OpenAI has released a new research preprint showing GPT-5.2 deriving and formally proving a new result in theoretical physics, marking a notable moment in AI-assisted scientific discovery.
The research focuses on a core particle physics concept known as a scattering amplitude, used to calculate the probability of particle interactions. According to OpenAI, the model identified a gap in long-standing assumptions and proposed a new formula for a gluon interaction previously thought not to occur under certain conditions. The result was later formally proven and reviewed with academic collaborators.
As the paper explains:
“The preprint studies a central concept in particle physics called a scattering amplitude.”
OpenAI says GPT-5.2 proposed the new formulation and generated a proof that was subsequently verified by physicists, highlighting a shift from AI as a research assistant to a more active contributor in theoretical work. In some cases, the model explored unconventional solution paths that human researchers had not previously pursued.
Why important?
Debate will continue around whether such results represent genuinely original ideas or highly sophisticated synthesis of existing knowledge. Either way, AI is increasingly being applied to challenge established assumptions in advanced scientific fields, and the pace at which it is contributing to research is accelerating quickly.
Sources