Elon Musk has recently announced that Tesla plans to rekindle its efforts on the development of Dojo, its third-generation supercomputing project. This marks a significant shift, as the Dojo initiative had previously taken a backseat while the company concentrated on advancing its onboard AI hardware for Tesla vehicles. Musk explained that the reason for this renewed focus is that the design of the AI5 chip—integral to the Dojo system—is now quite refined, providing a solid foundation to resume work.
The primary goal of the Dojo platform is to handle enormous volumes of video data and other information collected from Tesla's fleet of vehicles. This processed data feeds into the neural networks powering Tesla's Full Self-Driving (FSD) technology, enabling the cars to better understand and navigate their environment. However, last year, Musk emphasized that Tesla's resources would be better allocated. He stated that developing two separate AI chip lines—one for inference in vehicles and another for training—would be inefficient. Instead, Tesla will concentrate on the AI chips designed specifically for inference, which includes the AI5, AI6, and upcoming models, as these are capable of both inference and reasonably good training.
The AI chips under discussion currently reside inside Tesla vehicles and are optimized primarily for inference tasks, not for training neural networks. The next generation, AI6 chips, are planned to be installed at Tesla’s manufacturing facility in Texas, following Tesla's substantial $16 billion investment agreement. This move underscores the company's push towards vertical integration and in-house chip production.
But here's where it gets controversial: Musk has made numerous ambitious claims over the years—some of which turned out to be exaggerated or simply didn't materialize as expected. One particularly speculative idea he has floated is the concept of a 'space-based AI compute' system for Dojo3. Musk suggests that harnessing the vast energy potential of space—particularly sunlight—could provide an efficient power source, and the extremely cold temperatures in space might reduce cooling requirements for large-scale data centers. While this idea sounds groundbreaking and futuristic, many experts consider it highly speculative and perhaps impractical at this stage.
And this is the part most people miss: Musk’s ambitious visions often blur the line between visionary innovation and experimental speculation. Do you believe space-based AI computing will ever become a reality, or is it just another lofty idea that might remain in the realm of science fiction? Share your thoughts and join the debate—what do you agree or disagree with regarding Musk's bold predictions for Tesla’s future?
