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:
Welcome, thanks for coming over.

Just to clarify, by models we mean 3/Y vs X/Y. As you may have read in this thread, the hypothesis is that FSD is better in 3/Y compared to S/X.

To expand, there is a wide range of perceived quality (and thus usefulness) of FSD and I've been trying to figure out why. Some of the causes might be
- location differences (a lot of people think this is the cause)
- model differences (3/Y vs S/X)
- learning curve i.e. those who put in the effort to push the boundaries to learn how FSD behaves will understand its capabilities better and are able to use it more ... this is my contention.

ps : Expect a lot of bickering and debates here. Lots of folks have been testing FSD from the beginning.
Thanks for the summary 😀

Interesting discussion. Its really hard to test without the data, obvs. YouTube videos alone will not provide an answer, IMO. Personal anecdotal experience is too subjective.

My drives to/ from work have taught me that there a few spots the software interprets the same consistently. But each drive is unique so difficult to compare the drives.

I agree with you. In terms of maximizing the use of FSD beta, for me -it's about learning what it can/ can't do and being patient with it.
 
  • Like
Reactions: jebinc
I guess I'm one of the exceptions here, but I am glad that I purchased the FSD option when I bought my 2022 MSLR. At the time, it cost $10k, so a roughly 10% increase in the price of the car by adding the FSD. I realize that as a percentage of the car cost, the current $15k is much higher, especially on the base Model 3s and Ys.

To me, FSD was one of the main reasons I bought a Model S. I love cutting edge stuff, and I really appreciated getting over-the-air updates to see improvements to my car as time went by. Would my car be as enjoyable to me without FSD? Definitely not. The price of FSD, like anything in our capitalistic society, is set by the laws of supply and demand (tempered somewhat in this case by somewhat arbitrary decisions by Elon on pricing). No one is forced to buy FSD, and anyone who does even the smallest amount of research knows that Elon's over-the-top claims and promises are worthless (let the buyer beware).

I don't expect the software to be perfect, and I definitely curse at my car often when it does a really stupid maneuver, but overall it's fun for me to play with, and totally blows people's minds when I demo it for them (even as wonky as it currently is).

I guess it comes down to setting expectations properly. For those who don't take the time to look at Elon's FSD claims history and do some basic research before buying FSD, I'm sure their expectations are WAY too high. Oh well!
 
Yeah I’ll never buy FSD again, like you if I even stay with Tesla. With the exception of steam (I received just late enough to get steam but not the new lights), my journey has been a mirror of yours, disappointment on undelivered features for the refresh S. Though I do still love my S, I do miss it every time I’m forced to drive an ICE, I’m disappointed with how Tesla delivered it half finished and has made no effort to complete promised features.

My biggest and loudest piece of advice to anyone who asks for my recommendation on Tesla is not to, ever, under any circumstances purchase FSD.
I feel similarly, but I plan to possibly proceed differently. I most likely will not, I wouldn’t maybe say NEVER, but IF I buy another Tesla in the next 1-2 years, would not put FSD on that product. What I plan to do though is assuming that the value of my 2018 M3 LR WITH full FSD is probably only about ~30k, possibly lower in a couple years - but IF there is a chance that Robo Taxi isn’t more than from now a couple years out, I would consider keeping that M3 LR, putting it into a robotaxi fleet (which I would NEVER do with my personal daily driver) and so MAYBE per some simple maths that 2018 M3 could produce some monies to pay for itself and reduce the overall cost of a new MS LR, or something other.

But, I am glad that I only paid about ~7K for the full FSD, and could never think of paying 15K for it today. Frankly, I think upping the price of FSD to 15K has nothing to do with value or revenue generation for Tesla (as high margin as it is) but more to keep the dream alive with analysts.
 
Last edited:
  • Like
Reactions: pilotSteve
The stupid thing about FSD purchasing is that Tesla could actually solve this problem. Here's my solution: Just bind FSD to your Tesla at the car level with the option of removing it back out to the account level (one purchase needed per vehicle, of course) up until the point where it's "complete." At that point, it's bound to the vehicle and removed from the account level. That way, as some of us have done already, for right now - when we trade in the car to Tesla without getting what we paid for, we know we get the feature in the next Tesla. This would also encourage consumer loyalty to the Tesla brand, "guarantee" the next car will be a Tesla, too. I'd enjoy that far more than what I have here, which is having paid for FSD multiple times and having gotten pretty much zero value out of it.

