Sim2real Debate: Proper Sysid Makes It Easy, Vision Policies Are the Real Bottleneck
csgbwk · x · 2026-09-03
Roboticist csgbwk pushes back on Yacine's sim2real takes, arguing sim2real isn't that hard if you do proper system identification and avoid training vision policies—the hard part is vision. Yacine counters that the easiest path is training a generalist model spanning multiple robots of the same category, and dismisses sysid as unnecessary since RNNs learn it anyway. A substantive exchange on embodied-AI training methodology.
Related event: Robotics debate: what really makes sim2real hard?(3 posts)→
More from Embodied
- SimInfinity: open-source robotic simulation benchmark with 10K real-world tasks — joemeno · 2026-10-18
- Robi: a German/English voice companion demo built on Gemma, Tyger and Humaneness Voice TTS — realmrfakename · 2026-10-18
- Munich's Plim unveils wheeled home-care humanoid controllable from 9,000 km away — CyberRobooo · 2026-10-18
- Musk says Optimus and Cybercab will make up the vast majority of Tesla's value — XFreeze · 2026-10-18
- SAE weight editing steers robot behavior at inference time, IIIT Hyderabad paper lands IJCAI Safe Physical AI workshop — ponguru · 2026-10-18
- NTU-led survey reviews nearly 200 papers on affordance learning for embodied AI — jiqizhixin · 2026-10-18