We've been wanting to do this where I'm at for a while. While I hate to put human beings out of work, they really do make too many mistakes for shipping things perfectly.
That's understandable, but think that these people will probably find more fulfilling and productive jobs. As long as the layoffs are gradual, they will have time to retrain and find other jobs. Of course, if their skillsets are really obsolete and many are laid off at once, this can lead to structural unemployment. Eastern European countries went through that after the fall of communism. The transition was not pretty.
But smaller layoffs and career changes are just the cost of technical progress. Have a look at the Industrial revolution [1] and Luddites [2].
[1] http://en.wikipedia.org/wiki/Industrial_Revolution [2] http://en.wikipedia.org/wiki/Luddite
Also, here's my first law : "Anything that can be automatized WILL be automatized, whatever the costs."
You assume that there are jobs to be had, that retraining is feasible, and that technological progress does not obsolete people as workers. Capitalism will not survive the intelligence singularity.
Development of AI will simply stop before it reaches knowledge-worker-class strength? Why would it stop?
What do you think about Theodore Kaczynski's take on this?: http://news.ycombinator.com/item?id=435437
The point of automation is greater efficiency and efficacy, not some fetish with robots. Your "law" probably does not hold.
That humans are not as specialized as specialized robots was not contested. What about the "whatever the costs" part?
"... whatever the costs are now." or "... whatever the costs to society."