AI's Growing Role in Science Raises Questions About Trust and Integrity

A dispute among mathematicians has brought new attention to concerns that artificial intelligence could weaken confidence in scientific research. Last month, OpenAI said its AI agents had solved a long-standing problem involving the Navier-Stokes equations in 88 hours, and the body overseeing a related prize said the problem had "apparently been settled" pending review.

Hours before that announcement, Tristan Buckmaster of New York University raised concerns publicly. He and Levent Alpöge, a mathematician employed by rival company Anthropic, had been pursuing the same goal and kept drafts in OpenAI's Codex tool. Buckmaster asked whether his research could have fed into OpenAI's work, while stressing he was "not accusing anyone of anything." OpenAI said no one at the company, human or AI, saw the other team's work before publication, and later said Buckmaster's Codex prompts "could not have influenced the system in any way."

The episode highlights a subtler problem than fears of dangerous technology: AI could undermine faith in the scientific process itself. Chatbots can hallucinate false information, and there is evidence this is increasingly appearing in science. Journals are also struggling with AI-generated distortions, while studies have found AI fingerprints on a large share of biomedical papers and language-model modifications in peer review reports.

Some publishers are responding. Elsevier has developed about 100 signals that flag papers for closer human scrutiny, and other outlets restrict AI-generated images or penalize invented references. DeepMind's Pushmeet Kohli said turning away from AI would be "like saying you should be computing without using calculators," but accepted that human guardrails and accountability are needed.

Source: BBC News