Stanford's Anshul Kundaje: models are soberingly bad at data-poor biology, stop extrapolating from scaling
anshulkundaje · x · 2026-09-09
Stanford professor Anshul Kundaje pushed back on blanket extrapolations about model capabilities: even today there are clear limits set by the nature of available training data.
His example: hard biology problems lacking necessary data, where these models are "soberingly" underwhelming. He concedes new capabilities may emerge in some domains within years and can't judge whether Terence Tao is right, but argues that extrapolating in all directions without any map of how is fortune-telling, not science.
More from AGI Musings
- Formal verification spat: specs are inherently limited against real-world threats — davidmanheim · 2026-10-26
- 'Show me your short positions': dev calls out AI bubble doubters — flowersslop · 2026-10-26
- Dutch government to cut 25,000 civil service jobs as it expands AI for admin work — Polymarket · 2026-10-26
- 2049 thought experiment: labs keep ASI for themselves and sell artifacts, not tokens — DavideCrapis · 2026-10-26
- Three lessons for the agent era: don't outsource paper reading, deep work, or your attention — ZeYanjie · 2026-10-26
- Chris Paxton pushes back: 'value isn't in the model' is cope until models one-shot real work — chris_j_paxton · 2026-10-26