BY ELIZABETH GIBNEY
JOHN MARTINIS: Quantum builder
A physicist led Google’s first demonstration of a quantum computer that could outperform conventional machines.
When John Martinis was a graduate student in the mid-1980s, he went to a lecture that set the course of his scientific life. The famous physicist Richard Feynman discussed the idea of using particles’ quantum characteristics to make computers that could do things that are impossible on conventional machines. “It was clear to me that this was a great idea and that it would be wonderful to work on it,” Martinis says.
In October, Martinis took a big step towards Feynman’s dream. He led the work of a group of researchers at Google who announced that they had demonstrated a first: a quantum computer that could carry out a calculation faster than the best conventional computer. “Doing this experiment was the culmination of my career,” says Martinis.
The physicist, who works both at Google and at the University of California, Santa Barbara, has spent 17 years honing the hardware that underpins the firm’s quantum computer, named Sycamore. At its heart are tiny superconducting loops known as qubits, quantum systems that seem to exist in multiple states until they are observed. Physicists have long theorized that harnessing interactions between qubits could enable computers to excel at certain calculations, such as probing otherwise unsearchable databases and cracking conventional encryption.
A team of more than 70 scientists and engineers showed that, for a specific challenge — calculating the spread of outputs from a kind of quantum random-number generator — Sycamore could do in 200 seconds what they estimated would take the best supercomputer 10,000 years (although others argued that it would need only days).
The feat relied on improved hardware that lowered error rates and connected qubits in new ways. Some physicists debated the significance of the landmark, and the task has limited practical application. But Martinis says the experiment’s importance lies in demonstrating something fundamental: that physicists’ understanding of quantum interactions — learnt on small quantum systems — remains true at larger scales and complexity. “That’s really good news,” he says.
Hartmut Neven, who leads Google’s Quantum Artificial Intelligence laboratory, says that Martinis used to be a mountain climber, and that he applies that same careful, deliberate approach to building hardware, in which every sequence of moves must be thought out in advance. “John’s idea of a relaxing Sunday is to go into the lab and solder something together,” he says. “Life and work aren’t really separated.”
Martinis has many more ideas he’s hoping to pursue. His future priorities include making better quantum chips — including mastering methods to correct for errors caused by noise — and opening up Sycamore for use by outside researchers on a cloud system, to see whether there are useful algorithms that it could run. One idea is a method to verify that supposedly random numbers are truly random. “Physicists like me don’t retire,” he says with a smile. “We have lots of things to do.”