OpenAI announced three significant scientific results obtained with the help of its AI models. They concern mathematics, computer science and physics.
In the field of mathematics, the model proved the local smoothing conjecture in every spatial dimension of at least two. The result also covers the Kakeya conjectures and Stein's restriction estimates in the full range.
The company published the model's output directly on GitHub, following the experts' recommendations. Previously, the conjecture had been proved only in three dimensions.
In computer science, the AI refuted the Exponential Time Hypothesis. A deterministic algorithm is presented that exactly counts satisfying assignments for CNF formulas of fixed width in subexponential time.
Verification of the result will be led by an independent group under the direction of Russell Impagliazzo, Ramamohan Paturi and Ryan Williams. The materials were made public immediately.
In physics, a mathematically rigorous proof was obtained of the existence of certain interacting quantum field theories with gauge fields and fermions in four dimensions. Examples include quantum chromodynamics and a unified completion of the Standard Model.
The group of physicists will be led by David Gross, and it also includes Michael Douglas, Arthur Jaffe, Igor Klebanov and Nikita Nekrasov. They decided to postpone the full release in order to prepare their own interpretation.
OpenAI engaged external scientists as "shepherds" to work with the results. The company notes that different approaches to publication have their own trade-offs, including issues of access and information asymmetry.
Professor of Mathematics Alex Iosevich of the University of Rochester called the proof of the local smoothing conjecture a historic development. He expressed joy and optimism about the potential of AI in science, although he also noted mixed feelings among theorists.
How will the process of scientific discovery change when AI begins to propose more and more such results?
