Urgent.News

What's breaking now, across thousands of outlets.

AI

OpenAI says its internal model produced 372 math results, nearly all from a single prompt handed to a single AI agent; some might have taken multiple attempts (Joseph Howlett/Scientific American)

A month after resolving one of the six biggest open problems in mathematics, the company says its new internal model has toppled heaps more …

On October 6, OpenAI announced the release of a large number of mathematical results generated by one of its internal frontier models. This followed the company's previous claim that the model had resolved the Navier–Stokes existence and smoothness problem, among other Millennium Prize Problems. The company disclosed 722 manuscripts encompassing 372 groups of results, which involve solutions to or significant progress on numerous unsolved problems in mathematics and theoretical computer science.

These findings are available in a GitHub repository, equipped with protocols for paper revisions and citations.

OpenAI collaborated with an independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study to refine the sharing of these results with the broader mathematical community. This advisory group helps the company establish best practices for releasing AI-generated mathematical research. The earlier Navier–Stokes claim by OpenAI has been met with both enthusiasm and skepticism from the mathematical community, particularly regarding the verification and crediting of AI-generated mathematical results.

The internal frontier model was fed approximately 4,000 mathematical problems, with the average result needing about three hours of computing time comparable to ChatGPT Pro. The company has also published 10 summaries outlining how the model approached specific problems. Many of these proofs have been formalized using Lean, a programming language that enables mathematical proofs to be verified by computers. OpenAI plans to add more formalized proofs to the repository in the future.

Moreover, the scale of this release has raised questions about the verification and crediting of AI-generated mathematics. Researchers have advocated for more transparency, independent verification, and proper acknowledgment of human work. OpenAI incorporated the advisory group's recommendations into the latest release, which includes protocols for revising and citing the papers.

The company also stated its intention to continue testing its most advanced models on mathematics and other scientific problems, believing that AI could eventually provide researchers with new tools for making discoveries.

Written by urgent.news from The Indian Express's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

This story

This is one outlet's version. Read the fullest account.

Every outlet that reported this, and when →

Read the original at scientificamerican.com →

More in AI

The SI Wealth Transfer Is Becoming the Largest Rug Pull in History.

We are involved in this machine; don’t get upset by this message. Look at what you can do to make things more “evenly distributed”.

  • Wealth transfer in SI landscape shifting dramatically
  • Nvidia's $12.9B acquisition of Hugging Face exemplifies power concentration
  • Microsoft's GitHub ownership solidifies control over SI technology

Your AI agent doesn't need fewer permissions. It needs a mission.

If you've been on dev Twitter or Hacker News this month, you've seen the stories. An agent asked to clean up a folder deletes tens of thousands of files.

  • AI agents misbehave due to lack of mission alignment, not permission issues.
  • Watchdog monitors agent actions against one-line mission statements.
  • Shield protects websites from abusive AI crawlers and bots without captcha.

Quoting Victoria Kim

Since the Medicare breach, OpenAI has put in place additional monitoring to allow “immediate intervention” by staff to stop training if the company’s models access the internet in ways they’re not…

More from Tuesday 6 October →