The answers from U. Dreher, Seb Mirova and Eanes Torres Pereira cover major aspects including the vast field of AI and what AI really means. However, I think that the most difficult fields of operation of a robot are concerning ethics, moral (philosophy) and freewill. To get a better feeling of these aspects I recommend the books and contributions of Sir Isaac Asimov around the 3 laws of robotics. Just to give a few examples: what does a robot if any decision will hurt/kill lifes? In an airplane: will a robot try some (technically) forbidden action if this is the last chance to save lifes? What about intuition (translate an experience in one field into a completely different environment)? My feeling is, that any activity related to freewill, self-awareness and which is not based on statistics and probability will exceed the capabilities of a robot. However, if a robot could overcome these problems, acquire self-awareness and egoism as a major driver of evolution, and really become a "better version of mankind", why should he keep mankind alive as his most severe (but ethically incomplete and deficient) rival and competitor for resources? So let's hope that we are the better model in evolution and that natural intelligence (if it really exists somewhere in the universe) is beats AI.

Creativity.

Thinking ability

Innovation

Creativity and innovation.

While AI seems to do "magic", it is mostly bound to what it learned during the teaching stages - no magic at all. True creativity bases on experience but is capable to "leave" the basis, extrapolating and "dreaming".

Integrating different domains. It is natural for a human to make sense of various incompatible inputs: language, gesture, processing environment, sight, hearing, while remembering, comparing, judging, imagining, etc. We do all that simultaneously with ease and our brain consuming much much less energy than a computer.

Creativity is difficult because your definition of a creative task or output will differ. A deep neural network can reproduce, distort and 'create' art but not like the brain does. It simply uses statistical processes on pixel input. Ultimately we don't know how the brain does it.

Emotion recognition and expression. All human decisions are influenced by sentiments. If you agree that to have sentiments you must have a body with sense organs that are able to perceive and express emotions, than this is a great challenge.

The answers from U. Dreher, Seb Mirova and Eanes Torres Pereira cover major aspects including the vast field of AI and what AI really means. However, I think that the most difficult fields of operation of a robot are concerning ethics, moral (philosophy) and freewill. To get a better feeling of these aspects I recommend the books and contributions of Sir Isaac Asimov around the 3 laws of robotics. Just to give a few examples: what does a robot if any decision will hurt/kill lifes? In an airplane: will a robot try some (technically) forbidden action if this is the last chance to save lifes? What about intuition (translate an experience in one field into a completely different environment)? My feeling is, that any activity related to freewill, self-awareness and which is not based on statistics and probability will exceed the capabilities of a robot. However, if a robot could overcome these problems, acquire self-awareness and egoism as a major driver of evolution, and really become a "better version of mankind", why should he keep mankind alive as his most severe (but ethically incomplete and deficient) rival and competitor for resources? So let's hope that we are the better model in evolution and that natural intelligence (if it really exists somewhere in the universe) is beats AI.

@Ales Kralj, thanks for the link to the paper. I just went through and had a look to the references and saw the last conclusive sentence of the paper which I cite: "... Among the prefrontal regions, the FPC is likely specific to humans (31, 32), which suggests that the ability to jointly infer multiple possible causes of observed contingencies and, consequently, to test new causal hypotheses emerging from long-term memory isunique to humans." - End of citation!!!

Where is the intelligent robot?

By the way, the question whether "natural intelligence exists" was academic and a joke.

I think that one of the most difficult characteristics of the human work that will be most difficult to be performed by robots is: the way we see the world from a human perspective. This include many elements which, some of them, are difficult individually and more and more difficult together. Elements like at: feelings, knowledge, perceptions, emotions, common sense, creativity, relationships, understand abstract concepts, sentient reasoning, judgement and among others. So these traits are incredibly difficult to emulate together in a program.

Works involving genuine variable feelings. Later modification (I have to say I agree with the 2 previous colleagues, my first sentence was written befor reading their answer. Mine seems close to the idea of Claudia Mezo).

As computers and robots take over more and more of the tasks that we humans perform on a daily basis our interests and desires will extend into new areas and endeavors. The question of What would I like to do in my life? will become more prominent. This question cannot be answered by computers because it is, on the one hand, inspired by the enabling characteristics of the AI-based automation and, on the other hand, driven by our own distinctly human beliefs and aspirations.

I think intuition and clairvoyance might be some of the most difficult work for robots to do.