OpenAI's GPT-5.2 discovers new physical formulas for gluon interactions

Investing.com – OpenAI’s GPT-5.2 has made breakthroughs in theoretical physics, discovering that certain gluon interactions previously thought impossible can occur under specific conditions.

This research is detailed in a preprint titled “Single Negative Gluon Tree Amplitude Non-Zero,” challenging the standard textbook assertion that when one gluon has negative helicity and the others have positive helicity, the corresponding tree-level amplitude must be zero.

The authors include researchers from renowned institutions such as Harvard University, Cambridge University, Vanderbilt University, the Institute for Advanced Study, and OpenAI. They identified a specific “semi-collinear region” where gluon momenta follow special arrangements, allowing for non-zero amplitudes.

GPT-5.2 Pro initially hypothesized this formula after human authors computed complex expressions for smaller cases. The AI model then significantly simplified these expressions and discovered a pattern leading to a universal formula applicable to all cases. The internal version of GPT-5.2 took about 12 hours to develop a formal proof, which was later verified through standard methods.

Nima Arkani-Hamed, a physics professor at the Institute for Advanced Study, described these findings as “exciting” and noted that the paper demonstrates how modern AI tools excel in pattern recognition within physical formulas.

Nathaniel Craig, a physics professor at the University of California, Santa Barbara, characterized this work as “journal-level research advancing the frontiers of theoretical physics” and “a glimpse into the future of AI-assisted science.”

With the help of GPT-5.2, researchers have extended these findings from gluons to gravitons (particles mediating gravity) and are developing further generalizations.

This article was translated with artificial intelligence assistance. For more information, please see our Terms of Use.