Tesla Outlines Plan to Train Robots Without Human Drivers
The company is building a dedicated facility to have its humanoid units practice skills to generate necessary training data.
By Jonathan Hoffmann ·
Tesla is ramping up production of its humanoid robot while simultaneously devising a plan to teach it how to think.
The automaker has commenced manufacturing at a high-volume facility within its Fremont location, which previously housed the Model S and Model X assembly lines. This new operation is intended to produce up to one million units annually. According to the company’s third-quarter report, the initial units rolling off this line are the first to reach production-quality standards, confirming the manufacturing process is viable for commercial sale. Leadership indicated during the October earnings call that the necessary equipment to scale production is largely in place, with an eventual goal of reaching roughly 20,000 robots per week. However, the company acknowledged that this expansion will occur gradually, describing the growth trajectory as having a shallow slope initially before accelerating.
The primary challenge facing Tesla is generating data to train the robot's artificial intelligence. Unlike the company's fleet of millions of vehicles, which collect driving data, the humanoid robot does not have a pre-existing network to learn from. To address this, the company is establishing an Optimus Academy. This initiative will involve a group of 10,000 to 20,000 robots practicing a variety of tasks to create the data required for future software updates. Every unit produced at the Fremont facility will contribute to this training pool.
Future units are scheduled to be equipped with a new AI5 chip next year. Leadership stated that a video demonstrating the robot performing tasks autonomously, rather than following pre-programmed instructions, should be released within the next month or two. This move toward autonomous operation is intended to differentiate the product from competitor demonstrations that rely on scripted routines.
Production costs are being managed by working with specific suppliers, and the company recently certified several partners for component manufacturing. A larger facility for this project is currently under construction in Texas. While production costs are decreasing, the design team continues to focus on refining the robot's hands, which remain the most difficult aspect of the engineering process.
