first thoughts....
I saw a brief snippet on BBC Click (I've been reading about it for years), and the thought occurred to me spontaneously(as gushing over language queries was the subject):
"I wonder if watson can derive what is missing from all the datasets?"
As a science type guy this is one of the things that bugs me about this "connected world" we are told we live in. There is so much information squirreled away in either government, corporate or academic archives (classify,paywall,incompetence respectively) , it seems feasible that a powerful "inference engine" might be able to show where the cracks are.
Might this be useful in determining funding priorities...?
P.