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's called prophylactic wipes. That way the windshield has already been wiped when it rains, thinking ahead, neutralizing the problem of the wipers napping when it actually rains..........Simply brilliant! A complex system but give it a chance.........
Damndest thing, it’s still raining in my parking garage every afternoon. I need to get in touch with the office of the building about that.
 
Ironically, I don’t have working seat belts. Just received the recall notice in the mail about this one. 🤦🏻‍♂️

Your fault, you should have splurged for full self-seatbelted driving.........FSSDwtf.........
 
Actually, bad though this sounds, you CAN say this, and in fact you HAVE to. Nothing is ever 100% .. certainly not humans drivers and not even the best L4/L5 car we develop. If you wanted 100% safety you would have no cars at all, no airplanes, no staircases in homes (very dangerous), and live in a padded cell. What you want is an autonomous car that is SAFER than MOST human drivers, but what "SAFER" and "MOST" mean is of course tricky to define accurately.
Point well made and I agree to some extent but i'm referring to the use case that was provided where the tesla ran into the back of a parked emergence responder. I don't think you can say this would be "safer" than most people by any stretch of the imagination. The fact is that it still has issues in cases like this and also artifact images being misunderstood. This has gone on for a very long time with the promise of the next big leap in FSD solving all of these problems. However, I rarely go through a week where I don't have to disengage to prevent a potentially serious accident. You can believe Elon and influencers dangling the carrot of neural nets as solving these issues overnight but I don't buy it in a second. We are still a long ways off from even L3 at this point and you may disagree but my 2 yrs with beta FSD does not give me any confidence that the HW stack will ever be suitable for even L3. I see too many issues that cannot be solved through software alone. I end by saying i hope I am wrong about this.
 
  • Like
Reactions: drtimhill
Point well made and I agree to some extent but i'm referring to the use case that was provided where the tesla ran into the back of a parked emergence responder. I don't think you can say this would be "safer" than most people by any stretch of the imagination. The fact is that it still has issues in cases like this and also artifact images being misunderstood. This has gone on for a very long time with the promise of the next big leap in FSD solving all of these problems. However, I rarely go through a week where I don't have to disengage to prevent a potentially serious accident. You can believe Elon and influencers dangling the carrot of neural nets as solving these issues overnight but I don't buy it in a second. We are still a long ways off from even L3 at this point and you may disagree but my 2 yrs with beta FSD does not give me any confidence that the HW stack will ever be suitable for even L3. I see too many issues that cannot be solved through software alone. I end by saying i hope I am wrong about this.
You have to be careful when making statements like that, as there is data for it available. Humans crash into emergency vehicles a significant amount of times.

A recent study highlights the potential increase in crash risk when ambulances operate with lights and sirens. When an ambulance responds to an emergency call without using lights and sirens, the crash rate is 4.6 per 100,000 responses. The crash rate increases to 5.5 when lights and sirens are used. The increase in risk is even greater when the ambulance is transporting a victim. The crash risk without lights and sirens is 7.0 per 100,000 transports, and increases to 16.5 when lights and sirens are used throughout the transport.

The most recent report from the National Safety Council found that almost 200 people died in crashes involving emergency vehicles in 2021 alone.
 
You have to be careful when making statements like that, as there is data for it available. Humans crash into emergency vehicles a significant amount of times.
it's a moot point. and I say that because who really knows because tesla is not transparent in sharing data with the public. the crash rate for Tesla could be 100x higher than what you referred to or could be lower. It's all speculation. I just know from my 2 yrs in the FSD beta program that it is a much worse driver than me for the most part. the best that I can see for FSD would be L3 in the future but I don't think they get there with camera vision only. Of course that is an opinion based on my experience..
 
it's a moot point. and I say that because who really knows because tesla is not transparent in sharing data with the public. the crash rate for Tesla could be 100x higher than what you referred to or could be lower. It's all speculation. I just know from my 2 yrs in the FSD beta program that it is a much worse driver than me for the most part. the best that I can see for FSD would be L3 in the future but I don't think they get there with camera vision only. Of course that is an opinion based on my experience..
I seriously want to know if @STUtoday is a bot. No input and any disparaging remarks about EM or Tesla and its thumbs down. And not normal thumbs down. Weird, no counterpoint thumbs down. He’s more in sync with that thumbs down button than a pair of incestuous ice dancers.
 
A couple days ago I drove home from work, running some errands and making 3 stops along the way. Probably 20-25 miles total and 45-50 minutes of driving. The only time I had to take over was when there was some construction, half the road was blocked off and a flatbed semi was parked on the wiring side of the road. There was one spot where FSD drifted into a turn lane instead of going straight but was able to correct itself. This has been a perennial problem for FSD and was actually fixed on the last version but regressed with 4.7 so I’m not sure how to classify it. It doesn’t do it perfectly but if figures out it’s mistake and corrects it.

Either way, despite all the imperfections I can’t see how people say it’s worthless or junk or that no progress has been made.