On Oct. 6, the company released a slew of new mathematical results in the form of a massive GitHub repository — an info dump that will take hundreds of human hours to sift through. The company's internal frontier model provided solutions to more than 300 problems, but the company also implied they expect experts to take issue with them, debuting new processes for paper revisions and citations.

"For future releases, we are committed to further improving the quality of the papers via the citations, mathematical exposition, and presentation of the results for better understanding," wrote OpenAI.

Last month, the developer made an industry-rattling declaration that 10,000 AI agents had worked together to solve the Navier–Stokes existence and smoothness problem, a narrower portion of the unsolved Navier-Stokes equation that's part of the Millennium Prize Problems set. The robot mathematicians' "breakthrough" could have nabbed OpenAI a $1 million prize, but the company declined.

And possibly for good reason, academics later called out the company for solving a subset of the problem, not the whole thing, using a mathematical loophole — an approach that some labeled "unnatural."

Following OpenAI's announcement, the Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study⁠ said it did not endorse AI developers using proprietary models to test solutions without the scientific community's involvement. The developer announced a new advisory group to provide recommendations on future releases. But it appears OpenAI used the same unreleased model to solve the latest trove of equations.

Unlike its previous theoretical entries, however, OpenAI alleges the latest problems were solved with just a single prompt fed to a single AI agent, the developer told Scientific American.

Source link