Only for people, who don't know it: I think we need an Robot of the fourth gernation.
In the software-area of programming languages, existing different gernations of programming languages:

1. generation: machine languge

2. generation: Assembler

3. generation: a higher programming languiage like C/C++, Java, etc.

And I think, something like that existing for robots, too. I don't know if anybody have written it down, so I do it now here. The generrations of robots:

1. generation: the monotonic robots. The most robots in factories are so. They doing every time the same steps. You can reprogram them, to go other ways, but then they do everytime the new way.

You can also see it, in this InMoov-video:

http://www.youtube.com/watch?v=YwClOdjBQnc

The acoustic commands "Da Vinci", "Relax", "Perfect", "Close arms", "full speed" and "giving" causes reactions of an robot of the first generation.

2. generation: the way to the special goal can differ. It is in the video, where InMoov grabs the ball. Or looks at faces of persons. That is, what modern webcams an digital-cameras already do. And myrobotlab can that, too.

3. generation: the way to the special goal is not implemented. This is with genetic algorithms. Or you can implement it with the trial-and-error method.

So there is the goal and the robot have to find itself the best way and solution for it.

A goal can be, to find the best wings implementation for flying. Or to grab as fast as possible to a ball or so.

4. generation: no way and no special goal is implemented. This are the robots, which are currently only planed to create them. I think, that currently don't existis one of it.

For that robots "emotions" are important. I talk not about robots with consciousness (I am not sure, if they will ever have some) or real emotions. I talk about simulated emotions.

I also not say about a robot which is angry about an other robot and the other robot feels guilty. That is at the beginning to complex and is only important for communication. It can be added later.

At the beginning, I mean a robot with only one emotion: happiness!

It can be happy or unhappy in different nuances. Basically you can see it as an signed integer-number. A postive number is happy and a negative number is unhappy.

And there existing only one big goal for the robot: be happy as long as possible and the happiness as high as possible. But every happiness is only temporary.

And then there existing the other "goals", which makes the robot "happy":

At the beginng it will be like a baby. So, ineracting with the world, catching a ball, move something araound, etc - makes happy.

If the robot was praised/lauded, that it have done someting - makes happy.

Diskspace is rare. So the robot remebers at best solutions, which are emotional (which makes him happy or unhappy).

Later other little "goals" are more imporatant:

Helping individual humans: makes a little bit happy.

Bringing the mankind in a better world: makes much happy.

Going to a goal, which makes happy and have reached a middle step for it: makes happy.

... etc.

Do nothing. Or "seeing" that the result of its acts are negative for the mankind: makes unhappy.

Be the best in one are of interests: makes happy.

Explore the world: makes happy.

Invent somthing "useful": makes happy.

The hard thing would be, to let the robot recognize, what is for example useful and what is not. But how can a human recognize it?

Greatings

theuserbl