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:
Beta is available for all Tesla cars with HW3 that have paid for or subscribe to FSD. So there are cars from about 2017 onward. Why would you think 2022 cars are excluded and why ask? Also I have acquaintance who has a Model S Long Range that is (I'm reasonably sure) a 2022 he got Beta a couple of weeks ago.
Because I thought there was a freeze on adding new Beta testers. I purchased a Model S in January and had a 99 safety score for a long time but never got the Beta. Last month I opted out because I was tired of waiting.

Thanks for the update..
 
Beta is available for all Tesla cars with HW3 that have paid for or subscribe to FSD. So there are cars from about 2017 onward. Why would you think 2022 cars are excluded and why ask? Also I have acquaintance who has a Model S Long Range that is (I'm reasonably sure) a 2022 he got Beta a couple of weeks ago.
If by "available" you mean that the car can be added to the beta, then you are correct. If you mean that anyone with qualifying hardware/cars will get the beta, then no, you are not correct.
 
If by "available" you mean that the car can be added to the beta, then you are correct. If you mean that anyone with qualifying hardware/cars will get the beta, then no, you are not correct.
I mean available as in able to obtain. As an original (non influencer) late Oct 2021 Beta user I'm fully aware of the steps needed to acquire the "available" Beta.;)
 
Have been still regularly been getting into the beta this week? I had the beta on my Model 3 before getting my 2022 MS in late December and finding out that they had closed the beta for the time being. I opted into the queue about 10 days ago and have around 700 miles with a 97SS but still nothing yet. I had received a update notification last night, but was not the FSD Beta push.

Just curious to hear if they're pushing in waves or with major updates.
 
Because I thought there was a freeze on adding new Beta testers. I purchased a Model S in January and had a 99 safety score for a long time but never got the Beta. Last month I opted out because I was tired of waiting.

Thanks for the update..
I took delivery of my Model Y on Mar 30th, and had a safety score of 97 when I was accepted into the beta about 4 weeks ago. I have no idea how they determine who gets it, but I am glad to have it...the software is pretty amazing, even if you have to keep a close eye on it!
 
  • Like
Reactions: herrjon
With fsd beta 11 there will be a single stack . Currently on fsd 10.12 if I exceed max autopilot speed on the freeway, autopilot and fsd disengages for the remainder of the drive. It does not ding me for forced fsd disengagements allowing me to keep fsd. Will this change when there is a single stack? Will I get dinged on the fsd forced disengagement count?
 
I was kinda wondering why FSD is still on 2022.12 as that's March datestamp. Meaning there hasn't been any new releases in 5 months. I assume they are still working on FSD :) Maybe big update coming next?
About two months ago, Elon said the next release would be in about two weeks. We're still waiting. The last comment from Elon that I know of said they were working on it really hard-- rather a given, I think-- but that it just wasn't quite ready yet.
 
Looks like there's still network architecture changes happening with FSD Beta 10.13 to improve 3D vector representation of the environment:
- Improved lane position error by 5% and lane recall by 12% with a full vector lanes neural network update. Information bottlenecks in the network architecture were abated by increasing the size of the per-camera feature extractors, video modules, and internals of the GPT attention modules.​
- Improved lane position error of crossing and merging lanes by 22% by adding long-range skip connections and a more powerful trunk to the network architecture.​
10.11 introduced this vector space network architecture for lane predictions:
- Upgraded modeling of lane geometry from dense rasters ("bag of points") to an autoregressive decoder that directly predicts and connects "vector space" lanes point by point using a transformer neural network. This enables us to predict crossing lanes, allows computationally cheaper and less error prone post-processing, and paves the way for predicting many other signals and their relationships jointly and end-to-end.​
And while 10.12 improved lane predictions with more auto-labeled training of that new network, there were still significant errors needing a full update to the network in 10.13. Hopefully now this means 10.11's original note about "paves the way for predicting many other signals and their relationships jointly and end-to-end" can have meaningful progress towards unified vector space if the 10.13 architecture proves to be enough to predict lanes.

However with Elon Musk's response during 22Q2 earning call to a question about unified vector space for both static and moving objects
If they could be knit together at the neural net level, then you don’t need to reconcile them within C++ heuristics. That is an architecturally better way. That’s the most desirable outcome. I think it’s probably not necessary to achieve full self-driving, but it would be a slight improvement in the efficiency. It’s certainly something we want to get to. The nirvana situation is you have surround video auto-labeling of all static and dynamic objects, and then you have then surround video inference with spatial memory as well. I think we are almost certainly there before the end of the year.​

… sounds like unified vector space is not a FSD Beta 11 blocking requirement.
 
About two months ago, Elon said the next release would be in about two weeks. We're still waiting. The last comment from Elon that I know of said they were working on it really hard-- rather a given, I think-- but that it just wasn't quite ready yet.
It appears that Elon tweeted on 7/29/22 that FSD beta 10.13 is probably a week or so away.

I take that to mean that it probably will not be out before 8/6/22 or so.


I also am looking forward to an update, in part because the next release may include some of the improvements already in the non-beta releases, such as relocating the blinker-cam and blended braking when regen is limited.

SW
 
Elon Musk hinted / joked at the 2022 shareholder meeting that the upcoming 10.13 might get a FSD Beta 10.69 version label as it has so much architectural improvements deserving more than a minor version bump. So it very much seems like FSD Beta 11 "major version" is purely reserved for single stack highway even though there has been many technical milestones along the way on 10.x.