In May, Tesla (NASDAQ: TSLA) opened a new official Twitter account, @Tesla_AI, focused on reporting the company’s artificial intelligence (AI) progress around its FSD software and forthcoming humanoid robot, Teslabot.
On June 21, the AI-dedicated account shared another major development. Per the graphic published on Wednesday, the world’s largest electric vehicle (EV) maker’s much-anticipated Dojo supercomputer, is set to enter production in July 2023.
As observed from the chart, Tesla intends to continue building its compute until it becomes one of the top 5 most powerful on the globe by early 2024. The company plans to achieve this by taking advantage of its vast database of real-world driving videos, which it has been amassing for several years. By October 2024, Tesla expects its compute to reach 100 exaflops.
What is Dojo?
Dojo is Tesla’s proprietary custom-built supercomputer platform developed for AI machine learning. More specifically, the company will use Dojo for training neural nets using video data from its own vehicles. The supercomputer is expected to play a major role in advancing Tesla’s computer vision technology that powers its self-driving solution.
While Elon Musk’s EV manufacturer already has a powerful NVIDIA GPU-based supercomputer, the new Dojo offering uses chips and an entire infrastructure designed by Tesla itself.
In the extensive Twitter thread, Tesla’s AI account described how the automaker is building the foundational models for autonomous robots. Finally, it also discussed the company’s multi-model neural networks, which are already integrated into its fleet of electric cars.