HomeGreen TechnologyElon Musk Buried The Lede On Tesla AI Day

Elon Musk Buried The Lede On Tesla AI Day

By Roger Pressman, Founding father of EVANNEX

A lot has and might be written about Tesla’s AI Day—2022. Almost all of it is going to emphasize Tesla’s important lead in automotive autonomous driving functionality, FSD, it’s accelerating efforts in robotics with specific emphasis on its bipedal autonomous robotic, Optimus, and to a lesser extent, on its improvement of a supercomputer, DoJo, that in some ways could also be crucial facet of its sexier FSD and Optimus efforts. Hundreds of thousands of phrases might be used to research each technical bullet level and schematic. My intent on this publish isn’t to breed this reporting, however somewhat to supply a perspective from somebody who has inhabited the Tesla ecosystem from the very early days.

As some readers might know, I’ve been a software program and {hardware} engineer by my lengthy profession. In truth, my graduate analysis and work advocated and used early AI methods (the precursors to synthetic neural nets) for manufacturing (steel reducing, to be exact). This occurred manner, manner again within the Seventies—lengthy earlier than AI grew to become stylish and a long time earlier than the younger engineers on the AI Day stage had been even born!

The Tesla AI Day presentation is exclusive in lots of respects. By no means, so far as I do know, has any main tech firm (I agree with Elon Musk that Tesla is NOT a automobile firm), introduced three hours of technical element that describes its on-going strategic tasks. Positive, a tech big would possibly present a view from 10,000 ft with a definite advertising/PR really feel, however by no means have I seen a bunch of younger engineers, fronted by the corporate’s CEO, present a view from a floor stage that discusses technical structure and strategy. That in and of itself is spectacular. And by no means have I seen a CEO with Elon Musk’s grasp of technical element. Most CEOs are one tech query deep, two at most. Elon, it appears, can get into the technical weeds and would possibly even choose to remain proper there. That claims rather a lot about him and what he values inside the firm.

However what of the AI Day presentation itself—what are the takeaways that many summaries miss?

For my part, Tesla buried the lede in its AI Day presentation—discussing its supercomputer, DoJo, final when it ought to have come first. Why? As a result of in some ways, DoJo might be pivotal in its efforts to create Degree 4 autonomous driving within the subsequent few years, after which over the long term, stage 5 autonomous automobiles, together with its dream of autonomous fleets.

Petabytes of driving info are being gleaned from early customers of FSD, however all of that info (e.g., knowledge, video, exceptions) have to be correctly labeled and interpreted to boost machine studying. To this point, the duty of labeling and interpreted this info has been daunting—too time-consuming for speedy progress.

DoJo is a singular, highly effective, and extensible supercomputer structure that may change that. It’s, for my part, the enabling issue that may transfer FSD into prime time (and no, as a beta tester for FSD, I do NOT imagine the present FSD system is prepared for prime time, despite the fact that it’s an superior engineering achievement).

DoJo isn’t any much less vital for the event of Optimus. In truth, the identical guidelines apply. For an autonomous robotic to intelligently navigate the world, it should acknowledge a mindbogglingly massive assortment of conditions and objects, and be adaptable as it’s requested to function inside increasingly more use instances. To do this, the machine studying strategy for Optimus have to be iterative and quick, responding to extra petabytes of knowledge that have to be labeled and built-in within the studying structure. Dojo will allow that.

Lastly, there was a single remark about DoJo, made by Elon, which will have escaped your discover, if you happen to don’t comply with AI carefully. Elon steered that DoJo might finally allow “AGI.” That’s a mind-blowing assertion and a considerably ominous assertion. For the uninitiated, AGI stands for “synthetic normal intelligence”—the holy grail of the AI group. A fast Google search defines AGI this fashion:

“Synthetic normal intelligence (AGI) is the illustration of generalized human cognitive talents in software program in order that, confronted with an unfamiliar process, the AGI system may discover a resolution. The intention of an AGI system is to carry out any process {that a} human being is able to.”

That would come with the design and implementation of much more highly effective AGI programs, finally resulting in a “superintelligence.” It’s ironic that Elon is amongst many expertise leaders who’ve argued that unconstrained improvement of an AGI resulting in superintelligence may very well be harmful to people in some ways we are able to predict and in some methods which might be unexpected.

In his feedback about privateness and security, Elon alluded to this, suggesting {that a} “regulatory company” (just like the NHTSA or FDA) is required to manage the applied sciences that DoJo may unleash. But others, like Nick Bostrom in his ebook Superintelligence, argue that, “We’d need the answer to the [AGI] security drawback earlier than anyone figures out the answer to the AI drawback.”

Tesla builds the most secure vehicles the world has ever seen. Let’s hope that Elon and the younger engineers behind him have the knowledge to construct the most secure AI as properly.


Admire CleanTechnica’s originality and cleantech information protection? Take into account changing into a CleanTechnica Member, Supporter, Technician, or Ambassador — or a patron on Patreon.


Do not need to miss a cleantech story? Join each day information updates from CleanTechnica on e mail. Or comply with us on Google Information!


Have a tip for CleanTechnica, need to promote, or need to counsel a visitor for our CleanTech Discuss podcast? Contact us right here.





Please enter your comment!
Please enter your name here

Most Popular

Recent Comments