I agree, but I do think it's important to point out that a lot of human labor (like programming, for example) seems to be AI-complete. Once we have full-on AI, we'll have bigger problems than unemployment. As Bostrom points out, intelligence and morals are orthogonal. That is, while intelligent agents may converge on a provably optimal method of cognition, their goals will not converge. One can't construe conscience and capability. This is a bigger problem. There doesn't seem to be much point in worrying about unemployment when strong AI is on the table.