To many, Tesla is known for its production of electric vehicles, having developed cars on par with its competitors, including Ford and Volkswagen, and becoming the largest electric vehicle manufacturer in the world. However, following Tesla AI Day on September 30, 2022, the company portrayed itself as a catalyst for modern technology, particularly artificial intelligence. Throughout the presentation, Tesla showcased their progression in the development of Optimus, a humanoid robot designed to conduct a variety of household tasks, provided an update on Tesla’s self-driving software, and presented Dojo, a supercomputer tasked with training Tesla’s systems.
The highlight of Tesla’s event was the introduction of Optimus. For the first time, the robot walked onto the stage without support, steadily strolling forwards and waving to the audience. Though the robot was still in its early development, viewers were excited to see an improvement from last year, where an employee dressed in an Optimus costume walked on stage. Besides the live demonstration, several recordings were shown of Optimus conducting a variety of tasks, including picking up and placing objects and watering plants. Furthermore, Elon Musk shared his plans for the production of Optimus, claiming that Tesla plans to produce millions of robots which are “expected to cost much less than a car … probably less than $20,000.” Overall, although there is considerable progress to be made, the future of Optimus looks promising.
Additionally, Tesla presented its advancements in the company’s self-driving cars. With a significant increase in customers driving Teslas with self-driving software, the company has nearly 160,000 customers feeding data to the company. As a result, the company rigorously trains its AI with a massive supply of data, training more than 75,000 neural networks in just the past year. Slowly, but surely, Tesla’s self-driving cars are learning to act appropriately, even in unexpected scenarios.
However, with an incredible collection of data to train from, it would take far too long for Tesla to efficiently improve its self-driving software. Therefore, the company created Dojo, a set of three supercomputers comprising of 14,000 GPUs, 10,000 of which are used for training the AI and 4,000 for labeling. In return, the company has the ability to train networks which would have taken a month in simply a week. Moreover, Tesla plans to extend its use of Dojo beyond training self-driving cars, possibly providing services to other customers and companies as well.