OpenAI’s ‘embarrassing’ math | TechCrunch

“Lifted by their own GPTards.”

This is what Mater Chief AI Scientist Ian LeCun said described OpenAI researchers push after triumph over GPT-5’s supposed math breakthrough.

Google DeepMind CEO Demis Hassabis has been added“It’s embarrassing.”

Decoder Report In a since-deleted tweet, OpenAI VP Kevin Weil announced that “GPT-5 has found solutions to 10 (!) of Erdő’s previously unsolved problems and made progress on 11 others.” (The “Erdős problem” is famous conjecture (posited by mathematician Paul Erds.)

But the mathematician Thomas Bloom, who maintains Erdos problem websitesaid Weil’s post “A Dramatic Misrepresentation” — Although these issues were indeed listed as “open” on Bloom’s website, he said that only meant, “I’m not personally aware of a paper that addresses this.”

In other words, it is incorrect to claim that GPT-5 was able to solve previously unsolved problems. Instead, Bloom wrote, “GPT-5 found the reference, which solved this problem, which I personally did not know.”

Sebastien Bubeck, an OpenAI researcher who was also touting the achievements of GPT-5, then Recognized that “solutions have only been found in the literature,” but he suggests that this remains a real achievement: “I know how difficult it is to search the literature.”