Google has reportedly demonstrated for the first time that a quantum computer is capable of performing a task beyond the reach of even the most powerful conventional supercomputer in any practical time frame—a milestone known in the world of computing as “quantum supremacy.”
The ominous-sounding term, which was coined by theoretical physicist John Preskill in 2012, evokes an image of Darth Vader–like machines lording it over other computers. And the news has already produced some outlandish headlines, such as one on the Infowars website that screamed, “Google’s ‘Quantum Supremacy’ to Render All Cryptography and Military Secrets Breakable.” Political figures have been caught up in the hysteria, too: Andrew Yang, a presidential candidate, tweeted that “Google achieving quantum computing is a huge deal. It means, among many other things, that no code is uncrackable.”
Nonsense. It doesn’t mean that at all. Google’s achievement is significant, but quantum computers haven’t suddenly turned into computing colossi that will leave conventional machines trailing in the dust. Nor will they be laying waste to conventional cryptography in the near future—though in the longer term, they could pose a threat we need to start preparing for now.
Here’s a guide to what Google appears to have achieved—and an antidote to the hype surrounding quantum supremacy.
We still haven’t had confirmation from Google about what it’s done. The information about the experiment comes from a paper titled “Quantum Supremacy Using a Programmable Superconducting Processor,” which was briefly posted on a NASA website before being taken down. Its existence was revealed in a report in the Financial Times—and a copy of the paper can be found here.
The experiment is a pretty arcane one, but it required a great deal of computational effort. Google’s team used a quantum processor code-named Sycamore to prove that the figures pumped out by a random number generator were indeed truly random. They then worked out how long it would take Summit, the world’s most powerful supercomputer, to do the same task. The difference was stunning: while the quantum machine polished it off in 200 seconds, the researchers estimated that the classical computer would need 10,000 years.
When the paper is formally published, other researchers may start poking holes in the methodology, but for now it appears that Google has scored a computing first by showing that a quantum machine can indeed outstrip even the most powerful of today’s supercomputers. “There’s less doubt now that quantum computers can be the future of high-performance computing,” says Nick Farina, the CEO of quantum hardware startup EeroQ.
In a classical computer, bits that carry information represent either a 1 or a 0; but quantum bits, or qubits—which take the form of subatomic particles such as photons and electrons—can be in a kind of combination of 1 and 0 at the same time, a state known as “superposition.” Unlike bits, qubits can also influence one another through a phenomenon known as “entanglement,” which baffled even Einstein, who called it “spooky action at a distance.”
Thanks to these properties, which are described in more detail in our quantum computing explainer, adding just a few extra qubits to a system increases its processing power exponentially. Crucially, quantum machines can crunch through large amounts of data in parallel, which helps them outpace classical machines that process data sequentially. That’s the theory. In practice, researchers have been laboring for years to prove conclusively that a quantum computer can do something even the most capable conventional one can’t. Google’s effort has been led by John Martinis, who has done pioneering work in the use of superconducting circuits to generate qubits.
No. Google picked a very narrow task. Quantum computers still have a long way to go before they can best classical ones at most things—and they may never get there. But researchers I’ve spoken to since the paper appeared online say Google’s experiment is still significant because for a long time there have been doubts that quantum machines would ever be able to outstrip classical computers at anything.
Until now, research groups have been able to reproduce the results of quantum machines with around 40 qubits on classical systems. Google’s Sycamore processor, which harnessed 53 qubits for the experiment, suggests that such emulation has reached its limits. “We’re entering an era where exploring what a quantum computer can do will now require a physical quantum computer … You won’t be able to credibly reproduce results anymore on a conventional emulator,” explains Simon Benjamin, a quantum researcher at the University of Oxford.
Again, no. That’s a wild exaggeration. The Google paper makes clear that while its team has been able to show quantum supremacy in a narrow sampling task, we’re still a long way from developing a quantum computer capable of implementing Shor’s algorithm, which was developed in the 1990s to help quantum machines factor massive numbers. Today’s most popular encryption methods can be broken only by factoring such numbers—a task that would take conventional machines many thousands of years.
But this quantum gap shouldn’t be cause for complacency, because things like financial and health records that are going to be kept for decades could eventually become vulnerable to hackers with a machine capable of running a code-busting algorithm like Shor’s. Researchers are already hard at work on novel encryption methods that will be able to withstand such attacks (see our explainer on post-quantum cryptography for more details).
The main reason is that they still make far more errors than classical ones. Qubits’ delicate quantum state lasts for mere fractions of a second and can easily be disrupted by even the slightest vibration or tiny change in temperature—phenomena known as “noise” in quantum-speak. This causes mistakes to creep into calculations. Qubits also have a Tinder-like tendency to want to couple with plenty of others. Such “crosstalk” between them can also produce errors.
Google’s paper suggests it has found a novel way to cut down on crosstalk, which could help pave the way for more reliable machines. But today’s quantum computers still resemble early supercomputers in the amount of hardware and complexity needed to make them work, and they can tackle only very esoteric tasks. We’re not yet even at a stage equivalent to the ENIAC, IBM’s first general-purpose computer, which was put to work in 1945.
Besting conventional computers at solving a real-world problem—a feat that some researchers refer to as “quantum advantage.” The hope is that quantum computers’ immense processing power will help uncover new pharmaceuticals and materials, enhance artificial-intelligence applications, and lead to advances in other fields such as financial services, where they could be applied to things like risk management.
If researchers can’t demonstrate a quantum advantage in at least one of these kinds of applications soon, the bubble of inflated expectations that’s blowing up around quantum computing could quickly burst.
When I asked Google’s Martinis about this in an interview for a story last year, he was clearly aware of the risk. “As soon as we get to quantum supremacy,” he told me, “we’re going to want to show that a quantum machine can do something really useful.” Now it’s time for his team and other researchers to step up to that pressing challenge.