Now, I know Tesla's trade-in values can be kind of meh at times, so Tesla could dig deeper into solving that if sold privately with FSD removal (removed from the car and shifted back to the account level), so that the next car bought from Tesla gets FSD reinserted into that car, - again, up until the point where FSD is considered complete.
I think they could probably keep it in future fleets by just allow ppl for 5K to roll it into a next Tesla (possibly removing it from the prior vehicle, although if Tesla thinks that robotaxis are the future, Tesla is going to want all those cars back and they can just put FSD on them and generate 30-50K per car per year or more.
 
  • Informative
Reactions: APotatoGod
IMO it is high to discourage people from buying - they expect people to just subscribe. No deferred revenue (and legal liability) to carry on in the books ... Tesla isn't exactly cash-strapped now unlike even a couple of years back ($22B end of '22 compared to $6B end of '19).
Agreed. I think the honorable think would be to let anyone who bought the software to get a full refund/ switch to a subscription, given they are years away from the vision of the car being able to make money for the owner.

At this time, the feature is about safety and in my experience, FSD beta makes me a better and safer driver (albeit not quite 10x safer).

The car will not be a robotaxi while it's running on FSD beta. Means, it will not generate revenue while not being driven by a human.
 
The stupid thing about FSD purchasing is that Tesla could actually solve this problem. Here's my solution: Just bind FSD to your Tesla at the car level with the option of removing it back out to the account level (one purchase needed per vehicle, of course) up until the point where it's "complete." At that point, it's bound to the vehicle and removed from the account level. That way, as some of us have done already, for right now - when we trade in the car to Tesla without getting what we paid for, we know we get the feature in the next Tesla. This would also encourage consumer loyalty to the Tesla brand, "guarantee" the next car will be a Tesla, too. I'd enjoy that far more than what I have here, which is having paid for FSD multiple times and having gotten pretty much zero value out of it.

Now, I know Tesla's trade-in values can be kind of meh at times, so Tesla could dig deeper into solving that if sold privately with FSD removal (removed from the car and shifted back to the account level), so that the next car bought from Tesla gets FSD reinserted into that car, - again, up until the point where FSD is considered complete.
Now that the car profiles are cloud based, you should be able to, theoretically, use FSD in any car you’re actively driving. Maybe something like that could SOMEWHAT justify the unreasonable software price.
I guess I'm one of the exceptions here, but I am glad that I purchased the FSD option when I bought my 2022 MSLR. At the time, it cost $10k, so a roughly 10% increase in the price of the car by adding the FSD. I realize that as a percentage of the car cost, the current $15k is much higher, especially on the base Model 3s and Ys.

To me, FSD was one of the main reasons I bought a Model S. I love cutting edge stuff, and I really appreciated getting over-the-air updates to see improvements to my car as time went by. Would my car be as enjoyable to me without FSD? Definitely not. The price of FSD, like anything in our capitalistic society, is set by the laws of supply and demand (tempered somewhat in this case by somewhat arbitrary decisions by Elon on pricing). No one is forced to buy FSD, and anyone who does even the smallest amount of research knows that Elon's over-the-top claims and promises are worthless (let the buyer beware).

I don't expect the software to be perfect, and I definitely curse at my car often when it does a really stupid maneuver, but overall it's fun for me to play with, and totally blows people's minds when I demo it for them (even as wonky as it currently is).

I guess it comes down to setting expectations properly. For those who don't take the time to look at Elon's FSD claims history and do some basic research before buying FSD, I'm sure their expectations are WAY too high. Oh well!
It’s also cool to not get the production software features for several months only to possibly get an incremental junk update instead. My girl asks me about new features on her non-FSD Model Y, and I tell her I’m not sure, and would have to look it up, because I don’t have those features on my flagship Refresh Model S with a $10,000 FSD addon.
Just treat Tesla as a Games console. FSDb as a Game Software. I am too old to pass the fun!
Hell of a price for a game software lol
 
  • Funny
Reactions: jebinc
I’m really hoping 11.4.X that we end up getting fixes the massive nerf to the ability to actually engage FSD. On really nicely painted roads with wide lanes it refuses to engage constantly.

Has anyone heard or seen any comments about that getting changed? I’m not sure if they did that due to NHTSA whining or something.
 
I’m really hoping 11.4.X that we end up getting fixes the massive nerf to the ability to actually engage FSD. On really nicely painted roads with wide lanes it refuses to engage constantly.

Has anyone heard or seen any comments about that getting changed? I’m not sure if they did that due to NHTSA whining or something.
What have you done to diagnose, and what steps have you taken to self-resolve the problem?
 
