Walter Isaacson is probably the most successful biographer in recent times. He has written books on Einstein, Leonardo da Vinci, Jennifer Doudna (of CRISPR fame), Steve Jobs, Henry Kissinger, Benjamin Franklin (among others) and now Elon Musk. He has also led Time magazine and CNN, but it is as a biographer that he made his mark.  I have not read the book (it is waiting behind the many others I am struggling to get to), but I have listened to a number of podcasts featuring its author, so I have an inkling of how he treats his new subject.

Frequent readers of this column will know that I don’t like Musk, notwithstanding his extraordinary achievements. This is because he is famous for screeching at his underlings, arbitrarily firing them, and various other failings of emotional intelligence. Having once worked for a boss like that, I can tell you that it does not make for good leadership. It turns out that, not only are the reports of disrespect and denigration entirely accurate, but Musk completely owns his behaviour. That’s his style and he neither denies it nor apologises for it.  

As Isaacson recounts, when Musk was questioned about these interactions with underlings he retorted, “I am reinventing the automotive industry and colonising Mars. Did you really expect me to be a chilled guy?” (I slightly paraphrase Isaacson’s recollection here, but this is quite close to the quote).  

In one podcast with Lex Fridman here, Isaacson names other often abusive and narcissistic geniuses that he has written about — including Steve Jobs and Bill Gates — and he goes on to argue that great advances in human progress are most often driven by individuals, not groups, companies or movements as is sometimes assumed to be the case. There is always an obsessive and impassioned human being at the front energetically and impatiently pulling everyone along.  

Perhaps we should not expect them to be soft and squishy and polite at all. It seems as though my idea of great leadership is flawed and irrelevant, which is why I will never reinvent industries the way Musk has done.  

Which leads me to another Musk-related story. To Tesla, to be more exact. And, by my reckoning, this story is far more important than Musk’s offensive personality and management style.  

Tesla has pursued what has been called by one smart analyst a “dual flywheel” strategy. To simplify the concept, there are two major cogs in Tesla’s architecture — a completely new approach to automotive design, and a massive library of Tesla-designed software, a great deal of which has been fuelled by its ambition to reach its goal of completely autonomous driving. How to interpret sensor data in real time (particularly visual data from cameras), how to react and how fast to react, and how to deal with edge cases and anomalies. Basically, how to get from A to B more reliably than a human driver.  

On the software side, Tesla’s original approach was to write rules. Lots of them. Such as, when something suddenly crosses in front of the car, apply brakes urgently. If you see a traffic light which is not working, proceed as though it is a four-way stop. Look behind you before braking to ensure that whoever is there will have time to stop. I am making this up but you get the drift. Rules. Millions of them, constantly being updated and appended and extended as new data about their cars’ behaviour is evaluated. 

This approach to AI, sometimes called algorithmic or heuristic, has now been replaced by the sort of deep learning techniques which brought us ChatGPT. In short, this approach doesn’t rely on humans to dream up the rules, it gets the machine to find rules autonomously, using a huge pile of training data and applying statistical inference and other magic to facilitate useful predictions.  

Tesla has adopted this approach. So, what’s new about it? How is this different from all the generative AI models already out there? The difference is that Tesla’s training data includes billions and billions of frames of visual data pouring in daily from cameras (and other sensors) mounted on its millions of cars.   

No one else has this data. It belongs to Tesla. Generative AI requires data as its fuel and Tesla has access to a grade of fuel that no one else can access. 

Moreover, Tesla has built a supercomputer called Dojo to train its models. And it has designed a chip that will be six times faster than the Nvidia GPU, the kit used by everyone else for building these sorts of AI models.  

While the Dojo supercomputer sits at the centre of Tesla’s autonomous driving project, it has dramatically wider applications than navigating a vehicle along roads and through traffic. It can deal with any “real-world” data at scale — data from cameras, radar, lidar, temperature sensors and humidity sensors. Its applications outside of Tesla and self-driving vehicles are endless. While everyone else has been concentrating on bots and image generation, Tesla has quietly emerged as the only operating player in this new field of “real-world” data-driven AI machine intelligence. 

Which means that Tesla is as much an AI company as a car company and, if it chooses to turn out a version of Dojo as a separate business exploiting this new sector, it will be without competition.  

This is why Adam Jonas, the head of Global Auto & Shared Mobility Research at Morgan Stanley, released a 66-page report on Tesla last week which forecast mouth-watering growth for Tesla — a 60% share price increase in 12-18 months and a $600-billion increase in enterprise value. In Jonas’ words, “If Dojo can help make cars ‘see’ and ‘react’, what other markets could open up? Think of any device at the edge with a camera that makes real-time decisions based on its visual field.” 

My personal distaste for Elon Musk notwithstanding, I have, albeit reluctantly, to applaud both his vision and his execution. Except for Twitter/X of course, but that’s a story for another column. DM

Steven Boykey Sidley is a professor of practice at JBS, University of Johannesburg. His new book It’s Mine: How the Crypto Industry is Redefining Ownership is published by Maverick451. It can be ordered directly from the Daily Maverick Store here or on Kindle. It’s also available at bookstores.