Do as I Do: Human video as robot data — 'most robot data will soon be made without robots'

carlosdponx · x · 2026-11-07

Researcher Mahi will present "Do as I Do" at CoRL with an unreleased demo. He argues robots themselves are the bottleneck in scaling robot data — collecting at the speed of robots or humans will always lose to scaling at the speed of compute. The project builds a correspondence between chaotic human videos and sanitized teleoperation data for dexterous manipulation, and he predicts an inflection point where most robot data will be created without robots (or even humans) in the loop.

Original post →

More from Embodied

Embodied channel →