Two weeks ago, an unnamed robotics AI startup in New York posted a Craigslist ad seeking first-person videos of everyday activities — simple things like cooking dinner or doing laundry at home.
The reason? To train humanoid robots. According to the listing, the videos will be used to help robots learn “navigation, object manipulation, and household activities,” essentially teaching them how to perform basic human tasks.
As Business Insider reports, the robotics industry is booming, with venture capital firms pouring more than $12 billion into the field this year alone. Startups are scrambling to acquire “high-quality” training data — because unlike large language models, robots can’t just scrape the internet for examples of how humans perform physical actions.
“Robotics doesn’t have a ready-made dataset like text models do — you have to generate training data from scratch in the real world, which is far more complex,” said Ulrik Hansen, cofounder of data-labeling company Encord, in an interview with Business Insider.
Back in January, Bloomberg reported that AI companies were already buying up unused or unpublished creator footage, paying between $1 and $4 per minute for realistic, high-quality videos. These clips are crucial for developing robots capable of handling delicate, precise movements — skills that could one day be used for tasks as ambitious as building space habitats. (Jeff Bezos recently said at Italian Tech Week that he envisions robots constructing space colonies for “millions” of people.)
For now, though, those robots are just learning to sort laundry and tidy up the house.
According to Business Insider, the pay for this kind of data work can range widely — from about $50 an hour for general training videos, to as much as $150 an hour for recordings of more complex technical procedures like surgical equipment handling.
The Craigslist gig, however, is on the lower end, offering up to $20 an hour for two to four hours a day of recorded “specified tasks.” Along with cooking and laundry, the company is also seeking footage of participants cleaning, opening doors, assembling furniture, and even playing sports — all in the name of teaching the next generation of robots how to live like us.