AI Reportedly Proves New Physics Formula
A recent podcast claimed that OpenAI’s GPT-5.2 independently discovered and proved a new mathematical formula in theoretical physics. The AI reportedly produced a proof that corrects a decades-old consensus in particle physics within 12 hours, challenging skepticism about AI's creative and scientific capabilities.
- The specific area of theoretical physics involved is the study of gluon scattering amplitudes, which are crucial for calculating the likelihood of particle interactions in the strong nuclear force. - The AI's discovery challenges a "standard textbook argument" that predicted a zero result for a specific configuration of gluon interactions (where one gluon has negative helicity and the rest have positive helicity). - The breakthrough was a collaborative effort; human physicists from institutions including Harvard, Cambridge, and the Institute for Advanced Study first worked out complex expressions for a few cases before GPT-5.2 Pro simplified them, spotted a pattern, and proposed a general formula. - An internal, scaffolded version of GPT-5.2 then independently generated a formal proof for the conjectured formula. - The research paper, titled "Single-minus gluon tree amplitudes are nonzero," was submitted as a preprint to the arXiv repository for community feedback and is being submitted for formal publication. - Nima Arkani-Hamed, a professor at the Institute for Advanced Study, commented that the AI's strength in this context was its exceptional ability for pattern recognition within complex physics formulas. - Researchers have already begun using GPT-5.2 to extend this new understanding from gluons to gravitons, the particles that mediate gravity. - One physicist, commenting on the research, described it as potentially being a problem that "might not have been solvable by humans" alone due to the complexity of the expressions involved.
