How Powerful AI Models Are Changing Mathematics Research and Teaching

Summary

In a guest essay on Terence Tao's blog, Jennifer Taback considers what powerful AI models mean for mathematicians, especially those working at small liberal arts colleges. She says AI has not solved one of her long-standing problems, but it has helped identify an incorrect lemma, reorganize a paper, suggest unfamiliar directions, and improve her mathematical writing. For her, the models also partly fill the gap left by having no nearby collaborator, while their incorrect or confusing answers require active human questioning and verification. Taback argues that mathematics remains a human endeavor because machine-generated output needs substantial intervention before it can contribute to shared knowledge. She expects undergraduate teaching and assessment to change rapidly as students gain broad access to AI, but criticizes timed, proctored evaluation and says discussions dominated by research-intensive universities should include faculty at smaller institutions with heavier teaching loads and fewer assistants. A student’s AI-assisted animation of a World War II-era encryption device illustrates one constructive use. The essay also raises unsettled questions about disclosing AI use, standards for proof and publication, and how junior faculty will be evaluated under rules that have not yet been established. Taback remains optimistic that mathematicians and students can use stronger tools without giving up human curiosity, responsibility, or mathematical understanding.