OpenAI releases 722 mathematical papers at once

OpenAI has released 722 mathematical research papers addressing a range of problems, with some proving and others disproving them, ERR’s R2 Portal reported in a technology commentary by Kristjan Port. The scale is impressive but has left researchers perplexed, Port writes, adding that their reaction should also prompt the public to think more critically about intellectual productivity.

Port compares the case with suspicions that an artificial intelligence system may be behind a highly praised novel associated with French writer Thelyson Orelien. He also cites Nicholas Polson, a professor at the University of Chicago Booth School of Business, who has published more than 200 research papers this year and produced 14 short books.

Most of Polson’s research focuses on statistics and mathematics. His output raises questions about what intellectual productivity means, Port argues. Several critics suspect that artificial intelligence may be involved, but the commentary says they have no evidence to support that claim.

Port also examines the academic publishing system. Research papers are peer-reviewed by other scientists, and, he writes, most active researchers publish an average of one or two papers a year. Publishing three to five is considered a high collective pace.

He argues that the system relies heavily on publicly funded research and unpaid work by scientists, including peer review, while private companies own prestigious journals and sell the resulting research back to the public at high prices. The emphasis on publication counts, Port says, can turn publishing into a measure of success in itself, diverting resources from the search for knowledge.