OpenAI's latest language model has made an unexpected contribution to theoretical physics, proposing a novel formula for gluon amplitudes that researchers then formally proved and verified.
The discovery emerged from a preprint showing GPT-5.2 generating a mathematical expression describing how gluons, fundamental particles that bind quarks together, interact at high energies. Rather than simply retrieving existing knowledge, the model appeared to derive new relationships between these interactions.
OpenAI worked with academic collaborators to rigorously test the proposal. The team confirmed that the formula holds up under formal mathematical scrutiny, suggesting the AI had genuinely identified a previously unknown or overlooked relationship in particle physics.
The finding underscores a shift in how large language models are being deployed beyond text generation. Researchers increasingly tap these systems to explore mathematical relationships and theoretical problems where pattern recognition across vast datasets can surface insights humans might miss. The physics community has grown more open to computational assistance in areas like symbolic manipulation and proof exploration.
Gluon amplitudes sit at the heart of quantum chromodynamics, the theory explaining the strong nuclear force. New formulas for these amplitudes can streamline calculations in particle physics and inform experimental work at facilities like the Large Hadron Collider.
The result doesn't position AI as a replacement for physicists but rather as a tool for accelerating discovery when humans provide the right constraints and verification frameworks.
Author Emily Chen: "This is the kind of moment that separates real capability from hype, hard to dismiss when the math checks out."
Comments