@Hugo:
1 - Roboticization
I agree. I don't think it would ever get that far. Which is kinda why I posted it, because people like to talk about post-scarcity as if it happens suddenly. Instead you'd see a slow creep of more goods being produced by robots, until the point where you can put capital into a robot worker and it can be more profitable than using human workers. Then the self-interested companies all around the world want to each be most profitable, so they all transition to robots. People start getting laid off. Before we hit a slum economy there would still be a transition period where some companies use human workers. And people and governments would be able to see what's coming and do something to stop it.
I know that the docks in my hometown have resisted automation and mechanization because their union knows it would involve reducing man-hours. I think that's exactly what would happen on a larger scale. Not because the consumer refuses to buy robot-made goods, but because the workers resist the change.
2 - Advanced AI Teaching
It doesn't need to be advanced AI. You give me a book and I can learn from it. You make that book interactive and I can learn easier. At some point the book has so many tools and functions, and approaches to teaching, that it's just as easy to learn from as any college lecture. Eventually the book gets so good it's almost as good as one-on-one tutelage. That's the robot teacher I envision. I don't care about a robot making facial expressions.
I do agree on children though. I expect childrearing will be one of the last things mechanized.
3 - Social
Humans may value human contact, but if you could buy drinks at a bar without feeling the need to tip the bartender, and you get your drink pretty much immediately and always correct, and the drinks are the same price - you'll find a lot of people happier with that bar than a normal one.
//
Anyway, my overall point was, if because of economics we arrive at a point where technology is cheaper than hiring, and people have trouble finding employment, what will be the jobs that only people can do? We can disagree about why one field or another must be human, or whether the original mechanization shift will happen so completely.
I really think it will.
We had seamstresses complaining that sewing machines would reduce how many man-hours of seamstress work were needed - and it has. On the other hand movable type increased the consumption of printed material and probably more people were employed in printing than when it was a hand-illuminated manuscript. But at this point we're seeing printing go by the wayside with the advent of digital distribution. Are more people employed in digital information processing than were employed in paper information processing? Are there more database analysts than there were file clerks? Something tells me the workers are more educated and handle much more data but there are fewer of them.
Another example would be how offices used to have a secretarial pool for correspondence and dictation. Nowadays people do their own typing and, at most, will have one assistant who handles a lot of different work. A legal assistant in Washington for example can do everything a lawyer can short of appearing in court, representing a client, and giving legal advice. That means a lawyer should have a team of skilled legal assistants, personally oversee their work and make court appearances, and manage overall legal strategy. The lawyer doesn't have six secretaries - these are skilled legal workers who handle their own typing. The relationship is less like a secretarial pool and more like a team of "mini-lawyers" overseen by a more experienced manager.
EDIT: Antennas
That looks like a cool read. I'm still thinking it's a specific application so it acts more like a tool that produces and tests iterations until it sees improvement, then uses those to inform future iterations. Which is to say, definitely automated R&D, but you still need a human to produce the tool needed to invent the antenna.
That said, it could be baby steps toward R&D bots working on R&D bots.
I remember a programmable chip this guy had, that he wanted to get it to detect microwave radiation or something. It was a 64 bit chip and the bits could be flipped mechanically. The computer processed iterations until it got a chip that worked: when the stimulus was present it lit a light. It was also very efficient, much more than he could have done himself.
Problem was, the chip was set up very strangely. There was a section off to the side that wasn't even connected to the main logic area. Yet clearing those separate bits made it stop working. And copying the program to a second chip didn't work either. The computer was taking advantage of some manufacturing imperfections in the chip and the bits off to the side were actually communicating with the main block of bits through EM radiation. The program worked - but only on that chip.