OpenAI keeps bulldozing mathematicians

OpenAI has repeatedly made impressive mathematical breakthroughs but botched their announcements, alienating the mathematical community. In September, nine elite mathematicians announced the Advisory Group on Mathematics and Artificial Intelligence (AGMAI), an independent body meant to advise OpenAI and other frontier AI labs on reviewing and communicating emerging mathematical results. However, OpenAI's own announcement created widespread confusion about the group's true independence, with many mathematicians initially assuming it was OpenAI-appointed. AGMAI member Martin Hairer emphasized the group receives no financial or technical support from OpenAI and operates fully independently, but acknowledged OpenAI's phrasing "didn't really help." The article details a pattern of poor academic behavior by OpenAI, including poorly written manuscripts, scant engagement with existing literature, quiet document modifications without clear records, and vague claims about solving problems. OpenAI says its unreleased model has resolved more than 100 long-standing open problems, generating significant anxiety among researchers who fear their years of work could become irrelevant. Professors Álvaro Lozano-Robledo and Colva Roney-Dougal describe the situation as a "burden" on the community and express profound uncertainty about their own research directions. The article highlights the fundamental tension between AI companies producing solutions without the accompanying understanding traditionally valued in mathematics, and the continued need for human mathematicians to evaluate and contextualize results.

(Source：The Verge)