Indian Workers Paid to Film Manual Labor for Robot Training, Human Actions Become Fuel for Robots
According to reports, Indian workers are being paid to film themselves performing manual labor to train robots to learn human labor movements.
Workers wear head-mounted cameras or use smartphones to record first-person perspectives of daily or factory tasks, such as folding clothes, cutting fruits, sorting plastics, and sewing. This self-centered data is crucial for training humanoid or industrial robots.
Workers are paid approximately 150 to 250 rupees per hour. Companies involved include Objectways, which provides data to global robotics and AI firms. Some workers view this as an additional source of income.
The report notes that India, due to its scale and diversity of labor, has become an important location for such data collection, sparking discussions about future job replacements.
From a market mechanism perspective, event-driven data demand is driving low-cost labor to participate in the AI training supply chain, with funds flowing into data collection and annotation services. Beneficiaries include robot developers and data intermediaries, while manual laborers may face pressure from potential automation.
Source: Public Information
ABAB AI Insight
Global robotics and AI companies are turning to large-scale collection of human first-person action data to overcome the bottleneck in learning physical world tasks, with India becoming a key supplier due to its dense manual labor scenarios, forming a data collection network from factories to homes.
On the capital path, data companies hire workers at low hourly wages to generate training materials and sell them to overseas clients, motivated by the desire to reduce robot imitation learning costs through scale advantages and strategically build a moat around physical AI data.
Similar cases can be seen in other developing countries participating in the AI data annotation and content review supply chain, as well as the early image recognition training relying on crowdsourced labor, currently at a stage of surging demand for real action data on the eve of humanoid robot commercialization.
The structural judgment indicates a technological substitution: once human labor processes are converted into trainable data, it accelerates the replacement of robots for repetitive manual tasks, with the mechanism being that low-cost data collection shortens the cycle from demonstration to deployment for machines, thereby changing the value distribution in labor-intensive industries.
ABAB News · Cognitive Laws
Human actions become fuel for robots
Low-wage data feeds high-priced automation
Learn to be replaced before replacement