Fields Medalists Rebuke OpenAI After AI-Generated Navier-Stokes Breakthrough, Forcing Withdrawal from Caltech Mathathon

Academic–industry tension over machine proofs echoes the 1976 computer-assisted Four-Color Theorem and the 1998–2014 saga of Hales’s Kepler Conjecture formalisation: both cases sparked debate about whether results no human could check line-by-line still counted as mathematics. What is new is the profit-driven cadence and the data advantage of private labs, reminiscent of early-20th-century industrial research (e.g., Bell Labs vs. university physics in the 1920s) that reshaped norms of attribution and peer review. The current flare-up highlights two slow-moving but decisive trends: (1) automation is decoupling discovery from comprehension, and (2) control over fundamental knowledge is migrating from open academies to capital-intensive corporations. Over a 100-year horizon the episode may mark the fork where “human mathematics” becomes a curatorial craft while machine mathematics scales independently—much as navigation changed after the marine chronometer (1735) or as genomic science diverged after high-throughput sequencing. Whether this yields an era of unprecedented insight or a splintered, opaque knowledge base will depend less on any single proof than on how incentive structures are rewired now.