Wink Pings

OpenAI Publishes Three New 'Major' Math & Science Discoveries in Under a Month, This Time Assigns 'Shepherds' to AI Findings

After sparking academic controversy last month by releasing hundreds of AI-generated math solutions, OpenAI has now unveiled three so-called 'major' scientific breakthroughs in math, computer science, and physics. For the computer science and physics results, external experts are brought in as 'shepherds' to oversee the vetting and curation, while math findings are published as raw AI outputs, available for global researchers to independently verify and derive.

OpenAI logo overlaid on a complex math formula background, Image copyright: The Verge

It has been less than a month since OpenAI released hundreds of manuscripts claiming to solve long-standing math problems.

The publication did not include full peer reviews, sparking much debate in the math community. The core question was straightforward: could the complex proofs generated by AI hide logical flaws that would be difficult for humans to detect in a short period?

The company has recently announced three new 'major' research findings covering math, computer science, and physics. There has been a noticeable adjustment to the release rules this time.

OpenAI has brought in external experts for the findings in computer science and physics, officially defining these experts' role as 'shepherds' to guide the organization and verification of the results.

For the math field, a different approach has been adopted. External experts involved advised that the raw AI-generated outputs be published directly, leaving it to researchers around the world in the relevant fields to independently derive and supplement their own explanations and verification conclusions.

The relevant math results have been updated to OpenAI's public GitHub repository today. The corresponding entry can be viewed at math project history page. Currently, the first result listed on the page is a proof of local smoothness for the Euclidean wave equation. The public summary mentions that the proof uses an auxiliary parameter method, and the full derivation process is not displayed on the summary page.

You can access OpenAI official announcement page for full descriptions of the three findings. For background on the release of the first batch of manuscripts last month, refer to The Verge's previous report.

Information source: The Verge

发布时间: 2026-10-27 07:34