An AI System Found a New Kind of Physics that Scientists Had Never Seen Before

“Hearst Magazines and Yahoo may earn commission or revenue on some items through these links.”

Here’s what you’ll learn when you read this story:

For all the problems AI is causing society, one of its greatest benefits lies in the world of science.
A new study focused on the chaotic dynamics of dusty plasmas found that, when trained properly, AI can actually discover new physics all on its own.
By providing the most detailed description of this type of matter, the AI corrected long-held theoretical beliefs about how particles behave inside a dusty plasma.

In more ways than one, artificial intelligence is making the world worse. Generative AI now spews countless amounts of “AI slop,” and in classrooms, AI is slowly eroding critical thinking skills, which are… you know… critical. That’s not even mentioning AI’s unfortunate role as environmental decimator and job destroyer.

Luckily, some artificial intelligence and machine learning (ML) models have grander ambitions than ripping off beloved animators and mass-producing essays at an eighth-grade reading level. Take, for instance, a new ML model developed by a team of Emory University scientists. Typically, machine-learning algorithms are used as a tool to help scientists sift through immense amounts of data or optimize experiments, but this particular ML model actually discovered new physics on its own—at least, as it relates to dusty plasma.

You’re likely familiar with plasma—that fourth state of matter that actually makes up 99.9% of all ordinary matter in the universe. Dusty plasma is simply the same mix of ionized gas, but with charged dust particles. This type of plasma can be found throughout both space and terrestrial environments. Wildfires, for example, generate dusty plasmas when charged particles of soot mixed with smoke. In this new study—published in the journal Proceedings of the National Academy of Sciences (PNAS)—a team of researchers describes how their trained ML model successfully provided the most detailed description of dusty plasma physics yet, creating precise predictions for non-reciprocal forces.

“Our AI method is not a black box: we understand how and why it works,” Justin Burton, a co-author of the study from Emory, said in a press statement. “The framework it provides is also universal. It could potentially be applied to other many-body systems to open new routes to discovery.”

Put simply, non-reciprocal forces (as their name suggests) occur when forces exerted between two particles in a plasma are not the same. The authors describe the phenomenon as two boats impacted by the wake of the other—relative position can impact the particles’ attractive or repulsive forces.

“In a dusty plasma, we described how a leading particle attracts the trailing particle, but the trailing particle always repels the leading one,” Ilya Nemenman, another co-author of the study from Emory, said in a press statement. “This phenomenon was expected by some but now we have a precise approximation for it which didn’t exist previously.”

The ML algorithm was also able to correct some theoretical misconceptions about dusty plasma. For example, scientists thought that the charge of the particle was proportional to its size, but the model confirms that while a larger particle does contain a larger charge, it isn’t proportional, as it can also be influenced by density and temperature. They also found that the charge between particles isn’t only influenced by the distance between two particles, but also by the particles’ sizes.

One the trickiest parts of this project, according to the authors, was designing the ML algorithm in the first place. Generally, AI acquires its abilities by being fed (or training on) datasets—give AI one million pictures of a monkey, and it’ll get progressively better at identifying a monkey when it sees one. However, when it comes to discovering new physics, there isn’t much training data to go on. So, the team had to create a structure that allowed it to work with the data it did have while still giving it latitude to explore unknown physics.

“I think of it like the Star Trek motto, to boldly go where no one has before,” Burton said. “Used properly, AI can open doors to whole new realms to explore.”

You Might Also Like