That scam drove me to and from work again today. Sixty miles round trip of city/rural/interstate roads. Darn you, Tesla! Because of you, my driving skills are getting pretty rusty.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
$12K for FSD is insane
- Thread starter bjrosen
- Start date
I’d love to see some data on that. Retention rate etc.. My guess is it’s less $$. Eg my guess is (right now) many people would cancel the rental after it doesn’t live up to their expectations.
Anyway I like the rental idea. They can crunch the #s and determine if people are finding the features that come with it valuable or not.
dbldwn02
Active Member
Cool. Change out EAP for TACC. Still hits the brakes, 100% of the time.
stopcrazypp
Well-Known Member
I rolled my eyes when I saw the media make a big deal about the "jam" in the system. I watched the video and it was literally seconds, caused by there being some inefficiency in how the cars queue and park when loading/unloading passengers. It's nothing compared to a typical traffic jam, and inconsequential to the overall trip length.
I also saw a lot of people make snark, implying that a subway train was a much better solution. Putting aside that if Las Vegas went with a subway, not only would it not be ready for years and cost a ton more, subways also have similar delays. As someone that had ridden the subway daily for years, it's very common to have such delays when there are different trains stacked up waiting to enter the station. I wonder if the people making snark ever rode in a subway daily. It's going to happen to any system with many trains back to back where the station stop time is not fixed (the subway I rode in was manually driven, and the driver controls the stop time, so it's not fixed given the driver waits for everyone to board before closing the doors and departing; people can also hold the doors to prevent car from departing).
The Boring company certainly can improve their station design and loading/unloading procedures, but for a first effort and on time and on budget (which practically never happens with subways in the US), it's doing quite well.
Knightshade
Well-Known Member
Cool.
The manual explicitly addresses that too.
Your description of situation:
Model 3 owners manual said:
and
Model 3 owners manual said:
#RTFM
Sweeping right turn might not be quite the same as a winding road with sharp curves, but I don't know why the system would even be allowed to operate outside its apparent ODD. Well no, I can think of reasons why...
Another example of other companies being far more conservative in their deployment while Tesla just throws everything out there and puts the onus on the driver. Ford's Blue Cruise was heavily criticized for not functioning in curves, but I'd bet you it has very similar capabilities and Ford is just conservative in their ODD until it'll function 100% in these scenarios. Tesla sells the image of Autopilot, then it's on you to discover where it does and doesn't work.
Another example of other companies being far more conservative in their deployment while Tesla just throws everything out there and puts the onus on the driver. Ford's Blue Cruise was heavily criticized for not functioning in curves, but I'd bet you it has very similar capabilities and Ford is just conservative in their ODD until it'll function 100% in these scenarios. Tesla sells the image of Autopilot, then it's on you to discover where it does and doesn't work.
dbldwn02
Active Member
Look out!! Not sharp curves ahead.
So, one day it'll be able to FSD around that same, not sharp, curve...using already installed hardware that's capable of Tesla Taxi, since all Teslas sold since 2016 have hardware capable of FSD. RIIIIGHT....#GTFOH
Knightshade
Well-Known Member
Which part of this, specifically, do you not understand?
Model 3 owners manual said:
FSD, of course, is running an entirely different set of code than the basic AP/TACC you're using.
You don't have the beta, so you have no experience with this- but its behavior on such roads is very, very different.
That's not to say perfect currently... but unlike the code you're running it actually is meant to understand what oncoming traffic is.
It's certainly gotten vastly more capable driving on local roads than before I got the beta code (and improved over time as beta versions have revved).
Again though I've already mentioned, not that you seem to read half of the content of posts you reply to, I don't believe the current HW is sufficient for autonomous robotaxis. But it certainly seems capable of being a quite decent L2 driver aid on local roads- certainly far more capable than what basic AP owners have today.
dbldwn02
Active Member
It just started with you saying it doesn't slam on the brakes. I just said it does and so does the owners manual, as you highlighted.
I know FSD is level 2 and will probably only ever be that, even though many bought something that was advertised as more than level 2.
Knightshade
Well-Known Member
Braking, and "slams on the brakes" are quite different.
As I mentioned, when someone with an accelerometer actually measured their "phantom braking" is was 0.2-0.3g... about the same as regen.
"slams on brakes" is FAR harder deceleration than that.
