OpenAI announces three science breakthroughs and a new approach to working with academics
US AI company OpenAI has released results in mathematics, computer science and physics, this time involving outside experts in checking them. The computer science paper claims to refute the widely believed Exponential Time Hypothesis.
OpenAI, the US artificial intelligence company, has announced three discoveries it calls major, in mathematics, computer science and physics, and has posted part of the material on GitHub. Less than a month ago the company mass-released hundreds of manuscripts it said solved long-standing math questions. This time outside experts, whom OpenAI calls “shepherds,” are guiding the verification. The matter is significant because there are no settled norms yet for checking and crediting AI-generated research results.
The three results
In mathematics, OpenAI says its model proved the local smoothing conjecture for the wave equation in every dimension, which also resolves the Kakeya conjecture and Stein’s restriction conjecture in all dimensions. The company released the model output directly, on the advice of mathematicians, so researchers can work out their own explanations. A few months ago, NYU professor Hong Wang was awarded a Fields Medal for resolving the Kakeya conjecture in three dimensions.
In computer science, the paper describes a deterministic algorithm that exactly counts satisfying assignments of fixed-width CNF formulas in subexponential time. Its 3-SAT case is said to refute the Exponential Time Hypothesis (ETH), which was widely believed to be true. Verification will be led by UC San Diego’s Russell Impagliazzo and Ramamohan Paturi, who were among those who proposed the hypothesis in 1999, together with MIT professor Ryan Williams. At the group’s request, the write-up and Lean proof were made public on GitHub immediately.
In physics, OpenAI cites a mathematically rigorous existence proof of certain interacting quantum field theories with gauge fields and fermions in four dimensions, including Quantum Chromodynamics. The material has not been released in full; a group led by David Gross will shepherd the work.
Criticism of the approach
OpenAI admits this approach may have problems. The AGMAI panel of math advisers it assembled said a day earlier that AI companies should not designate certain people to receive output first, citing unequal access, borrowed reputation and biased selection. The panel recommended releasing output on scholarly repositories with Lean-formalized proofs. OpenAI also published an open letter asking the community about early access, credit and support for researchers.
