Robotics training debate: sim distillation vs. human behavior cloning
KyleMorgenstein · x · 2026-08-26
Kyle Morgenstein argues that distillation from single-task policies trained in simulation differs fundamentally from large-scale behavior cloning (BC) with human data, as single-task policies cannot scale the way BC + human data does. Chris Offner questions if multi-expert distillation represents a single policy performing multiple skills, noting that "athletic intelligence" remains capped at mimicking human motions or myopic MPC-like lookahead, failing to scale to other morphologies.
Related event: Debate: Sim Distillation vs Scalable Human Data in Robot Learning(2 posts)→
More from Embodied
- How Microsoft optimized Windows for Nvidia's Arm-based RTX Spark laptops — The Verge AI · 2026-10-15
- John Deere's second-gen autonomous tillage: 16 cross-checking cameras beat sensor drift — lukas_m_ziegler · 2026-10-15
- Europe's Application-Layer Bet: Can the Rotterdam Humanoid Center Deliver? — ingliguori · 2026-10-15
- VC thesis 'Data at the Edge': cheap sensors plus AI are rebuilding the internet's data flywheel offline — rebeccakaden · 2026-10-15
- PingpongDuo shows humanoid robots playing doubles table tennis — arankomatsuzaki · 2026-10-15
- Robotics data gap: just 300K hours of manipulation data globally vs 100M+ needed, $3B spend projected — 0xSammy · 2026-10-15