Elon Musk has announced that Tesla will restart work on its Dojo supercomputer project, a move that signals a renewed focus on in-house AI training capabilities. In a post on X, Musk stated that the company is reviving the Dojo initiative with Dojo3, the third generation of its supercomputer. This follows a period last year when the Dojo team was reportedly disbanded as Tesla prioritized the development of its AI chips for onboard vehicle use. The impetus for the reboot, according to Musk, is that "the AI5 chip design is in good shape."
Dojo's Purpose and Previous Pause
The primary objective of the Dojo project is to leverage the vast amounts of video data collected from Tesla vehicles to train the neural networks underpinning its Full Self-Driving (FSD) software. This data-intensive process requires significant computational power, which Dojo was designed to provide. However, last year Musk indicated a shift in strategy, suggesting it was inefficient to divide resources between two distinct AI chip designs. The focus then shifted to the Tesla AI5, AI6, and subsequent chips, which would be optimized for inference—the process of making predictions using a trained model—while also being "at least pretty good for training."
Space-Based Ambitions and Skepticism
Adding another layer of intrigue, Musk also stated that Dojo3 will be "space-based AI compute." This reflects his belief, shared by some, that data centers in orbit offer advantages over traditional land-based facilities. According to The Verge, the theoretical benefits of space-based data centers include easier access to solar energy and the potential for reduced power consumption due to the cold temperatures. However, as Engadget reports, this concept remains largely speculative, and many experts harbor doubts about its feasibility.
The AI6 chips, crucial for Tesla's ongoing AI efforts, are slated to be manufactured by Samsung at its Texas factory, following a substantial $16 billion agreement with Tesla. This partnership underscores the importance of advanced chip technology for Tesla's ambitions in autonomous driving and AI. As always, it's crucial to approach Musk's pronouncements with a degree of caution, given his track record of ambitious claims that don't always materialize. But the renewed commitment to Dojo suggests a significant shift in Tesla's AI strategy, one that could have far-reaching implications for the future of autonomous driving and AI compute. The revival of Dojo, especially with the space-based component, certainly sets a bold, if uncertain, course for Tesla's technological trajectory.