OpenAI

OpenAI Researchers Use GPT-5.2 Pro to Discover New Quantum Gravity Result


Executive Summary

OpenAI, in collaboration with researchers from several universities, has published a preprint detailing a new mathematical discovery in quantum gravity. The research demonstrates that a class of graviton interactions, known as single-minus amplitudes, are non-zero under specific conditions, contrary to standard assumptions. This finding was achieved with the assistance of OpenAI's GPT-5.2 Pro model, which successfully extended a previous result from gauge theory to gravity, highlighting a new paradigm for AI-assisted theoretical research.

Key Takeaways

* Product/Initiative: A research preprint titled "Single-minus graviton tree amplitudes are nonzero."

* Core Finding: The paper shows that single-minus helicity graviton amplitudes, long assumed to vanish at the tree level, are in fact non-zero when particle momenta are in a special "half-collinear" alignment.

* AI's Role: GPT-5.2 Pro was given a similar prior result for gluons and tasked with constructing the corresponding amplitudes for gravitons. The model solved the problem using a novel technique (the directed matrix-tree theorem) and produced a preliminary draft of the paper.

* Significance: The result connects to an infinite-dimensional symmetry known as 'w-(1+infinity)', which is considered central to quantizing gravity.

* Methodology Shift: This project suggests a change in the scientific discovery process, where AI can generate initial conjectures, allowing human researchers to focus their efforts on verification, consistency checks, and formal write-ups.

Strategic Importance

This announcement demonstrates the capability of OpenAI's frontier models to contribute to fundamental scientific discovery, moving beyond data analysis to genuine theoretical reasoning. It positions AI as a powerful collaborative tool for researchers in highly complex fields like theoretical physics, accelerating the pace of discovery.

Original article