Ted Xiao: Robot Progress Is Limited by Verification—One Policy Can Take 100+ Hours to Validate
lukas_m_ziegler · x · 2026-11-04
In an exclusive interview, Ted Xiao argues the bottleneck almost nobody names in physical AI is verification: a humanoid learns to walk in 15 minutes in sim, but validating a single real-world robot policy can take 100+ hours. Progress moves at the speed of verification, not intelligence.
More from AGI Musings
- Would You Still Pay for an App If It Was Fully Vibe Coded? — prasenx · 2026-11-04
- Debate: Scott Alexander opposes AI pause because he believes in Singularity utopia — zetalyrae · 2026-11-04
- Coinbase launches sign-up page for AI trading agents, plus an agent trading contest — LexSokolin · 2026-11-04
- Dev who tested AI design tools argues AI for Hardware revenue won't match AI for Software — bookwormengr · 2026-11-04
- Sam Altman says AI is entering the RSI era: how smart should a system be allowed to be? — ChrisGPT · 2026-11-04
- AI vs bureaucracy will be bloody: 30-minute work sitting in a 7-day queue — airkatakana · 2026-11-04