called It 'Quantum Supremacy' | Quanta Magazine (by John Preskill)
- Google's 'Quantum Supremacy' Isn't the End of Encryption | WIRED
