Beyond Navier-Stokes: uninterpretable AI proofs raise the question of who controls scientific discovery

hugobowne · x · 2026-09-12

Hugo Bowne-Anderson's essay examines what AI-generated mathematical breakthroughs mean for human understanding. At an MPI anniversary keynote, Oxford's Michael Bronstein argued biology should generate data optimized for ML—prioritizing scale over interpretable measurements—and that black-box models will yield equations with hundreds of uninterpretable parameters, sacrificing insight for outcomes. The essay asks: in a world of verified proofs no human understands, who controls scientific discovery, and what does it mean for knowledge work?

Related event: After AI Cracks Navier-Stokes, Who Controls Scientific Discovery?(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →