Once again, Elon Musk has dropped hints on the capabilities of the inbound Dojo supercomputer from Tesla that it has been building for a while now.
It is said that this supercomputer helps optimize Tesla’s autonomous algorithms for autopilot, enabling Teslas to completely self-drive.
Now, Elon Musk has tweeted that access will be offered as a web service to the Dojo supercomputer so that individuals can train their machine learning algorithms that will make the supercomputer even smarter.
In reaction to a comment, he tweeted. “Yeah, we will open Dojo for training as a web service once we work out the bugs,”.
Yeah, we will open Dojo for training as a web service once we work out the bugs
— Elon Musk (@elonmusk) September 20, 2020
Musk has already demonstrated that it has the ability to conduct floating-point operations of one quintillion per second, which is approximately 1,000 petaFLOPS.
In order to crack the exaFLOP barrier, the Dojo will be on point. Fugaku is actually the world’s fastest supercomputer, and Fujitsu is in Japan.
It is based on the ARM architecture and has a 415,530 TFLOPs output from Linpack. The Dojo is probably not going to be as hard, but it will use a similar underpinning architecture driven by the designs of ARM. Nvidia recently acquired ARM, which used to supply autonomous technology to Tesla chips.
As Tesla’s FSD chips are also built on the same chipset architecture, Dojo is possibly also built on ARM’s technology.
“Dojo uses our own chips and a computer architecture optimized for neural net training, not a GPU cluster. I could be wrong, but I think it will be best in the world,” Musk said earlier.
This is also a snide assault on Nvidia’s technology, fuelled by its patented GPU technology, which has proved to be very powerful in the last decade for machine learning and AI, usurping Intel and AMD’s dominance of x86-based CPUs.
The vision of Musk for a vertically integrated Tesla moved him towards designing his own architecture of chipsets.
In the past two decades, ARM has been the most neutral chipset network, with more than 22 billion devices built on its technology, including iPhone shipping.
Last week’s acquisition of ARM by Nvidia has opened it up to scrutiny where many ARM customers believe Nvidia might change the business model and start competing or hampering its customers.