GoKawiilDimensional pays $20/hour to walk robot dogs for AI training data in San Francisco
AI startup Dimensional is hiring people in the Bay Area to walk its robot dogs through city streets for four- to six-hour shifts at $20 an hour, according to The New York Times. The dogs, roughly 33-pound Unitree units of which Dimensional owns more than 100, are fitted with cameras and sensors that record the walks, and walkers must be prepared to carry the robot back if it overheats.
GoKawiil's interpretation of the reporting above, not reported fact.
Dimensional is building software meant to let AI agents operate robots for physical labor like hauling or construction, and founder Stash Pomichter says the technology is 'still very bad at controlling hardware,' which suggests the walking gigs are meant to generate real-world movement data to improve that control. The setup highlights an odd transitional phase where human labor is needed to teach machines intended to eventually replace manual work.
- Dimensional pays $20/hour for humans to walk its robot dogs in San Francisco.
- The walks generate camera and sensor data meant to train AI control of hardware.
- Dimensional owns over 100 robot dogs, using $2,000 Unitree models from China.
Unitree Go2 Robot Dog — This is the very same quadruped robot platform referenced in the article as the hardware Dimensional uses for its robot dog walkers. It's a fascinating entry point into consumer robotics, letting you explore AI-assisted navigation and mobility tech firsthand.
See Unitree Go2 Robot Dog on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.Source: entrepreneur.com — Jonathan Small, 2026-10-30
Published there as: “This San Francisco Startup Is Hiring Humans to Walk Robot Dogs for $20 an Hour. No Experience Required.”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.