A San Francisco startup recently sought human handlers to walk robotic quadrupeds through city streets at a rate of $20 per hour, forming part of an initiative to train artificial intelligence systems to better navigate complex outdoor environments.
Dimensional, operating out of the city’s Dogpatch neighborhood, published a job advertisement seeking operators to take four-legged machines on four- to six-hour walks along designated outdoor testing routes in San Francisco. The October 5 listing indicated that no prior robotics experience was necessary and that paid training would be provided, offering immediate compensation following each completed shift.
Human operators were tasked with steering the robotic dogs while on-board cameras and sensors gathered environmental data. Because the machines occasionally experienced technical malfunctions or overheating, workers were required to be comfortable carrying equipment weighing approximately 33 pounds back from routes.
Dimensional founder Stash Pomichter explained that the field tests yielded clear insights into current technological limitations. Human intervention remains vital because autonomous agents continue to struggle with hardware management.
Dimensional is currently developing an open-source operating system intended to enable AI agents to control robotic hardware directly. To support development, the firm has acquired more than 100 robot dogs and 50 humanoid robots from manufacturers located in the United States, China, and Germany. The quadrupeds utilized for the San Francisco street trials were manufactured by the Chinese robotics firm Unitree and cost approximately $2,000 each, operating via handheld remote controls or mobile applications.
The company aims to eventually deploy its technology across industrial and commercial settings, including factories, data centers, construction sites, and security operations. Pomichter noted that the long-term vision is to automate heavy physical labor currently performed by humans, such as transporting bulky packages.
