Tech companies desperately want to film you doing chores

Shift, an AI training startup, is cleaning New Yorkers’ homes for free in exchange for video footage of cleaners performing chores like scrubbing dishes, wiping counters, and mopping floors. The company plans to expand into London and has already paid tens of thousands of people across 15 countries to record their activities through its app. The data aims to help robots learn physical tasks, a challenge due to the complexity of real-world environments compared to digital AI training. In India, home services platform Pronto has faced backlash after using clients’ homes to film chores like cooking and laundry for AI training, despite claiming explicit customer consent. Rival startups denied similar practices, highlighting ethical concerns over privacy and compensation. Meanwhile, Silicon Valley-based Human Archive partners with service companies to collect first-person video data using camera-equipped hats worn by gig workers. Companies like Shift and Human Archive are addressing a critical bottleneck: high-quality physical data for robotics is harder to obtain than digital data. Unlike chatbots or image generators, robots require real-world interactions—understanding space, motion, and materials—which demands extensive, controlled footage. Some startups, such as those in California, are creating staged data farms where workers repeat mundane tasks like folding towels or carrying boxes while sensors capture every movement. The practice raises questions about labor exploitation and privacy, as workers may not fully grasp the implications of their recorded activities being used to train AI. Shift’s expansion into London and other cities signals growing competition for physical AI data, with companies racing to refine robotics by leveraging human labor in unconventional ways.