Elon Musk announced on X that Tesla is set to recommence its development of Dojo3, marking the third iteration of the company’s proprietary supercomputer initiative. The Dojo team, previously disbanded last year, had its resources redirected towards the AI chips utilized in Tesla vehicles. According to Musk, the revival of the project comes at a time when the design of the AI5 chip is progressing well.
The primary aim of the Dojo project is to analyze video recordings and other data collected from Tesla cars to enhance the neural network powering the Full Self-Driving (FSD) software. Musk previously argued against splitting resources between two different AI chip designs, emphasizing that Tesla’s focus would remain on the AI5, AI6, and subsequent chip developments, which are poised to deliver strong performance for both inference and training tasks.
Notably, the AI chips mentioned are currently designed for real-time processing within Tesla vehicles and are less suited for extensive training operations. The AI6 chips are anticipated to be produced at Samsung’s Texas facility, following a substantial $16 billion partnership established with Tesla.
Adding to the intrigue, Musk has suggested that Dojo3 could introduce “space-based AI compute,” a concept proposing that data centers in orbit may outperform terrestrial facilities. This theory posits that space offers better solar energy access and cooler environments, which could significantly reduce energy demands. While this notion is gaining traction among some futurists, skepticism exists within the expert community regarding its feasibility.
Key Highlights:
– Tesla restarts the development of its Dojo3 supercomputer.
– Focus shifts back to Dojo after progress on the AI5 chip.
– Dojo aims to enhance Full Self-Driving software by analyzing vehicle data.
– AI6 chips to be manufactured in Texas, part of a $16 billion deal with Samsung.
– Musk introduces the concept of space-based computing for future AI developments.
