Bio researchers: AI is bad at open-ended brainstorming — and that may prevent misuse

kesvelt · x · 2026-09-26

Stanford bioinformatics professor Anshul Kundaje and other researchers debated frontier models' brainstorming ability in biology. Kundaje argued models show a "pretense of understanding" in open-ended domains without closed verification loops. Others countered that with enough hours and expertise, models are useful partners — and the fact that non-experts struggle to avoid rabbit holes may be a natural protection against AI misuse.

Related event: Stanford Professor Finds Frontier Models Fake Understanding in Open-Ended Domains(8 posts)→

Original post →

More from AGI Musings

AGI Musings channel →