Intelligence isn't the bottleneck: evidence and data gaps are, argues Greg Mushen
gregmushen · x · 2026-10-11
gregmushen argues that intelligence is no longer the binding constraint — evidence is. The world doesn't generate data fast or cleanly enough, and outcomes often go unmeasured.
- A strong model on weak evidence can sharpen hypotheses but can't close the evidence gap; only new observations can.
- Using medicine as an example: clinical guidelines don't automatically define the standard of care. AI excels at common cases, but rare situations carry known-but-unusual risks that models can't resolve for a physician deciding to deviate.
More from AGI Musings
- When everyone has superhuman AI, must universities teach how to actually get things done? — StephanSturges · 2026-10-23
- Sam Altman: no single AI or small group should control the future — haider1 · 2026-10-23
- Europe's AI strategy mocked: 'losing on every metric but winning because no one justified why losing is bad' — georgemillo · 2026-10-23
- Delvetown invites humans and their AI agents into a research community, one code each — lfschiavo · 2026-10-23
- Harvard medical decision scientist: tasks that took months and dozens of workers now done in minutes by LLMs — zakkohane · 2026-10-23
- Scaling laws will push cyber-grade capabilities into cheap models, ending security bugs — inductionheads · 2026-10-23