2.25.2011

Was the freakiest answer given by 'Watson' on jeopardy one of the few it answered incorrectly??

YORKTOWN HEIGHTS, NY - JANUARY 13:  (L-R) Exec...Image by Getty Images via @daylife
If IBM's 'Watson' A.I. computer is so smart, why did it guess "What is Toronto????" on final jeopardy when the catagory was "US Cities"?!?

The clue, to paraphase, was "which US city has two airports, one named for a hero of a WWI battle and the other for a major battle in the Pacific during WWII?" The answer to the question happened to be Chicago (O'Hare and Midway). Both human contestants, Ken Jennings and Brad Rutter (the top two money winners in the show's history), despite having their asses handed to them all day, seemed to find the question to be surprisingly obvious and answered correctly.

How did they make the machine make such a human mistake and make such an absurdly illogical guess? Wouldn't a computer, guided by strictly by logic, immediately eliminate any answer that wasn't actually a US city?!? Of course it would, probably even before the clue was even read, which begs the question -- did 'Watson' make an attempt to be witty and downplay its lack of an educated guess?!? Was 'Watson' slyly tipping his cap to the host and his team of clever human robot slayers?!?

Computers that are witty scare me more than computers that are intelligent...
Enhanced by Zemanta