September 27, 2023
Don't count on Tesla's Dojo supercomputer to start the AI ​​revolution

You’d have to be very brave to bet against the idea that applying more computing power and data to machine learning – a recipe that gave birth to ChatGPT – won’t lead to some kind of further progress in artificial intelligence. Nevertheless, you would still be bold to bet that the combo will deliver specific progress or success on a specific timeline, no matter how desirable.

A report released last weekend by investment bank Morgan Stanley predicts that a supercomputer called Dojo, which Tesla is building to boost its work on autonomous driving, could deliver huge profits in car making, robotaxis and sales. This could add $500 billion to the company’s value. Software for other businesses.

The report added more than 6 percent, or $70 billion, to the EV-maker’s market cap as of September 13 – roughly the value of BMW and significantly less than the price paid by Elon Musk for Twitter – to Tesla’s share price as of September 13.

The 66-page Morgan Stanley report makes interesting reading. This makes a meaty case for why Dojo, the custom processor that Tesla developed to run machine learning algorithms, and the massive amounts of driving data the company is collecting from Tesla vehicles on the road, should pay huge dividends in the future. Can do. Morgan Stanley analysts say Dojo will deliver breakthroughs that will give Tesla an “asymmetric” advantage over other carmakers in autonomous driving and product development. The report also claims that the supercomputer will help Tesla enter other industries where computer vision is important, including health care, security and aviation.

There are good reasons to be cautious about those grand claims. You can see why, at this particular moment of AI mania, Tesla’s strategy might seem so attractive. Thanks to a remarkable leap in the capabilities of the underlying algorithms, ChatGPT’s mind-bending abilities can be traced back to a simple equation: more computation x more data = smarter.

The wizards of OpenAI were early followers of this mantra, betting their reputation and millions of their investors on the idea that supersizing the engineering infrastructure for artificial neural networks would lead to major breakthroughs, including language models like ChatGPT. . In the years before the founding of OpenAI, the same pattern was seen in image recognition, with larger datasets and more powerful computers leading to a remarkable leap in the ability of computers to recognize – even if at a superficial level – what an image shows.

Walter Isaacson’s new biography of Musk, liberally cited last week, describes how the latest version of Tesla’s optimistically-branded Full Self Driving (FSD) software, which guides its vehicles on busy roads, is working hard. -Relies less on coded rules. And much more on neural networks trained to mimic good human driving. This sounds similar to how ChatGPT learns to write by consuming countless examples of text written by humans. Musk has said in interviews that he hopes Tesla will have a “ChatGPT moment” with FSD in the next year or so.

Musk has made big promises about breakthroughs in autonomous driving several times before, including a prediction that there would be one million Tesla robotaxis by the end of 2020. So let’s think about this carefully.

By developing its own machine learning chips and building out Dojo, Tesla can certainly save money on training the AI ​​system behind FSD. This could further help it improve its driving algorithms by using real-world driving data collected from its cars, which its competitors lack. But whether those improvements will mark an inflection point in autonomous driving or computer vision seems almost impossible to predict.


Leave a Reply

Your email address will not be published. Required fields are marked *