Elon Musk has decided to impress the world by diving headfirst into a field where Tesla hasn’t been particularly active before: supercomputer development. Work on the new supercomputer, codenamed “Dojo,” is gaining momentum, and the project could radically transform the approach to machine learning in the automotive industry.
Zooming In
So, what exactly catches our attention here? “Dojo” isn’t just a regular computer for homework. According to Tesla, its processing power will exceed 1 exaflop. For those unfamiliar with mathematical jargon: this is the major league of computing, capable of performing one quintillion (yes, that’s a lot of zeros) operations per second.
Instead of spinning disks, this machine will spin deep learning algorithms that Tesla uses to improve its self-driving system — and possibly give us a little more hope for safer roads. As Tesla admits, “Dojo” is seen as a crucial tool for achieving the elusive goal of full vehicle autonomy.
Its sheer size is also stirring curiosity: up to 500 millimeters in height, this beast will still be compact enough for Tesla’s engineering-optimized storage spaces. Each card of this supercomputer, housing up to 354 chips designed specifically for neural networks, acts as the device’s “brain cells.”
For now, according to Musk, completion is expected by mid-2024. Tesla doesn’t see “Dojo” as a commercial product but plans to use it solely to enhance its own autopilot technologies.
What could this mean for us, the inhabitants of planet Tesla? Potentially — safer and more precise self-driving systems, as well as lower computing costs in the production of such components.
In any case, Tesla is making a bold move toward a new technological frontier — but only time will tell whether it’s truly revolutionary and effective.
