Workers in Tamil Nadu film daily tasks such as slicing mangos and tying shoes using headband cameras, earning about $2.40 per hour to supply training data for AI-powered robots. The footage, known as egocentric video, is processed by companies like Objectways, founded by CEO Ravi Shankar, which supplies US tech clients including Amazon. Goldman Sachs forecasts the humanoid robot market will reach $38 billion by 2035, driving demand for this type of human-context data. The practice has sparked debates over worker privacy and pay equity, as low-cost labor in India feeds high-margin robotics products in the US.
Workers Film Daily Tasks to Generate AI Training Data
In southern India, individuals record themselves performing ordinary activities — slicing mangos, tying shoes, making coffee — using smartphone cameras strapped to their heads. The resulting first-person footage captures how hands move through tasks, mapping intent, motion, and environment in a single stream. This egocentric video format is prized by AI developers because it teaches robots imitation without requiring step-by-step coding. Workers are paid approximately $2.40 per hour for recording these home and workplace tasks. One example cited in the source is Nagireddy Sriramyachandra, who films mango-slicing in Tamil Nadu.
Objectways Supplies Egocentric Video to US Tech Clients
Objectways, a data-labeling company founded by Ravi Shankar, operates as a hub in the AI training supply chain. Employees film hundreds of micro-tasks inside staged homes and factory mockups, then colleagues annotate the frames into machine-readable steps. According to the company, the output is delivered to US tech clients building household robots and warehouse systems, including firms like Amazon. The business model relies on producing large volumes of clean, real-world behavior data to train humanoid and mobile robots. Goldman Sachs projects that spending tied to humanoid robots could exceed $38 billion by 2035, assuming hardware costs decline and general-purpose AI models continue to improve.
Privacy and Pay-Equity Concerns Emerge from Data Collection Practices
The recording of daily activities in kitchens, living rooms, and factory floors raises privacy questions. Some workers avoid filming bedrooms or family members, and others seek clear rules on data retention, licensing, and whether their footage will be used in future commercial models without additional compensation. Pay equity is another concern: the gap between robots trained on low-cost labor and premium products sold in the US has drawn scrutiny from policymakers and customers. The question of whether datasets enabling high-margin robotics should command higher wages for contributors parallels earlier debates around ride-hailing and content moderation. Despite these issues, US teams continue to require diverse hands, lighting, and environments to avoid brittle AI models, sustaining demand for the footage.
FAQ
What tasks do workers in Tamil Nadu film for AI training?
Workers film ordinary daily activities such as slicing mangos, tying shoes, and making coffee using headband cameras. The footage is used to train AI-powered robots to perform similar tasks.
How much are workers paid for recording egocentric video?
Workers earn about $2.40 per hour for filming home and workplace tasks that supply training data for AI robot development.
What privacy concerns arise from this data collection?
Footage often comes from kitchens, living rooms, and factory floors, raising questions about retention, licensing, and whether content will be used in future commercial models without ongoing worker compensation. Some workers avoid filming bedrooms or family members.
