Artificial city

This is the artificial intelligence system with which Tesla wants to impose itself on the autonomous car

Currently, Tesla is World leader in electric carsHowever, the American company is not satisfied with this and wants to become the leader in one of the main battles of the automotive industry: autonomous vehicles.

Tesla made public for the first time artificial intelligence technology that it was developed for use in its self-driving cars. With this, he hopes to revolutionize the transport industry and be the first to offer these vehicle models.

Tesla’s AI system is called dojo and expanded its operations at the Hot Chips conference. Dojo has assembled hundreds of its giant D1 chips exodusThe system is responsible for analyzing the video of the fleet of Tesla cars that are currently circulating on the roads. How driving works in the real world,

in detail InternetThis artificial intelligence is the basis of your system fully autonomous drivingWith which Tesla hopes to drive its cars through freeway intersections, parking lots and traffic lights.

Even then, Tesla struggled to offer FSDs to customers, many of whom paid for it years ago. Developing this AI is much more complicated than Tesla expected, which slowed it down. Dojo’s vast computing power is finally here to make self-driving Teslas a reality.

FSD is currently in limited beta testing and requires continued human monitoring Disqualifies Tesla as a true self-driving carBut that doesn’t stop Elon Musk from insisting that this technology will pave the way for self-driving vehicles that don’t need a human.

OK, FSD is giving Musk a headache. Just a few days ago, it announced that the price of the service would increase from $12,000 to $15,000. “Autonomous driving is pervasively a difficult problem that requires a large amount of real-world AI to solve,” Musk tweeted in 2021.I didn’t expect it to be so hardBut the difficulty appears in retrospect.

Even then, dojo something that is not commonthe American brand has already started talking about this technology in 2021. But so far, Tesla has not detailed how these D1 chips work and how they are interconnected by the dozens or hundreds in a vast computer fabric.

,We need to speed up these AI processorsDescription: Ganesh Venkataraman, who leads Tesla’s autonomous vehicle team. To achieve this, manufacturers start by designing chips according to their needs, mainly in the processing of video data that adapts to the changing environment of the car. tracks.

D1 chips are packaged in tray-wide groups of 25 in a single square “training room”. These cells join the edges to form a grid with their “neighbors”. Data moves from square to square, with cars traveling through city streets or, in the case of long journeys, using a network that resembles a train.

To speed up the process and adapt it to your needs, Tesla plans to move the AI ​​process from nvidia processor for the dojo, but we do not yet know at what stage this change is. “We’ve had the equipment for a long time and we’re running it in the lab,” explains the engineer. MaterialEmile Talpes.

As complicated as the dojo can be, basically This is Tesla’s biggest asset against traditional brands, The rest of the manufacturers rely on an extensive network of component suppliers to be able to develop their technologies, which slows down the process. In other words, Musk’s company applies vertical integration,

This integration is on the rise as it allows companies to have tighter control over their products and services so they can work together for customers. Moreover, it means that these companies are not required to share the profits with anyone else. In the case of Tesla, this allows them to regularly update their SoftwareA service for which he charges $10 per month.