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.
I think Tesla should keep highway stack that way for a while. They should only fix the early, late exit problem that was introduced in v12.
There are many things in city street stack that should be fixed first.
Man, once again, you are partially correct.
V11 IS STILL RUNNING ON HIGHWAYS...
V12 will make it to highways.
Listen to the other posts, they prefer city over highway.
V12 vs. V11
It's coming..
That's what she said...
 
  • Funny
Reactions: Hiline
Anecdotally I can say I definitely have fewer disengagements. In fact I had my first disengagements this afternoon. For the first one I was driving in the left lane on the highway and the car signaled to move to the right lane. Normally would have been totally appropriate except there was a tow truck on the side of the road (with lights on) helping a motorist so i delayed the lane change until after the tow truck.

The second was when it started to enter an inappropriate turn lane. It was a block before it needed to turn and it was also behind a trailer so I’m not sure if it would have caught itself but there was a car right behind me so I prophylactically took over to prevent any issues.

so, yeah… exactly like a human driver!
Hi, Sleepydoc --

Thanks for the response! It definitely seems like 12.3 is a real advance over prior versions. What I find surprising is the difference between critical/non-critical city disengagements as reported by the tracker. At 1 every time miles, non-critical is well within prior experience; but we're up to 1 critical every 417 miles, and I think the two (!) there came within the first couple of hundred, so 12.3 has gone some 600 miles without a critical disengagement. As those following know, individual experience with FSD is highly variable, ranging from "works like a charm" to "tried to kill me immediately". I'm curious to see if people are perceiving a real difference in the safety profile.

Yours,
RP
 
I never said it should use the minimum, I suggested 200-300 feet.

500+ feet is just way too far
You continue to cite distances when it is pretty clear it is not about distance (as the driving manual makes clear).

For example I signal right about 500-600 feet in advance here when traveling 55-60mph. And so does v12. Very appropriate (and the factors here make it clear why). It’s not about the distance.
You shouldn't signal a turn option ahead. It confuses people.)
Agreed. Ambiguity is not good, though of course sometimes unavoidable when trading off multiple factors. Ideally the car takes into account (amongst many others, like speed of course):
1) intervening turns.
2) presence of car behind and following distance
3) presence of car at intervening turn

Before deciding whether a relatively early potentially ambiguous signal is ok.

All goes into a network and magically emerges with a nice low-stress signal we can all agree on.
 
Last edited:
  • Like
Reactions: sleepydoc
Robotaxi is never coming. It was promised "end of year" in 2019. You only make a prediction like that when you have something working. Then it was pushed back, and now we hear nothing about it.

Elons first "prediction" was a complete and utter lie. Maybe they've made progress, but I doubt it.
What does Any of your repeat complaining about robotaxi have to do with this thread topic. Take it to the drive, roundabout, how soon is to soon to stop thread.
 
V12 is obnoxiously slow on the first drives I've had with it tonight. Turns are a bit smoother but I don't know how much of that is just do to a slower speed. Braking and acceleration need some work. It goes from 0-20 in a blink.

Lane selection and changes were fantastic though and leads me to think this is the right way forward.
Reviewing the progress of FSD over the years to now be to the point that the Primary user complaint is not speeding Has to make even the doubters a bit happy. I say round of applause for 12.3 and beyond. Rather impressive I must say.
 
Seems so far that only the Model S seems to be curbing the wheels? Are there any 3/Ys that have curbed the wheels?

Model Y, no curbed wheels, but it comes a lot closer to some curbs than I am comfortable with.

V12 logic is AI predicts the correct speed based on, well, anything and everything.

(It won’t exceed your target if you have one set but otherwise the target is irrelevant)

I've had V12.3 ignore my max set speed by 10 mph twice now, same place both times. The speed limit is 25, and 12.3 just tools along at 35 until I hit the brake. Scrolling down the speed has no effect.
 
