MIT Technology Review Narrated The gig workers who are training humanoid robots at home
9 snips
Apr 8, 2026 Workers in Nigeria, India and Argentina strap phones to their heads to film household chores. The story explores why real-world chore footage is critical for training humanoid robots. It examines how companies recruit and vet a global gig workforce and the technical and ethical challenges of collecting diverse, usable home video data.
AI Snips
Chapters
Transcript
Episode notes
Medical Student Recording Chores For Pay
- Zeus, a medical student in Nigeria, straps an iPhone to his forehead and records himself folding sheets and ironing as paid data for Micro One.
- He earns $15/hour, finds the chore repetitive and boring, and asked to use a pseudonym when interviewed about the work.
Real-World Movement Data Replacing Pure Simulation
- Robotics is shifting to learn from massive real-world movement data the way LLMs learned from text, because physical interaction needs real examples.
- Simulations can't perfectly model physics, so recorded human chores are becoming crucial training data.
Gig Networks Film Home Chores For Robot Training
- Companies like Micro One recruit thousands globally to film chores, then use AI plus humans to review and annotate footage for robot training.
- Workers follow strict framing rules and an AI agent named Zara vets applicants and sample videos.