OpenAI’s recent mathematics release has sparked both celebration among some academics and concern for researchers still pursuing those same breakthroughs. NYU mathematics professor Tristan Buckmaster told CNBC’s “Squawk Box” Thursday that the company’s release has disrupted the work of early-career mathematicians.
“Entire research programs were wiped out on Tuesday by their data dump,” Buckmaster stated. On Tuesday, OpenAI published more than 700 mathematical manuscripts on GitHub, characterizing them in an announcement as outputs from an internal frontier model.
The firm also provided computer-verifiable versions of numerous proofs.
For certain mathematicians, the worries go beyond simply being scooped on solutions. Northwestern University mathematics professor Bryna Kra told Themoneytimes that while AI could propel research forward, she cautioned that releasing such a massive volume of work simultaneously might erode the collaborative culture that enabled those advances.
“Mathematics is an exceptionally collaborative discipline,” Kra remarked. “No one at this company could possibly deliver a talk or field technical questions about any of these findings.” Kra noted that researchers regularly exchange unfinished ideas through talks and discussions, and she fears that concern over those ideas being ingested by AI before publication could drive mathematicians toward greater secrecy.
Buckmaster also questioned whether unpublished work fed into AI tools might have fed into OpenAI’s mathematical outputs. OpenAI did not reply to a request for comment.
“You have to understand, they have access to all our research proposals,” Buckmaster explained. “When you submit a research proposal, a panelist reviews it and must write a summary, and often, the way they craft those summaries, they feed them into the [LLM],” Buckmaster added.
In its announcement, OpenAI stated that the mathematical results came with revision and citation protocols, and that it intends to enhance the papers’ presentation and sponsor workshops to help researchers comprehend AI-generated findings. Rutgers University distinguished mathematics professor Alex Kontorovich sees grounds for optimism.
Kontorovich told Themoneytimes that mathematical training could grow more valuable as AI compels academia to reconsider hiring practices and awards. “I don’t believe we have any interest in awarding prizes to someone who merely pressed a button,” Kontorovich said.
“Mathematics fundamentally teaches people to think clearly and deeply about major problems for extended periods,” Kontorovich added. “And I believe there will be even greater demand for such skills in the future, which are hard to acquire and require extensive training.”