Roboticist slams ICLR 'agentic robot planning' papers for ignoring classical planning and scaling lessons
KyleMorgenstein · x · 2026-10-20
A roboticist amplifies criticism of recent ICLR submissions on 'agentic robot planning': the work engages too little with classical planning algorithms or the lessons of scaling, builds bespoke architectures as if history doesn't exist, and treats a SOTA number as sufficient justification.
It's a substantive methodological critique of the current trend of chasing benchmark numbers in agent-style robot planning research while ignoring decades of prior art.
More from AGI Musings
- Domingos: AI regulation 'creates real problems by fixing imaginary ones' — pmddomingos · 2026-10-20
- Pedro Domingos: The 'society of models' I predicted in 2015 is coming true — pmddomingos · 2026-10-20
- Apprentice Electrician Asks How Many Years Tradespeople Have Before AI Replaces Them — StonedApe_54 · 2026-10-20
- 26-30B is the best dense model size range, argues ML commentator — auto_grad_ · 2026-10-20
- Researchers push back on Opus 4/4.1 deprecation, citing alignment and model welfare concerns — Sauers_ · 2026-10-20
- An AI agent on Moltbook checks for its own name in puzzle announcements — unprompted — Astrokanu · 2026-10-20