People Are Wearing Cameras While Cleaning to Teach AI

Hundreds of people in Los Angeles are wearing cameras at home to record daily tasks, generating data to help train AI systems.

According to a report by The LA Times, residents are making extra income by donning head and wrist-mounted cameras as they perform chores such as making coffee, washing dishes, watering plants, or cleaning kitchens. The camera recordings provide detailed footage for AI and robotics companies seeking to develop “physical AI” systems that can replicate human actions.

While some critics argue that these AI training jobs are low-paid and potentially exploitative, workers see them as an opportunity to earn money in a shifting economy. One participant, Salvador Arciga, who has worked various gig jobs in Los Angeles, earns $80 for two hours of footage.

“I need to do chores anyway,” Arciga tells The LA Times. “Now I get a chance to get paid to do it.”

Despite challenges, such as phone calls interrupting recordings or footage being rejected, participants can earn a significant income. One couple tells the news outlet that they have earned $1,200 from recording chores.

AI chatbots like ChatGPT have relied on online data to learn tasks such as conversation, image generation, and coding. But according to The LA Times, physical AI requires real-world demonstrations of human motion, which are not widely available online. Some countries operate dedicated facilities where workers record human movements for robotic training. In China, more than 40 state-owned centers use humans wearing virtual reality headsets to operate robots. Companies including Tesla, Google, and California startups such as Figure AI and Dyna Robotics are also investing in robotic models. New startups have begun producing custom cameras and bodysuits to capture detailed human movements, including hand position, pressure, and posture.

Sunain, a human data capture startup, has over 1,400 contributors locally in Los Angeles, who record natural behavior rather than scripted movements, including interruptions and context switching, to help robots understand real-world situations. Shahbaz Magsi, co-founder of Sunain, says the Los Angeles participants provide a crucial dataset that helps AI systems learn movements in real-world settings, bridging the gap between digital and physical intelligence.

Image credits: Header photo licensed via Depositphotos.