Three new discoveries in mathematics, computer science, and physics

1013 points · 560 comments on HN · read original →

Points and comments are a snapshot, not live.

OpenAI announces AI-discovered proofs in math, CS, and physics.

OpenAI is releasing AI-generated results in three fields with different release strategies. In mathematics, the models proved the local smoothing conjecture in all dimensions, implying the Kakeya and Stein restriction conjectures. In computer science, they refuted the Exponential Time Hypothesis (ETH) with a subexponential algorithm for k-SAT, verified in Lean. In physics, they produced an existence proof for Quantum Chromodynamics and a unified Standard Model completion, building on Balaban's approach. External expert groups will shepherd each result. OpenAI commits funding for conferences and workshops.

What commenters are saying

Commenters focus heavily on the ETH refutation, with many calling it the most striking result. Some explain that ETH being false does not imply P=NP, only that SAT has subexponential algorithms. Several note the constructive algorithm is surprising. A few question whether this shifts priors toward P=NP, but others argue the gap between subexponential and polynomial is vast. One commenter clarifies the Lean proof files: the 'sorry' file is just the problem statement; the actual proof is in the Algorithms folder.