I found no noticeable difference before and after the recall with either of my cars. Much ado about nothing for me.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
New AutoPilot is horrible after update
- Thread starter Tezla1
- Start date
airborne spoon
Active Member
jasonh4451
Member
FSD was smoother and even more human like in my first drive today. Did not notice any difference with nags.
Handsome Jack
Member
That's actually a very good description how exactly it works now! (on my car, before the update)
And I find it as a good compromise, i.e. to write an essay on the phone while driving is huge NO! No L2 system should allow it, but 3-5 seconds off the wheel (while glancing on the road and being aware of the situation, where it's not dangerous to do so) to choose your next podcast is OK.
But the largest question, is the newest update changed that? Like 2 second not looking on the road causes red alert and banning from autopilot or forcing you to jiggle steering wheel each 5 seconds no matter the situation, and if you didn't jiggle within 1 second - ban for a week?
It clearly has not gotten that bad for me… Can still go 2 minutes or so between nags on highway
sublimejackman
Member
Context does matter. So here's a little:
The problem the NHTSA has with AP is almost exclusively related to it being engaged on roads other than limited use highways (backroads, cities, exit ramps, etc.) and its confidence inspiring marketing. Tesla could have addressed these issues five years ago by further limiting where and in what conditions AP could be used and/or by adding some sort of safety check when approaching interections.
Tesla chose to do niether of these things. Instead they doubled down on automation, AI and that AP is "safer than a human driver" (make extraordinary claims, you need extraordinary evidence). Which I guess could be true in certain situations on limited use highways, but almost certainly isn't true in adverse conditions or when approaching intersections.
So between the the first page of the forum here and Reddit containing ways to "defeat nags" and the Tesla CEO bragging about how the "cars can drive themselves", here we are.
You don't like the nags, stop trying to defeat them and drive safely. This isn't to mention the obvious: STOP BOASTING ABOUT TEXTING WHEN DRIVING PUBLICLY. CNN can read forums too.
People are dying when AP is engaged and at a rate that other car makers aren't seeing when their similar systems are engaged. How about we stop throwing gasoline on the fire?
Last edited:
Citations needed for both of these claims.
In particular I strongly disagree with the first one. Most AP usage (in terms of usage time and miles) is on highway, and Tesla has previously had issues with handling stopped vehicles (fire trucks) at highway speeds. NHTSA obviously cares about highway driving and is going to be concerned about distracted drivers (as they should be)
sublimejackman
Member
Sigh... I already spent a week combing through the raw data and posted it elsewhere... here's the link to that interaction.
I probably went too far saying "almost exclusively". I should have said "a large concern for the NHTSA"
The TLDR is other than when AP plows into emgerncy vehicles, the majority of the collisions of concern to the NHTSA were on backroads, in snow, at intersections, stop lights, heavy rain, fog... etc. Everyone mostly agrees that AP works pretty well on limited use highways, when the driver is paying attention. It's the other roads. Hell a pedestrain was hit and killed earlier this year.
Remember somone needs to file legal action for the NHTSA to respond. They are reactionary, not precautionary. The reaction this time was to lawsuits, mostly regarding people getting injured or killed on backroads.
GM and Subaru also telemetricly report their level 2 related collisions to the NHTSA. Between the two of them they reported fewer than 100, Tesla reported over 900 this year!
Just because we all want it to be safe, doesn't mean it is. Just because a computer does it, doesn't mean its safe. Just because you can use AP safely doesn't mean its safe. Any Jane or Jim who can get a $30k auto loan can buy one of these now, and they do. And many use it recklessly.
Raw data: https://static.nhtsa.gov/odi/ffdd/sgo-2021-01/SGO-2021-01_Incident_Reports_ADAS.csv
Other thread: Tesla recalls 2 million vehicles to limit use of Autopilot
Last edited:
gtae07
Member
I got many nags on the interstate yesterday (just the "apply torque message") even though I was doing it right (applying pressure, eyes on the road). Sometimes just moving my hand to hit the turn signal would do it.
But then other times I can go hands off and stretch or whatever and it doesn't say anything. I don't get it.
It did freak out one time when a guy drifted over but I might have disengaged it myself.
But then other times I can go hands off and stretch or whatever and it doesn't say anything. I don't get it.
It did freak out one time when a guy drifted over but I might have disengaged it myself.
Do they report telemetrically for all of their ADAS capable vehicles or only some of them? Only people with an active On*Star subscription?
sublimejackman
Member
That's a great question and I'm not 100% sure. The NHTSA is a nebulous backhole if you want to sure up the data. Also it's probably only GM and Subaru cars built since 2016. Not going as far back or as large of a fleet size as Tesla.
But I think the key distinction here is 1. Where these ADAS systems are being used? and 2. Do these collisions result injury and/or death?
In 2023, there have been about 200 collisions reported by Tesla, many with severe injury and/or death, occuring in places and conditions we would all probably agree is inappropriate for AP. This just isn't something the raw data is showing for other car companies.
This is why I don't think the 2023.44.30.4 is the complete recall. Both Tesla and the NHTSA mentioned that AP use would be further restricted and/or additional nags would be put in place around intersections and back roads. I have 2023.44.30.4 and haven't noticed anything new with regards to use on backroads or any additional nags around intersections.
Last edited:
But we know the data is incomplete for all other OEMs, so we really can't make that kind of comparison. (We also don't know miles travelled on ADAS, or even # of vehicles with an active ADAS system to try to adjust the numbers for population/usage.) The data is literally worthless for comparing collision rates between OEMs. (Which NHTSA specifically calls out.)
