Pieter Abbeel Weighs In on Humanoid Robot Training as Authors Show Real-Hardware Validation
pabbeel · x · 2026-09-06
UC Berkeley professor Pieter Abbeel asked @SongShuran whether training a single robot across enough tasks could generalize toward humanoids, calling one example "a pretty convincing stochastic gradient step" in that direction. The authors replied that, at reviewer 2's request, they also validated the design on real-world hardware.
Related event: Model auto-optimizes ALOHA robot design, validated on hardware(3 posts)→
More from Embodied
- Viam launches free 90-minute robotics course: build a palletizing robot arm app in simulation — SucceededMind · 2026-09-07
- Testing VibeCAD on a two-part mold: hand-designed backup ready if AI fails — IanPritchard · 2026-09-07
- First Tesla CyberCab spotted in New Jersey with Florida plates, sparking cross-state testing speculation — ChrisUniverse · 2026-09-07
- AI-generated 20 kN methalox engine goes from spec to hot-fire in about three weeks — creatoroff · 2026-09-07
- MIT and EPFL build a sub-300g robot that swims underwater then flies out — lukas_m_ziegler · 2026-09-07
- IDUN Technologies eyes smart glasses integration with its brain-sensing earbuds at IFA — melnykowycz · 2026-09-07