early next year, IBM’s supercomputer “watson” will take on the two best jeopardy players in the show’s history, ken jennings and brad rutter (the guy who beat jennings in the jeopardy tournament of champs). we know computers can win or draw against master players in competitions requiring brute force calculations when the game complexity is low enough.
but playing in this format is something different. how well the comp does will depend greatly on the quality of it’s database obviously, but the structure of the clues requires more than simple brute force calculations. to find the correct replies, contestants have to parse the sometimes intentionally opaque grammar of the clue correctly, place it in the context of the category, understand various subtle references, puns, ironies, double entendres, and relationships (near/far, little/big).
imagine trying to come up with the algorithm that could provide the question to this recent clue in the category “hmm… pronoun trouble”:
A sovereign may use the “royal” this pronoun in formal speech; are you not amused?