  • Funny
Reactions: Ramphex
Has anyone seen any recent updates on 11.4 release? Seems DBA; where “B” = “Before”.
*cough*


Read what he said...no safety "critical" interventions. Just regular safety interventions, I imagine. Because, you know, it doesn't work.
 
*cough*


Read what he said...no safety "critical" interventions. Just regular safety interventions, I imagine. Because, you know, it doesn't work.
Looks like more junk inbound, albeit late…

IMG_7431.gif
 
I guess I'm one of the exceptions here, but I am glad that I purchased the FSD option when I bought my 2022 MSLR. At the time, it cost $10k, so a roughly 10% increase in the price of the car by adding the FSD. I realize that as a percentage of the car cost, the current $15k is much higher, especially on the base Model 3s and Ys.

To me, FSD was one of the main reasons I bought a Model S. I love cutting edge stuff, and I really appreciated getting over-the-air updates to see improvements to my car as time went by. Would my car be as enjoyable to me without FSD? Definitely not. The price of FSD, like anything in our capitalistic society, is set by the laws of supply and demand (tempered somewhat in this case by somewhat arbitrary decisions by Elon on pricing). No one is forced to buy FSD, and anyone who does even the smallest amount of research knows that Elon's over-the-top claims and promises are worthless (let the buyer beware).

I don't expect the software to be perfect, and I definitely curse at my car often when it does a really stupid maneuver, but overall it's fun for me to play with, and totally blows people's minds when I demo it for them (even as wonky as it currently is).

I guess it comes down to setting expectations properly. For those who don't take the time to look at Elon's FSD claims history and do some basic research before buying FSD, I'm sure their expectations are WAY too high. Oh well!
same. this thread has devolved into a bitch fest and there's no benefit to continuing to subscribe. I'm interested in updates and the improvements as the releases arrive. the last update greatly improved stop sign behavior for me which was welcomed.
 
Most of the AV forum is just entertainment nowadays. Tesla already has fsd in the bag, so now we're just along for the ride, bitching on the way there.
And waiting for the next release…

“In the beginning…” releases would occur every two or three weeks. Now, wide releases seems to be 5-6 months apart...

So, while we wait…
 
I guess I'm one of the exceptions here, but I am glad that I purchased the FSD option when I bought my 2022 MSLR. At the time, it cost $10k, so a roughly 10% increase in the price of the car by adding the FSD. I realize that as a percentage of the car cost, the current $15k is much higher, especially on the base Model 3s and Ys.
...
I don't expect the software to be perfect, and I definitely curse at my car often when it does a really stupid maneuver, but overall it's fun for me to play with, and totally blows people's minds when I demo it for them (even as wonky as it currently is).

I guess it comes down to setting expectations properly. For those who don't take the time to look at Elon's FSD claims history and do some basic research before buying FSD, I'm sure their expectations are WAY too high. Oh well!
It reflects well how I’ve been thinking about it.

I treat system as work in progress. Usefulness and reduction in stress on the road come from using it at times, when it performs well or when even its limited capabilities make my process of driving easier and safer (ADAS / augmented driving). I don’t expect full end to end autonomy as I know FSD is not there yet.

In my everyday drives, the car absolutely keeps making mistakes. New builds bring improvements and regressions. Less traffic definitely improves the behavior but I drive in dense traffic about half of the time. Busy all way stops are a problem - hesitation, slow reaction but most of all lack of ability to ‘read’ other drivers and predict what they will do. So are multilane roundabouts and occasional unprotected left turns. There are also issues not related to other drivers that especially annoy me, like when my car changes lane before the intersection and ends up with its rear stuck two feet into the old lane and at the same time there is plenty of space ahead of it.

I’ve learned to expect those mistakes and I intervene either by gentle press of accelerator, sometimes continued press through a situation. I also have to adjust speed using the wheel, to keep the car flowing with the traffic. When I’ve someone tailgating me in a potentially problematic situation, I proactively hover foot over accelerator or disengage in advance.

Despite all of those frustrations I still choose to have it enabled most of the time on most of my drives as I find it adds up to my overall awareness of a situation on the road and helps me be a safer driver.

However, I don’t think current price, or 10k I paid for it, is a good value for most people. I also strongly believe the transfer of FSD beta on car upgrade should be allowed. In my eyes, Tesla overpromised capabilities and underestimated the rate of progress. They should at least compensate for it by allowing transfer of any FSD beta license to a new car at no cost.