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.
Total conspiracy theory here, but what if early v12 was E2E, but subsequent versions are adding back bits of hard code to solve some issues, like the low speed?
Sounds at least plausible. Seems that would be a way to "smooth" out the speed problems to start.

Also referencing Chucks ULT. It seems one of the biggest regressions is the inability to pull into the turn lane at an angle and stop (this is at best a legally ambiguous move). For instance this may benefit from hard code. Elon must be steaming after the latest "embarrassing" video. May even put Chuck back on the naughty list. 🤣
 
  • Funny
Reactions: nonStopSwagger
So yesterday on my way home in a snow storm, I came across a traffic light which was dead due to power failures. It treated the traffic light just like a stop sign which is what it should have done. Was not expecting that, impressive to say the least.
Before owning a tesla we went through one of those at night while on back roads during a road-trip. In the heavy rain and flashes of lightening, we didn't see the stop line, which was our clue there was a controlled intersection, until it was too late to stop. The driver's focus was on the road and keeping the car on it. Of course, had there been a car in the cross direction, we would have seen its lights and hopefully would have slowed more (and thus made the stop safely).
 
You have a good amount of time to shake the steering wheel or move the scroll wheel if it sees you not looking, you must not have noticed the screen flashing for 10 seconds?
Hey, I went a good month or two after taking delivery and using AP, and getting a bunch of disengagements, before I realized that the screen flashes at all. I was oblivious until all hell broke loose with the red warning. It's because when I'm driving I'm looking at the road.
 
Hey, I went a good month or two after taking delivery and using AP, and getting a bunch of disengagements, before I realized that the screen flashes at all. I was oblivious until all hell broke loose with the red warning. It's because when I'm driving I'm looking at the road.
It should make a low single beep noise when the blue flashing starts and a louder noise when red flashing starts
 
I wonder what data we have so far about reactions to V12 from first-time Tesla buyers and first-time FSD users. I would expect general disappointment, but I’m curious how big of a disappointment.
No data of course but if you follow X the reviews appear to be mostly positive. Surprisingly positive so based on the feedback here. My son and son-in-law own Model Y's. Neither thought FSD was worth the money based on their FSD drives with me have taken Tesla up on the free trial. Feedback from both is really positive and they admit that surprised them. The true test of course will be if they sign up for the subscription service. If the cost was $100/month they say yes but $200/m maybe not.
 
I wonder what data we have so far about reactions to V12 from first-time Tesla buyers and first-time FSD users. I would expect general disappointment, but I’m curious how big of a disappointment.
It’s all about expectations.

Anecdotal:

We’re first time FSD users. Husband thought we should not bother with the v11 stack so we’ve been waiting for v12 to rent occasionally. Husband has been following the progress in detail and when we conveniently got this free trial I did some reading and watched the latest videos. We were both thrilled when we finally tried FSD - just amazed at how well it did on our typical errand running routes. Almost no driving required on our part.

Someone else on another forum expressed disappointment that they were required to pay attention and have hands on the wheel. Obviously they assumed L5 capability not understanding that it’s L2. Their attitude was more that it’s not useful if they have to pay attention. Hard for me to understand as I found not having to deal with all the driving manipulations/details enormously freeing and I was very happy to just supervise.
 
Last edited:
I sometimes wonder if a future hardware version will (or should) include an articulated telephoto camera, with dedicated sub-networks for aiming the camera and interpreting the telephoto view. Map data is far too often wrong for FSD not to override it based on prima facie evidence, such as road signs or painted road markings.
Tesla has the infrastructure in place to build the most accurate road maps that the world has ever seen. I keep wondering why they don't pursue that. Do they have a contractual agreement with map data providers to not pursue their own?
 
  • Like
Reactions: Ben W and lzolman
Yeah, ridiculous. Who does this, and why are people saying stops are normal and well executed? Do they care about their credibility? For some people is this truly not happening (that would be interesting)? Or do they think this is how people drive (look around, it's not, people are usually within about 2-5 feet of the line, not 5-10 feet)?

Exhibit A (this is a good result, by the way; it was probably only 4-5 feet behind the line at the "NHTSA stop" (FYI: this is not an NHTSA stop)):

Rant: Why do people keep saying NHTSA stop? It's so infuriating! NHTSA literally has absolutely nothing to do with this!!!! Tesla needs to stop to 0mph at the line. It's not a freaking big deal. What they are doing in this picture is ILLEGAL (if there's a bicycle in front of the car you need to stop at the line). You need to stop at the line - that's what the whole NHTSA recall or whatever was about!

View attachment 1035739

Tesla's FSD solution is a crutch with simplicity seasoning. The system needs lots of dwell time for high confidence and net training success improves when they combine scenarios or neuter it by one-size-fits-all training to the most demanding scenario. It makes training much easier but the end product is a short-cut 'sugar' show.

