Robot Takeover Enters Phase Two: Jeopardy!

Back in 1997 I.B.M.’s chess-playing program Deep Blue beat chess-playing human Garry Kasparov. It was an interesting milestone, but it didn’t exactly usher in an era of thinking machines. Deep Blue won by brute force, playing out zillions of possible future scenarios and picking the moves most likely to lead to a win. It didn’t exhibit the flexible, creative thinking and understanding that makes human minds so unique.

I.B.M.’s latest challenge is considerably more difficult though: they’re working on a computer program, called Watson, to compete against human contestants in Jeopardy! Watson will get questions by text and answer in a synthesized voice (which we can only hope sounds like HAL 2000 or Johnny Five). It’ll be powered by a Blue Gene supercomputer.

From The New York Times:

To approximate the dimensions of the challenge faced by the human contestants, the computer will not be connected to the Internet, but will make its answers based on text that it has “read,” or processed and indexed, before the show….

I.B.M. will not reveal precisely how large the system’s internal database would be. The actual amount of information could be a significant fraction of the Web now indexed by Google, but artificial intelligence researchers said that having access to more information would not be the most significant key to improving the system’s performance.

The official promo video:

Artificial intelligence demonstrations have generally failed to impress even when they’ve succeeded. When a program or robot finally beats a human at a board game or navigates a set of stairs successfully, it only makes us realize how much more we were expecting. Watson could win and still disappoint.

The program will probably have an unforeseen, and boring, advantage in certain categories (think dictionary challenges like “Words containing ‘cat’”). And it won’t pass the Turing Test—the holy grail of AI—because it won’t be engaged in a free-flowing dialogue. Moreover, it will almost certainly make one or two absurd, boneheaded blunders that give the lie to the idea of an intelligence behind the avatar.

But it’s still exciting. If Watson does well with the questions that make use of the subtleties and ambiguities of English, it would have to be able to understand language, at least insofar as answering a nuanced question requires understanding. That could be really impressive—and lead to plenty of potential real-world applications.

Photo from Wikimedia Commons.