Mathematicians express concerns over AI dependence

Powerful artificial‑intelligence models are increasingly being identified by scholars as a potential existential threat to the field of AI research itself. Recent studies highlight that the rapid scaling of large language models and multimodal systems has amplified concerns about uncontrolled capabilities, concentration of expertise, and the difficulty of reproducing or auditing results, thereby jeopardizing the long‑term sustainability of scientific inquiry in the domain.

Despite these risks, the same models continue to dominate research pipelines because of their unmatched performance on a wide range of tasks, from natural‑language understanding to scientific discovery. Leading institutions report that the efficiency gains, data‑driven insights, and competitive advantage offered by these systems make them indispensable for advancing both academic and commercial projects, creating a paradox where the tools that could undermine the field’s foundations remain central to its progress.