Tesla CEO, Elon Musk announced in 2020 that the company is working on a supercomputer named “Dojo” for video data processing on Twitter:
Tesla is developing a NN training computer called Dojo to process truly vast amounts of video data. It’s a beast! Please consider joining our AI or computer/chip teams if this sounds interesting.
— Elon Musk (@elonmusk) August 14, 2020
Andrej Karpathy, Tesla’s head of AI, in a recent CVPR2021 (Conference on Computer Vision and Pattern Recognition) announced the company’s new supercomputer, which allows the manufacturer to replace radar and lidar sensors with high-quality optical cameras in self-driving cars.
With vision-only technology, computers must respond to novel settings with the same speed and acuity as humans, according to Karparthy. However, this necessitates AI training on a large dataset and the use of a powerful supercomputer to crunch the data. “Dojo,” a next-generation model with 1.8 exaflops of performance and 10 petabytes of NVME storage running at 1.6 terabytes per second, is one of them.
Explaining the insane supercomputer, karparthy said: The cluster has 720 nodes, each with eight Nvidia A100 GPUs (the 80GB type), for a total of 5,760 A100 GPUs in the system. This accelerator power is backed up by ten petabytes of “hot tier” NVMe storage with a 1.6 terabytes per second transfer rate. This “very fast storage” is “among the world’s quickest filesystems.” So, He told Techcrunch:
This is a massive supercomputer and I actually believe that in terms of flops this is roughly the number five supercomputer in the world
Further, NVIDIA presently holds the fifth position with their Selene cluster, which has a very similar architecture and a similar amount of GPUs (4480 vs. 5760, so a little less).”
Karpathy shared a few examples of how Tesla’s supercomputer uses computer vision to correct bad driver behavior, including an emergency braking scenario in which the computer’s object detection kicks in to save a pedestrian from being hit, and a traffic control warning scenario in which the computer can identify a yellow light in the distance and send an alert to a driver who hasn’t yet begun to slow down.
The company believes that a supercomputer will ultimately help automobiles achieve advanced self-driving capabilities, but the best way to wait is to see what is next.