Tesla has made significant advances in self-driving so far; To the extent that it wants to remove radar sensors from its electric cars. The head of Tesla’s artificial intelligence, Andre Caparti, stated at the CVPR (Computer Insights and Patterns) Workshop 2021 that this is to be achieved with the Dojo supercomputer.
“When using this technology, the speed at which a computer detects and responds to new environments should be like a human,” Caparti explained. This requires training artificial intelligence in a huge data set and a powerful supercomputer. Tesla’s new supercomputer has this feature.
The Dojo is the company’s next-generation supercomputer with 1.8 exaflop processing power and 10 petabytes of NVME memory running at 1.6 terabytes per second. According to Carpathians, the Dojo is one of the fastest supercomputers in the world, ranking fifth in the world if we count the total number of flaps; The site currently belongs to Nvidia’s Selene supercomputer cluster.
Tesla’s new supercomputer collects recorded video from eight in-car cameras to teach the system. The speed of each of these cameras is 36 frames per second.
Tesla’s new Dojo supercomputer
This system has been able to show acceptable performance in secluded environments. However, Tesla knows that using such a system on busy streets is fraught with challenges. Apparently, the Dojo also has the ability to detect pedestrian collisions and misuse the pedal (pressing the pedal instead of the brake).
Tesla’s supercomputer is still in development. However, Ilan Mask, the company’s CEO, has also expressed hope for the project, calling it a major achievement in improving the autopilot system.