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:
It has been overcast here in middle Tennessee the past couple of days and I have been experiencing the dry wipe problem many have had. I have to admit it is very frustrating to have the wipers operate on a dry window.
I've been happier with my wipers since I turned off auto sensing. The main problem I had was that they would often go into high speed mode unnecessarily. Very annoying.

Luckily I didn't have the dry wipe problem. that would have pissed me off. I'm pretty anal about trying not to scratch my windshield. If my windshield is dirty and dry I only clean it with glass cleaner and a soft rag. If it starts raining I don't turn on the wipers without squirting washer fluid for about five seconds. One of my old cars had a very wiper scratched windshield and I'm trying to prevent that from happening again.
 
The Charlatan continues his shenanigans this Thanksgiving 🦃 holiday.


Notice the now added v13 qualifier, “over time”…. So much for v12 not being beta…. Nonetheless, I’m sure it will be “mind-blowing”!


🤣😂😔💦

View attachment 993254

I’m sure V12 will be mind blowing - even if it performs at the current 11 level. Because it’s end-to-end trained NN.

May be not from a consumer perspective- but from a nerd perspective.

ps : Similar to how mind blowing GPT is. Not because it’s better than humans but because it is somewhat close.
 
Another example of B-Pillar obstructed view and how dangerous it can be. FSD tried to pull out using Picture 1 B Pillar camera view without being able to see on coming cars so I had to disengage. I could see just fine- Picture 2.

Picture 1- B Pillar Obstructed View

View attachment 993258

Picture 2- Drivers View Leaning Forward

View attachment 993259
What do larger vehicles with a larger front - like a truck do ?

Anyway, any road with no shoulder will have this issue.
 
What do larger vehicles with a larger front - like a truck do ?

Anyway, any road with no shoulder will have this issue.
The vast majority of trucks surprisingly don't have this problem. From the front of the truck to the steering wheel is not 9' which is how far the front of a Model Y is to the B-Pillar camera is. Check out any pick up truck and you'll see this and the same goes for most large truck cabs. I was surprised by this when I started comparing. And you won't find a single passenger vehicle where the length is 9' either. Of course there are exceptions but not nearly as many as people believe. As you said a key factor is whether there is a shoulder to the crossing road. Where I live there often isn't one so creeping out for full visibility is harder. I just find that FSD often pulls out in front of oncoming traffic at obstructed view intersections when the typical human would wait. Why? because FSD doesn't see any vehicles coming :)
And just wait until snowbanks get in the way.
 
Last edited:
  • Informative
Reactions: heapmaster
Another example of B-Pillar obstructed view and how dangerous it can be. FSD tried to pull out using Picture 1 B Pillar camera view without being able to see on coming cars so I had to disengage. I could see just fine- Picture 2.

Picture 1- B Pillar Obstructed View

View attachment 993258

Picture 2- Drivers View Leaning Forward

View attachment 993259
From the shadow of your car, I can see that your driver view picture have the car more forward! Also, the B pillar camera have a more higher angle and be able to see over the fence further out!
 
The vast majority of trucks surprisingly don't have this problem. From the front of the truck to the steering wheel is not 9' which is how far the front of a Model Y is to the B-Pillar camera is. Check out any pick up truck and you'll see this and the same goes for most large truck cabs. I was surprised by this when I started comparing. And you won't find a single passenger vehicle where the length is 9' either. Of course there are exceptions but not nearly as many as people believe. As you said a key factor is whether there is a shoulder to the crossing road. Where I live there often isn't one so creeping out for full visibility is harder. I just find that FSD often pulls out in front of oncoming traffic at obstructed view intersections when the typical human would wait. Why? because FSD doesn't see any vehicles coming :)
And just wait until snowbanks get in the way.
It's downright dangerous for FSD to throw the dice like that. At the least FSD should alert the driver when it's about to make a decision at an intersection with obstructed views.
 
  • Like
Reactions: jebinc
I’m sure V12 will be mind blowing - even if it performs at the current 11 level. Because it’s end-to-end trained NN.

May be not from a consumer perspective- but from a nerd perspective.

ps : Similar to how mind blowing GPT is. Not because it’s better than humans but because it is somewhat close.
I think that V12 will be a lot better driver than the average human, but only when it can "see" critical areas. The present mix of cameras on the Tesla fleet is insufficient to handle a wide range of ULTs and URTs, because of FSD's inability to gauge cross traffic behavior in all circumstances. This is never going to change unless V13... include ESP as part of the NN.
 
I think that V12 will be a lot better driver than the average human, but only when it can "see" critical areas. The present mix of cameras on the Tesla fleet is insufficient to handle a wide range of ULTs and URTs, because of FSD's inability to gauge cross traffic behavior in all circumstances. This is never going to change unless V13... include ESP as part of the NN.
This is neither here nor there.

The point is that from a technological advancement perspective an end to end trained NN doing the driving is mind blowing.
 
This is neither here nor there.

The point is that from a technological advancement perspective an end to end trained NN doing the driving is mind blowing.
That puts V12 right up there with the Samsung Galaxy Note 7. The media consensus regarding the Note 7 was that it was a technological marvel and probably the best smartphone ever. Of course, that was before the first Note 7 caught fire. Let's hope that Teslas's lack of cross traffic sensing isn't V12's equivalent fatal flaw.