Textile Workers Wear GoPros to Train AI Robots Without Knowing
In Karur, India, textile workers like 40-year-old Shoba Kirubakaran have been told to wear GoPro cameras and smart glasses while packing aprons, unaware that the footage is being used to train AI robots to take over their jobs. Kirubakaran, who earns $90 a month at the Texworth factory, told NPR that nobody informed her of the data collection purpose.
The Push for Egocentric Data Across India
India’s large population, low wages, and openness to technology have made it a primary hub for gathering “egocentric data”—video captured from a human perspective that provides AI models with the physical motor skills and intricate hand movements they currently lack. Prathosh A.P., a professor at the Indian Institute of Science in Bengaluru, explained to NPR that artificial intelligence currently functions as a “brain without a body,” requiring human motion recordings to bridge the physical gap.
While some workers receive hourly pay to record their tasks, others like Kirubakaran capture footage without explicit explanation. Astha Kapoor, co-founder of the Aapti Institute, described the practice to NPR as labor exploitation.
The Power Constraints Facing Automated Factory Floors
Despite the rapid collection of training data, widespread deployment of industrial and domestic robots faces steep infrastructural hurdles in India. Kapoor noted to NPR that robots are expensive to run and consume massive amounts of power in regions where electricity can be costly and unreliable.
Simulating Domestic Tasks and Handling Materials
Ramaswamy demonstrated to NPR how workers wear head-mounted cameras while manipulating objects like water bottles or kitchen carrots to teach AI systems object recognition and physical dexterity. Ramaswamy noted that domestic training requires region-specific data, explaining that an AI trained to handle stainless steel plates in India might apply too much force to delicate ceramic plates commonly used in the United States, smashing them.
Worker Reactions and Future Outlook
For employees participating in the data collection for hourly wages, the transition carries mixed expectations. Sneha Jegadish, a 24-year-old earning $2.60 an hour to record household chores, told NPR she expects robots to eventually take over domestic drudgery. Similarly, 42-year-old textile worker Kannagi Periyasaamy acknowledged to NPR that her recorded packaging tasks are directly training her future robotic replacement, prompting her to consider transitioning to cattle farming on family land.
