Judging from today state of the art AI - creativity would be automated much earlier than dumb physical jobs.
We already have pattern matching and theorem proving software, I think that throwing sheer computing power at hard problem where strict criteria for correctnes are defined would be cheaper than human problem solver.
On the other hand mechanical and interfacing side of robot revolution seems to lag after AI research. Teaching robot to run is hard problem, brain has many layers of hardware for doing this, and it's easier to make new human, than it is to make new robot.
All in all I predict the exact opposite - unskilled workers will have jobs, creativity in many cases would be automated.
We already have pattern matching and theorem proving software, I think that throwing sheer computing power at hard problem where strict criteria for correctnes are defined would be cheaper than human problem solver.
On the other hand mechanical and interfacing side of robot revolution seems to lag after AI research. Teaching robot to run is hard problem, brain has many layers of hardware for doing this, and it's easier to make new human, than it is to make new robot.
All in all I predict the exact opposite - unskilled workers will have jobs, creativity in many cases would be automated.