With a smartphone strapped to her head, Indian housewife Nagireddy Sriramyachandra films herself slicing mangoes. The footage isn't for social media — it's being sold to AI companies to teach robots how to perform domestic tasks. She's aware of the irony. "I may get a robot," she has noted, referring to the technology her own labor is helping build.
According to reporting by multiple outlets including Moneycontrol, thousands of Indians are participating in this kind of data-labeling work, recording themselves performing routine household chores so that AI-powered robots can learn to replicate those movements.
The arrangement reflects a wider pattern in the global AI industry: workers in developing economies are paid modest fees to generate the training data that powers increasingly sophisticated automation systems — systems that could, in time, displace the very kinds of jobs those workers depend on.
Data labeling has long been a quiet but essential pillar of AI development, typically involving tasks like tagging images or transcribing audio. Embodied AI — training robots to navigate physical space and manipulate objects — takes that a step further, requiring detailed, real-world demonstrations of physical tasks.
The story matters because it puts a human face on one of the thorniest questions in the AI era: who bears the cost when the people building tomorrow's automation are the same ones most exposed to it.