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:
Like this?


Like this?

Started having the flashing Autosteer icon today when it was never a problem before. I'm been on 2022.45.12 for several weeks with no issue until this morning, so something was pushed out to start this problem. Tried rebooting several times but no improvement.

If I'm going really slow or at a stop sign and the icon pops ON and I engage Autosteer really quick, it does work just fine. Then I can activate the route an FSD takes over.

Also notices I didn't receive the front parking distance warning today. It's always worked before - '21 MS+ that does have the ultrasonic sensors?
 
Or you have uberhuman standards for driving maybe? People tell me my Tesla has "trash build quality" all the time and I don't even know what "build quality" is. It's fine to me.
Well observed and the same I see when reviewing the responses. Yes no question there is absolutely work to be done but the “barking” for perfection or the perception of what one views as perfection is so very deviated on the FSD topic. I use it every day as I have for 35k miles and there is no question to me it is continuing to improve. Maybe not at the pace a few wish but improving. IF there was a comparable example of another company solving this software problem (on ALL roads US) to judge quality or time frame I would be with the barkers but for now this is still a never before accomplished task so I continue to enjoy what has been completed warts and all. IF these threads began to change in dialog to “I had an FSD accident today” I may change my perspective but in three years I have yet to see that happen. If Anyone expects the next or next after update to miraculously solve this monumental software problem you might as well tap out now and enjoy life without this so troubling anxiety. Life is too short to spend it complaining.
 
my working theory that my nags have greatly reduced due to my high mileage with beta
Here's 100 consecutive nag timing seconds with 11.3.4 where I drove on a mix of city streets, interstate highways and local highways:
Code:
43 41 42 34 45 58 41 52 41 44
40 43 47 25 34 34 40 47 57 25
51 29 26 26 28 53 58 29 28 30
27 29 28 33 28 28 44 27 30 34
38 29 48 22 20 34 28 29 16 14
29 25 16 32 32 25 17 27 14 31
41 15 25 29 13 13 13 22 14 13
15 13 19 18 19 18 26 33 46 47
57 37 31 33 48 46 26 31 51 18
27 28 19 44 15 21 25 20 58 29

That's over 50 minutes of continuous driving without disengagement. It also includes a 15 minute portion where law enforcement was right on my tail following 1 second behind on a local curvy mountain highway with people making unprotected turns ahead of us. 😅 (The only potentially confusing thing was a single blink of an unnecessary turn signal at a fork.)

But back to the numbers, there seems to be a wide range of nags down to 13 seconds and up to 58 seconds in this case. I believe the longest intervals were on the interstate with a lead vehicle following 1.5 seconds behind on Chill. But even on the local highways with plenty of curves and intersection, the nags could be 30+ seconds apart.
 
When the car is trying to run off the road, yes, I'm anticipatory of an update that fixes it.
It’s just interesting to contrast the difference between some people who are dying to install it the second they get it, and a small number of people who wait around for reviews and possibly skip updates or aren’t able to install for some reason.
 
Well observed and the same I see when reviewing the responses. Yes no question there is absolutely work to be done but the “barking” for perfection or the perception of what one views as perfection is so very deviated on the FSD topic. I use it every day as I have for 35k miles and there is no question to me it is continuing to improve. Maybe not at the pace a few wish but improving. IF there was a comparable example of another company solving this software problem (on ALL roads US) to judge quality or time frame I would be with the barkers but for now this is still a never before accomplished task so I continue to enjoy what has been completed warts and all. IF these threads began to change in dialog to “I had an FSD accident today” I may change my perspective but in three years I have yet to see that happen. If Anyone expects the next or next after update to miraculously solve this monumental software problem you might as well tap out now and enjoy life without this so troubling anxiety. Life is too short to spend it complaining.
Fortunately most FSDb drivers are on guard for potential accidents. Here's a 11.3.6 overlap cut-in where FSDb totally missed the ball.

 
Well observed and the same I see when reviewing the responses. Yes no question there is absolutely work to be done but the “barking” for perfection or the perception of what one views as perfection is so very deviated on the FSD topic. I use it every day as I have for 35k miles and there is no question to me it is continuing to improve. Maybe not at the pace a few wish but improving. IF there was a comparable example of another company solving this software problem (on ALL roads US) to judge quality or time frame I would be with the barkers but for now this is still a never before accomplished task so I continue to enjoy what has been completed warts and all. IF these threads began to change in dialog to “I had an FSD accident today” I may change my perspective but in three years I have yet to see that happen. If Anyone expects the next or next after update to miraculously solve this monumental software problem you might as well tap out now and enjoy life without this so troubling anxiety. Life is too short to spend it complaining.
I would say it depends how egregious the driving behavior is. I've never experienced it until this latest update (11.3.4) repeatedly violently merge into turning lanes on the right-hand side without warning. Two steps forward and three steps back is a hard pill to swallow.
 
  • Like
Reactions: Ramphex
Fortunately most FSDb drivers are on guard for potential accidents. Here's a 11.3.6 overlap cut-in where FSDb totally missed the ball.

Drivers should always be on guard. In that example until they can teach ”mind reading for goof balls” there will never be a solution. There will Always be crazy unpredictable drivers forever!
 
Here’s an example of a cut-in, which could easily have been gracefully handled.

This is a more severe example.

Note my reaction time was one second faster than FSD Beta. This makes all the difference on a smooth compensation for the soon-to-be slow leading vehicle. And note the movement of the brake pedal, and the very large slowdown immediately before disengagement.

Why didn’t it anticipate and slow down earlier, or change lanes (what I would have done earlier if I had been driving - to be clear, I was driving…)?


While an extreme example, much more minor cut-ins show the same behavior where the car does not react until the vehicle actually enters the lane. Unacceptable. This is a VERY common occurrence on the freeway.

I have had about 6-7 instances of unnecessary jerky speed adjustments due to this sort of thing in the brief testing I have done over the last week.

Everything must be butter smooth.
 
Here’s an example of a cut-in, which could easily have been gracefully handled.

This is a more severe example.

Note my reaction time was one second faster than FSD Beta. This makes all the difference on a smooth compensation for the soon-to-be slow leading vehicle. And note the movement of the brake pedal, and the very large slowdown immediately before disengagement.

Why didn’t it anticipate and slow down earlier, or change lanes (what I would have done earlier if I had been driving - to be clear, I was driving…)?


While an extreme example, much more minor cut-ins show the same behavior where the car does not react until the vehicle actually enters the lane. Unacceptable. This is a VERY common occurrence on the freeway.

I have had about 6-7 instances of unnecessary jerky speed adjustments due to this sort of thing in the brief testing I have done over the last week.

Everything must be butter smooth.
So evasive decelerate or irrational jump out of lane hoping it’s clear. I would have let it decelerate from 70 to 65 as the correct action personally.
 
  • Like
Reactions: STUtoday
Legit automation in one state > overly supervised junk in 50 states
Have you used this L3 system in Nevada? If not, how can you judge that against something many of us use in the other 47 contiguous US streets and Nevada and parts of Canada?

I've seen some videos of Mercedes on highways and I think FSD is definitely as good and in most cases several magnitudes of order better. Especially V11. Autopilot was already better than Mercedes and other commercial use ADAS systems. FSD V11 is another evolution of this.
 
  • Like
Reactions: Yelobird
Well observed and the same I see when reviewing the responses. Yes no question there is absolutely work to be done but the “barking” for perfection or the perception of what one views as perfection is so very deviated on the FSD topic. I use it every day as I have for 35k miles and there is no question to me it is continuing to improve. Maybe not at the pace a few wish but improving. IF there was a comparable example of another company solving this software problem (on ALL roads US) to judge quality or time frame I would be with the barkers but for now this is still a never before accomplished task so I continue to enjoy what has been completed warts and all. IF these threads began to change in dialog to “I had an FSD accident today” I may change my perspective but in three years I have yet to see that happen. If Anyone expects the next or next after update to miraculously solve this monumental software problem you might as well tap out now and enjoy life without this so troubling anxiety. Life is too short to spend it complaining.
 
  • Funny
Reactions: FSDtester#1
jump out of lane hoping it’s clear.
Huh? Why would it do that? I knew it was clear (since I was driving and had a second to double check) so presumably it did too!

I would have let it decelerate from 70 to 65
I would have too, except it didn’t slow down at the time I reacted!!! It had an eternity. And it should be faster than me.

I bet Alan Subie is a pro race car driver or something
Lol, definitely not. I like driving though.