Key to AI is clear. But, which is the challenge? Are we talking about predicting actually? Or is it rather about building a huge case-based reasoning system tha

Key to AI is clear. But, which is the challenge? Are we talking about predicting actually? Or is it rather about building a huge case-based reasoning system that can "replicate" known patterns? Most recent research is focused on neural nets and deep-learning to that end, this answers the question wrt prediction, but this seems to be a partial solution.