ChatGPT Spits Out Surprising Insight in Particle Physics
A recent article in Science has highlighted a preprint publication authored by three IAS scholars from the School of Natural Sciences: Alfredo Guevara, Roger Dashen Member; Andrew Strominger, Member (1982–87); and David Skinner, IBM Einstein Fellow (2012–13).
The article describes how the scholars joined forces with Alexandru Lupsasca of Vanderbilt University and Kevin Weil of OpenAI to show that an interaction widely assumed to be impossible may actually occur "somewhere deep inside the murky guts of protons and neutrons."
The team was studying scattering amplitudes, the mathematical expression used to compute probabilities for particle interactions. For decades, a particular gluon-scattering configuration was widely treated as having zero amplitude, meaning it was assumed not to happen.
But the researchers noticed "a potential loophole." The "forbidden" configuration might be allowed if the particles are all moving almost in the same direction. However, "the calculations ended up being cumbersome and time consuming," until Guevara "finally discovered a pattern in the team’s scattering amplitude formulas. [...] The team hoped to generalize the formula and show that the interaction could happen for any number of gluons, but the resulting expression was dozens of terms long and essentially unworkable."
"After some first attempts to probe the model, the theorists asked OpenAI’s latest and most advanced public model, ChatGPT-5.2 Pro, to simplify the expression for four gluons, which it did in about 20 minutes." They then used another internal OpenAI model to generate a proof that withstood human scrutiny. The result was announced earlier this month at the annual meeting of the American Association for the Advancement of Science (AAAS).
Guevara noted in a statement to IAS: "It’s very hard to precisely document the role of AI in a theoretical physics paper. But I now believe it was crucial in elucidating cases which can escape the usual physics roadmap. I regard it as a collaborator that explores unconventional lines of thought, which is always welcome!"
Reflecting on the significance of the result, which will be submitted for peer review in the coming weeks, Nima Arkani-Hamed, Gopal Prasad Professor in the School of Natural Sciences, stated in a blog post for OpenAI: "To me, 'finding a simple formula' has always been fiddly, and also something that I have long felt might be automatable by computers. It looks like across a number of domains we are beginning to see this happen; the example in this paper seems especially well-suited to exploit the power of modern AI tools. I am looking forward to seeing this trend continue towards a general purpose 'simple formula pattern recognition' tool in the near future."
Read the article in full via Science.
