In the southern Indian city of Karur, a cavernous textile factory floor operated by Texworth has become a focal point for the evolving intersection of human labor and artificial intelligence. Three men walk across the factory floor, distributing GoPro cameras and smart glasses to employees, including 40-year-old Shoba Kirubakaran. She is currently tasked with packing yellow aprons decorated with chickens into plastic wrapping for the brand Designs By Danica. Kirubakaran, who has worked at the factory for 10 years, six days a week, from 9 a.m. to 6 p.m., earns $90 a month. She notes that this is the third time she has been instructed to wear a GoPro, and she relies on her income to help her husband, a rickshaw driver, raise their two sons. "I only know how to do this job," she says.
The practice of recording workers became a subject of public concern in India this past April after a video showing textile workers wearing headsets went viral. News outlets such as News9Live warned that the scenes were not from a science fiction film but were happening on factory floors across the country. AsiaNet News similarly cautioned that Indian workers are training AI robots to take their jobs. India has emerged as a key location for this data collection due to its large population, relatively low wages, and openness to new technology. Developers require millions of hours of what the industry calls egocentric data—footage gathered from the perspective of a human—to train AI robots.
Prathosh A.P., a professor at the Indian Institute of Science in Bengaluru, explains that current AI systems are essentially "brains without a body." He notes that developers must provide these systems with motor skills to perform the intricate motions that humans execute using their physical bodies. While some workers sign up for the task because they are paid by the hour, others like Kirubakaran are often made to record their work without detailed explanations. Astha Kapoor, co-founder of the Aapti Institute, which studies the impact of technology on people, describes the practice as "exploitation of labor."
The broader impact on India’s workforce—the world's second-largest, with 600 million people working or seeking work—remains a point of concern. Amit Basole, an economics professor at Azim Premji University, questions whether the country has the policy agility to train, retrain, and absorb workers into new occupations. However, some experts argue that a future dominated by factory and domestic robots is not immediately imminent. Kapoor points out that AI robots are expensive to deploy and consume significant amounts of power, which is both costly and unreliable in many parts of India. "We may be able to make the robots, but they're bloody expensive to run, right?" she says. "Humans are cheaper to run than robots."
This power crisis is visible in Karur, where the Indian tech company Objectways gathers egocentric data for AI developers. The company, which appears to be one of the town's largest white-collar employers, occupies at least four building blocks and relies on massive, rumbling generators to ensure a seamless electricity supply. Ravi Shankar, the CEO of Objectways and a Karur native who migrated to Arizona decades ago, agrees that AI robots are currently more expensive than human labor. He explains that his clients are interested in designing robots for jobs that are difficult to fill, geographically isolated, or physically grueling, such as tapping rubber from trees, dusting solar panels in rural areas, or destemming chiles.
Inside Objectways, employee Rajeshkumar Ramaswamy demonstrates how they collect exocentric data—footage from the perspective of a robot—to train systems for repetitive industrial tasks. Young men with GoPro cameras strapped to their heads record themselves manipulating grippy robot claws to perform tasks like screwing bottle lids on and off. Ramaswamy explains that they repeat this process about 100 times to teach the AI to identify various shapes and sizes of water bottles. "The robot is like a child," Ramaswamy says. "We are the humans. We are teaching the robots what needs to be done."
The company also trains robots for domestic tasks, such as folding towels, putting away books, and unpacking bags, selling this data to companies developing domestic robots that can retail for as much as $20,000. On a floor designed to resemble an Ikea-style three-bedroom house, a woman named Rani N. records herself chopping onions. Because many clients are based in the U.S., Objectways also pays Americans to record themselves performing housework to ensure robots are trained on American-specific data. Ramaswamy explains that a robot trained on Indian stainless steel plates might use too much pressure for American ceramic plates, causing them to shatter.
To gather this data, the company pays hundreds of people across India, including 24-year-old Sneha Jegadish, who lives on the edge of town. Jegadish clips a smartphone to a headset and records her hand movements while chopping carrots or folding clothes. She earns about $2.60 an hour, totaling around $120 a month. Although she notes the work is drudgery, she hopes to have a robot of her own one day. "The work we are doing now is going to be done by robots in the future," she says.
This sentiment is shared by 42-year-old textile worker Kannagi Periyasaamy, who has been recording her packaging and quality control work for the past three months. She is one of the few workers who is paid $2.60 an hour and understands the purpose of the recording. "We are training them on how to package the aprons, how to do a quality check, how to fold the aprons," she says. Convinced that a robot will eventually take over her job, she is considering taking up farming on a small plot of land her family owns. In the meantime, she views the AI as her student, adding, "I'm proud I can teach it to do a good job."
