Wednesday, February 7, 2024

AI: Tesla Installing Second Dojo Supercomputer In New York Gigafactory (TSLA; NVDA)

Mr. Musk says if the company can get financial returns from the computers they will keep building them.

From TechCrunch, January 26:

Tesla to spend $500M to bring its Dojo supercomputer project to Buffalo factory

Tesla will spend $500 million to build one of its so-called “Dojo” supercomputers at its Buffalo, New York factory, the state’s governor Kathy Hochul said Friday during a news conference just days after CEO Elon Musk called the project a “long shot.”

Tesla’s decision was “informed by New York’s reliable power supply, strong talent pipeline and availability of usable space for the project,” according to Hochul’s office.

Dojo, which was first announced at Tesla’s “AI Day” event in 2021, is a supercomputer meant to help advance the company’s still-unrealized goal of building a self-driving car. Tesla plans to use the supercomputer to process reams of video data that come off of its electric vehicles in order to train the AI that now powers its most advanced driver assistance software, which it calls Full Self-Driving Beta. Musk said last year that Tesla plans to spend “well over $1 billion” on Dojo.

Bringing the Dojo project to Buffalo is the latest shift in Tesla’s priorities for the location, which has turned into something of a boondoggle for New York state. Once dubbed “Gigafactory 2,” Tesla took over the factory from SolarCity when it acquired the troubled solar panel company in 2016. The state had already committed $750 million to the plant by that point. Tesla promised to make Solar Roof tiles there, but struggled to produce the product at scale. Its partner, Panasonic, pulled out of the plant in 2020, and Tesla pivoted to employing people who labeled training data for its less-advanced Autopilot software.

Musk said last April that he believed the Dojo supercomputer project was a “long shot bet” that could “pay off in a very, very big way… in the multi-hundred-billion-dollar level.”

He reiterated the point this week on a call with analysts. “It’s not, like, a sure thing at all, It’s a high-risk, high-payoff program,” he said. “We are scaling it up, and we have plans for Dojo 1.5, Dojo 2, Dojo 3, and whatnot. So, you know, I think it’s got potential, but the kind of size enough high risk, high payoff.”....

....MUCH MORE

And that, for all the young business brainiacs out there, that is why we keep going on and on about competitive flywheels and hyper-pareto distribution of profits. It's not 80% of the loot going to the top 20% of companies; it is 95% going to the 5% that can afford to sit at the table. 

If interested see September 2023's:

HBR—From Pareto To Hyper-Pareto: "AI Is Going to Change the 80/20 Rule" 

And the links we embedded in that post: 

This type of information advantage is more and more accruing to the biggest and richest of corporations. It is a type of rich-get-richer advantage akin to the flywheel effect:

And related, now that we see what is happening, what, if anything, should society do about it?

...Much more important than the direct monetization of big data is the strategic advantage it can bestow over time.
In a winner-take-all economy, as in a horse race, small differences in superiority are rewarded all out of proportion to the actual advantage. A top thoroughbred may only be a couple fifths of a second faster than the field but those two lengths over the course of a season can mean triple the earnings for #1 vs. #2.
In commerce the results can be even more dramatic because rather than the 60%/20%/10% purse structure of the racetrack the winning vendor will often get 100% of a customer's business.....

Posted by at