Why AI could eat quantum computing’s lunch

There is a caveat: since the ground states are actually found by trial and error rather than by explicit calculation, they are only approximations. But that’s precisely why this approach can help make progress on a problem that seemed intractable, says Juan Carrasquilla, a researcher at ETH Zurich and another co-author of the study. Science reference paper.

If you want to accurately track all interactions in a highly correlated system, the number of calculations you need to perform increases exponentially with the size of the system. But if you’re happy with a good enough answer, there’s plenty of room for shortcuts.

“There may be no hope of capturing it accurately,” Carrasquilla says. “But the hope is to collect enough information to cover all the aspects that physicists care about. And if we do this, it will be virtually indistinguishable from the true solution.”

Although highly correlated systems are generally too difficult to model classically, there are notable cases where this is not the case. Projected to 2023, this includes some systems relevant to modeling high-temperature superconductors. paper in Natural communications.

“Because of exponential complexity, there will always be problems for which there is no shortcut,” says Frank Now, a research manager at Microsoft Research who has led much of the company’s work in this area. “But I think the number of systems for which you can’t find a good shortcut will become much smaller.”

No magic bullets

However, Stephanie Cishekassistant professor of physics at the University of Ottawa, says it can be difficult to predict what problems neural networks can realistically solve. For some complex systems they work incredibly well, but for other seemingly simple ones, the computational costs unexpectedly increase. “We don’t really know their limitations,” she says. “No one yet knows exactly what conditions make it difficult to represent systems using these neural networks.”

Meanwhile, according to scientists, significant advances have been made in other classical quantum modeling methods. Antoine Georgesdirector of the Center for Computational Quantum Physics at the Flatiron Institute in New York, who also contributed to the recent Science reference paper. “They are all successful in their own right and also complement each other very well,” he says. “So I don’t think these machine learning methods are just going to completely displace all other methods.”

Quantum computers will also occupy their niche, says Martin RoettelerSenior Director of Quantum Solutions at IonQ, which develops quantum computers built from trapped ions. While he agrees that classical approaches will likely be sufficient for modeling weakly correlated systems, he is confident that some large, highly correlated systems will be beyond their reach. “The exhibitor will bite you,” he says. “There are cases with highly correlated systems that we cannot consider classically. I firmly believe that this is the case.”