A 25-year-old woman named Rani, living in Chennai, spends most of her day doing the same household chores expected of millions of Indian women: washing dishes, cleaning the kitchen and tidying her home. The difference is that she does it with a smartphone strapped to her forehead, recording every moment.
If she cleans the kitchen today, she may be asked to fold clothes tomorrow. If she performs a task from one angle, she may be asked to repeat it from another. The footage is then used to train artificial intelligence systems and humanoid robots to perform everyday human tasks. For this, she earns around ₹250 an hour.
Speaking to AFP, Rani said, “No one gives you money for doing household work. “But here you are earning around ₹250 an hour.”
Rani has a point. This may be just a simple daily chore for her – providing some data to companies and receiving some income in return. And for many, that is enough. In a country where stable employment remains difficult for many people, earning ₹250 an hour for household work can seem like a good opportunity.
Across India, thousands of people are being paid for recording themselves carrying out everyday tasks so the footage can be used to train artificial intelligence systems. They film themselves washing dishes, folding laundry, cooking, cleaning and performing other routine activities, allowing machines to learn how humans move, respond and interact with the world. Unlike most jobs, however, the purpose is not simply to complete a task. It is to teach AI how to perform independently.
To teach AI how people carry out everyday tasks, workers are often required to repeat the same task multiple times from different angles. For example, a kitchen cleaned today may need to be cleaned again tomorrow with a camera positioned differently so the system can capture every movement in greater detail. This helps AI analyse every movement and action clearly.
Speaking to AFP, Rani said the process can be frustrating. As she works, the device frequently interrupts her with automated alerts repeating, “Hands are not detected; hands are not detected,” forcing her to stop, reposition herself and adjust her movements so the AI can capture them correctly.
Workers may need to slow down or change the way they perform a task so that the camera can capture their movements clearly. If their hands are not visible or a movement is not recorded properly, they may have to stop and do it again. The worker is still doing an ordinary task, but at the same time, every movement is being watched, recorded and turned into data that can be used to teach a machine.
This constant surveillance to train artificial intelligence is not limited to household workers like Rani. A recent investigation by The Guardian found that garment workers near New Delhi, India’s capital, were also asked to wear small cameras on their heads while working in factories. The cameras recorded how workers folded clothes, handled products, and carried out routine tasks so that the footage could be used to train AI systems and robots.
Many of the workers said that they became more conscious of their movement once the recording started. Some also said they talked less with their co-workers because they knew they were being filmed. One worker explained how the atmosphere in the factory changed after the cameras were introduced. The technology has not replaced workers, at least not yet, but it is already changing the way they work. They are no longer simply doing their jobs; they are doing them while knowing that their movements are constantly being watched and recorded.
According to an Al Jazeera report, researchers such as Aditi Surie of the Indian Institute for Human Settlements noted that services built around collecting human data for artificial intelligence are likely to expand as the industry grows. In a country where millions continue to struggle with unemployment and underemployment, it is not difficult to understand why many people agree to do this work. The immediate opportunity to earn money can outweigh concerns about being constantly recorded or about how the footage may eventually be used.
Yet many argue that the discussion cannot stop at the pay cheque.
Madhumita Dutta, an associate professor at Ohio State University, told The Guardian that workers involved in these projects are not simply selling their labour. They are also creating “a valuable digital asset”. The concern is that while workers are paid for the hours they spend recording themselves, the data they generate can continue to be stored, reused and incorporated into commercial AI systems long after the payment has been made.
Similar concerns are raised by other researchers, such as Sarayu Natarajan of the Aapti Institute. She argues that these recordings capture workers’ “bodily knowledge”, the skills, instincts and experience built through years of practice. In other words, workers are not merely completing tasks. They are transferring human knowledge to machines.
And this raises a much bigger concern.
People should earn money from this work, and AI will continue to grow anyway. Those are not the issues. The real question is whether we are paying enough attention to the costs that would come with it in the long run.
Rani may be right that AI and robotics are the future. But a future built on constant monitoring deserves far more public debate than it is currently getting. The cameras being worn today may be training machines, but they are also training workers to accept being watched all the time.
The income is real, and the opportunities are real. But so are the unanswered questions. How much surveillance should workers accept in exchange for a pay cheque? And what happens when the machines have learned enough from the very people training them today?
The problem is not that people like Rani are being paid to help train AI. For many workers, this income can be useful and important. The real concern is what may happen if being watched all the time slowly becomes a normal part of earning a living. Today, a worker may agree to wear a camera for a few hours in exchange for money. But the footage recorded during those hours may be stored and used for years, long after the worker has been paid.
This is why an hourly wage should not be the end of the discussion. Workers should know where their recordings go, how long they will be stored, who can use them and whether they will be used again in the future. If AI companies are learning from people’s movements, skills and experience, then the people providing that knowledge should have a say in what happens to it. As AI grows, the question should not only be how quickly machines can learn from humans, but also how fairly the humans teaching them are treated.
The answers may take much longer to become clear. But if we fail to ask critical questions now, the consequences may arrive much sooner than we expect.
This article is republished from The Secularist under a Creative Commons license. Read the original article.
Karuna Kumari Kandregula is an independent writer and researcher from Andhra Pradesh, India, whose work focuses on social issues, politics, education, rural systems, gender, child protection, and climate.
