Gig Workers Train Humanoid Robots with Real-World Data

Gig workers worldwide now record daily tasks to train humanoid robots, a growing trend impacting robotics development. Micro1, a US firm, employs thousands globally to capture videos of activities like folding laundry and cooking using smartphones. These recordings feed data-hungry AI models for companies such as Tesla and Agility Robotics, which invested over $6 billion in humanoid robots in 2025. Workers earn $15 per hour in countries like Nigeria, where this income supports local economies despite the repetitive nature of tasks. Real-world data collection addresses simulation limits in teaching robots physical object manipulation, crucial for construction and manufacturing automation. The demand for such data exceeds $100 million annually, with firms like Scale AI and Encord also recruiting data recorders. This shift signals a new gig economy sector intertwined with AI and robotics advancements, offering operational insights for AECM professionals integrating automation on projects. Upcoming milestones include expanding data collection networks and refining AI training models to enhance robot dexterity and task versatility in industrial environments.