It only feels hard to you because it's unexpected. Humans are notoriously terrible measuring devices.
Indeed, the fact Tesla changed the FSD product to explicitly only promise L2 back in March 2019 has been pointed out by any number of folks as having a reason, myself included.[
Will it be a real L2 system, or will it be L2 in name. Eg like auto-wipers/highbeams. If it’s real (as opposed to a marketing trick) I’ll buy it. I’d like to be able to turn it on and be able adjust the radio etc. I’m going to adjust the radio no matter what and I need to be safer and more relaxed with it engaged as opposed to not engaged.
Right now it’s the opposite. I turn on fsd and it requires my total attention. I turn it off to adjust the radio.
ucmndd
Well-Known Member
This is the best description of current reality I’ve seen so far. Almost all of Tesla’s driver “assistance” features currently demand more from the driver than not using them at all.
Only the True Believers can convince themselves otherwise. You just gotta have faith.
But, I mean, not TOO much faith. Because that sugar will kill someone. Maybe you.
It doesn’t matter. The human is the one that needs to like the thing and if it some way it overly bothers the human(s) they will turn it off. And complain. And tell their friends not to buy it. Or if they work in the press, make some article about how it upset them etc…
The unexplained & abrupt shedding of speed, on the highway it does right now is a problem. It doesn’t matter if it actually slammed on the breaks or did some sort of regen. It’s a significant defect.
Knightshade
Well-Known Member
The fact owners continue to rack up billions of miles with the system on makes it clear outside of an overly-sensitive few, this is not happening.
Apparently most owners find it more helpful than not.
Again, clearly nonsense given the fact it remains heavily used.
It makes my wonder if you've ever even used it properly?
AP on highway trips significantly reduces driver effort and fatigue- with tons of owners reporting arriving far better relaxed and rested at the end of long trips than they did in legacy vehicles. Some in this very thread.
If people found it worse than not having it they wouldn't be using it in droves.
Those words literally don't make sense in that order.
Neither wipers nor high beams, auto or otherwise, would have any classification in the SAE J3016 system, as neither has anything to do with the dynamic driving task.
And an actual driver assist system is either L2 or it's not, there's no such concept as "fake L2" and the classification is pretty clear.
That's how AP already works. And has for years. You still need to pay attention and be prepared to take over, which is in the definition of L2 though.
You appear to be confusing the beta FSD with basic AP.
A lot of that going around in this thread.
Last edited:
Fair point. For me I think it’s a regression. I wasn’t paying specific attention but believe AP worked for me originally. And now after fsd-beta & vision-only it has an issue. I suspect vision-only because others have posted that they’ve owned older & newer cars and have said the newer vision-only ones have the phantom breaking problems.
If these are historical billions of miles, or miles being racked up by radar enabled cars, then the volume of radar-enabled data might masking that there’s an issue. It would be good know how it’s going on the newer radar-less cars. Have they broken that data out? Since they have both types of data they could compare relative usage rates between the before & after cars.
ucmndd
Well-Known Member
I’m sure it does. You seem to be deep deep deep in the depths of confirmation bias.
I have driven 135,000 miles over 5 years in my Model S. How about you? I think I’m pretty familiar with the ins and outs of autopilot at this point. I’ve been at the table for the entire evolution of the “FSD” product since before the beginning.
It is a textbook Silicon Valley fake it til you make it scam. The safety statistics you blindly accept at face value to support your narrative don’t stand up to the most basic scrutiny. When directly confronted about that you pretend my posts don’t exist. Your visible contortions to paint everything in the most favorable light possible are downright dishonest.
So yeah, I’m sure you do wonder if maybe I’m “holding it wrong”.
Last edited:
Knightshade
Well-Known Member
FWIW I agree entirely with that.
Having driven both I believe the vision-only highway system has not yet caught up to parity with the former vision+radar system for example... (and we have clear admission of this from Tesla themselves- given the max speed is 80 mph instead of 90, and min follow distance is 2 instead of 1)
This isn't surprising... AP2 took a good while to get to parity with AP1 as well.
Where things will get significantly more interesting is when FSDBeta V11 comes out, since it will allegedly finally integrate city and highway driving to use the same single stack... (versus what is largely pretty old highway code at this point- and even the city code will be changing a fair bit)
Tesla will need to retrain their NNs largely from scratch because it's a pretty massive reworking of the system... so there might well be additional regression before we see large improvement.
