Tesla needs to produce between 10,000 and 20,000 Optimus humanoid robots to train them to perform tasks through repeated practice, the company’s chief executive, Elon Musk, said during its third-quarter earnings call on 21 October.
“In order for Optimus to get the training data we need, we actually need to make a lot of Optimi,” Musk said.
The company has recently begun production on a new Optimus manufacturing line in Fremont, California, as it steps up efforts to develop its humanoid robot programme.
Musk said the robots would practise a wide range of tasks at a facility he described as an “Optimus Academy”.
Data generated through the training process, known as self-play, would then be used to improve the artificial intelligence systems controlling the robots.
He likened the approach to raising a child, arguing that robots must learn through practical experience and repeated interaction with their environment rather than depend entirely on pre-programmed instructions.
“They initially are sort of wobbly and they don’t know how to do things, and eventually they learn to wipe their own ass, which is a big milestone,” he said.
“They learn to read and do all sorts of things… And they learn this by playing. And we actually need Optimus to play at scale in order to have it be really good at tasks in the real world,” Musk said.
Tesla already collects data from its vehicles to improve its self-driving software, but its Optimus humanoid robots do not yet have a comparable fleet capable of generating real-world training data.
Ashok Elluswamy, head of Tesla’s AI software team, said the robots would initially be trained by observing humans performing various tasks.
“We have to bootstrap it by making the robots and having them do self-play, which is really how a human learns to navigate the world,” he said.
Tesla has installed most of a production line designed to eventually manufacture one million humanoid robots annually, according to Musk. However, the company will need at least another year to reach a production rate of 20,000 robots per week, as some aspects of the design are still being refined.
“There’s like 10,000 things that have to go right for production of the robot, and it’ll move as fast as the slowest, least likely of the 10,000,” he said.