Assuming FSD can even do it, they need to add more than one default stop sign response. That's the way humans drive so why aren't they using the collected data? Perhaps the probabilistic nature of NNs can't be trusted in these high risk scenatrios? Until then, FSD will continue to approach all stop signs as if groups of school kids and little old ladies with walkers are walking in the cross walk.
 
Two interesting V12 things happened today on the way to the dry cleaners that I thought would be worth sharing.

First, I think Tesla must be using Tesla Vision as the source for traffic data. V12 turned a corner and came across a road that was entirely blocked by construction. Unfortunately, it just stopped at the stop sign nearest the construction, and then wouldn't continue moving. But as I disengaged to turn the corner, that road in front of me immediately registered as dark red. To my knowledge, there are no traffic cameras there, and Tesla would have no other way of knowing that road was blocked other than the data from my car. So my V12 disengagement there likely stopped other Teslas in the immediate area from taking that route.

Second, V12 exhibited some delightfully polite human behavior. While stopped at a red light, another car wanted to squeeze past me to make it to a turn lane. The light didn't change, and the traffic ahead of me did not move, but the car attempting to squeeze behind prompted V12 to move up a little bit to give them room.
 
After 3 years of chazman sending bug reports for his left turn, and Tesla sending a swarm of cars to gather data, Chuck's latest video shows FSD still can't do it safely.If they can’t fix one left turn, how will $TSLA ever get FSD to work on the millions of other left turns?

Tesla always has the option to avoid UPL turns that are clearly difficult. Just like UPS and other delivery companies do every day. Anecdotally there have been comments that some people who live inn Chuck's area turn right due to the dangerous crossing traffic. With U-turn support now this becomes an option in some situations.

Over the last year I have sent emails to Tesla on a dangerous UPL near my house. Navigation now has me avoid that turn. Whether it was my emails (doubtful) or not the fact is FSD avoids the nasty turn now.
 
Even with the prior versions which were not nearly as good, it was very common for people to feel that the software got better after a week or two of use. Give it a little time, get used to it and if you've got any open-mindedness at all you'll probably enjoy it more after a little use. Also don't give up on it to the extent that you won't give the next version, and the one after that, a decent trial and learning curve to see what's happening. In any case thanks for the report and for continuing to work with it as long as you can.

It's funny you mentioned this because that is EXACTLY what has happened the last few days.

2 things are happening:
I'm understanding the limitations of FSD and only using it in situations that I know it works well in.
And I'm getting used to it's quirks like sudden speed changes and hugging too close to the line.

I would be more impressed by it if I had the chance to experience the previous versions so I could be excited by it's progress, it's easy to be impressed if you watch something grow from nothing but if you stand back and look at it with fresh eyes it's not as impressive since you have nothing to compare it to.

I'm a Tesla fan boy, I have Tesla solar, batterywalls, two Teslas and $TSLA, I WANT FSD to work well.
 
Yeah it's exactly what I predicted. Quote below. I'm not surprised. Not my first rodeo!

The beer is mine (4/7 or 5/7 depending on whether you count driving at 25mph on a ~45mph road a failure), though I maybe will not collect it if Chuck does less than ten attempts (@Daniel in SD and I will have to examine the fine print of this rule in the 10.69 thread). It's a moral beer if nothing else. The best type of beer. 🍺

Do remember that 12.3 and 12.3.x does not incorporate any additional unprotected left training - that all occurred after the 12.3 release (before 12.3.3 but that isn't relevant since it doesn't appear to have been retrained).

NEXT time I will finally lose. I am betting against FSD 12.4 as well. Any perfect performance with 7 or more attempts is ok by me. The march of the 8.6s to 10.0s.
The bet was for FSD (beta) but this was FSD (supervised). I think the safety drivers were testing FSD (unsupervised) (beta) which probably has a couple nines of performance by now. FSD (supervised) is a driver assist so it's going to require driver interaction. Maybe they're trying out ways to make an EU legal version? Chuck was merely confirming actions that FSD (supervised) was already taking.
 
This morning's Tesla news shocker... Replacing the low cost EV effort with robotaxis. Maybe they need to add another neural net or two and viola? :)


Screenshot 2024-04-05 084302.png
 
Do you know if this is true for AI or are you just speculating? And just because it may be harder doesn't mean it won't be done. I certainly don't know so we'll wait and see what happens.
I don’t know if it applies to AI or not but my understanding is that the neural net is taught by showing it examples of ‘good’ driving. If that’s the only teaching method then the only way to learn what not to do is by inference. I could be wrong and maybe someone with more knowledge on NN training can chime in.