Gary Marcus: AI is strong in constrained domains, weak in the open physical world
GaryMarcus · x · 2026-09-11
Gary Marcus reiterates his decade-old argument: AI excels in constrained, simulable domains like chess, Go and math, but remains weak in open-ended physical worlds. He warns that pushing cognitive techniques without solving alignment was unwise, and robotics must not repeat that mistake.
More from AGI Musings
- Mario Draghi in FT: AI is Europe's best shot at sustaining its social model — paulnovosad · 2026-09-11
- US school districts restrict screens and AI as low-tech education movement gains momentum — Polymarket · 2026-09-11
- AI Fear Is Baked Into OpenAI and Anthropic's Culture, Says Rob LeClerc — robleclerc · 2026-09-11
- The Waymo Effect: how AI is quietly making research less collaborative — tw1st3d_m3nt4t · 2026-09-11
- If AI out-researches humans, who will spend 10 years becoming a scientist? — MustBeSomethingThere · 2026-09-11
- Researcher quits Anthropic after 3 years in pretraining at OpenAI and Anthropic, citing reckless race to superintelligence — thisdudelikesAI · 2026-09-11