AI has made more headway in other branches of physics research, however. Loukas Gouskos, a particle physicist who joined Brown’s physics department in 2023, has been a pioneer of AI-based research for the Large Hadron Collider (LHC), the revolutionary particle accelerator used to discover the Higgs boson in 2012. For decades, the Higgs boson was the last unconfirmed particle in the Standard Model, our best current theory for how the universe operates on the microscopic scale. Gouskos has led the development of various AI algorithms to more comprehensively study this particle, among other computational pursuits, to utilize the full power of the LHC, the largest machine of its kind, and potentially break through current shortcomings in physics.

An AI-based collaboration between Gouskos, fellow Brown physics professor Stephon Alexander ’95 ScM, ’97 ScM, ’00 PhD, and their students is ambitious on this front. Gouskos, with his experimental background, and Alexander, primarily a theoretical physicist, hope that by combining their research perspectives, they can create new models to probe physics’s biggest mysteries. “ We are trying to explore the unknown using techniques like anomaly detection, teaching our algorithm to learn very well the physics that we know, and if we see something that deviates, it could be a hint for new physics,” Gouskos says. He and Alexander see particular promise in using this technology to characterize dark matter, a hypothetical form of matter that, despite it making up an estimated 85 percent of the universe’s total mass, is nearly impossible to research because it doesn’t interact with light.

The application of AI to physics is intrinsically interdisciplinary. That’s evident by its historical roots, which Alexander himself helped germinate as a graduate student at Brown. “It was the mid-’90s,” he remembers. “Aside from some lone guys like John Hopfield and Geoffrey Hinton, back then, AI and physics was like, what are you talking about? I came up with a collaboration between a cosmologist in the physics department, Robert Brandenberger, and Leon Cooper, who was doing machine learning. It’s interesting because now the interface of AI and astrophysics is a big field. But this was the very first project back in 1995.”

The project, which would’ve examined the large-scale distribution of galaxies, never came to fruition because the computers accessible back then weren’t powerful enough. Alexander, now a tenured professor, sees promise in reviving the idea.

Nevertheless, Alexander, like Pober, recognizes AI’s limits. “Do I think that AI is going to be able to make breakthroughs in theoretical physics? Not at the current moment. I like to think of it as a soundboard,  the same way I may talk to a colleague outside of my field, and they may say the silliest thing to me, but it may be useful.”

In fact, physicists, who by the nature of their work are comfortable with uncertainty and the bounds of human knowledge, are perhaps the best equipped of all of us to understand AI holistically, with all its groundbreaking potential and ongoing limitations. AI is just a tool, and it’s one that breaks.

So look up and know that, within this infinite cosmic fabric, under long-dead stars whose light is just gracing our planet, AI is small and it knows little—a lot like us.