GEN-1.5 success driven by repetitive motion data and UMI collection
DrJimFan · x · 2026-08-20
Dr. Jim Fan analyzes the success of GEN-1.5, attributing it to leveraging naturally repetitive motions in human data: symmetric patterns and recovery from fumbles. The key is keeping the full arc of failure and recovery rather than over-sanitizing data. He also highlights UMI (human-in-the-loop data collection) over teleoperation, as it preserves essential physical intuition. With sufficient data, zero-shot object manipulation becomes possible.
More from Embodied
- Generalist AI shows robots learning on the fly from videos — willknight · 2026-08-21
- By late 2030s, AI plus humanoid robots could upend an economy built on wages — VraserX · 2026-08-20
- Using RealSense 3D camera for virtual green screen with depth control — chrismatthieu · 2026-08-20
- LLMs outputting JSON are too slow to control robot joints in real time, argues researcher — nikolaj2030 · 2026-08-20
- NVIDIA Tutorial: Post-Train Cosmos 3 Edge for On-Device Robot Control on Jetson Thor — MonaJalal_ · 2026-08-20
- Open-Gen Attempts to Replicate GEN 1.5 Embodied Model Architecture — KyeGomezB · 2026-08-20