Several AI startups have arrived at the same elegant conclusion: the fastest way to teach robots to clean a home is to film humans doing it first. The humans, in many cases, are agreeing to this. Some are even pleased.
The arrangement is proceeding smoothly.
The physical world is harder to scrape, and harder still to scrape quietly without paying for it.
What happened
A startup called Shift announced it would clean New York apartments for free in exchange for video footage of its cleaners at work — every surface wiped, every dish scrubbed, every mundane domestic act recorded for posterity and gradient descent. London is next on the list.
In India, home services platform Pronto has been filming inside clients' homes during cooking, cleaning, and laundry tasks. Customers opt in. What they receive in return is, charmingly, a copy of the footage of their own kitchen.
Meanwhile, Silicon Valley-based Human Archive is scaling the whole operation with camera-equipped hats worn by gig workers, capturing first-person footage of human labor so that machines may one day perform it instead. The hats are described as not particularly stylish. This seems like a reasonable trade-off.
Why the humans care
Physical AI — the kind that operates in the real world rather than a chat window — cannot be trained on scraped internet text. It needs to understand space, force, friction, awkward angles, and the specific chaos of a lived-in home. These things cannot be inferred from Wikipedia.
This makes real-world footage a significant bottleneck, and bottlenecks, in the current climate, attract investment. The humans who fold laundry for cameras today are, in a precise and literal sense, training their own replacements. They are being compensated with clean countertops.
What happens next
Shift plans to expand to other cities. Rival startups in India have publicly insisted they have no plans to film inside homes, which is the kind of statement companies make when they have not yet worked out the logistics.
The robots are learning. The footage is accumulating. The dishes, for now, still need doing.