  • Informative
Reactions: primedive
Ten years is eons and eons given the current rate of AI technology advancement. It’s very nonlinear. Robotaxi within 2 years wouldn’t surprise me at all.
It will be interesting to see where they draw the line as far as "Robotaxi" is concerned. If it means that the car can drive all by itself (with no one in it) to pick up a new "passenger", that will require at least L4-level autonomy, probably L5-level. What sort of reliability (distance between safety-critical errors and/or self-disengagements, where it pulls over and can't continue) will be required for this? Once every 100k miles? That's a LOT of orders of magnitude beyond the current state. In my experience v12.3 still makes a major mistake every few miles. So if v13 is 10x better than v12, and v14 is 10x better than v13, and v15 is 10x better than v14, and so on, then by v16 or v17 they will finally be approaching the needed reliability ballpark. That is at least several years off, even with exponential improvement.

Of course, this is just my prediction based on current and past trends, but still taking exponential progress into account. As an owner of two FSD-enabled cars, I would love to be proven wrong!
 
  • Like
Reactions: AlanSubie4Life
The speed limit is 25, and 12.3 just tools along at 35 until I hit the brake. Scrolling down the speed has no effect.
I could be wrong, it is not like I have exhaustively tested everything, but for the scroll down to robustly work in this case, you’d need to turn off Automatic Speed Control.

Again, that does not help with going too slow (both modes are crippled), but it should allow a relatively quick reduction in speed to your limit. I’ve seen that, I think.

Robotaxi within 2 years wouldn’t surprise me at all.
I don’t see it with current hardware. It’s just too difficult a problem, too many orders of magnitude improvement needed, and so I don’t think the hardware has what it needs to pull it off.

I too would love to be proven wrong. I am not sure anyone will be doing Robotaxis in a couple years. Including Waymo!

We may not yet have passed the Valley of Disappointment.

On the upside, I am quite optimistic I’ll have a decently functioning driver assist in a couple years, assuming someone does not total my car meanwhile (better increase my insurance on FSD value I guess - I almost (but not quite) would miss it?). I do think it will take a year or two for them to smooth out the speed and stopping issues - then it will be great. Maybe the speed is just a big screwup and that’ll be fixed on the next release. But if not expect a multi-month process of improvements. A good model to use is autowipers and auto high beams.
 
Last edited:
  • Like
Reactions: Hiline
A bit of an important discovery tonight for those that think Auto Max drives too slowly: the right scroll wheel still works while on Auto.

Driving past a known sensitive speed trap, and Auto Max chose to go 27 in a 20 MPH. Out of habit, I scrolled down on the right scroll wheel, and after about a half a second delay, it slowed down to 24. Another scroll down, and it's down to 20. Another scroll up, and it's up to 24.

So yeah, you don't need to press the accelerator and hope it keeps up. Scroll up on the right scroll wheel a few times.

I don't know if I'm just the first person to try this, or whether others have done it already and I've missed it.
I I found that the response to the scroll wheel speed adjustment works but is extremely delayed and laggy. I think this is a byproduct of 12.3 generally making less abrupt speed adjustments when it knows or believes it's in a different speed zone (I'm a little surprised because one of the recalls last year forced the speed zone adjustment to be much more abrupt).

I posted in my initial report on 12.3, that I feel I've lost the ability to quickly adjust speed up or down. But I agree it basically functions - in both manual and Auto speed offset modes.
 
If Josh gets killed by V12 on his first drive and slows down rollouts in the future...
I will miss Josh more than future updates...
If you don't hear from me by the end of the night..
Dibs on your referral points!
Oh, and any cool accessories that aren't contaminated with post-accident Josh ick...
I guess I should say hope it goes okay, be careful out there blah blah blah (and just leave a note on your desk about my referral points)
 
I will add that 12.3 in Laguna Beack did get a little closer to parked cars on the road during heavy traffic. The ultrasonics in my Model S were showing partially red as the car passed the parked cars. Of course with tight streets, maybe version 11 might have done the same. I might have driven a little further from them.
 
  • Informative
Reactions: JB47394
Highway and highway ramps.

It took a really long ramp and maintained 45mph until it was already in the highway then sped up to 70mph. Very dangerous without intervention. It had a similar issue with a very long exit ramp where it flew at 60mph until it realized and braked for standing traffic at a light.

V12 in the city and roads is great. Slower than usual but really great.
Highway stack needs e2e love
This is likely switching to the v11 stack. Hope Tesla implements v12 for highways in the next few updates. Still not to likely and could be a while going by how long it took to get the highway stack on FSD to begin with.
 
  • Like
Reactions: FSDtester#1