Summary of the Core Content
Thousands of low-income individuals in India—housewives, street vendors, factory workers, etc.—are earning a meager hourly wage of around $2.6 by recording their own actions while performing household chores, handicrafts, and assembly line tasks from a first-person perspective. This data serves as a crucial resource for global AI giants to train humanoid robots. These robots are designed to learn how to operate effectively in the real world, potentially replacing the very jobs that these people depend on for a living. The essence of this arrangement is that the lower-income groups are using their current low incomes to gamble on future employment opportunities, while the benefits of AI are being monopolized by Silicon Valley giants and capital owners, leaving them facing a precarious existence with no safety net.
1. AI Motion Trainers: Recording Household Chores for $2.6 per Hour
You may not have heard of “AI motion trainers,” but Srilamaya, a housewife in India, is one such worker. She ties her phone to her head and films her hands as she cuts mangoes, washes vegetables, and organizes kitchen utensils. She earns 250 rupees (about $2.6) per hour—an “extra income” that she can earn from home, since “who would pay for doing household chores?”
Not only her, but also Ponni, a 55-year-old street vendor who makes flower wreaths, and factory workers who record their assembly line operations, as well as street vendors who record the process of carrying goods, are all involved in this labor. Their work is highly repetitive: they have to re-record their actions if their posture is incorrect or if their hands are blocked, until the data meets the standards required for AI training. They lack both the technology and the power to negotiate their working conditions; they simply sell their labor.
2. Why Do AI Systems Need Real People’s Actions? Virtual Data Isn’t Enough
AI can write articles, draw pictures, and engage in conversations, but it still needs real-world examples to learn how to perform tasks like chopping vegetables, sweeping floors, or making flower wreaths. For instance, while AI understands the concept of “cutting mangoes,” it doesn’t know how to hold a knife securely, how to avoid the pit, or how to handle different textures of the fruit. These details can only be taught through real people’s first-person footage. Developers argue that first-person videos are essential for training robots to replicate human behavior. Tasks like chopping vegetables, making flower arrangements, and sewing cannot be generated by AI; they must be recorded by humans.
3. India as the Global AI Data Factory
India has become the largest AI data production hub due to its large population and cheap labor: the hourly wage of $2.6 is significantly lower than in Europe and America. Additionally, housewives and street vendors can work part-time without the need for a fixed workplace. This makes India a vital source of data for companies in Europe and America, which use this information to advance their robotics technologies. The household robots you might buy in the future could very well be trained using data collected from Indian women.
4. A Cruel Paradox: Teaching Robots to Take Their Jobs While Having to Keep Working
Srilamaya is aware that the data she records will enable robots to perform her tasks, potentially replacing her job. Ponni is even more realistic: “Our generation relies on skilled labor, but the next might have no jobs left.” Yet they continue to work because the money they earn now is too important. For these people, it’s a situation of being trapped between poverty and unemployment: if they don’t do this part-time job, they have no income; if they do, their jobs could be lost to robots in the future. They joke about buying robots for themselves in the future, but they know that the money they earn might not even cover the cost of one.
5. The Imbalanced Distribution of AI Benefits
In this scenario, Silicon Valley giants profit billions from these data, while the lower-income groups receive only a tiny portion of the benefits. The majority of India’s workforce (490 million) consists of informal workers who lack the skills and opportunities to transition to other jobs, often engaging in tasks that are easily automated by robots (household chores, handicrafts, simple assembly line work). Indian government think tanks have long warned that discussions about AI-induced unemployment focus on white-collar workers (programmers, analysts) while ignoring these marginalized individuals. They have no support for learning new skills or securing a future. This is the most unfair aspect of the AI era: the wealthy benefit from technological advancements, while the lower-income groups are forced to sacrifice their livelihoods to facilitate capital growth.
The Final Question
When robots learn all basic human tasks, where will these workers go to find new jobs? This may be the most overlooked survival crisis of the AI era.