CoRL reflections: 'robot data is expensive' argues for PhD topics, not startup bets
joemeno · x · 2026-11-20
A CoRL attendee argues the robot learning community hasn't internalized the bitter lesson: 'collecting robot data is expensive' is a good argument for PhD topics but poor for company bets—data is the prior. For companies, uncertainty costs more than data collection since operator output is measurable while research timelines aren't. Architecture work still matters, as recent LLM research shows.
More from Embodied
- US imports half its machine tools: the Russian-doll bottleneck of physical AI — chrisgrayson · 2026-11-20
- Open-source AgnesHarness gives agents dual-system brains, cutting task time 32.9% — 量子位 · 2026-11-20
- Musk says Tesla Cybercab could win California approval as early as next year — mitchdeg · 2026-11-20
- AI writes all code to make a real robotic arm play chess, open-sourced — FinanceYF5 · 2026-11-20
- KAIST's VSpark Speeds Up Autoregressive VLA Robot Models 2.1x with Speculative Decoding — kaist-ai · 2026-11-20
- WAM-Policy Brings Language-Conditioned Loco-Manipulation to Arm-Free Quadruped Robots — Maincode · 2026-11-20