-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
2019.40.1.2 on TeslaFi
- Thread starter kkillebrew
- Start date
diplomat33
Average guy who loves autonomous vehicles
diplomat33
Average guy who loves autonomous vehicles
I saw a video on this and its questionably functional. If you go to the Tesla site and go through the motions of buying a model 3 when it gets full self driving, it says city driving later this year. By my calculations there are 3 weeks left to later this year.New feature alert: Stop sign warning!
It appear to warn as you go through or are committed to go through. It should warn while you are a 1/4 sec away from safely stopping at the stop sign. This is quite similar to @verygreen stop light recognition. You had to be entering the intersection in auto steer mode when the light was red.
diplomat33
Average guy who loves autonomous vehicles
Yeah, stop sign warning seems to be a bit iffy right now. But remember, Tesla releases the feature as just a warning in order to collect data to make it more reliable. Then, when it is reliable enough, we'll get the feature where the car stops on its own. Tesla did the same with traffic light warning. They released it where it just warns of a red light, when it was still unreliable but have presumably been collecting data to improve it. Once, both traffic light and stop sign warning gives them enough data to make the features reliable enough, we'll get the feature where the car can actually stop for red lights or stop signs. But Tesla can't release that until it is reliable enough.
I do think that Tesla will miss their end of year deadline though. But stop sign warning is a good sign that Tesla is moving ahead with collecting data to make it more reliable and is therefore getting closer to being able to release traffic light and stop sign response.
Mine never recognized a stop light. Has/does yours? I got 40.2.1 yesterday. My ref to "later this year" was as of yesterday. They are a publicly owned company materially misrepresenting their product. This is concerning!
diplomat33
Average guy who loves autonomous vehicles
I have not personally tested it yet. So right now, I am just going off of what I see in videos.
Part of the problem is that right now, the only way to know if the car is seeing the stop sign is if we deliberately blow through a stop sign on AP and it gives us the warning. So the car may be seeing a stop sign but we have no way of knowing it sees it because it won't tell us it sees it if we stop on our own at the stop sign. I wish the display would just show us what the car is seeing. Then we would know if it is seeing a stop sign without having to blow through it.
To your last point, if Tesla knows that they don't have traffic light and stop sign response but are still promising by end of this year, yes, that would be misrepresenting their product. But we don't that is the case. It could very well be that Tesla has traffic light and stop sign response that is not quite reliable yet and they are simply behind schedule. If so, that would not be misrepresenting their product.
[
And the model they set is they, in fact, do that with the left side of the monitor. Even if you are not in AutoSteer or NoA, you see the traffic. So, yes, they should show what they see and take action on it only if in AutoSteer or NoA.
The only vid(s) I've seen on traffic light recognition is from @verygreen to my recollection. And that was on an acknowledged hacked software. I tried even to the extent of all but running a red light on AutoSteer. Again, it should show the traffic light even if it's green or amber.
Last Feb Musk told Ark 100% certainty development would be done by years end. He doubled down in April at Investor Day with FSD, no geo fencing, level 5 by end of year. It's not impossible there will be a big release on New Years Day but I think that incredibly unlikely at this point, even requiring the same attentiveness as AutoSteer and NoA require today, that being it'll do the work but the driver must continually demonstrate he is paying attention to the car and is PIC.
Disclaimer: No one should infer I am bad-mouthing Tesla or Musk. I bought the car for FSD, I'd be thrilled if they deliver it on time or very minor delay. Again though, as of yesterday they are advertising in print city driving by end of year.
And the model they set is they, in fact, do that with the left side of the monitor. Even if you are not in AutoSteer or NoA, you see the traffic. So, yes, they should show what they see and take action on it only if in AutoSteer or NoA.
The only vid(s) I've seen on traffic light recognition is from @verygreen to my recollection. And that was on an acknowledged hacked software. I tried even to the extent of all but running a red light on AutoSteer. Again, it should show the traffic light even if it's green or amber.
City driving which they advertise "this year" not only includes traffic light/sign, speed limit signs, but most importantly actually turning at an intersection. Turning right at an intersection when the light is red unless it says no right on red. So that's another sign it MUST recognize for city driving.
