Dojo Supercomputer Unveiled: Elon Musk’s announcement on a conference call reveals the Dojo supercomputer’s specialized design to handle extensive data, including Tesla car videos.
During an investor call, Elon Musk expressed his confidence as he stated that the Dojo supercomputer has a distinct design specifically crafted to handle extensive data, including video content from Tesla vehicles. The importance of this design stems from its critical role in propelling the development of autonomous-driving software. Bloomberg cited Musk’s statement, highlighting its significance.
Tesla’s owner emphasized that the company possesses a “staggering amount” of video footage, primarily because customers eagerly use Autopilot, a camera-based driver-assistance system, and the related Full Self-Driving Beta feature. This collection has gathered over 300 million miles of data.
In their latest earnings release, Tesla announced the initiation of production for their Dojo training computer. The company outlined four essential technology pillars for achieving vehicle autonomy at scale: an extremely large real-world dataset, neural net training, vehicle hardware, and vehicle software. Tesla is committed to developing each of these pillars in-house, with the Dojo training computer representing a significant advancement in faster and more cost-effective neural net training.
Although Tesla already possesses one of the world’s most potent NVIDIA GPU-based supercomputers, the new Dojo custom-built computer incorporates Tesla-designed chips. Elon Musk clarified that the reason for this choice is NVIDIA’s inability to supply more of its advanced A100 tensor core GPU clusters, likely due to high demand from numerous customers. Musk stated that if NVIDIA could provide enough GPUs, the need for Dojo might not have arisen.
