OpenAI claims mathematical advances as proof validation remains pending

OpenAI claims its internal AI models have produced major results in harmonic analysis, computational complexity and quantum field theory. The claimed advances include proving the local smoothing conjecture in all spatial dimensions, deriving Kakeya and Stein restriction results, and finding a subexponential-time algorithm for fixed-width Boolean satisfiability problems. None of the results has undergone independent peer review.

Natural language papers and Lean-formalized code are publicly available for the mathematics and computational complexity claims. The SAT algorithm would refute the Exponential Time Hypothesis if validated, but would not establish that P equals NP because its runtime remains nonpolynomial. An external expert panel has examined the computational complexity work, although broader academic acceptance remains pending.

The quantum field theory claims concern rigorous existence proofs for four-dimensional Yang-Mills theory, QCD and extensions incorporating standard model fields. Neither the proofs nor the associated Lean code are public, making these claims the least independently verifiable. The disclosed existence claim alone does not resolve the Millennium Prize Problem, which also requires a mass gap. Across the work, Lean can check logical derivations, but experts must still determine whether formal statements match the original problems and explain the methods. OpenAI has also acknowledged concerns about unequal access when selected researchers receive unpublished results first.