Could you take seriously an overlord called "Watson"?

There is a lot of talk about Watson's performance on Jeopardy! in the tech world this week.  Stephen Wolfram wrote a fascinating, respectful (and inevitably biased) comparison of Watson to his own Wolfram|Alpha prior to the 14/16th Feb showmatches.  Now that Watson has so convincingly vanquished the Grandmaster human opposition ($77147 to $24000 and $21600), it is worth returning to the comparison, and considering what the two differing approaches tell us about themselves and the nature of AI.

Whereas Watson threshes masses of text for statistical correllation, W|A instead translates language into an internally defined abstract representation of meaning. This latter is described by its proud parent as representing true "linguistic understanding"... but neither he nor David Ferrucci of IBM would yet go so far as to call their creations true Thinking Machines.  It is still too early for the eager references to John Henry or HAL that journalists also made the last time a computer beat a human at the top of a human game (Deep Blue of course), but the question remains; of the two differing approaches, which is more intelligence-like?

Ok so I am not a Cognitive Scientist.  This is a problem.  But I can at least guess at some possible answers.  One is that each approach represents a different aspect of intelligence, and that in a true AI the two would work together - the brain of course contains many differentiable structures that do just this (and which have also been considered at a higher level of abstraction, if you are concerned that the algorithms Watson and W|A use would be mapped more accurately to these structures).

Another is that intelligence is a more universal term than the above modular explanation suggests.  I dipped a toe per Stephen Wolfram's suggestion into his grandly titled New Kind of Science where his "Principle of Computational Equivalence" seems to imply (in my crude interpretation) that all roads lead to Rome, and that intelligence could potentially be defined as any system that appears to act as such.  Not in the Turing Test sense that has the feel of strongly stated Cultural Relativism (who are we to say that what a computer calls intelligence is not so?), but rather in the sense that all complex manipulations of data are in some way equivalent. That the details of the implementation are irrelevant, not because we should judge by results rather than process, but because no one process is inherently "smarter" (in an abstract sense) than any other.

The only third explanation I can think of is that big-i Intelligence is just beyond our understanding at this time, that we are not subtle enough yet to properly describe it.  This last feels like a cop-out, the second is perhaps neat to the point of triteness and the first implies that the future of AI is in OOP.  None of these are particularly satisfying to me.  I do however feel that whatever it is, if/when it does arrive, it will thankfully be better able to explain itself that I am, and if the events of this week are anything to judge by, it may do so from Oprah's couch.