Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

The FSD v9 showstopper

This site may earn commission on affiliate links.
We are not talking about Tesla going from 0% progress on BEV to 100% progress in one update (there is no 100%, at least not pre-Singularity).

We are talking about:

  • more vision NNs being upgraded from 2D projection to BEV

  • better BEV net architectures (e.g. ones that solve the challenge of internally representing uncertainty and probability)

  • better trained BEV nets (e.g. through larger/more diverse/better curated datasets)

  • BEV nets that have been debugged, use compute more efficiently, or have otherwise been refined
An analogy is the Autopilot rewrite that happened starting sometime after Andrej Karpathy joined (and Chris Lattner) departed in June 2017. The rewrite was a long process, not a one-time event. However, there were watershed updates like the 2018.10.4 update in March 2018 where a lot of the effort of the rewrite was concentrated in one update.

In that one update, a lot of Tesla drivers reported that Autopilot went from ping ponging between lane lines all the time to lane keeping smoothly. It could handle a lot of sharp turns and hills that it couldn't before. And automatic lane changes went from sloppy to reliable.

All I'm suggesting is that we'll see a similarly dramatic improvement in FSD from v9 and that it will be mainly because a lot of the progress from the 2.5D-to-4D rewrite will be concentrated in that v9 update.
 
Tesla is removing radar because of improvements in vision. Those improvements in vision are achieved through “significant architectural changes”.
With Elon saying that radar is being removed in production this week doesn’t that make this wrong and invalidate a huge part of your thesis?


You also posted a tweet of Elon saying V9 will bring fundamental improvement. But he also said the third beta version had “fundamental improvement”. This is simply the hype language that Elon uses, whether it’s “significant, gigantic, massive, blow your mind, fundamental improvement, etc”


Before he said 8.3 had 1,000 improvements. This is another hype statement. Cruise for example said they have hundreds of pull requests daily which goes through extensive simulation to make sure it’s only improving the software and not regressing. Imagine if cruise came out each week to say they had 3,000 improvements? This is what Elon is doing.

All I'm suggesting is that we'll see a similarly dramatic improvement in FSD from v9 and that it will be mainly because a lot of the progress from the 2.5D-to-4D rewrite will be concentrated in that v9 update.

My question to you is what will the disengagement rate look like with this new v9 that is supposed to provide a leap in computer vision according to you?

The way things are looking like is that V9 is released, safety disengagement on city goes from 1-5 miles on average to 15-25 miles on average (that’s after 8 months of development since first release). Elon right after tweets 9.1 is gonna have double the improvement of 9.0 then we are back to square one of the hype.

Rinse and repeat.

It’s what’s gonna lead to the qualitative leap in vision performance from FSD beta v8.2 to v9.0.

So what should the safety disengagement rate look like with this “dramatic improvement”, “qualitative leap in vision performance”?
 
Last edited:
It redirects to a YouTube link. I, for one, am never going to give up TinyURL links. They're never going to let you down.
61FF2767-EDAF-4A38-BE50-90A635F8481F.png
 
... Before he said 8.3 had 1,000 improvements. This is another hype statement. Cruise for example said they have hundreds of pull requests daily which goes through extensive simulation to make sure it’s only improving the software and not regressing. Imagine if cruise came out each week to say they had 3,000 improvements? This is what Elon is doing. ...
Yeah, I've done pull requests for logging improvements.