Daniel Lemire Says Agentic AI Could Disrupt Academic Mathematics

Summary

Mathematician and computer scientist Daniel Lemire argues that agentic AI could soon allow a capable non-specialist, such as a top high-school student with a sufficiently large token budget, to produce work comparable in output to a good mathematics PhD thesis. He presents this not simply as a breakthrough in problem-solving, but as a disruption to academia's sorting mechanism, where papers and credentials help determine access to prestigious academic jobs. Lemire says the pressure is especially significant because only about 10% of PhD holders currently obtain tenure-track or tenured positions, with the share falling. He argues that the same disruption could affect fields beyond mathematics, including mechanical engineering and sociology. Referring to his 2024 writing, he says AI could flood the literature with irrelevant theoretical papers if evaluation continues to reward publication and passage through peer review rather than problem-solving and real-world impact. His proposed response is to move away from treating papers as the final product and reward people for producing meaningful results regardless of the tools used. Lemire also rejects the idea that AI has already replaced human researchers: despite using many AI accounts, he says he still works about 50 hours per week because completing one task with AI creates additional work.