The electric vehicle company at one of the forefronts of autonomous driving, and is using the real world as its testing bed.

Latest Stories

Elon Musk says Tesla will hire AI talent with whom he'll "meet/email/text almost every day"

February 3rd

The Basics

Tesla, with its publicity stunts and unpredictable CEO, is building something unique in the AI world.

Like other AI applications, autonomous driving needs a wealth of training data. To teach a car to drive itself, developers need millions of examples of driving behavior, as well as data showing what vehicles, pedestrians, and other obstacles look like. While Tesla's competitors are creating this data inside simulations and controlled test environments, Tesla's Autopilot system is pulling it from Tesla vehicles around the world.

Musk says using real world data is their key to developing Autopilot, and that driving with it is twice as safe as driving with it off.

Most software companies release products and improve them using real world data, but the risks posed by autonomous vehicles have caused many to question his approach.

"Musk’s decision to put Autopilot in the hands of as many people as possible amounts to an enormous experiment playing out on freeways all over the world," Zachary Mider wrote in Bloomberg.

The approach couldn't be more different from the way cars have been traditionally built and sold. While most manufacturers today use over-the-air updates to upgrade key parts of vehicle software, Musk is promising to fundamentally change features on already-sold vehicles over time.

One of those ways is to let owners hire out their cars as autonomous taxis. While Musk is known for giving overly ambitious timelines, he has reason to want to develop his own taxi fleet.

Tesla's competitors, like Waymo, WeRide, Uber and Yandex, have logged millions of miles of their own. Most are conservative, operating in designated neighborhoods or test roads. But not all of them are -- WeRide operates autonomous taxis across Guangzhou, and Yandex showcased its own self-driving taxi in Las Vegas in 2019.

Musk bets that his approach -- with years of real world data under his belt, testing Autopilot in real time -- may give him an advantage in the impending autonomous taxi race. Launching the Tesla fleet could be as simple as a software update, as well.

"The fleet wakes up with an over-the-air update; that’s all it takes," he said in 2019.

Go deeper

Is AI research headed toward a dead end? Let's look at past AI winters for clues

February 24th