Google’s “quantum supremacy” usurped by researchers using an ordinary supercomputer – TechCrunch


Back in 2019, Google proudly announced that they had achieved what quantum computing researchers had been searching for for years: proof that esoteric engineering can surpass traditional ones. But this demonstration of “quantum superiority” has been challenged by researchers who claim to have overtaken Google on a relatively normal supercomputer.

To be clear, no one is saying Google lied or misrepresented its work — the painstaking and groundbreaking research that led to the 2019 announcement of quantum supremacy is still hugely relevant. But if this new paper is correct, the competition between classical and quantum computing is still anyone’s business.

You can read the full story of how Google took quantum from theory to reality in the original article, but here’s the very short version. Quantum computers like Sycamore are no better than classical computers at anything, with the possible exception of one task: simulating a quantum computer.

It sounds like a way out, but the point of quantum supremacy is to show the feasibility of the method by finding even one highly specific and weird task that it can do better than even the fastest supercomputer. Because that puts the quantum foot in the door for expanding that task library. Maybe all tasks will end up being quantitatively faster in the end, but for Google’s purposes in 2019 it was only one, and they showed in great detail how and why.

Now a team from the Chinese Academy of Sciences, led by Pan Zhang, has published a paper describing a new technique for simulating a quantum computer (specifically certain noise patterns it outputs) that appears to be a tiny fraction of those estimated for classical computers It takes time to calculate this in 2019.

Since I am neither an expert in quantum computing nor a professor of statistical physics, I can only give a general idea of ​​the technique that Zhang et al. Second hand. They represented the problem as a large 3D network of tensors, with the 53 qubits in Sycamore represented by a node lattice and extruded 20 times to represent the 20 cycles the Sycamore gates went through in the simulated process. The mathematical relationships between these tensors (each a separate set of interconnected vectors) were then computed using a cluster of 512 GPUs.

An illustration from Zhang’s article showing a visual representation of the 3D tensor array they used to simulate Sycamore’s quantum operations. Photo credit: Pan Zhang et al.

Google’s original paper estimated that this simulation scale would take about 10,000 years to run on what was then the most powerful supercomputer (Summit at Oak Ridge National Laboratory) – although, to be clear, their estimate for 54 qubits was 25 cycles; 53 qubits making 20 is considerably less complex, but would still be on the order of a few years, they estimate.

Zhang’s group claims to have done it in 15 hours. And if they had access to a real supercomputer like Summit, it could be done in seconds – faster than Sycamore. Her work is published in the journal Physical Review Letters; You can read it here (PDF).

These results have yet to be fully verified and replicated by those in the know, but there’s no reason to believe it’s some kind of bug or hoax. Google even conceded that the baton could be passed back and forth a few times before dominance is firmly established, since quantum computers are incredibly difficult to build and program while classical computers and their software are constantly being improved. (Others in the quantum world were initially skeptical of their claims, but some are direct competitors.)

Google offered the following comment acknowledging the progress here:

In our 2019 paper, we said that classical algorithms were improving (in fact, Google invented the random circuit simulation method used here in 2017, and the methods of trading fidelity for computational cost in 2018 and 2019) — but the key point is that quantum technology is improving exponentially faster. So we believe that this classic approach, despite significant improvements in recent years, will not be able to compete with quantum circuits in 2022 and beyond.

As University of Maryland quantum scientist Dominik Hangleiter told Science, this is by no means a black eye for Google, or a knockout blow for quantum in general: “The Google experiment did what it was supposed to do, start this race.”

Google could certainly hit back with new claims of its own – it hasn’t stood still either. But the fact that it’s even competitive is good news for everyone involved; this is an exciting area of ​​computing, and work like that of Google and Zhang continues to raise the bar for everyone.


Comments are closed.