The Rise of ‘Data Freelancers’ Training AI with Everyday Tasks

This summer, Taylor and her roommate took on a unique job.For a week, they wore GoPro cameras on their foreheads while painting, sculpting, and even doing chores. They weren’t making a documentary; they were training an AI vision model. It involved carefully syncing their footage, giving the system multiple angles of the same actions. It wasn’t easy, but the pay was good – and it let taylor spend her days creating art.

“We’d wake up, do our usual routine, then strap on the cameras and sync the times,” Taylor explained. “Then we’d make breakfast and clean up. After that, we’d each work on our art.”

They aimed to deliver five hours of synced footage daily, but Taylor quickly realized she needed seven. this allowed for breaks and recovery time. “It gave you headaches,” she said. “You’d take it off and have a red square on your forehead.”

Taylor, who preferred not to share her last name, was working as a data freelancer for Turing Labs, an AI company. They connected her with this possibility through TechCrunch. Turing isn’t trying to teach AI to paint; they’re focused on developing skills like sequential problem-solving and visual reasoning. unlike large language models, Turing’s vision model learns solely from video – and moast of that video is collected directly by Turing.

Turing is hiring artists like Taylor, but also chefs, construction workers, and electricians – anyone who works with their hands. Sudarshan Sivaraman, turing’s Chief AGI Officer, told TechCrunch that manual data collection is the only way to build a diverse enough dataset.

“We’re collecting data from so many different blue-collar jobs,” Sivaraman said to TechCrunch. “This ensures we have a wide variety of data during the pre-training phase.”