Word puzzles may go the way of chess and Jeopardy! next weekend at the American Crossword Puzzle Tournament in Brooklyn.
Human prowess at chess -- with world champion Garry Kasparov as its representative -- fell to the computer Deep Blue in 1997. But okay, it's chess, a game of logic and skill, with no genuflections to the wrinkles of chance, the challenges of limited information, or the subjectivity of language or knowledge. But soon those barriers fell as well, and in the decade-and-a-half since Deep Blue's victory, computers have emerged victorious in games that engage these other elements -- for examples, poker in 2008 and Jeopardy! last year.
This coming weekend a computer program named Dr. Fill (get it?) will take a gander at the tough word puzzles at the American Crossword Puzzle Tournament in Brooklyn. Dr. Fill won't actually be a contestant in the event, but anyone who beats it -- *if* anyone beats it -- will receive an "I Beat Dr. Fill" button from New York Times puzzle editor Will Shortz. The program is the work of Matt Ginsberg, a computer scientist and longtime puzzle-lover from Eugene, Oregon, Ben Zimmer writes in The Boston Globe.
Zimmer reports that Ginsberg "conservatively guesses" that Dr. Fill can place in or near the tournament's top 30. In simulations of 15 recent tournaments, Dr. Fill beat all its human competitors three times. Though the computer may not match humans in accuracy, it can whip them in speed, solving puzzles in 30 seconds and then editing its answers for another 90. Not even the top humans can complete the tournament-level puzzles in two minutes. The computer's dictionary database currently contains 10 million entries.
But no matter the machine's speed and the breadth of its resources, certain puzzle tricks are out of its reach. Zimmer relates a lovely example: A 2010 clue, "Apollo 11 and 12 [180 degrees]." The answer? SNOISSIWNOOW -- gobbledygook that means nothing to a computer (and to most humans). But to an adept puzzler, the answer is a thing of beauty: MOONMISSIONS, rotated upside-down.
Ginsberg is not entirely comfortable with the machine he has created. Zimmer writes that there was a moment when an update to Dr. Fill's database rendered "most Times puzzles child's play and [put] even the hardest tournament puzzles in its grasp. 'That was sobering,' Ginsberg said. 'Being in the top 50 is fine, but being in the top one is different. It was a little bit bittersweet. I was very surprised to feel that way.' " Zimmer seems to agree, writing that machine intelligence's shortcomings are "cold comfort, perhaps, for Team Human, as we prepare to lose another battle to computers on our home turf."
Top puzzlers may not be pleased by competition from artificial intelligence, but for those of us who never expected to be among the best, the Sunday puzzle will continue to be a pleasure, whether we are outdone by humans -- or by code.
This article available online at: