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:
2023.44.30.8 (11.4.9)

Just a quick sample showing the issues with stop signs as requested. The car will only proceed when it gets to 0mph. That is fine, no issues with that - it is great. However, it does not use the brakes to get to 0mph! It brings the speed down rapidly and not too smoothly from 10mph with mostly regen (hard to see friction brake use in video, it is very slight use, mostly regen), to 1mph. Then it just coasts to 0mph, which takes significant time - over one second, from 0:22 to past 0:23 is spent between 1 and 0mph. For comparison, from :21 to :22 takes it from 7mph to 1mph (does not seem extreme, though arguably higher than needed - it is 0.27g). If friction brakes were used the halt the vehicle at the appropriate position instead, time would be saved (I would estimate 2-3 seconds in total - with ~1 second from the drift alone - with other modifications to the profile, which is a very large amount of time) and the stop would feel more perceptible.
I wonder if the issue is the car wanting to stop in exactly the right location? So it slows down to approximately correct location, then drifts? I don’t think so, but could be one reason for this bizarre behavior.

If I get a chance I’ll do a human stop at this stop sign for comparison. But I think in the video, the issue is clear.

Finally, proof. Let me know if others see better behavior (with video if possible!). I think some people have claimed they don’t see this behavior (though many have indicated it is definitely an issue so I am not alone).
The 2 issues you have shown in your videos are the issues I've been mentioning for many months. It is very poor at stopping. This is exactly why I seldom allow FSD Beta to brake anymore. I typically gently tap the brake pedal to disengage so that the report goes to Tesla as a disengagement rather than cancelling with the stalk. It is so awful how it slows with regen to 10 MPH then abruptly slams on the friction brakes. When I have let it stop with passengers in the car ppl who don't know I'm using FSD Beta comment about the poor stopping. This happens both for red lights & stop signs. What I've observed is that the stop line on the display dances around a lot. I think it's because the car isn't sure exactly where to stop so it comes in fast, then when it realizes where it needs to stop it has to hit the brakes hard. It has issues with perceiving where to stop when there's no painted stop line (most intersections around here) and also when a stop sign is of a non-standard size or non-standard placement.
 
typically gently tap the brake pedal to disengage so that the report goes to Tesla as a disengagement rather than cancelling with the stalk
Perhaps I am misremembering, but I think disengaging with the stalk usually prompts for a voice report too. Sometimes there is location dependence though. I think it still counts as a disengagement in most cases.

I also typically voice report all the time (as I said, I was not doing so, so that I could enjoy viral YouTube glory with my riveting FSD video). About once every 2-3 miles, though in light traffic conditions it might be less.

It is very poor at stopping.
Agreed. It is possible to be much smoother and faster.

It is also not great at going, to be honest. Sometimes it is reluctant to get up to speed. I am not sure why. It drives like a typical Tesla driver, it seems. 😞

Also, at turns, PICK A LANE, ANY LANE. At certain turns I nearly always disengage to avoid infuriating drivers behind (something not all FSD drivers do apparently - unfortunately, because the disengagements would be good info!).
 
Last edited:
Perhaps I am misremembering, but I think disengaging with the stalk usually prompts for a voice report too. Sometimes there is location dependence though. I think it still counts as a disengagement in most cases.

I also typically voice report all the time (as I said, I was not doing so, so that I could enjoy viral YouTube glory with my riveting FSD video). About once every 2-3 miles, though in light traffic conditions it might be less.


Agreed. It is possible to be much smoother and faster.

It is also not great at going, to be honest. Sometimes it is reluctant to get up to speed. I am not sure why. It drives like a typical Tesla driver, it seems. 😞

Also, at turns, PICK A LANE, ANY LANE. At certain turns I nearly always disengage to avoid infuriating drivers behind (something not all FSD drivers do apparently - unfortunately, because the disengagements would be good info!).
No voice reports on legacy cars so we don't have that option.
 
I wonder if the issue is the car wanting to stop in exactly the right location? So it slows down to approximately correct location, then drifts? I don’t think so, but could be one reason for this bizarre behavior.
I always assumed this was a logic it sometimes chose to make it safer in light of potential perception errors, processing frames at 1 mph is easier than 10. Without radar there is no direct distance and velocity estimate, has to be done only through estimation of successive video frames and that will have error, particularly with monocular only vision (another mistake by Tesla).
 
The 2 issues you have shown in your videos are the issues I've been mentioning for many months. It is very poor at stopping. This is exactly why I seldom allow FSD Beta to brake anymore. I typically gently tap the brake pedal to disengage so that the report goes to Tesla as a disengagement rather than cancelling with the stalk. It is so awful how it slows with regen to 10 MPH then abruptly slams on the friction brakes. When I have let it stop with passengers in the car ppl who don't know I'm using FSD Beta comment about the poor stopping. This happens both for red lights & stop signs. What I've observed is that the stop line on the display dances around a lot. I think it's because the car isn't sure exactly where to stop so it comes in fast, then when it realizes where it needs to stop it has to hit the brakes hard. It has issues with perceiving where to stop when there's no painted stop line (most intersections around here) and also when a stop sign is of a non-standard size or non-standard placement.
Such a steal, for $12,000….
 
  • Funny
Reactions: Pdubs and kabin
I always assumed this was a logic it sometimes chose to make it safer in light of potential perception errors, processing frames at 1 mph is easier than 10. Without radar there is no direct distance and velocity estimate, has to be done only through estimation of successive video frames and that will have error, particularly with monocular only vision (another mistake by Tesla).
Maybe.

This profile occurs for every stop (not as noticeable on uphills since the car stops faster!) I have seen, regardless of whether vehicles are present.

To me this looks like a byproduct of a weird stopping profile. It does not make a lot of sense to not use the friction brakes when you are moving 1mph.

Certainly I have seen very early slowing followed by cautious creeping when other vehicles are present. That is a separate scenario and behavior and may be for safety.
 
Or because they have to fix an issue asap.. or comply with a NHSTA order
Oh yeah. I bet there are still a lot of cars running on a firmware version that did not have the Autopilot restrictions imposed to follow the NHTSA recall order. But Tesla can't force the user to install the update, can they? They can force only the download.
 
  • Like
Reactions: Pdubs
This isn’t actually good for Tesla. I think they like at least two versions out in large numbers at the same time, so they can do A/B testing and gather enough data for a while to improve future versions faster than on a single release at a time (which is eventually better for us too).
I suspect they don't need to do that much any more. I imagine they are going to do most of the work training/testing in the dojo, and just validate in the cars. In other words They can compare what they see in th real world with what they trained/tested and that's your A/B testing
 
2023.44.30.8 (11.4.9)

Just a quick sample showing the issues with stop signs as requested. The car will only proceed when it gets to 0mph. That is fine, no issues with that - it is great. However, it does not use the brakes to get to 0mph! It brings the speed down rapidly and not too smoothly from 10mph with mostly regen (hard to see friction brake use in video, it is very slight use, mostly regen), to 1mph. Then it just coasts to 0mph, which takes significant time - over one second, from 0:22 to past 0:23 is spent between 1 and 0mph. For comparison, from :21 to :22 takes it from 7mph to 1mph (does not seem extreme, though arguably higher than needed - it is 0.27g). If friction brakes were used the halt the vehicle at the appropriate position instead, time would be saved (I would estimate 2-3 seconds in total - with ~1 second from the drift alone - with other modifications to the profile, which is a very large amount of time) and the stop would feel more perceptible.
I wonder if the issue is the car wanting to stop in exactly the right location? So it slows down to approximately correct location, then drifts? I don’t think so, but could be one reason for this bizarre behavior.

If I get a chance I’ll do a human stop at this stop sign for comparison. But I think in the video, the issue is clear.

Finally, proof. Let me know if others see better behavior (with video if possible!). I think some people have claimed they don’t see this behavior (though many have indicated it is definitely an issue so I am not alone).

It sucks for sure but is it cause or effect? I always assumed that extra programmed time was a crutch to add dwell time at stop signs/UPL/intersections. Gawd knows there still isn't enough dwell time given all the instances of FSD pulling into oncoming traffic.

As I recall this delay logic was added about the same time as Chuck's unprotected left and creep line were added. Prior to that FSD would wig-out at stop signs with numerous psychotic indecisive starts/stops as it got more and more into the path of oncoming traffic.

But also, I suspect critical moments, like intersection stops, they want to minimize pixel smearing caused by brake induced suspension dive, speed changes, excess steering input as they can contribute to poor object estimation of range, speed, and path.
 
  • Like
Reactions: Pdubs and jebinc
The way TSLA stock price is heading these days Elon will need to pull a rabbit out of the hat with a v12 release if at all possible. Earnings release is Jan 24. He'll also need the next FSD PR distraction for v12's potential fall from grace.

Yearly options expire on Jan 19th with a fairly significant inflection point at 225 creating an incentive for market makers to push below that level. I think most downward pressure right now is somewhat artificial.
 
How is this possible?

The parking brake is only on two of the four wheels. Those two wheels didn't provide enough friction to keep the car from sliding down the driveway. (Likely inappropriate tires for parking on an incline covered in ice.)
 
How is this possible?

Yeah, it just slid. Very common with vehicles that have only two locked wheels (and can happen in vehicles with all four braked).

I warned my brother about this and he had to experience his car at the end of his driveway before he took it seriously.
 
The way TSLA stock price is heading these days Elon will need to pull a rabbit out of the hat with a v12 release if at all possible. Earnings release is Jan 24. He'll also need the next FSD PR distraction for v12's potential fall from grace.
Nah, he will deflect by making up some sort of BS about an upcoming product ($25k car, FSD v13, etc.) Smoke, mirrors and other charlatan behaviors we have all seen before. Wait for it....