Cleantech & EV'sNews

Tesla changes course with Full Self-Driving to introduce end-to-end AI in v12, says Musk

Tesla is planning to update its Full Self-Driving package with end-to-end AI in its FSD Beta v12, says Elon Musk.

FSD Beta already makes plenty of use of neural nets powered by artificial intelligence, but it is used for the perception part of the equation – meaning for the vehicle to detect its environment.

When it comes to the decision-making and driving input (steering, accelerator, brakes, etc.), Tesla relies on regular programming.

CEO Elon Musk previously said that Tesla wouldn’t need AI for that part, but he appears to have now changed his mind.

In a new comment hyping FSD Beta improvements in the upcoming v11.4 update, Musk talked about another upcoming v12 update that would include “end-to-end AI”:

Arguably, v11.4 should be v12.0, as there are so many major improvements. v12 is reserved for when FSD is end-to-end AI, from images in to steering, brakes & acceleration out.

The CEO didn’t offer a timeline for this particular update. The automaker is currently starting to expand the release of v11.4 beyond its internal fleet.

Recently, Musk said that he believes that Tesla can achieve “full autonomy” this year – though, as we reported, it’s not clear what he even means by that at this point.

Electrek’s Take

It wouldn’t be the first time that Musk changed his mind about Tesla’s self-driving effort. This one makes sense though. I think that the best part about FSD Beta right now is its perception. It detects its environment extremely well.

If we can get that kind of performance on the decision-making side of things, I can see FSD Beta significantly improving. Maybe that’ll even be what is going to let me see a path to Tesla finally delivering on its promises of full self-driving.

That said, I still have doubts because I’ve had FSD Beta for over a year now, and I saw very few improvements over that period of time. It makes it hard to think that FSD Beta could become a true fully autonomous system any time soon.


Author: Fred Lambert
Source: Electrek

Related posts
AI & RoboticsNews

Medical training’s AI leap: How agentic RAG, open-weight LLMs and real-time case insights are shaping a new generation of doctors at NYU Langone

AI & RoboticsNews

OpenAI’s ChatGPT explodes to 400M weekly users, with GPT-5 on the way

AI & RoboticsNews

Together AI’s $305M bet: Reasoning models like DeepSeek-R1 are increasing, not decreasing, GPU demand

DefenseNews

Army Stinger missile replacement competition heads into flight tests

Sign up for our Newsletter and
stay informed!