A photo provided by Google shows, from left, Sundar Pichai, the chief executive, and a Google researcher with Google's quantum computer machine.GOOGLE/The New York Times News Service
In 200 seconds – the time it takes to poach an egg – a quantum processor has performed a numerical task faster than any known computer system on Earth.
Exactly how much faster is up for debate, but the breakthrough achievement led by researchers at Google AI in Mountain View, Calif., is the most convincing demonstration yet of a feat referred to as “quantum supremacy.” In essence, it means that a quantum computer – a device that leverages what Albert Einstein once described as “spooky” physics at the atomic scale – has attained a level of performance that is impractical, if not impossible, for conventional computer systems to emulate.
What’s more, researchers said, the design of their quantum processor is one that can be improved and scaled up. If that assessment proves correct, the device points the way to a game-changing technology that could revolutionize and disrupt a range of computer-related fields and applications.
“It’s really exciting because the physics came through. The physics was right,” said John Martinis, the senior scientist on the project, during a conference call with reporters on Wednesday. “There’s clearly a lot of hard things to solve … but we feel there’s a really great future ahead of us."
Though still a nascent technology, quantum computers offer a powerful advantage over standard digital computers in one key respect. While an ordinary computer operates based on bits of information that are designated by either 1 or 0 (corresponding physically to “off” or “on” in a circuit on a microchip), a quantum computer runs on “qubits” that quantum physics allows to hold a mixture of those values.
Qubits can also be linked up so that they can influence one another in ways that produce rapid short cuts for certain types of problems. Future applications of quantum computers range from cybersecurity and drug development to modelling financial markets.
The Google team developed a 53-qubit device, nicknamed Sycamore, and fed it a set of random instructions. They then sampled the answers the device provided in the form of strings of numbers one million times, effectively embedding the quantum characteristics of the system in the answers. While the exercise is neither commercially useful nor difficult for the device – the scientists compared it to writing a program that reads out “Hello world" – it requires a massive series of calculations to do the same thing without qubits.
In describing their experiment in the journal Nature, Dr. Martinis and his colleagues said it would take any existing state-of-the-art supercomputer about 10,000 years to reproduce Sycamore’s output.
The claim was quickly disputed by a team at IBM who issued a result showing that a supercomputer can be optimized to solve the problem in 2½ days, reducing the Google claim of quantum supremacy to the equivalent of winning by a nose rather than by a mile.
“Scientifically, both papers are right,” said David Poulin, a physicist at the University of Sherbrooke and co-director of a research program in quantum information science supported by the Canadian Institute for Advanced Research.
The IBM response illustrates that quantum supremacy is a moving target, Dr. Poulin added, thanks to constant improvements in the hardware and algorithms that underpin the world’s biggest computers. But while “that may take some of the gloss” off the Google announcement, he said that what matters more is that Sycamore has demonstrated that a quantum computer can be designed to operate at a scale that some theorists had speculated might be impossible in principle.
“It shows that quantum computing is here for real and a significant player on the computing landscape,” said Joseph Emerson, a faculty member at the University of Waterloo’s Institute for Quantum Computing. Dr. Emerson added that, going forward, each small increase in the capability of quantum computers will require exponential leaps by conventional computers to keep up.
But quantum computers still face big challenges. Among the biggest is the fact that as more qubits are added, it becomes increasingly difficult to maintain the conditions needed for reliable quantum computations. To get around this, quantum computers can have a portion of their qubits allocated to error correction. Dr. Emerson is the founder of Quantum Benchmark, a Waterloo startup that has developed a way to assess and improve the effectiveness of quantum error correction. The company is collaborating with Google and is set to play a role in future versions of its quantum chip.
On Wednesday, Dr. Martinis said that he expected that his team would have a working 1,000-qubit device in about a year, a scale at which quantum error correction would likely become essential.
Some of the team members behind the Google project were trained at the Waterloo institute or have other ties to Canada’s quantum computing ecosystem. Alexandre Blais, another expert in the field at the University of Sherbrooke, was one of the key originators of the interconnected superconducting loops that serve as the qubits for the Google system.