The gig workers who are training humanoid robots at home

2026-04-01 · Source: MIT Technology Review · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Cybersecurity & Data Privacy · Depth: Fundamental Awareness, medium

Summary

Micro1, a US company based in Palo Alto, California, is employing thousands of gig workers across more than 50 countries, including Nigeria and India, to collect real-world movement data for training humanoid robots. These workers, like medical student Zeus, strap iPhones to their heads and record themselves performing household chores such as folding laundry, washing dishes, and cooking. This data collection addresses the challenge of training humanoid robots, which require vast amounts of complex physical interaction data that virtual simulations struggle to provide accurately. Robotics companies, including Tesla and Figure AI, are investing over $6 billion in humanoid robots, driving a booming gig economy for data recording. Micro1 pays workers around $15 an hour, a significant income by local standards, but the work raises concerns about privacy, informed consent, and the quality and variety of data collected from diverse home environments.

Key takeaway

For CTOs and VPs of Engineering exploring humanoid robot development, recognize that real-world data collection is a critical, albeit complex, bottleneck. Your teams should prioritize robust data acquisition strategies, acknowledging the ethical implications of privacy and consent in gig-economy data sourcing. Be prepared for the significant time and investment required to gather the vast, varied datasets necessary for effective robot generalization, as this challenge is more profound than for large language models.

Key insights

Gig workers globally are recording daily chores to generate real-world movement data for training humanoid robots.

Principles

• Real-world data is crucial for humanoid robot manipulation.
• Data variety is essential for robot generalization.

Method

Workers record chores with head-mounted iPhones; videos are reviewed by AI/humans, then annotated to label actions for robot training.

In practice

• Consider real-world data for complex physical AI tasks.
• Implement AI and human review for data quality control.

Topics

• Humanoid Robots
• Gig Economy
• Robot Training Data
• Data Collection
• Privacy Concerns

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Scientist, Robotics Engineer, AI Ethicist

Related on AIssential

• The Download: gig workers training humanoids, and better AI benchmarks — AI training and evaluation methods require significant evolution to meet real-world demands and ethical considerations.
• Startup offers free home cleaning—if it can record it all for robot training — Startups are offering free services or payments for user-generated video data to train embodied AI robots.
• China is training a robot future — one folded shirt at a time — China is overcoming robotics data scarcity by locally crowdsourcing real-world task videos, leveraging labor costs and public support.
• Data Standards for Humanoid Robotics: The Missing Infrastructure for Physical AI — Data standards are essential for enabling cumulative physical experience and overcoming data silos in humanoid robotics.
• Tech companies desperately want to film you doing chores — Physical AI development faces a critical data bottleneck due to the complexity and cost of acquiring real-world interaction footage.
• Operating a Humanoid With Your Body Is a Hot Job in China’s Hardware Capital — Human teleoperation of humanoid robots collects training data for eventual autonomous functionality in workplaces.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by MIT Technology Review.