I've been reading RayKurzweils predictions lately of an 'AI revolution' set to happen this century, whereby most jobs currently done by humans will be done by robots.
What do you regard as being at present the major technological impediments to this revolution currently? Is 3D object recognition the main hurdle? Voice/speech recognition? Object manipulation not good enough yet? Power supplies not yet practical?
No one yet knows how to make a machine learn new tasks as fast as a human does.
Learning how to produce complex behaviors in response to complex high dimension stimulus signals is what our robots can't do like a human. That includes all forms of recognition and behavior production, such as the two limited examples you gave of 3D recognition and voice and speech. Currently, programing a robot to perform complex tasks is way too hard and way too expensive - the the point that the real advanced stuff we humans can do have never been programmed.
If we had the software problem solved, then issues about power sources and reliability and other factors might come into play, but so far, it's all about the intelligence. The tasks our current machines can be used for are always limited by the machines intelligence, not their physical size or strength.
I too believe that this software problem will be solved this century (probably long before the end of the century), and when it does, humans won't be able to find work anymore because there will be nothing left that we can do better than the machines. It's going to cause some fundamental changes to society.
Some people think the robots will become part of our society (as equals at some level with humans). I don't think we will build machines like that. It think they will continue to be slaves to us, just like our machines now our slaves to us. They will just be a lot smarter so that they won't need humans to operate them. We will tell them what we want and they will do it.