LLMs Are Near Random Chance at Robot Hand Control, Argues Researcher in Dexterity Debate
adam_karvonen · x · 2026-10-24
adamkarvonen argues LLMs alone won't solve robotics: basic pick-and-place works fine with xyz and open/close commands well suited to LLMs, but controlling a 22-DOF hand or a 100+ DOF humanoid body for delicate tasks like balancing and installing awkward objects is a different problem — model Astra is near random chance at robot hand control.
He guesses the ultimate solution is an LLM providing high-level guidance to a strong humanoid robot model. A replier pushed back that "LLMs can't do X yet" takes have a bad track record, but the author held that dexterity is genuinely different.
More from AGI Musings
- Critic slams deliberately handicapping AI search as dumb as the clock-speed wall — JosephJacks_ · 2026-10-24
- Ross Taylor: studying LLM behavior will increasingly resemble social science — rosstaylor90 · 2026-10-24
- New Yorker cartoon: "pick a field that will still require humans when you graduate" — docmilanfar · 2026-10-24
- Most "is AI conscious?" debates boil down to two people talking past each other — granawkins · 2026-10-24
- 20VC: Naveen Rao says inference is a $10T market and robotics 100x bigger than coding agents — 20VC · 2026-10-24
- Hot take: Gen Z may lose more to AI than Gen Alpha, argues Reddit post — SoftAndWetBubbles_78 · 2026-10-24