OpenAI Claims Three AI Scientific Breakthroughs and New Review Plan

OpenAI announced three scientific discoveries attributed to its artificial intelligence models, covering mathematics, computer science, and physics. The company says its models resolved long-standing mathematical conjectures, refuted an important hypothesis about computing limits, and achieved a theoretical advance in particle physics.

The mathematical claim involves the local smoothing conjecture, proposed in 1991 and related to wave behavior and energy concentration over time. OpenAI says its models produced a valid proof in all spatial dimensions from two onward, also resolving the Kakeya conjecture and the Stein restriction conjecture in all dimensions. The demonstration uses an auxiliary partial differential equation, and OpenAI released both the model-generated proof text and a formalized version in Lean, a computer-verifiable mathematical language.

“OpenAI does not seek academic credit for the discoveries produced by our internal models.”

In computer science, OpenAI claims to have disproved the Exponential Time Hypothesis, which concerns whether 3-SAT can be solved in polynomial time. The company made the proof available and organized an external group to examine it.

In physics, OpenAI says it has a proof of certain interacting quantum field theories containing gauge fields and fermions in four dimensions, including Quantum Chromodynamics and a unified completion of the Standard Model of particle physics. The experts responsible decided to prepare the work before releasing the AI-generated materials.

OpenAI also introduced a new strategy to involve independent researchers in analyzing the results, following criticism over how it released hundreds of AI-generated mathematical works in early October. The company said the results still need to be turned into understandable and verifiable scientific knowledge and should not be treated as definitively established conclusions by the academic community.