[warning: unfinished thoughts]
On the far end of the "technological unemployment" worry spectrum are those who really believe that AI techniques will yield robots that can do essentially all human work better/faster/cheaper than a human could.
When discussing that possibility, I wonder if our current statistics and vocabulary are really sufficient? After all when you talk about employment, wages, productivity, etc we're thinking about governments and corporations. We cannot forget that governments and corporations are human power structures and one of the biggest forces pushing them forward is the desire of the rich and powerful to become richer and more powerful.
If AI really begins to threaten the supremacy of its human masters, don't you think those same masters will suppress the AI and try to keep the status quo? After all a CEO with a billion dollars in the bank doesn't come to work just to collect his/her paycheck, they come to work to be the boss.
Modern humans have been shown to be strongly, and possibly entirely, motivated by relative wealth and power. The millionaire among billionaires feels poor, the wage earner among the unemployed feels rich.
So maybe, however inefficient, we will continue to employ each other only because it feels good for the employer to retain power over the employee? Maybe we will reject a super-efficient robot-produced future because we don't know how to form a society unless it's founded on a human power structure. Maybe we'll find things for each other to do just so that we can make each other do them.
I agree that if you enumerate all of my current material desires it's possible that a future robot will be able to fulfill them. But I cannot anticipate the societal evolution of desires and how much power over other people plays into that.