Indian factory workers train AI systems that may replace them

The story celebrates that a dataset was built - the footage, the motion capture, the thousands of hours of skilled hands stitching, cutting, pressing, fed into models that will learn to replicate the gesture without the hand. But a made thing does not stop where its maker’s attention stops; it goes on acting in a world no lab contains. The question the launch skips is the only one that lasts: who is answerable for what this does after release, and what did its makers fail to imagine?

The made thing here is not the camera, nor even the model. It is the substitution - the system that watches a worker’s expertise, learns its rhythm, and then renders the worker superfluous. The makers are the factory management who authorised the recording, the supervisors who stood on the floor in India and placed the cameras on workers’ heads, the AI companies (unnamed in the telling but present in the architecture) that will train on this corpus and sell the result back as automation. The moment the story treats creation as completion is the first time a supervisor handed a garment worker a head-mounted camera and called it a task, not a transfer. That handoff was the launch. Everything after is the long tail.

Consider the worker who puts on the device. She knows the seam allowance by heart; her fingers find the fabric’s grain without looking. The camera records what her body knows - the micro-adjustments, the tension judgments, the speed that comes from repetition. This is not raw data. It is embodied knowledge, extracted. The makers call it training. The worker calls it her livelihood, filmed from her own forehead. When the model later guides a robotic arm through the same motion, no one will ask whether she consented to training her replacement. The consent was assumed in the handoff. The debt was incurred in the same gesture.

The stakes are not abstract. The workers’ jobs are at risk of being replaced by automation, and their current labor is being used to train the AI systems that may replace them. This is the maker’s debt in its purest form: the creation acts upon the very people who made it possible, and the makers have not remained to answer for the consequence. The contested point - the specific financial compensation or future support for workers who are replaced by the automated systems they are helping to train - remains unresolved. Unresolved is the polite word. Unacknowledged is the accurate one. The makers have not yet decided whether the debt exists. They are still celebrating the dataset.

There is a particular cruelty in the geometry of it. The camera sits on the worker’s head, seeing what she sees. Her gaze becomes the training signal. Her skill becomes the algorithm’s capability. When the factory floor reconfigures - fewer stations, more robots, a supervisor monitoring screens where dozens of workers once stood - the camera will not have moved. It will have done its work. The makers will have moved on to the next deployment, the next dataset, the next launch. The worker will remain, or she will not. The system does not care. The makers have arranged not to care either.

This is not a failure of foresight. It is a refusal of stewardship. The cleverness to build the extraction pipeline was real; the wisdom to be accountable for where the pipeline leads was never budgeted. The gap between those two achievements is where the harm lives. A supervisor handing a camera to a seamstress in a factory in India is a small scene. But the scene contains the whole architecture: the maker who releases the made thing into a world they do not control, and then pretends the release was not their hand that opened the door.

The ledger remains open. The cameras are still recording. The models are still training. The workers are still stitching. No one has yet told them what comes due when the substitution is complete.