Ever since Tesla rolled out its Full Self-Driving Beta last week, the lucky group of individuals who have been sharing the new system’s capabilities are proving that Autopilot has improved significantly. Previously challenging tasks for the 2.5-dimension Autopilot versions are no longer a tough task thanks to a 4D comprehension of surroundings, which are a preview of what is to come with Tesla’s upcoming “Dojo” Supercomputer.
Tesla owners who have had FSD for some time know that the capabilities of the self-driving suite were somewhat limited. Everyone who purchased FSD knew it was a work in progress, and by driving with the capability activated, it was becoming more sophisticated with the help of Tesla’s Neural Network. However, the Tesla Artificial Intelligence team knew what had to be done: the amount of information that could be processed needed to be greater, and the vehicle’s comprehension of its surroundings needed to be more complex. Therefore, Tesla is developing Dojo.
Dojo is Tesla’s Neural Network training program that aims to begin breaking down data in 4D instead of “~2.5D,” which is what the automaker’s Autopilot was previously using.
Musk detailed the need for a more complex autonomy system during the Q2 2020 Earnings Call:
“Well, the actual major milestone that’s happening right now is really a transition of the autonomy system or the cars, like AI, if you will, from thinking about things in — like two-and-a-half feet. It’s like think — things like isolated pictures and doing image recognition on pictures that are harshly correlated in time but not very well and transitioning to kind of a 4D, where it’s like — which is video essentially.”
The issue with previous FSD and Autopilot builds was that not enough information was being transmitted through pictures. There needed to be timestamps and more accuracy through an increasingly fluid comprehension of the surroundings. The key was to transition from images, or 2D, as Musk called it, to video, or 4D.
“So what we’ve been doing, thus far, has really just been like 2D — mostly 2D, and like I said, well correlated in time. So just hard to convey just how much better a fully 4D system would work — does work. It’s capable of things that if you just look — looking at things as individual pictures as opposed to video — basically, like you could go from like individual pictures to surround video, so it’s fundamental. So the car will seem to have just like a giant improvement.”
Roundabout Navigation
One way to show how the new system is operating more efficiently is a Tesla’s navigation of a roundabout. Musk stated that it would be able to handle roundabouts “not perfectly at first,” but it would be able to navigate through them.
Previous versions of Autopilot have had difficulties navigating through roundabouts, and very rarely did they manage to get through one without human intervention. An example can be seen in a July 2019 video from YouTuber Dirty Tesla, who showed his Model 3 attempting to go through the tricky stretch of roadway. At the 3:25 mark of the video, you can see the Model 3 doesn’t do a great job of making it through, and the driver is forced to intervene with the vehicle.
Tesla’s FSD Beta is proving that an increase in comprehension is just what Tesla Autopilot needed to function more accurately. A video from fellow Tesla Model 3 owner James Locke, who received the FSD Beta, shows the navigation through a roundabout with relative ease. Even Locke was impressed and stated that the maneuver required no intervention from him, and Autopilot took care of the entire process independently.
Dojo’s coming release in conjunction with the new FSD Beta could prove to be the answer to all of the issues that Tesla had previously. With a new, more complex system that takes in more information on terrain, surroundings, and obstacles, Autopilot is more accurate than ever before. The increase in capability is being displayed daily as new videos of the FSD Beta are being rolled out regularly.