SHOWDOWN AT THE CHALKBOARD! A spat among mathematicians is fuelling fears that AI could undermine trust in science itself, the BBC reports.
Last month OpenAI said its AI agents had solved a major unsolved problem tied to the famous Navier–Stokes equations, cracking it in 88 hours. The body that runs a prize for the problem said it had “apparently been settled,” but that its review would take time.
Hours before the announcement, NYU mathematician Tristan Buckmaster went public with concerns. He and Levent Alpöge, who works by day at rival Anthropic, had been chasing the same goal and kept their drafts in OpenAI’s coding tool, Codex. Buckmaster asked whether that work could have fed OpenAI’s, while stressing he was “not accusing anyone of anything.”
OpenAI says nobody there, “human or AI,” saw the other team’s work before it was published. It first said it couldn’t “rule out” that anonymised product data helped its models, then said days later that Buckmaster’s Codex prompts “could not have influenced the system in any way.”
The bigger worry is the flood of AI in research. A Cornell team found about 147,000 phantom citations to research that doesn’t exist last year, and a Stanford analysis found up to 17% of peer-review text at four big computer science conferences substantially modified by a language model. “Generative AI has basically disrupted the whole scientific process,” says science journalist Dalmeet Singh Chawla.
Featured image: AI-generated editorial illustration by Bugle Blast.
Sources: BBC News
Leave a Reply