I don't understand the leap from "it's difficult and there are many special circumstances" to "it's never going to happen". I don't get why you think that any of these problems are unsolvable given enough time.
Here's one way to get to the "it's never going to happen" outcome:
The grand-parent comment argues that we need AI that works in a much broader set of circumstances to "solve" self driving. In particular it would need to understand how other humans would react to its actions in novel situations. That is approximately a description of having a Theory of Mind. Some argue that you can't have a Theory of Mind without being conscious. We might ban captive conciousnesses for ethical reasons.
I think you could invent lots of other scenarios that yield the "never going to happen" outcome. They probably all sound ludicrous, because having an AI that understands the workings of human minds sounds ludicrous (and frightening).
To me this is a technocratic stance - believing that technology will always find a solution and absolve us of all problems.
Maybe. Maybe not. What I read from his post is much closer in spirit to "we cannot simulate real physics down to the atom" when modeling protein folding so what we do instead is building models (good enough estimations) and AI approaches to recognize complex (but still top level) patterns.
These reduced views on reality help our problem space to a large degree but they can never account for the full scope of the physical reality and they largely work on assumptions which might be proven wrong at any point in time.
The point is people who are proponents of self driving have a blind spot. Its construed as a code problem, ignoring entirely the massive social aspect of human behavior.
Driving isn't locomotion. You can't assume that the people around you are going to be reasonable, sane or predictable. More data doesn't fix that, unless you start building a "civilzed behavior" model and then add that to your movement model.
I really want to see how honest people are when they build a fair representation of human civilization, warts and all.
Self driving as has been defined assumes some absurd things about the world in which humans live.
No one is going to buy a car which wont know to run when they are about to get robbed.
To be blunt, the blind spot assumes you live in America (and probably california), not that you live in Brazil or India or Egypt.
Because no matter how much time you give it, there are social problems that need to be solved first; and said social problems are not the kind that are amenable to mere technical solutions.
Literally the case of: If the technically perfect implementation existed tomorrow, we still would not be ready to flip the switch because of how drastic the reorg of societal norms would be. It'd be a complete refactor.
The grand-parent comment argues that we need AI that works in a much broader set of circumstances to "solve" self driving. In particular it would need to understand how other humans would react to its actions in novel situations. That is approximately a description of having a Theory of Mind. Some argue that you can't have a Theory of Mind without being conscious. We might ban captive conciousnesses for ethical reasons.
I think you could invent lots of other scenarios that yield the "never going to happen" outcome. They probably all sound ludicrous, because having an AI that understands the workings of human minds sounds ludicrous (and frightening).