I myself believe that the most interesting thing will not happen in the part of unstructured, and in part structured: the structuring of different ontologies, different compactification knowledge, different versions of the same recognition of real world objects (associative memory, work with analogies, etc.), encoded different, different speed algorithms and formal derivation of heuristics for such algorithms, and so on. For Watson showed that with "unstructured" can operate at no worse than the world champions in the rapid understanding of the issues. Then you need to realize is not so much "ponimalku-vspominalku" taking advantage of all the world's knowledge as "dumalku." There is still strewn with very, very few horses.