Well put. It is all very hard, especially reading the minds of other drivers. Every 4 way stop we do this. I think they might be better off doing things piecemeal and at least get the emergency braking better.it is such a hard problem, to ensure that auto-braking is fast when you need it, but does not trigger falsely, otherwise.
I bet you could have a whole company do research on this and they would still have trouble getting it right enough to be good enough for daily use.
so many areas are harder than it seems. lane centering has a lot to deal with, planning for what is near-ahead, mid-ahead and far-ahead. trying to read minds of other drivers. on and on.
this is going to be something even harder than a moonshot. there were no random crazy drivers on the way to the moon - amiright?
seriously, one single company trying to write all their own software. FAIL. this just is a mankind-wide effort, more so than even the moonshots of the past.
and yet, I dont see any real collaboration. everyone thinks they can solve this very hard problem on their own.
but tesla, especially, in their arrogance, are taking big steps instead of safer small ones.
I hope that someday we can pool resources - as a whole human race - and divide this up and solve it in pieces. and it will still take all of mindkind to think about this and work on it for years, maybe decades.
baby steps are what we need. lane centering is a huge step forward. and it mostly works. braking - its not even alpha quality let alone beta. they need a whole division working on that.
and NN is not the answer. we need to find deeper levels of understanding than statistics, which is all NN really is. that's NOT understanding and that's not enough to drive a car in a crowded random human-filled environment.
One bite at a time.