Figure May Have Found a Scaling Law for Robot Pretraining, Predicting 8x-Data Runs
rohanpaul_ai · x · 2026-09-18
Figure appears to have found a scaling rule for robot pretraining: with model and task training fixed, adding more of its Index human-behavior dataset makes action-prediction loss fall in a clean, predictable way. Smaller runs predicted the 8x-data run almost exactly, letting robotics firms estimate gains before spending compute — though the curve predicts action loss, not real-world task success, so predictable gains in robot reliability remain unproven.
More from Embodied
- Mask R-CNN author Georgia Gkioxari argues closed-loop self-improving perception is AI's next leap — CSProfKGD · 2026-09-18
- World's first robot vs human fight set for tomorrow night in SF — cixliv · 2026-09-18
- UChicago neuroscientist decodes brain signals to restore touch and movement via brain-machine interfaces — plopesresearch · 2026-09-18
- Camera-free AI glasses are easy for Meta; shaking the 'pervert glasses' reputation is the hard part — Glittering_Royal6799 · 2026-09-18
- Figure's Home Robot Math: $400-600/Month Lease With a 56% Task Completion Rate — yongqianme · 2026-09-18
- China now has a humanoid robot factory designed for 10,000+ units a year, one every 10 minutes — VraserX · 2026-09-18