Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register
This site may earn commission on affiliate links.
The next big milestone for FSD is 11. It is a significant upgrade and fundamental changes to several parts of the FSD stack including totally new way to train the perception NN.

From AI day and Lex Fridman interview we have a good sense of what might be included.

- Object permanence both temporal and spatial
- Moving from “bag of points” to objects in NN
- Creating a 3D vector representation of the environment all in NN
- Planner optimization using NN / Monte Carlo Tree Search (MCTS)
- Change from processed images to “photon count” / raw image
- Change from single image perception to surround video
- Merging of city, highway and parking lot stacks a.k.a. Single Stack

Lex Fridman Interview of Elon. Starting with FSD related topics.


Here is a detailed explanation of Beta 11 in "layman's language" by James Douma, interview done after Lex Podcast.


Here is the AI Day explanation by in 4 parts.


screenshot-teslamotorsclub.com-2022.01.26-21_30_17.png


Here is a useful blog post asking a few questions to Tesla about AI day. The useful part comes in comparison of Tesla's methods with Waymo and others (detailed papers linked).

 
Last edited:
There's a huge difference between Elon Musk and Elizabeth Holmes. Holmes never had any actual working technology; everything was a fraud. Elon and Tesla have produced quite a bit. I'm not trying to argue that he has been anything close to accurate with his statements, simply pointing out that the two are not at all in the same category.
So my idea of perfecting FSDb with a single drop of blood is not going to work? I can't see why not..........
 
Oh that is good information. Forgot you had done the trade in.

So certainly narrows it down to the new release. (Now I don’t need to worry about trying to clean the camera portion of the windshield.)

We can only hope it wipes more reliably when raining, as a consolation prize. 😂
It doesn't. Last week it was raining and my wipers didn't come on at all.
 
I have to agree. My son has a Ford Mach E GT and the fact that he can drive on the highway without ANY steering wheel nags is very nice. Why is Tesla so late to party on this capability? Even if it’s only on Highway it would be a huge help.
I have advocated for L3 on highway from Tesla for a while now. I would even renew my FSD Subscription based solely on that!!
 
  • Like
Reactions: FSDtester#1
  1. Tesla could easily fix Autosteer. But they don't. Even +10mph would be enough that I no longer feel like a road rage target.
  2. Tesla could easily fix TACC. But they don't. Dumb it down so it only slows when the car in front of you slows. Not for FSD obv. - I understand the need for extra caution there but when in just TACC (or TACC+Autosteer), it is just cruise control.
  3. Tesla could easily fix FSDb. But they don't. Just add a "Lane Change Confirmation" option and it would be pretty sweet. Phantom braking is largely predictable so that doesn't bother me nearly as much as bouncing back and forth and jumping into the wrong lane just before a turn.
Everything looks easy when it's not your responsibility.
 
  • Like
Reactions: sleepydoc
Everything looks easy when it's not your responsibility.
I mean, some of it is easy, like getting blind-spot monitoring to work in a car. Oh look, here's the new Model 3 with a blind-spot light on the... driver's side A-pillar speaker, that's going off with oncoming traffic because Tesla is using Tesla Vision for this, because Tesla. Because it's not like we didn't have a better way of doing this before...

It's pretty hysterical every time a car passes, the blind-spot light illuminates (4:40 onward).

Is this not the most Tesla thing you've ever seen?

1694126600131.jpeg



 
Last edited:
I mean, some of it is easy, like getting blind-spot monitoring to work in a car. Oh look, here's the new Model 3 with a blind-spot light on the... driver's side A-pillar speaker, that's going off with oncoming traffic because Tesla is using Tesla Vision for this, because Tesla. Because it's not like we didn't have a better way of doing this before...

It's pretty hysterical every time a car passes, the blind-spot light illuminates (4:40 onward).

Is this not the most Tesla thing you've ever seen?
I have had blind spot monitoring for decades on many cars and on ANY car I drive within 2 minutes.

 
I mean, some of it is easy, like getting blind-spot monitoring to work in a car. Oh look, here's the new Model 3 with a blind-spot light on the... driver's side A-pillar speaker, that's going off with oncoming traffic because Tesla is using Tesla Vision for this, because Tesla. Because it's not like we didn't have a better way of doing this before...

It's pretty hysterical every time a car passes, the blind-spot light illuminates (4:40 onward).

Is this not the most Tesla thing you've ever seen?

View attachment 971792


A joke and more junk coming at ya!
 
  • Funny
Reactions: Ramphex
The point of the V12 demo was that, apparently, the crowd at Tesla had found a new and different way of doing things and that the benefits of using that approach resulted in better CPU performance. When Musk was talking about being able to get rid of 500k lines of C++ code; that's code that doesn't need to run. And using NN, according to them, had better performance.
From what I've understood so far, the foundation of v12 is a "no rules" approach; all of the hard coded rules are now gone. Instead, the system will be trained from the ground up by watching training clips of drivers "doing it right," with very few (if any) hard coded rules. The system will learn that it needs to stop at a red light because all of the training clips it was trained on showed drivers doing exactly that, and not because of a hard set of rules that say "thou shalt stop at a red light." This may be a bit of an extreme example, as I imagine there will probably still be a few hard coded rules (particularly for general red lights), but you get the idea.

While there are a lot of obvious advantages to this method, there are undoubtedly many unforeseen ones as well. And of course, the corresponding disadvantages. One outright glaring issue is curation of training data now needs to be spot on, as they'll need to use data of drivers doing it right every time. This is going to potentially make auto-labeling a mission unto itself.

Since all of my popcorn poppers are now broken from so much use since August of 2019 (when I joined this foray into the future), I find myself prying open the wallet to purchase yet another industrial popper capable of producing the tons of popcorn I'll need for this latest effort.

Oh yes, I do believe that Elon et al will eventually solve FSD, but I think we haven't looked past the last "false horizon" yet.
 
I mean, some of it is easy, like getting blind-spot monitoring to work in a car. Oh look, here's the new Model 3 with a blind-spot light on the... driver's side A-pillar speaker, that's going off with oncoming traffic because Tesla is using Tesla Vision for this, because Tesla. Because it's not like we didn't have a better way of doing this before...

It's pretty hysterical every time a car passes, the blind-spot light illuminates (4:40 onward).

Is this not the most Tesla thing you've ever seen?

View attachment 971792


That light was not on because the oncoming truck that is traveling in the far left lane (shown in your screenshot), but because of the previous truck that was in the lane immediately adjacent.

If you rewind the video 2 seconds you'll see that first truck pass in the closest lane to the new 3. Only after it passes and is in the 3's blind spot does that light flash on for a fraction of a second.

Now, the question rightly posed is: why does one need to know a vehicle traveling in the opposite direction is in one's blind spot?
 
  • Informative
Reactions: jebinc
all sorts of issues on a simple "city streets" drive this evening - but as it seems that most of our reports are just deleted, I didn't bother reporting any.
I used to report stuff every single day. As expected, they go no where, or get prioritized by more often reported by more users, to be addressed, as the other ones just die in some report landfill and I keep experiencing the same issues in the same areas. I’ve even written emails to FSDbeta’s email. If you were wondering about the response time, they never responded.

I think it would be better to not have a report button, than have a report button that does nothing, only making the users waste their time to falsely feel more relevant “contributing data to the coolest thing ever”