But yet Tesla, and NHTSA, specifically say that this version resolves the recall.
Have you tried approaching an intersection without hands on the wheel greater than a minute before you get to the intersection? Did you commonly do that before the update to know if it is different?
This. This is a huge factor. Tesla has a big fleet of vehicles with automation, and if people generally like AP they are likely to use it more. Accident rate, not absolute count, is the only fair way to assess
To make a point I am just going to guess at numbers since we don't have actuals for anyone but Tesla:
- Tesla ~2 million vehicles with AP reported 900 collisions
- GM&Subaru: ~200k vehicles with ADAS reported 100 collisions.
- As of the end of 2022 GM had only built 40k vehicles with SuperCruise. (I think that 5x of that in one year is being generous.)
We just don't have enough data to make valid conclusions like this.
sublimejackman
Member
Yeah. Probably a bad move on my part to directly compare it. Tesla is reporting via telemetrics regularly and for the whole fleet, something the other OEMs aren't doing. But my point still stands, AP is being used in some sketchy conditions (to say the least) and people have died. Probaby best for the Tesla CEO to walk back their claims a bit and put further restrictions on its use in these conditions.
They said that it did for some models but not all. I don't have an interior cabin camera but I do have HW3. So maybe I'm just not in the recall yet?
I have approached an intersection, with my hands on the wheel. If I hadn't slammed on the brakes (I looked beforehand, no one was behind me) it would have just blown the light without warning. Bahavior that could be changed fairly easily if Tesla was really focused on using automation to increase safety (and years ago). Not to say give us EAP for free but it could do the red beep beep and stopped the car if I hadn't and counted as a suspension strike.
Last edited:
sublimejackman
Member
Yeah I agree, I'll stop making these direct comparisons.
My conclusion still stands. Many Tesla drivers abuse AP, use it unsafely, and use AP in conditions it is not equipped to handle. That coupled with the company CEO making repeated claims that the cars are able to safely drive themselves have lead to hundreds of collisions and multiple deaths.
stopcrazypp
Well-Known Member
I made a similar point in the other thread. Subaru didn't even have an L2 system until the 2020 model year (the standing order only applies to L2 vehicles).
Tesla recalls 2 million vehicles to limit use of Autopilot
Wow, the recall applies to non-FSD capable cars as well, so AP1 cars aren't safe from the upcoming nerf: Cars don't have to have FSD to have Autopilot.
teslamotorsclub.com
Both GM and Subaru has subscriptions for their accident reporting systems and relatively short included trials. It is unclear if the car still sends data about a crash if that subscription is not paid for.
It's hugely misleading to use the data the way it's being used now, but apparently that isn't stopping people from continually doing that.
Ok, so it "knew" you were paying attention, so no need to nag you.
That would be an option, but the vehicle will almost always obey the driver, even if it means breaking a traffic law or running in to something. You were telling it to proceed, by intervening. (Maybe you wanted to run the light?) Just like it will let people slam into a building if they floor the accelerator. (When it can it will try to reduce acceleration and warn the driver, but it will still let them do it.)
I think it should put up a big red warning and beep loudly when attempting to run a red light or stop sign, even if driving manually. (But given most people don't actually stop at stop signs, it would be going off all the time.)
sublimejackman
Member
This cuts both ways tho. Tesla makes safety claims about their AP system all the time vs human drivers. However, the AP system is mostly used on highways, where there are way fewer accidents per mile driven overalll (human or overwise). They compare that data directly to all miles driven by humans. Which is super misleading.I made a similar point in the other thread. Subaru didn't even have an L2 system until the 2020 model year (the standing order only applies to L2 vehicles).
Tesla recalls 2 million vehicles to limit use of Autopilot
Wow, the recall applies to non-FSD capable cars as well, so AP1 cars aren't safe from the upcoming nerf: Cars don't have to have FSD to have Autopilot.
Both GM and Subaru has subscriptions for their accident reporting systems and relatively short included trials. It is unclear if the car still sends data about a crash if that subscription is not paid for.
It's hugely misleading to use the data the way it's being used now, but apparently that isn't stopping people from continually doing that.
As I stated above, I agree that making direct comarisons, using the NHTSA reported numbers, was a bad idea on my part. Won't happen again.
That being said, my conclusion still stands. Many Tesla drivers regularly abuse AP in conditions it is not suited for. Conditions Tesla knows its not suited for but it's CEO keeps saying it works on. That is leading to collisions and deaths.
Until we get raw data from all cars on the road, we will never really know if AP is safer. I guess Tesla could report the number of miles driven in different conditions/road tpyes and we could compare those, but something tells me they will never release that data...
stopcrazypp
Well-Known Member
That's definitely true, although it's unclear if the prevalence of this is more or less than when using other systems or when driving in general.
I say this is a strawman and largely irrelevant (same deal with the naming of it). The drivers know the car can't drive itself, they just use this as an excuse when they get into trouble. I've yet to see a case of an accident where this actually played a factor.
The only scenario I can think of is when using it the first time and the driver have not experienced their first nag yet. However, such drivers are naturally cautious and untrusting of the system and are unlikely to get into an accident as a result. In contrast, all the cases seen so far seem to be from experienced users that know the limitations of the system extremely well.
Elon also seems to make it clear when he is doing a forward looking statement (that he thinks it will eventually be able to do that) vs the production car already doing that at the moment.
Similar threads
