In a sprawling textile factory in Karur, a city in the southern Indian state of Tamil Nadu, a quiet experiment in automation is unfolding — and the workers at its center say they were never told what they were really participating in. According to reporting by NPR, employees at the facility, operated by the company Texworth, have been asked to strap on GoPro cameras and smart glasses while they go about their daily tasks, capturing every movement of their hands as they fold, package, and prepare products for shipment.
The stated purpose of this data collection is to train artificial intelligence systems and robots to eventually perform the same jobs. The workers themselves, however, have largely been kept in the dark about that goal. One of them, forty-year-old Shoba Kirubakaran, was recorded on camera while packing yellow aprons decorated with chicken motifs for the brand Designs By Danica. When NPR journalists informed her during an interview that the footage was being used to teach machines to take over her role, she was visibly surprised. "Nobody told me anything about it," she said, adding that it was already the third occasion on which she had been instructed to wear a GoPro.
Her reaction captures a tension that is becoming increasingly common as industries around the world race to adopt AI-driven automation. The push to collect human movement data is not limited to manufacturing. Similar efforts are underway in warehouses, agriculture, food processing, and logistics, where companies hope that detailed recordings of human labor can serve as training material for robotic systems that never tire, never ask for raises, and never call in sick.
The logic behind the approach is straightforward. Robots have historically struggled with tasks that humans find simple — picking up irregular objects, adjusting to slight variations in a product, or applying just the right amount of pressure when wrapping something delicate. Traditional programming cannot easily account for the infinite small decisions a worker makes over the course of a shift. By recording real people doing real work, engineers can build datasets that capture those nuances, allowing machine learning models to imitate human dexterity and judgment over time.
For the companies involved, this represents a potentially enormous efficiency gain. Labor costs in textile manufacturing are significant, especially in countries like India where the industry employs millions of people. If robots can be trained to handle packing, folding, and quality inspection, the savings could be substantial. Texworth, like many of its competitors, is under constant pressure to keep prices competitive for international brands that source products from South Asia.
Yet the human cost of this transition is difficult to ignore. Workers like Kirubakaran are, in effect, helping to design their own replacements. They are supplying the raw data — the gestures, the timing, the subtle adjustments — that will one day allow a machine to do what they do, possibly faster and at a fraction of the cost. And in many cases, they are doing so without fully understanding that this is what they are contributing to.
Advocates for workers' rights argue that this lack of transparency is a serious ethical problem. They point out that consent obtained without a clear explanation of the consequences is not truly informed consent. Employees who agree to wear cameras on the assumption that they are being monitored for quality or safety purposes may not realize they are participating in research and development for automation that could eliminate their positions.
The Indian context adds further complexity. Labor protections in the country vary widely depending on the sector, the size of the employer, and whether workers are formally employed or engaged through contractors. In many factories, union representation is limited, and employees may feel they have little power to refuse a request from management. Asking questions about new technology or expressing concerns about job security can be risky, particularly for workers in low-wage positions with few alternative employment options.
There is also the question of what happens to the data itself. Footage of workers' hands and movements could, in theory, be used for purposes beyond robot training. It could feed into performance monitoring systems, be shared with third-party AI developers, or be retained indefinitely. Without clear regulations governing how such data is collected, stored, and used, workers have little ability to control what happens to recordings of their own bodies and labor.
Economists studying automation have long debated whether new technologies ultimately create more jobs than they destroy. Optimists point to historical precedents: the mechanization of agriculture displaced farmworkers, but many of them eventually found employment in factories; computerization eliminated some clerical roles while generating entire new industries. They argue that AI will follow a similar pattern, automating routine tasks while creating demand for new kinds of work.
Skeptics counter that this time may be different. Previous waves of automation tended to replace specific physical or repetitive tasks, leaving humans to perform roles that required judgment, creativity, or social interaction. AI systems, by contrast, are increasingly capable of handling cognitive and adaptive tasks as well. For workers in sectors like textile manufacturing, where the work is already routine and the margins are thin, the transition could be far more abrupt than in past decades.
For Kirubakaran and her colleagues, the future remains uncertain. The GoPro cameras continue to record. The data continues to accumulate. And somewhere, engineers are using that data to build machines that may one day stand in the same spot on the factory floor, folding aprons with tireless precision. Whether those machines will create new opportunities for the workers they displace — or simply leave them behind — is a question that neither the companies nor the governments involved have yet answered.
(0)