FSD v12.x (end to end AI)

[Dewg]

Like @Izolman I also live in Mass and the foliage delay he refers to at Stop signs is real since the car takes an extra 2-4 seconds to creep because of the foliage. Once I can see potential cross traffic leaning forward the car always creeps another 2-3 feet for the B-pillar camera before proceeding. No foliage intersections the delay is cut by almost 50%. With the long delay I have also increased my use of the accelerator. And sometimes FSD never moves.
I sometimes wonder how much Tesla is tracking accelerator usage.

Today had wheel of death 3 times

1. Auto Max switch to highway stack. I've had this a bunch of times.
2. Heavy rain on the highway. After about 30 minutes of working fine I got the wheel. Up until then FSD worked admirably in bad conditions.
3. Exiting highway during rain storm. Auto Max did not slow down enough on the exit ramp with an abvious large puddle of water. Spray from puddle covered the windshield cameras. I thought about taking over but wanted to see what would happen.

Well that just sucks... Guess I'd be fitting my car with gallons of Roundup and spray nozzles, and deforesting the intersections around there.

 

[mgs333]

Prediction: 12.4 in June. lol

whoa, really going out on a limb there!

 

[Todd Burch]

How about Robot Driver in Training?

I got a bumper sticker that just says:

"I creep. A lot."

People (women in particular) seem to keep their distance--I'm assuming so that if the car makes any strange maneuvers there won't be a collision.

 

[rlsd]

I got a bumper sticker that just says:

"I creep. A lot."

People (women in particular) seem to keep their distance--I'm assuming so that if the car makes any strange maneuvers there won't be a collision.

Just say "I have 8 cameras".

 

[SSDTester#1]

Are there any OGs on 2024.3.25, got pushed to the latest version in this forum?

Still no sign of any updates in my case...

Still on 3.25 here

 

[aronth5]

Well that just sucks... Guess I'd be fitting my car with gallons of Roundup and spray nozzles, and deforesting the intersections around there.

Please start with this intersection You may need a shovel too!
Oh and I'll provide the beer

 

[zoomer0056]

Yes there is ONLY ONE direction the car can go. So a signal to say you are going the ONLY direction you can go is redundant.

Interesting take. More than half of all people don't use turn signals. I've always wondered why. This post answers one particular case. "No need to, can't you tell we're merging?" I erroneously thought it was some sort of entitlement move. Or is that the same thing. TMC is continously pushing open the curtains of life's little mysteries.

 

[bimmer88]

Anyone else think that FSD should have a box you can check that says, “Do not stay behind 18 wheelers”?

 

[AlanSubie4Life]

More than half of all people don't use turn signals.

Evidence needed. People like to make this claim; I have never seen it backed up. Obviously 100% of drivers don’t use their turn signals sometimes. But that’s of course not the claim being made here.

 

[DanCar]

Evidence needed. People like to make this claim; I have never seen it backed up. Obviously 100% of drivers don’t use their turn signals sometimes. But that’s of course not the claim being made here.

Varies by state. California: seems police stop enforcing and people don't bother with turn signals much. Arizona: more people use turn signals according to my experience. From video's of Oklahoma, lots of people use turn signal.

 

[stopcrazypp]

If you turn on your blinker but no one’s around to see it, does it actually blink?

It's kind of like the stop sign logic previously. If there were no cars in the intersection, it would roll the stop (like most human drivers do). Of course the law doesn't allow for that.

For the record, I stop at stop signs and also signal even if there are no apparent cars around. Better safe than sorry given there are no exceptions in the law (a cop hiding in a bush or corner can easily write a ticket if they wanted to).

 

[AlanSubie4Life]

California: seems police stop enforcing and people don't bother with turn signals much.

I’ve paid attention to this before (recently!) and it is a remarkably high percentage of people who use turn signals when needed to signal their intent to turn. Certainly way more than half! Just good drivers in my area I guess.

I am curious what the actual numbers are.

 

[swedge]

Well that just sucks... Guess I'd be fitting my car with gallons of Roundup and spray nozzles, and deforesting the intersections around there.

I have a right turn at a stop sign to leave our neighborhood. It seems to be that FSD takes far longer to proceed if no one is coming from the left than is does is there is someone coming but far enough away to safely proceed. It is as if a moving car is quicker to see than no car. This sort of makes sense, but is odd. Nothing is hard to see than something?

 

[swedge]

Still on 3.25 here

I was hopeful, but no, I'm still on 2024.3.25 as well.

We purchased FSD with the MY in late 2021, and our first FSD update was 2022.12.3.20 / fsd 10.12.2 on 6/9/22, delayed due to deer strike repair parts delay and FSD beta slow rollout.

 

[OldManCan]

Not all of 2024.3.25 is testers. Last time I checked 15 of cars still on that version only around 50% were testers. The rest appear to be updating to 2024.14.X.

