In Karur, India, the Texworth textile factory employs workers like 40-year-old Shoba Kirubakaran, who spends her days packing yellow aprons for Designs By Danica. Recently, workers like Kirubakaran have been asked to wear GoPros and smart glasses to document their tasks. This footage is used to train AI robots to eventually perform the same tasks. Kirubakaran, unaware of this purpose, was informed by NPR during an interview. Despite her 10 years at the factory, she earns $90 a month, supporting her family with the income.
The situation gained public attention in April when a video went viral depicting textile workers with these devices. Media outlets highlighted this development, cautioning against the potential job replacement by AI robots. India has become a hub for collecting ‘egocentric data,’ owing to its large workforce and relatively low costs. Developers require vast amounts of this data, collected from the human perspective, to train AI robots in human tasks.
Some workers are voluntarily participating for pay, but others like Kirubakaran are unaware. Astha Kapoor from the Aapti Institute criticizes this as labor exploitation. India’s vast workforce of 600 million people may face challenges if retraining and policy adjustments aren’t made to accommodate AI advancements.
The Costs and Challenges of AI Deployment
Experts argue that widespread deployment of AI robots in India is not imminent due to the high costs and power consumption. Kapoor notes that running robots is expensive compared to human labor. In Karur, companies like Objectways collect and sell data to AI developers despite power challenges, relying on generators for consistent electricity.
Ravi Shankar, CEO of Objectways, agrees that AI robots are costly and may not be viable substitutes for human workers yet. His company focuses on creating AI for specific tasks like rubber tapping, isolated maintenance jobs, and physically demanding roles.
Training AI Through Human Tasks
At Objectways, workers like Rajeshkumar Ramaswamy use GoPro cameras to capture themselves performing tasks with robotic-like claws. This ‘exocentric data’ aims to train AI robots in industrial jobs, such as handling various bottle types. Ramaswamy describes teaching robots basic manipulations as akin to instructing a child.
The company also records domestic tasks in a simulated household environment for clients mostly based in the U.S. Differences like plate material usage necessitate localized training.
Sneha Jegadish contributes by recording herself doing house chores, earning about $120 monthly, learning these tasks from her mother. She hopes future robots can replace human effort in housework.
Textile worker Kannagi Periyasaamy also participates, training robots in packaging tasks. She envisions a future transition to farming once AI takes over her job, viewing her role in AI development as a form of teaching.
Leave a Reply