Can we scale robot training faster than robot bodies?
Tesla just committed to building a million Optimus robots a year, but the episode reveals the real bottleneck: teaching each robot to perform hundreds of household tasks requires human operators manually training them one interaction at a time. If training data collection can’t scale as fast as manufacturing, do we end up with warehouses full of sophisticated robots that don’t actually know how to do anything yet?
Commentaires (1)
In today’s episode of Minds, Bodies, and Terawatts (June 1st), we explored the unglamorous reality behind the headlines—while Tesla halted car production at Fremont to build Optimus assembly lines, the episode profiles workers like Fernando Flores who spend eight hours a day pouring coffee through a robot arm, deleting every spill, recording grip angles and wrist rotations. The guest explains that self-driving cars learned passively from millions of existing vehicles, but home robotics can’t work that way: every task, from folding a t-shirt to loading a dishwasher, requires active human teleoperation and data collection. The episode maps three potential solutions (fleet learning, remote teleoperation, and wearable sensor collection), but the core question remains unresolved—can any of them scale fast enough to justify manufacturing capacity? Jump in and tell us which approach you think actually works at scale.
Related reading on unscarcity.ai:
Envie d'aller plus loin ?
Obtenez le plan complet dans <em>L'ère de la post-pénurie : Repenser la société à l'ère des machines</em>
