Tesla say it will. It clearly doesn't at the moment, but if you're saying it CAN'T work rather than it DOES NOT CURRENTLY work you're going to have to explain your reasoning. You are able to drive (and park) a car despite only being able to look in one direction at a time and with even more blindspots. Why can't a computer?
At the moment, in most respects yeah. But if you're keeping score then you have to note that it can already identify curbs, which the sainted USS, even after YEARS of refinements in the technology, could not. If you want a car to park itself (in the general case, not just in specific parking bays, etc) , not hitting curbs is quite important.
You keep saying this with considerable confidence. Presumably you are old enough to have owned cars without USS on them, and presumably you were able to park them, so I ask again why you're so confident that a computer can't do the same.
I have stereoscopic 3D vision that gives me depth measurement, and I can bob my camera mount around to get further instant depth information from parallax. I also have an insanely well optimised NN AI that makes fantastic use of all that camera input plus extensive training data to make a damn good persistent model of my surroundings. Even with all that I still can't accurately measure the distance to surrounding objects as well as USS can, particularly in poor lighting conditions and where the car itself obscures my cameras. Hence the usefulness of USS.
TV doesn't even come close to my own vision capabilities, never mind to myself augmented with USS. So yes I do indeed say not only does TV
currently not perform well enough to park a car, or depart/navigate from very close proximity to other objects, but also that it most likely
can't with the current hardware. Let's see what the next few years brings!
This is just lunacy. Other manufacturers are no use at all for a view of where cars are going. Tesla may have a fault of being hugely ambitious and dragging customers along with them (whether they want to go or not) but the other manufacturers would quite happily crank out the same vehicles with incremental tweaks until the end of time.
The only race where the established manufacturers are winning is in the nosedive towards bankruptcy, because they can't respond to the manufacturing process improvements Tesla have made and so (even after the recent price cuts) can't get anywhere near Tesla's profit margin. If Tesla could snap their fingers and bring new production capacity in to existence overnight they have the headroom to cut so much off the price of new cars that the other manufacturers would have to sell them below cost price to remain competitive.
I think what the car industry (i.e. other manufacturers) is doing is actually a pretty damn good view of where cars are going!! And they are all sticking with multi-sensor systems and, as I said, are slowly feature by feature leaving Tesla behind - working auto park, working summon, L3 driving features, etc. The only time what other manufacturers are doing is not a good indication of the future is of course when some major paradigm shifting disruption is afoot. Elon/Tesla absolutely has been this genius disruptor in many areas of the industry - the whole EV powertrain, packaging, manufacturing, etc. But in terms of autonomous driving aids I think it is already being shown vision-only is sadly a dead end. It initially looked to have some promise, but legacy auto has it right this time, and as all the various sensors get better and cheaper, the knowledge to merge them into a cohesive system matures, and more and more cars get features that work better than Tesla's, that will only be more apparent.
You correctly point out benefits and capabilities of cameras, but are failing to put forwards an argument for cameras-
only! Except for cost saving, which you claim has nothing to do with it! I'm not arguing USS
or cameras, I'm arguing for USS
and cameras. Give me one good reason for cameras-only that isn't cost saving. And no you can't have 'but sensor fusion is difficult' because everyone else is starting to show it can be done.