I am standing in front of a gigantic touch screen in a garagelike laboratory at Google’s facility in Goleta, Calif., using my finger to move little squares containing symbols—an X, a Y, an H and other, more arcane glyphs—across the display. The squares represent functions that can be performed on a quantum bit—a qubit—inside a large, silvery cylinder nearby. Of the myriad functions on offer, some cause the bit to flip from 1 to 0 (or from 0 to 1); one makes it rotate around an axis.
Another square on the display reveals the state of the qubit, represented by what looks like a lollipop moving around inside a sphere, its stick anchored in the center. As it moves, numbers beside it oscillate between 1.0000 and 0.0000. This is one of the strengths of qubits: they do not have to be the all-or-nothing 1 or 0 of binary bits but can occupy states in between. This quality of “superposition” allows each qubit to perform more than one calculation at a time, speeding up computation in a manner that seems almost miraculous. Although the final readout from a qubit is a 1 or 0, the existence of all of those intermediary steps means it can be difficult or impossible for a classical computer to do the same calculation. To the uninitiated, this process may appear a bit like magic—a wave of the hands, a tap of a touch screen and, presto, a rabbit is pulled from a quantum hat. Google has invited me here—along with a select group of other journalists—to pull back the curtain on this wizardry, to prove it is not magical at all.
On the right half of the screen, squiggly lines display waveforms that correspond to the functions being performed on the qubits. Next to that section is a box about the size of a desktop printer, which sends those waveforms as electrical pulses through wires and into the silver cylinder. If the cylinder were open, one would see a series of six chambers, arranged in layers like a wire-festooned upside-down wedding cake. Each chamber is chilled to a temperature significantly colder than the one above it; the bottommost layer is a frigid 15 millikelvins, nearly 200 times as cold as the depths of outer space. Wires passing through the successive stages relay control signals from the warm outside world and pass back results from the chamber.
On supporting science journalism
If you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
That chamber is in vacuum, shielded from the light and heat that would otherwise disrupt the delicate qubits, which sit on a chip at the end of all the wires, isolated in the dark and cold. Each qubit is about 0.2 millimeter across, big enough to be visible through an ordinary microscope. But chilled and hidden away from external influences,each becomes a superconductor that lets electrons flow freely, acting as if it were a single atom so that the laws of quantum mechanics scale up to dictate its behavior.
Gentle pulses of microwaves cause the qubits to vibrate. And when two neighboring qubits reach the same resonant frequency, they become entangled—another quantum-mechanical property meaning that measuring the state of one tells you the state of the other. Electromagnetic pulses at a different frequency cause the bit flips. The quantum computer is rather like a box containing a bunch of pendulums, says Craig Gidney, a quantum software engineer at Google. I and others outside the chamber sending the signals into it are pulling on the strings of the pendulums, changing their swings to perform different logical operations.
All this chilling and vibrating, Google’s quantum team says, has allowed it to achieve quantum supremacy, the point at which a quantum computer can do something that an ordinary classical computer cannot. In a paper published this week in Nature—but inadvertently leaked last month on a NASA Web site—Google engineers describe a benchmark experiment they used to demonstrate supremacy. Their program, run on more than 50 qubits, checks the output of a quantum random-number generator. Some critics have complained this is a contrived problem with a limited real-world application, says Hartmut Neven, manager of Google’s Quantum Artificial Intelligence Lab. “Sputnik didn’t do much either,” Neven said during a press event at the Goleta facility. “It circled the Earth. Yet it was the start of the space age.”
David Awschalom, a condensed matter physicist specializing in quantum-information engineering at the University of Chicago, who was not part of the research, agrees that the program solved a very particular problem and adds that Google cannot claim it has a universal quantum computer. Such an achievement would require perhaps a million qubits, he says, and lies many years in the future. But he believes the company’s team has reached an important milestone that offers other scientists real results to build on. “I’m very excited about this,” Awschalom says. “This type of result offers a very meaningful data point.”
