Cleantech & EV'sNews

Elon Musk says Tesla is now working on ‘final piece of the FSD AI puzzle’

Elon Musk says Tesla is now working on the “final piece of the Full Self-Driving (FSD) AI puzzle,” and apparently that is “vehicle control.”

Despite his best efforts to stop making predictions about self-driving, the CEO has recently still predicted that Tesla would achieve full autonomy by the end of the year.

For many of us who use Tesla’s FSD Beta regularly, it is hard to imagine that Tesla can make the jump from the current state to a level 4 or 5 of autonomy where the automaker would take responsibility for the system and enable drivers to use it without monitoring in just a few months.

But that’s what Musk is suggesting – pending regulatory approval, obviously.

On Twitter, the CEO said yesterday that Tesla is now working on the “final piece of the Tesla FSD AI puzzle.” He says it is “vehicle control”:

Vehicle control is the final piece of the Tesla FSD AI puzzle. That will drop >300k lines of C++ control code by ~2 orders of magnitude.

It sounds like Tesla plans to more heavily rely on neural nets for vehicle control rather than direct coded instructions.

The CEO added that Tesla is currently training these neural nets and that “training compute” is the limitation at this point:

It is training as I write this. Our progress is currently training compute constrained, not engineer constrained.

Top comment by CMG30


Liked by 43 people

At this point, it’s ‘show me’. It doesn’t give me a lot of confidence that they’re just starting to work on some parts of the puzzle years after Elon claimed that FSD would be ready.

View all comments

Musk has been talking a lot lately about how Tesla has been securing as much compute power from NVIDIA as possible on top of deploying its own Dojo supercomputer program at the same time.

Electrek’s Take

As you know, I’m certainly not the most optimistic person when it comes to FSD, but the last of my optimism has been linked to Tesla’s Dojo program and the potential to accelerate the rate of improvement of the system, which I haven’t been impressed with over the last two years.

Tesla has started to deploy the supercomputer this month, and it could potentially accelerate the training of Tesla’s many neural nets powering FSD.

Obviously, this could just be hopeful thinking and there could be bigger underlying problems, but I have some hope.


Author: Fred Lambert
Source: Electrek

Related posts
AI & RoboticsNews

Nvidia and DataStax just made generative AI smarter and leaner — here’s how

AI & RoboticsNews

OpenAI opens up its most powerful model, o1, to third-party developers

AI & RoboticsNews

UAE’s Falcon 3 challenges open-source leaders amid surging demand for small AI models

DefenseNews

Army, Navy conduct key hypersonic missile test

Sign up for our Newsletter and
stay informed!