There is no reason to believe that a future robot couldn't do a better job of creating a robot than could man. Obviously, that's not biological evolution, but it certainly could lead to an ever evolving progression of greater capability in any way that is important. Emotions are an important part of human capability. Whether one would implement emotions as a way of solving the same problems faced by a robot is a question. I would suggest that allowing robots that create robots to share evolutionary imperatives is a scary thought. We're going to need to be careful about the objectives robots are allowed to have. We have some time,obviously, as right now objectives tend to be stuff like rapidly spot welding a car into existence or ordering food that has been removed from the reefer. And, we don't give them latitude in methodology - or the ability to wonder if it wouldn't be more efficient to simply prevent humans from removing food in the first place.
Please, we are talking about AI in the sense it co-exists and is aware. In those terms all the machine would need to view is the basics. Humans need extensive land to grow food, machines don't need food. Humans require clean air and water, machines do not, or the most they would need is basic filters if anything...
