Why IBM Thinks Google Hasn't Achieved 'Quantum Supremacy'

Google’s claim to “quantum supremacy” has finally been published in Nature, a month after its research was accidentally leaked online.

But quantum supremacy has definitely not been reached, according to IBM’s quantum computing research team, who argued in a blog post that it’s not even really a thing.

IBM’s resounding rejection of the much-hyped quantum computing milestone came two days before Google's paper was published on Wednesday. The scientific community has been waiting for its publication with bated breath, and a good amount of skepticism, since a draft paper appeared online last month.

In a blog post published on Monday, IBM researchers Edwin Pednault, John Gunnels and Jay Gambetta disputed Google’s claim that it would take a state-of-the-art classical computer around 10,000 years to complete the sampling task Google used to demonstrate quantum supremacy on its Sycamore quantum computer. "Supremacy" here is the point at which a quantum computer can quickly complete tasks that would take a non-quantum computer more than a human lifetime to do.

The researchers instead claim that IBM’s Summit supercomputer could perform effectively the same job in just 2.5 days, by using hard drive storage and “performance-enhancing techniques,” which Google allegedly did not consider in its estimation.

“The concept of 'quantum supremacy' showcases the resources unique to quantum computers, such as direct access to entanglement and superposition,” IBM's blog post states. “However, classical computers have resources of their own such as a hierarchy of memories and high-precision computations in hardware, various software assets, and a vast knowledge base of algorithms, and it is important to leverage all such capabilities when comparing quantum to classical.”

In an email, a Google spokesperson said that they welcome new work on advanced classical simulation techniques, like those proposed by IBM, but emphasized that it's crucial to test these techniques on actual classical supercomputers, which IBM has not yet done. The spokesperson also reiterated the company’s claims to quantum supremacy.

“With Sycamore we’ve demonstrated that we’re now in the NISQ [Noise Intermediate-Scale Quantum] era, performing on real hardware a computation that’s prohibitively hard for even the world’s fastest supercomputer, with more double exponential growth to come.

“We’ve already peeled away from classical computers, onto a totally different trajectory,” they said.

The focus on whether or not Google has achieved quantum supremacy has arguably overshadowed the real innovations it did demonstrate in the research paper. Regardless of how long it would take a classical computer to simulate the same task, Google’s quantum system is the first superconducting quantum computer of this scale which is programmable, which means it can receive instructions for many different tasks, rather than being designed to execute one specific task.

Google also showed it had achieved an important level of monitoring and control over the quantum computer, turning off one faulty qubit—the quantum equivalent of a computer bit storing 1s and 0s—out of the original 54 to reduce errors resulting from other qubits interacting with it.

Even IBM gave Google props for this. "Google’s experiment is an excellent demonstration of the progress in superconducting-based quantum computing, showing state-of-the-art gate fidelities on a 53-qubit device," the blog said, while maintaining that it still should not be considered proof of supremacy.

“I regard the quality of the fabrication and the way the control systems work as a superb piece of engineering,” said Peter Knight, a professor of quantum optics at Imperial College London and advisor to the UK Research & Innovation’s (UKRI) Quantum Technology Strategic Advisory Board.

“Other people are getting there," he added, "but it really is quite difficult to reduce the noise down to achieve this level of control”

According to Knight, there’s also much to be said for reducing the power and resources it takes to run computational tasks. Even if, as IBM claim, a classical computer can achieve similar results in 2.5 days, Google’s ability to complete the task in just 300 seconds has huge implications. Knight mentioned, as an example, the “computationally intensive” problem of protein folding—a biochemical process thought to cause diseases like Alzheimers when it goes wrong— which quantum startups are already working on.

Yet IBM has not provided enough detail on its own claims about classical simulations for it to be independently verified, Knight said, though some of his own colleagues have already started that work.

The dispute over quantum supremacy runs much deeper than the current back-and-forth over whether supremacy has been reached, however. IBM’s criticism goes even further, rejecting the very notion of quantum supremacy.

Like many others, including the inventor of the term "quantum supremacy," IBM prefers to talk about "quantum advantages," referring to the value that quantum computers deliver without comparing them to classical computers. IBM proposes alternative milestones: quantum computers which use a fraction of the memory required by classical computers, and quantum computers which can do tasks that are not possible for classical computers to do even within thousands of years.

“It is well known in the quantum community that we at IBM are concerned of where the term 'quantum supremacy' has gone,” the IBM blog states. “A headline that includes some variation of “Quantum Supremacy Achieved” is almost irresistible to print, but it will inevitably mislead the general public. First because by its strictest definition the goal has not been met. But more fundamentally, because quantum computers will never reign 'supreme' over classical computers, but will rather work in concert with them, since each have their unique strengths.”

For all the hype, quantum "supremacy" is not the only way to evaluate quantum computing progress, and arguably it’s not the most useful either.

“The supremacy experiment/benchmark is not tied to practical applications/implications,” Lieven Vandersypen, a professor at QuTech, a quantum research lab at the Delft University of Technology in the Netherlands, said. “Quantum supremacy as claimed by Google may give the wrong impression that we can now use quantum computers to solve relevant problems. That is in fact still many years away. The supremacy claim pertains to a very artificial and useless problem.”

Vandersypen said he is still extremely impressed by Google’s technical achievement, highlighting the fact that the system had more than 50 qubits but less than a 1% error rate. But he believes other ways of evaluating quantum progress are more meaningful and practical, including IBM’s published benchmarking system.

In response to a request for comment, IBM directed Motherboard to a company blog on measuring quantum computing power, which lays out their approach.The approach is much more closely tied to commercial and practical considerations than the more abstract notion of "supremacy." It measures four key features of the qubit chip (relaxation, dephasing, two-qubit error rates, and single-qubit error rates) and defines a new metric called Quantum Volume to evaluate overall performance of the computer, including errors, device connectivity and computer efficiency.

Conversations around how to validate and measure quantum computing innovation are still at a very early stage, Knight said, since benchmarking has only become a relevant consideration with the latest quantum breakthroughs.

And, despite the caveats and arguments, this is why he is so impressed by Google’s recent work, he said. The quality of its quantum device has shown, even without any comparison to classical computers, that we’re a step closer to really useful quantum machines.