Three math students affiliated with MIT reported last week that 76 percent of the 268 graduate students and postdoctoral researchers they surveyed felt less motivated to do research “as a result of AI developments.”
Nearly 76 percent also said they were increasingly unlikely to stay in academic math after finishing their degrees, according to the paper, titled “Early career researcher responses to AI in math” and posted to arXiv. About 85 percent of the respondents were graduate students and about 15 percent were postdocs. They came from almost 60 institutions in at least five countries, and all of them answered anonymously.
Their sharpest anger was aimed at their own institutions. In the survey’s free-response section, respondents called academic leadership wayward, useless, meek, silent and ineffectual, Business Insider reported.
“It’d be great if institutes gave guidance instead of platitudes on this subject,” one respondent wrote. Another wrote: “We are lacking high-level leadership from across the academic community.”
The authors, Anna Brandenberger, Hanna Mularczyk and Dora Woodruff, ran the poll through a Google Form, with multiple-choice and free-response questions about AI’s growing ability to do the work they have spent years training for. The response window closed just after OpenAI’s Oct. 6 release of AI-generated math results, and the authors cautioned that the findings capture the mood at “only a small window of time.”
They didn’t hide their own stake. “The past couple of months have been exhausting,” they wrote. “We’re being forced to reckon with existential questions about a field we love, all while incredibly uncertain about our own futures.”
OpenAI moved into mathematics in September, when it published solutions to the Navier-Stokes Millennium Prize question, Business Insider reported. On Oct. 6 the company said an internal model had written more than 700 manuscripts and produced almost 400 other mathematical results, Euronews reported. The Association for Human Mathematics accused OpenAI of plagiarism and of “total disregard for the norms of scientific research.” “Mathematicians did not ask for this work to be done,” the group said.
Nine days later, the Simons Foundation said it had teamed up with OpenAI and Anthropic on a combined $15 million in funding and resources to help mathematicians check AI-assisted results, learn the new tools and set shared standards for AI use. OpenAI said it was putting in $5 million for “workshops, conferences, and special programs,” along with model access and computing power. The announcement didn’t say how the $15 million was divided. OpenAI also set up an advisory group to judge its math work.
Some mathematical councils have issued rules on how AI should be used. Nothing obliges OpenAI to follow them.
The project started smaller. In September the authors surveyed only MIT’s math graduate students and discussed the results in an open letter to the department, then widened the poll to ask how early-career mathematicians use AI and which ethical questions worry them. The paper argues that Fields medalists, AI labs and senior professors have dominated the debate, even though the personal stakes are highest for people just starting out.
Students told the authors that the money and the advisory group had done little for them. “It is difficult to do research and go to seminars when it feels like none of the individual work will matter much in the future,” one respondent wrote.