Essentially each time Tesla has been through this they've hit some local maximum in what the system was capable of, and the rewrite initially was worse, but capable of eventually improving far more than the previous design.
some more discussion on that here if you're interested in a bit deeper dive
FSD AP improvements in upcoming v11 from Lex Fridman interview
The recent Lex Fridman interview of Elon had some interesting tidbits about the next major FSD beta version (https://www.youtube.com/watch?v=DxREm3s1scA). First, something simple. Tesla’s AI inference chip (each car has two of them) has an image signal processor that processes each image frame...
teslamotorsclub.com
To my knowledge they haven't-- that said, the data they publish each quarter is just for that quarter-- And we can guesstimate roughly how many cars right now are vision vs vision+radar
Vision only should be roughly 350k or so rough ballpark (Freemont switching over in May, Shanghai in November)
Total production of radar+vision cars is basically 2017 through 2020 in full plus remaining ~600k in 2021 so total of about 1.81 million cars.... so right now it's pretty heavily weighted to radar in the data set.
Of about 2.15 million cars, 350k are vision only, so roughly 16.28% vision to 83.72% vision in the fleet.... or about 6 radar cars for every 1 vision car very roughly.
That's not a TON... but over 16% of fleet is enough I'd think we'd start seeing WORSE safety results if vision was actually more dangerous-- rather than just maybe a little more annoying with unneeded braking. Instead we continue to see safety results improve... so either vision isn't actually worse for safety, or the software is improving faster than the vision % of the fleet can bring down the #s.
But... We expect Tesla to produce something like 1.5ish million vision only cars in 2022. So by end of this year vision only should be a small majority of the overall fleet.... if safety #s continue to not get worse it'll be pretty tough to claim vision is less safe, even if it's less smooth.
(FWIW I'm discounting AP1 cars here, I don't actually know if they're included in the data set, but they're an increasingly tiny % of the fleet either way)
Of course everyone expects Tesla to be significantly improving the vision code during this year- especially as they go to single stack... so should be interesting to see what changes we see in the data they do publish.
Last edited:
Knightshade
Well-Known Member
I do not think those words mean what you think they mean
Though I admit you and the other guys admissions you use AP on local roads does confirm the point that claiming "All those AP miles are ONLY highway miles so we can't compare them to local miles!" is an increasingly silly argument to try making. Especially when it's the guys admitting they don't only use it on highways trying to make that argument.
More words you don't appear to be using correctly.
That's weird since I directly replied to a bunch of em.
Maybe check your glasses?
Still waiting for you to provide any stats, of any kind, showing AP is less safe of course....rather than just continuing to stamp your feet and insist you don't believe every stat we DO have available showing it improves safety, because TACC slows down in a spot the owners manual warns you it might slow down in
I've explicitly stated I don't believe the company is capable of delivering the FSD system they promised everyone who bought it from launch until they vastly reduced the promised features in 2019 on existing HW.... but I'm "painting everything in the most favorable light possible"?
I agree someone is being downright dishonest, but it ain't me.
With it not being tied to the owner, but the actual car that will be out of service by the time FSD is actually FSD, paying anything extra for it is absurd.
If you’re paying for what it might become, it should be tied to your Tesla account and every Tesla you ever own in the future. Any other arrangement is just paying Teslas development costs for them.
If you’re paying for what it might become, it should be tied to your Tesla account and every Tesla you ever own in the future. Any other arrangement is just paying Teslas development costs for them.
CDennis100
Member
Hahah…we need to rename this post something to the effect of, “watch adults acting like kids debating”. We’ve sooo gotten away from topic.
Here’s my take. I love and use basic AP features. It’s good enough for me and meets my expectations. I use as I should (highway) and it allows me a few seconds to take my eyes off the road and look at a hot chick driving next to me, or to text and drive, which those that do without automation are idiots.
With a net worth over 2 million I can easily afford FSD but…
It’s not worth it. Not at 12k, not at 10k. I’d pay maybe 2k for the features it does offer that I’d use. I’ll be damned if I pay 7, 10k or more. Not worth it. My personal opinion is true FSD is at least 4-6 years away and I won’t own my current M3 then, and since I can’t transfer it….it’s a NFSD for me!
Similar threads
