AIxBio builder: people build the models they can, not the models they should
hardimanjames · x · 2026-10-29
James Hardiman offers a pointed critique of AI-driven biology: most teams build the models they can build, not the ones they should — illustrated with the classic streetlight cartoon of searching where the light is, not where the keys are.
More from AGI Musings
- Invoking the Lucas Critique: agent models fail without modeling environment feedback loops — soumitrashukla9 · 2026-10-29
- Alignment researchers deliberately avoided understanding deep learning, argues jd_pressman — jd_pressman · 2026-10-29
- Harvard's toughest humanities class is suddenly popular as AI revives interest in the humanities — begusgasper · 2026-10-29
- Does AI shrink the 'Gini of abilities'? An ideological clash over AI egalitarianism — zetalyrae · 2026-10-29
- jd_pressman: Old-School Alignment Researchers Self-Censored Understanding Deep Learning — jd_pressman · 2026-10-29
- Reddit debate: make AI perfect or teach people to reason — most users can't tell when AI is wrong — amyowl · 2026-10-29