I think the assumption is that the cars won't literally circle the streets forever and that some parking will remain, just that it can be more efficiently allocated when the proximity from parking-space to destination isn't as-critical. [1]
Though, like you, I am highly skeptical about the curb-side-pickup fairy tale, due exactly the same sorts of situations you describe: concerts, theaters, last call -- even something as routine as picking the kiddies up from school.
People don't seem to realize that if automated cars delivered themselves to the curb at the end of a concert, a 200 space lot would more than wrap an entire city block. [2]
[1] Some breathless fantasists seem to think automated parking will somehow result in more-cars-per-structure, simply assuming valet stands translate into efficiency -- though almost certainly never having been a car porter or valet and knowing it's primarily just a convenience. You're just not going to see the 50-100% space-efficiency gains they seem to imagine.
[2] Assuming a city block of 1/8 of a mile and 20' per parallel space. That's 33ish spots per side of the block, not counting curb-cuts, hydrants, padding for intersections, etc. In reality, a 200 space lot would probably wrap a block twice or more.
Why would I pay for parking? The streets are free. "Only SUCKERS pay for parking" will be the motto.
People will be hacking their vehicles, right? They can program any behavior they want, right? NYC's off-street parking rates can easily exceed $20/hour. It is CHEAPER to tell your car to drive around aimlessly waiting for you than for it to park. Sure, traffic is miserable. But you aren't in it, so it's okay.
Find 15-minute loading zones. Park there for 15 minutes. Move to another one (your spot immediately taken by another empty vehicle). Repeat.
Park next to fire hydrants at no cost. Use the same visual scanners you use to navigate to detect approaching police and parking enforcement, and if one is detected, drive away.
There are limitless opportunities for self-serving behavior that ends up being hugely anti-social when any significant portion of the population does it. Basically there are three choices:
1. self-driving cars are locked down as all hell, The Man controls them all (proactive enforcement)
2. all the bad stuff is fined out the wazoo - Arab sheiks and Steve Jobs just pay the tickets, but no one else does - (reactive enforcement)
3. self-driving cars create a miserable traffic situation for everyone, in a massive tragedy of the commons problem (no enforcement)
Are the streets really free? Don't property and gas taxes pay for them? Is it really that hard to imagine a system that charges per-mile for driving on them? I don't think the three options you list are the only three and I also don't think this is an intractable problem.
I think the assumptions in the article are most applicable to an urban context, where car ownership is already nowhere near 100%, and alternative modes already exist that accommodate concerts, movies, etc. I'm reasonably certain, for example, that none of the movie theaters in the city where I live (Washington, DC) have surface parking lots. People are already used to not driving here, and I don't think the sudden arrival of self-driving cars will mean every single person will want curb-side pickup from the movie theater that's already half a block from a metro station.
Though, like you, I am highly skeptical about the curb-side-pickup fairy tale, due exactly the same sorts of situations you describe: concerts, theaters, last call -- even something as routine as picking the kiddies up from school.
People don't seem to realize that if automated cars delivered themselves to the curb at the end of a concert, a 200 space lot would more than wrap an entire city block. [2]
[1] Some breathless fantasists seem to think automated parking will somehow result in more-cars-per-structure, simply assuming valet stands translate into efficiency -- though almost certainly never having been a car porter or valet and knowing it's primarily just a convenience. You're just not going to see the 50-100% space-efficiency gains they seem to imagine.
[2] Assuming a city block of 1/8 of a mile and 20' per parallel space. That's 33ish spots per side of the block, not counting curb-cuts, hydrants, padding for intersections, etc. In reality, a 200 space lot would probably wrap a block twice or more.