I think you will find a View link on each line which will bring up details on that particular vehicle. Lower on that detail page is the update history. Looking at that, I expect you will find that these vehicles first got FSD after 5/22/2023, version 2023.12.10, FSD 11.3.6 or later. I think that was when FSD was made available to anyone who paid or subscribed. Since then, every main branch update has included FSD. But some of us had received FSD 10.3.6 with 2022.45.15 as much as two months before on 4/8/2023. These folks are what I call the testers, and we are being left behind on 2024.3.25.

This is how it appears at this point. I have checked 20 cases like your highlights (random among but with on 50,000 miles and in the US. to increase the likelihood of finding a tester) and not one of these appeared to be a tester. I would check yours for you, but I ran out of clicks earlier today trying to find one. ;-)

Very informative. Thank you! I'm a FSD monthly subscriber on a 2023 MYLR with EAP. Certainly not an early FSD tester but hopefully they will include me in the tester group. The fact that I have not yet been updated to 2024.14.X is promising! Fingers crossed!

 

[Dewg]

I have a right turn at a stop sign to leave our neighborhood. It seems to be that FSD takes far longer to proceed if no one is coming from the left than is does is there is someone coming but far enough away to safely proceed. It is as if a moving car is quicker to see than no car. This sort of makes sense, but is odd. Nothing is hard to see than something?

My guess is that its straining to see something vs not see something.

For example, if you're people watching, it's a very different level of concentration and brain engagement than if you're looking out for a specific person in a crowd. You might see someone wave at you, and you think it's them and wave back only to find out it wasn't them.

So the delay in FSDS may be the equivalent of you squinting to see "is that a car coming?"

 

[swedge]

Very informative. Thank you! I'm a FSD monthly subscriber on a 2023 MYLR with EAP. Certainly not an early FSD tester but hopefully they will include me in the tester group. The fact that I have not yet been updated to 2024.14.X is promising! Fingers crossed!

It is impossible to know (without info from Telsa), but I think that no testers have been added since Dec of 2023. If you look up your updates, you are not a tester if you have had multiple updates of FSD 11.4.4 . When that was released to the main branch, testers were on 11.4.8, and there is no going backwards.

Tester or no, updates will come when they come. Main branch gets feature updates which testers do not get, while testers get buggy FSD updates which the main branch does not. Every one in a while they merge, and confusion reigns. Unless one looks up their own update history...

 

[joelkfla]

Evidence needed. People like to make this claim; I have never seen it backed up. Obviously 100% of drivers don’t use their turn signals sometimes. But that’s of course not the claim being made here.

I must be in the 0.001% then. It was drilled into me in bus driver CDL training. And there was no reason to risk points on a CDL.

 

[swedge]

To which I say, "if you're not going to pay attention, don't drive a effin' car!"

Every time I turn on FSD I listen for the double chime and glance down at the screen to verify that it's engaged. Every time I turn on TACC I listen for the chime, check the set speed on the screen and hold my foot over the accelerator until I've verified that it's engaged. Every time I use cruise control on our other car I do the same.

When I'm using FSD I'm watching the road with my hands next to the wheel so I can take over if it does something stupid. If it disengages it gives an audio signal and the screen changes and oh, I happen to notice as the car starts to drift out of its lane and I take over immediately.

Now what they've done is make it so you can't use TACC at all but you can still disengage FSD just as easily. One can argue that it's even more dangerous since it will suddenly stop in the flow of traffic and since speed control is so horrible with 12.3 you just assume it's messing up anyway.

Sorry, I have no sympathy for those who don't pay attention and can't figure out what the car is doing.

I've said this before: I think that steering wheel torque as a disengagement cue was a mistake. A contact sensor on the wheel would be OK, I think, but torque from the driver should mean "change turning radius".

In an airplane autopilot, pushing the controls overrides the autopilot and changes the plane's attitude, and when pressure is released, the autopilot returns to the set heading and altitude. Some cars work the same way. But since the power steering senses torque already, Tesla saw a "free" way to sense attention. But this created mode discontinuities: In manual mode, an increase of torque on the steering wheel causes change in turning radius, while in autopilot, a slight torque means you are awake, but more torque means disengage autopilot. Disengagement this way causes swerving, as does stalk or brake commands disengagement during a turn. In either of these cases, disengagement causes abrupt, and usually inappropriate steering changes.

If instead, steering wheel torque merely overpowered the autopilot momentarily, one could dodge a pothole or other traffic, and letting go of the wheel would let the autopilot return to the lane center smoothly.

But Tesla has chosen the "free" sensor approach. Perhaps, if the interior vision proves effective, they will eventually rely on that rather than torque.

Till then, it is what it is.

 

[AlanSubie4Life]

I must be in the 0.001% then. It was drilled into me in bus driver CDL training.

Even I (who thinks most people use their turn signals, because laws and such and general decency and desire to not get in an accident) believe that every driver has failed to use their turn signal where required at least once in their driving career.