A screenshot taken from a Tesla promotional video showing Optimus being trained via motion capture.

The company is paying people to train its humanoid Optimus bot via motion capture, but the task may require vast amounts of data.

Tesla is employing people to help train its humanoid Optimus robot by wearing motion capture suits and mimicking actions that it will be expected to perform. The position, advertised as a “Data Collection Operator” on Tesla’s career site, pays up to $48 per hour and requires walking for over seven hours a day while carrying up to 30 pounds and wearing a VR headset for “extended periods of time.”

Employees must also be between 5’7’’ and 5’11’’ tall — likely to be in a similar range to Optimus’ projected 5’8’’ height. The official Tesla Optimus X account has previously teased what the role entails, and Business Insider reports that Tesla has hired over 50 workers for the position over the past year.

Motion capture is a common and typically cost-effective way to train robots to perform human-like movements, but Tesla is one of the first to do so at this scale. While some online videos have shown Optimus performing various tasks, Animesh Garg, a senior researcher at Nvidia Research, told Business Insider that Optimus may require millions of hours of data before it’s fully ready to work in Tesla’s factories.

“The amount of data collection you’d need would easily be half a billion dollars and the real question is ‘Even if you do that, do you succeed?’ Because there is no guarantee of success,” Garg said.

That’s pretty typical for humanoid robots and the artificial intelligence systems running them, which all require an arduous amount of training. While Tesla CEO Elon Musk has promised to have “genuinely useful” bots in production next year (a timeline he himself has admitted is mostly guesswork), Optimus currently doesn’t seem any closer to completion than the rival offerings from Boston Dynamics, Figure, and Apptronik being tested by other vehicle manufacturers.