OpenAI shares new AI research claims and adds outside reviewers

In short: OpenAI says its AI produced three major research results, and it is involving outside academic experts to help review and explain them.

What happened

OpenAI announced three new research claims in math, computer science, and physics. This comes less than a month after the company released hundreds of AI written manuscripts that it said included solutions to long running math problems.

This time, OpenAI says it is changing how it works with academics. For the computer science and physics results, it named outside experts as “shepherds,” meaning they will guide how the work gets checked and communicated (like editors helping organize a complex draft for a wider group).

For the math claim, OpenAI says outside advisers recommended a different approach. Instead of trying to publish a fully polished explanation right away, OpenAI is releasing the AI’s output directly so other researchers can work through it and write their own explanations.

In computer science, OpenAI says its work shows the Exponential Time Hypothesis is false. This is a widely believed idea about how hard certain problems are to solve as they get bigger. OpenAI posted materials, including a Lean proof, on GitHub. (Lean is a tool that lets you check steps in a proof the way a calculator checks arithmetic.)

OpenAI also says it has a math result related to “local smoothing” for waves, with knock on implications for other famous conjectures. The physics work has not been fully released, but OpenAI says a group led by physicist David Gross will shepherd the review.

Why it matters

These claims are hard to verify quickly, even for experts. How OpenAI releases AI generated research, who gets early access, and how the wider research community checks it could shape whether people trust results that come from AI.

Source: The Verge AI