Last Feb Musk told Ark 100% certainty development would be done by years end. He doubled down in April at Investor Day with FSD, no geo fencing, level 5 by end of year. It's not impossible there will be a big release on New Years Day but I think that incredibly unlikely at this point, even requiring the same attentiveness as AutoSteer and NoA require today, that being it'll do the work but the driver must continually demonstrate he is paying attention to the car and is PIC.
Disclaimer: No one should infer I am bad-mouthing Tesla or Musk. I bought the car for FSD, I'd be thrilled if they deliver it on time or very minor delay. Again though, as of yesterday they are advertising in print city driving by end of year.
Feature complete meant they'll have some lab version somewhere that kind of did everything they said it would; not that it would be released to the public. The full feature set for the public is likely still years away irrespective of their heavy marketing making it sound like everything will be ready. Those of us who've been hearing and reading the rhetoric about autopilot since late 2016 know better.
We've been through this on a million other threads. Do we really need to rehash it on every release thread? Can we stick to the topic?
No, categorically NO. It means short of QA bug, dev is done. Yes, I get a bunch of ppl on here with no software development experience pontificate on what tesla software dev is doing. Sound and fury signifying nothing. Look at my bio. Investor Day Musk talked about the 2020 robotaxi fleet. That implies releasing to the public. Don't go all histerical on this. I merely suggested Tesla appears to be materially misrepresenting their product saying city driving would be out this year, aka next 3 weeks. Seems you are vehemently agreeing with me.
I'm just going to add one thing real quick that came to me yesterday. It's mid Dec (all but), we only have access to an October release 48-Dec, 44-Nov, 40-Oct. It could well be 2019.48.x.y does have city driving and intersection management. One can hope. I'm excited at the prospects of 2020.0.x.y!
diplomat33
Average guy who loves autonomous vehicles
Good point. Yes, I know the website promises the features will be released by end of the year but if 2019.48 has "traffic light and stop sign response" or "automatic city driving", Tesla could at least say that they had the features done in development by the deadline even if they were not released by the deadline.
I am definitely excited for what comes in 2020.x,y updates. IMO, we should definitely see some form of "city autopilot with driver supervision" sometime in 2020.
boonedocks
MS LR Blk/Blk 19”
Tesla has said things like that many times. If it isn't in our hands and useable, everything else is just not believable. I am a fanboy #1 but also more than a bit skeptical where all things ElonSpeak is concerned. If he espouses "Feature Complete" and we aren't using it in our cars....in mass....then it is just more ElonLipService. I do wish Elon had a "handler"
The reality is they already have that, not just in the developer version some employees run, including Musk, but in NoA. It does interchanges and has for some time. If you approach a red light and not the first car, the car slows and stops. When the car in front of moves, so do you. So instead of a slower moving car the trigger is a red/green light. The software knows how to turn by itself, NoA does that all the time with exits and interchanges.
So what do they really need to do? Very little. sign/signal recognition I think is, if not all, most of the remaining work. It has to understand it's at an intersection and clear left and right. So if you are in AutoSteer and someone / something pulls out in front of you... Oh yeah, they seem to have done a poor job, to date with that one, so add it to the list. And, frankly, that's just bad design. They have radar and visual. The phantom braking, be it for shadows or overpasses just mean they need confirming input. If radar shows something does visual confirm. If visual shows something, does radar confirm. Maybe that's the issue. Radar will echo off a belt buckle but not a deer. So, for visual, is it moving with respect to the vehicle? Is it getting larger over time, as in you are approaching it?
The issue is, and every Tesla owner has skin in this game. This is NOT Elon Musk being a crazed cheerleader. He is CEO / CTO. He is an officer of the company. If he were just a developer or janitor fanboy hyperbole would be innocent. As an officer of the company it is materially misrepresenting the company's product. That has serious SEC implications. There are already law suits trying to force Tesla to turn off anything hinting at automation. There are also efforts to remove him from management.
There are people who approach this as general population pontificating on Tesla's inability to have flying cars or whatever. I approach this from a development design/implementation perspective, what do they have, what do they need, and what competencies need to be married. They are dangerously close. There actually isn't that much left to do. So, yeah, being code complete by years end, definitely within the realm of possibilities.
As for "feature complete" you aren't only wrong, you are way wrong. That is a very specific software industry term, with a very specific meaning. Again, I did this for 45 yrs. I am not taking issue for the sake of argument. SO please don't take offense, I am just approaching this from a different perspective.
Similar threads