Google’s quantum computing chip, dubbed Sycamore, achieved its results using exactly 53 qubits. A 54th one on the chip failed. Sycamore’s aim was to randomly produce strings of 1’s and 0’s, one digit for each qubit, producing 253 bit strings (that is, some 9.700199254740992 quadrillion bit strings).Because of the way the qubits interact with one another, some strings are more likely to emerge than others. Sycamore ran the number generator a million times, then sampled the results to come up with the probability that any given string would appear. The Google team also ran a simpler version of the test on Summit, a supercomputer at Oak Ridge National Laboratory, then extrapolated from those results to verify Sycamore’s output. The new chip performed the task in 200 seconds. The same chore, the researchers estimated, would have taken Summit 10,000 years.
Yet a group of researchers at IBM, which is also working to develop quantum computing, posteda preprint paper on arXiv.org earlier this week arguing that, under ideal conditions and using extra memory storage, Summit could accomplish the task in two and a half days. “Because the original meaning of the term ‘quantum supremacy,’ as proposed by [California Institute of Technology theoretical physicist] John Preskill in 2012, was to describe the point where quantum computers can do things that classical computers can’t, this threshold has not been met,” the scientists wrote in a post on the IBM Research Blog. Perhaps, then, Google’s achievement might be better labeled “quantum advantage.”
But Scott Aaronson, a theoretical computer scientist at the University of Texas at Austin, who sometimes collaborates with the Google researchers, says it is not really correct to say quantum supremacy has not been achieved—even if it is not an unambiguous “man on the moon” sort of result. After all, Sycamore was still far faster at the task than Summit. And as the number of qubits in Google’s setup grows, its computing power will expand exponentially. Moving from 53 to 60 qubits would give the company’s quantum computer the equivalent computational heft of 33 Summit supercomputers. At 70 qubits, a Summit-like classical supercomputer would have to be the size of a city to possess the same processing power.
Aaronson also suspects that what Google achieved might already have some unintended practical value. Its system could be used to produce numbers verifiably guaranteed by the laws of quantum physics to be random. That application might, for instance, produce far stronger passwords than humans or classical computers are able to come up with.
“I’m not sure the right thing is to argue whether it’s ‘supremacy’ or not,” Awschalom says. The quantum computing community has yet to agree on the best ways to compare different quantum computers, he says, especially those built on different technologies. Whereas both IBM and Google are using superconductors to create their qubits, another approach relies on trapped ions—charged atoms suspended in a vacuum and manipulated by laser beams. IBM has proposed a metric called “quantum volume,” which includes factors such as how fast qubits perform their calculations and how well they avoid or correct errors.
Error correction is, in fact, what quantum computer scientists must master to make truly useful devices—ones containing thousands of qubits. At that point, researchers say, the machines could run detailed simulations of chemical reactions that might lead to new drugs or better solar cells. And they could also quickly crack the cryptographic codes most commonly used to protect data on the Internet.
To reach that kind of performance, however, a quantum computer must self-correct, finding and fixing errors in its operations. Errors can arise when a qubit flips from 1 to 0 spontaneously or when its quantum superposition decays because of interference from the outside world. Google’s qubits currently last about 10 microseconds before decaying. “They have a finite lifetime,” says Marissa Giustina, one of the project’s researchers. “They’re very fragile. They interact with their surroundings, and we just lose the quantum information.”
Classical computers tackle error correction with redundancy, deciding whether a digital bit is on or off by measuring not a single electron in a capacitor but tens of thousands. Conversely, qubits are, by nature, probabilistic, so trying to clump them together to perform one bulk measurement will not work. Google is developing a statistical method to correct errors, and John Martinis, a physicist at the University of California, Santa Barbara, who teamed with the company to develop Sycamore, says the tentative results so far have revealed no fundamental aspect, no showstopper, that would prevent error correction from getting better and better. The show, it seems, will go on.
Meanwhile Google’s engineers will be working to improve their qubits to produce fewer errors—potentially allowing many more qubits to be interlinked. They also hope to shrink down their large, desktop-printer-sized control boxes—each can handle 20 qubits and associated circuitry, so three are needed to run Sycamore’s 53 qubits. And if their system grows to reach about 1,000 qubits, its cooling needs will exceed the capacity of those large silver cylinders.
Julian Kelly, who works on quantum hardware and architecture at Google, says the company’s announcement is an engineering achievement above all else, but it is one that could open up unexplored terrain. “We’ve demonstrated that the quantum hardware can do something that is extremely difficult,” he says. “We’re operating in a space where no one has been able to experiment before.” What the outcome of that progress will be, he says, is something “we don’t know yet, because we’ve just got here.”
