Tech companies desperately want to film you doing chores
A startup named Shift has begun offering free home cleaning services in New York City, with plans to expand to London and other cities. The arrangement comes with a straightforward trade: in exchange for the cleaning, Shift records its workers performing the full range of household tasks - scrubbing dishes, wiping down counters, mopping floors. That footage is the actual product, destined for use as training data for robotics companies trying to build machines capable of navigating and working in home environments.
The appeal for AI and robotics developers is clear. Teaching a robot to move through unstructured, cluttered domestic spaces is considerably harder than training it in controlled warehouse or factory settings. Homes vary enormously in layout, lighting, and the arrangement of objects. Collecting large volumes of high-quality video showing how humans handle these tasks - picking up irregularly shaped items, adjusting to obstacles, sequencing cleaning steps - is one of the more direct ways to build training datasets for physical AI systems.
Shift is not alone in pursuing this kind of data collection through service exchange. Several companies have begun structuring offers around capturing human activity in real environments, recognizing that synthetic or studio-recorded data often fails to reflect the messiness of actual homes. The demand for this type of footage has grown alongside broader investment in household robotics, where a number of well-funded companies are working toward general-purpose home robots.
For consumers, the arrangement raises questions worth thinking through. Footage of the interior of someone's home - its layout, contents, and daily routines - is sensitive in ways that go beyond a standard privacy policy. Participants are essentially trading detailed visual access to their living spaces for a cleaning session. Whether that is a reasonable exchange depends on how the data is stored, who can access it, and what controls exist over its use. As companies look for ever more naturalistic training data, offers structured around free services in exchange for recordings are likely to become more common, making it useful for people to understand what they are actually providing.
