For the first time ever, a quantum computer has performed a computational task that would be essentially impossible for a conventional computer to complete, according to a team from Google.
Scientists and engineers from the company’s lab in Santa Barbara announced the milestone in a report published last week in the journal Nature. They said their machine was able to finish its job in just 200 seconds — and that the world’s most powerful supercomputers would need 10,000 years to accomplish the same task.
The task itself, which involved executing a randomly chosen sequence of instructions, does not have any particular practical uses. But experts say the achievement is still significant as a demonstration of the future promise of quantum computing.
William Oliver of MIT compared the feat to the first successful flight by the Wright brothers. “It is what the event represented, rather than what it practically accomplished, that was paramount,” he wrote in a commentary that accompanied the study.
Google’s scientists are hailing the achievement as the first demonstration of what’s known as “quantum supremacy.” The phrase was coined in 2012 by John Preskill, a theoretical physicist at Caltech, to describe the point at which quantum computers can do things that classical computers simply can’t.
Not everyone agrees that Google’s announcement represents true quantum supremacy. Computer scientists at IBM have countered that their most powerful supercomputer, called Summit, could complete the same task in 2.5 days rather than 10,000 years.
“A computation that would take 10,000 years on a classical supercomputer took 200 seconds on our quantum computer,” study co-author Brooks Foxen, a graduate student researcher in physics at Google AI Quantum in Mountain View and the University of California, Santa Barbara, said in a statement.
“It is likely that the classical simulation time, currently estimated at 10,000 years, will be reduced by improved classical hardware and algorithms, but, since we are currently 1.5 trillion times faster, we feel comfortable laying claim to this achievement,” Foxen added.
Quantum computers store information using subatomic particles, which behave according to very different rules than the ones that govern our macro world. For example, quantum particles can exist in a “superposition” of two different states at the same time, and particles can be separated by light-years yet still be “entangled,” affecting each others’ properties.
This weirdness is key to the incredible potential power of quantum computing. Because of the superposition phenomenon, quantum computers can store and manipulate far more information per unit volume than can traditional computers, which encode information in a binary way using 0s and 1s. (The basic unit of information in a quantum-computing system, by the way, is known as a qubit, which is short for “quantum bit.”)
The new study gives us a taste of that power. The research team, led by Frank Arute of Google AI Quantum, used a quantum computer called Sycamore, which featured 53 functional qubits (plus one that didn’t work properly).
The scientists entangled those 53 qubits into a complex superposition state, then had Sycamore perform a task akin to random-number generation. The results were then compared with simulations run on the Summit supercomputer at Oak Ridge National Laboratory in Tennessee.
“Summit is currently the world’s leading supercomputer, capable of carrying out about 200 million billion operations per second,” William Oliver, a physicist at the Massachusetts Institute of Technology, wrote in an accompanying “News and Views” piece in the same issue of Nature.
“It comprises roughly 40,000 processor units, each of which contains billions of transistors (electronic switches), and has 250 million gigabytes of storage. Approximately 99% of Summit’s resources were used to perform the classical sampling,” added Oliver, who was not involved in the new study.
As Foxen noted, Sycamore finished in about 3.5 minutes, and the Summit work suggested that even the most powerful traditional supercomputer would have to chew on the problem for about 10,000 years.