OpenAI is being piled on by its rival labs over appearing to make claims that GPT-5 had solved some hitherto-unsolved Erdos problems.
After Google DeepMind CEO and Nobel Laureate Demis Hassabis had called OpenAI’s claims “embarrassing”, Meta’s AI Chief Scientist Yann LeCun has also mocked the claims. LeCun’s comments came after OpenAI employee Sebastien Bubeck had deleted his post where he’d appeared to make the assertion that GPT-5 had solved Erdos problems.
“I deleted the post, I didn’t mean to mislead anyone obviously, I thought the phrasing was clear, sorry about that. Only solutions in the literature were found that’s it, and I find this very accelerating because I know how hard it is to search the literature,” Bubeck said.
To this, Yann LeCun replied with the zinger: “Hoisted by their own GPTards”. It was a play on the expression “hoisted by your own petard”, while means being harmed by one’s own scheme or plot. A Petard was a small bomb invented in France in the 1500s, and LeCun, who is French, cleverly replaced it with GPTard.
“Science acceleration via AI has officially begun: two researchers found the solution to 10 Erdos problems over the weekend with help from gpt-5,” Bubeck had earlier posted.
But as it turned out, he’d used the phrasing “found the solution” rather liberally. Most people on reading the post thought that GPT-5 had independently come up with a solution to the problems, but as the owner of ErdosProblems clarified, GPT-5 had merely successfully searched the internet for solutions to these problems that had already existed. Bubeck’s claims had been called out by none other than Google DeepMind CEO Demis Hassabis in the harshest terms, with him saying “this is embarrassing” in response to Bubeck’s post. OpenAI has thus far used X masterfully to hype and market their products, but every now and then, these efforts also seem to backfire.