OpenAI Releases New Scientific Discoveries in Mathematics, Computer Science, and Physics

OpenAI has announced three significant scientific discoveries across mathematics, computer science, and physics, following its release of hundreds of manuscripts last month claiming to solve longstanding mathematical questions.

The company's latest findings mark a departure from its previous approach, as it now involves external experts guiding the results in certain fields. In computer science and physics, OpenAI has designated these experts as "shepherds," who are overseeing the outcomes. However, for the mathematical problem at hand, the shepherds advised OpenAI to release the AI's output for researchers to examine and provide their own explanations.

One of the key findings is related to the Euclidean wave equation, where a proof has been developed using an auxiliary partial differential equation. This breakthrough establishes Sogge's local smoothing for waves in every spatial dimension of at least two. The result also yields significant implications for other conjectures, including the Kakeya maximal and dimension conjectures, as well as Stein's spherical restriction estimates.

OpenAI is making its model output available on GitHub, mirroring a similar release made last month.

OpenAI is making available on GitHub the materials related to a proof that has significant implications for computer science theory, specifically the Exponential Time Hypothesis (ETH). The ETH, widely believed to be true, suggests that certain computational problems require an exponential amount of time to solve.

The proof, which refutes the ETH, was obtained through the work of researchers led by three prominent computer scientists. To help verify and explain this groundbreaking result, OpenAI has assembled an independent working group consisting of Russell Impagliazzo, Ramamohan Paturi, and Ryan Williams.

As a gesture of transparency and collaboration, OpenAI is making the materials public immediately, allowing other experts to examine and review the proof alongside the working group. This move acknowledges the importance of peer review in verifying significant scientific discoveries.

However, OpenAI's approach has raised some concerns. The company admits that designating certain individuals to receive the output first could lead to unequal access, borrowed reputation, and biased selection.

The research team has made the AI output and Lean proof available on GitHub for public access. This move aims to facilitate transparency and scrutiny of the claims presented by the AI model.

In addition to sharing the research findings, the group has issued an open letter soliciting feedback from the academic community on several key issues. They are seeking input on how to handle early access to AI-generated results, disclosure of potential conflicts of interest, and the use of unpublished materials for future research.

The letter poses three fundamental questions that require consideration: How should credit be assigned to both new discoveries and the work involved in verifying and extending those findings? What steps can be taken to support researchers and students as methods and research directions evolve?