No, just no. Visualization is irrelevant and pointless. A distraction. It will never be what the car is seeing because it can't be. So all it can ever be is a misleading security blanket. Spending any effort on it distracts from progress on actual FSD work.
The driver must be prepared to take over always. So of course it's true for this particular case. It's all eye candy. Keep your eyes on the road and drive.
no.
bingo. We are not at this point yet.
You car just entered opposing traffic lane overtaking another car, nothing immediately in front of you, do you abort and manually merge or no? No visualizations for you, so you don't know if the car is planning to merge back (part of an overtake maneuver) or not (car thinks it's a lane you can keep on traveling in).
Not really, unlike what you or other devoted Tesla fans think, stating Tesla is not perfect does not require a burning desire to complain about Tesla. I'm just stating some facts about visualizations that the rest of the autonomous vehicle industry has figured out - you know, like the guys that actually have cars driving private passengers as part of ride sharing today, not just staged fake videos.
Let me tell you, so you don't have to strain your imagination any more. I still own two of four Model S I bought because they are a dream to drive (manually, nothing to do with autopilot) and while they are early adopter cars requiring TLC from Tesla (each and every one of the 4 required service to fix manufacturing or engineering issues), service used to be awesome - I could talk to a live person, they'd provide loaners, they'd make everything right - the time spent servicing was a worthy trade-off for owning and driving the car, especially they that after few post delivery visits they have held up well for us (except for the yellow screen which showed up months later). Yea, Tesla tricked me into paying $25K extra for car speced at 691hp, and only admitted two years later that the car would need a 50% power increase to meet that, but without it I still would have bought one, just not the performance model - service was still stellar back then. However, given the service experience deterioration and continual profit chasing changing of warranty terms, I am no longer interested in buying a new one. So, I am now stuck with the Teslas because I can't find a replacement that drives as well (again, nothing to do with AP). I hope this answers your dilemma.
Yep, I think this is where we disconnected. My opinion on visualization was based on whether they are needed for FSD features (which according to Elon are coming any day now). Given no FSD feature (as per AP2 definition since we are talking here about people upgrading from MCU1, not the 2019 neutered version) has been released today, yes, they are not required.
I never said you need 1:1 representation, come on - that would be AR. I simply said visualization showing objects such as obstacles, road markings, signs, traffic lights, pedestrians, cones, and anything else which AP recognizes and is aware of when making the decision, combined with an overlayed planned path, where is AP planning on going. If you know AP is planning an overtake maneuver because it shows you an arrow going into oncoming traffic lane, and you see a truck coming in the opposite direction which doesn't appear on the visualization, you abort it before it steers the car into oncoming traffic lane just before the truck passes you. Without visualization you have a split second to react, with visualization you have many times that.
Of course it is purely for the benefit of the driver, in order to allow the driver to more safely supervise FSD. Technically avoiding an accident is also benefits FSD hardware, as a crash would potentially destroy it, but that is not the primary reason for visualizations - it is to keep the driver safer, not the FSD hardware. To be clear, visualizations are for humans to make decisions, AP does not read back the visualization and make decisions - visualization are machine-to-human interface, not machine-to-machine.
What's clear is that you think that. But nobody else does. It's insanity. Humans must make decisions based on actual road conditions. The visualizations are strictly eye candy, there only to let people see the auto-pilot is actually seeing something, and cool to show your friends. And since it's useless, it's a coding task you give to interns -- nice if it works and no matter if it doesn't. Your imagined purpose for this is bunk.
If you truly believe that knowing whether or not autopilot sees an obstacle, or another car, or lane marking, or traffic light (or that it thinks it's green when you can see it's red) ,or knowing what it's planning on doing (go straight, turn, stop at a light, pass a car, switch lanes, etc) is completely useless for a human who must supervise it for safety reasons, then I think we should just agree to disagree on that and call it a day - I see no point trying to argue. I concede that you think you know better than all those other companies working autonomous driving just wasted their time implementing such confidence views.
Tell me something I said about Tesla that wasn't true. In this case I wasn't even talking spefically about Tesla, rather about the need for confidence view in autonomous vehicles. If you're curious as for why I still own them, see my prior post (#2904). I used to be Tesla's biggest fan too, telling everyone who would listen to buy one, until I got burned, so now warning people, while still telling them they are best driving car you can buy today.
It must be a new feature of Tesla to warn you about things it can't see. All those Teslas which plowed into stopped fire trucks could have used that feature.
The confidence view is to to keep on looking at while you're driving. It's to look at when you're not sure why the car is doing something that concerns you. It is not a replacement method for situational awareness, it is just another tool to add to your situational awareness. It is not meant to replace the driver looking around and being aware of their surroundings. It is there so the driver can confirm that the car sees an obstacle or maybe that it's missing that giant firetruck in its lane, so they know to brake right away and/or switch lanes. Increased situational awareness means generally safer driving.
I don't understand. How can they come in serialized by the VIN of the car to be installed into unless the owner of that VIN has already purchased the upgrade? Did you actually purchase the upgrade? If not, how would Tesla ever know you were intending to do so, such that it would "serialize" an upgraded MCU for you?
I think they have a range of VINs that are allowed. If you submit a request and your VIN is in the allowable range, then they can do the upgrade. If your VIN isn't in the allowable range, you have to wait until the VIN range is expanded.
