I doubt that Watson can be "aware of" what you might call high level contradictions. Many people have trouble sensing these.
Suppose you ask Watson: "Given cancer X in condition Y, what is the best treatment?" It goes through the literature and comes up with an answer. Next question: "Among public health epidemiologists, what is the consensus of opinion regarding the efficacy of cancer treatments on longevity?" Two things: 1. I would not be surprised if Watson can parse this sentence correctly and search the literature for a viable answer. That is just the kind of thing they are training Watson to do. Their parsing algorithms are awesome. They are science fiction by the standards of 1976, which is when studied this kind of thing in college. 2. I would not be surprised if Watson came with: "The benefits of all known treatments are negligible. [CITE NCHS]" (only in more detail). The thing is, Watson would feel no contraction between this response and the one he gave a minute before. He would not recognize that if the second response is true, it invalidates the first one. That is too high a level of abstraction for Watson, and for many people. You can think of Watson as resembling an incompetent mid-level officer in a hopeless war. The other day I read about a colonel in Afghanistan who was asked something like, "what do you see as the long term goal of this war?" His answers would befit a sergeant, but no one higher up. He kept talking about specifics such as building an airfield, or securing this many villages or educating that many replacement troops. He had no "big picture" notion of what to do or how to win. He saw only specifics and short-term, concrete goals. There were many Japanese officers like that in WWII. They knew how to win locally, and they were efficient at doing their jobs, but they did not have a clue how to win the war because, obviously, there was no way for them to win. They didn't want to think about it. Watson would be incapable of thinking about it. He can only focus on the level of the question he is asked to look into. I do not think he is capable of treating it as a meta-question about larger issues. Imagine you ask a Japanese officer for a plan of attack tomorrow. It would never occur to him to say: "Let's not attack, because even if we win the Americans will only send in more troops and beat us next time. Let's do nothing and let these fine young men of ours go home alive after the war." - Jed
