IBM's Watson Routs Humans on Jeopardy!

NEW YORK ( TheStreet) -- IBM ( IBM) won a big battle Tuesday evening. Its Watson computer system, the glass-encased series of servers that so far seem to be excelling in an attempt to "understand" the verbal intricacies of the English language, handily beat its human competitors in a round of Double and Final Jeopardy!.

In what almost seemed to be a fixed match due to the speed with which the computer buzzed into answers, Watson repeatedly bulldozed the two Jeopardy! champs and ended up capturing both Daily Doubles.

The audience, made up of some of IBM's biggest corporate clients, according to Frank Balboni, IBM's global leader for business analytics and optimization, laughed at the odd dollar amounts Watson bet for the Daily Doubles -- $6,435 and $1,246.

While the numbers seemed random, Watson uses a specific formula to calculate how much to bet to either get ahead -- or as was the case last night, stay ahead. According to IBM researcher Dr. Gerald Tesauro, Watson uses an "in-category Daily Double confidence" model to pick his bet, which is based on Watson's history of accuracy of answering the Daily Double in thousands of practice tests. He also takes into consideration his opponents' scores, the value of the remaining clues and the number of remaining Daily Doubles.

The system worked incredibly well last night; at the time of Final Jeopardy!, Watson was up $31,281 over Brad Rutter, who was in second place with $5,400.

And then came the final question, which duped Watson into submitting his silliest answer yet. The category was "U.S. Cities," and the clue was "Its largest airport is named for a World War II hero; its second largest for a World War II battle."

While both humans answered correctly -- what is Chicago? -- Watson's computerized placard, which showed the computer's lack of confidence with a series of question marks after his answer, read: "Toronto????"

IBM's lead scientist for the Watson project, David Ferrucci, explained how the question tricked Waston on IBM's Smarter Planet blog. Because the category was phrased "U.S. Cities" instead of "What U.S. City," Watson didn't give U.S. cities as much weigh in determining his answer -- he has already learned that Jeopardy category names only weakly suggest the kind of answer expected.

Secondly, there are cities in the U.S. named Toronto, which the computer took into consideration. "So this is just one of those situations that's a snap for a reasonably knowledgeable human, but a true brain teaser for the machine," said the blog.

In the end, it didn't matter: Watson had wagered only $947. He won with $35,734, compared to Brad Rutter's $10,400 and Ken Jennings' $4,800.

The final installment of the man vs. machine Jeopardy challenge airs Wednesday at 7 p.m. ET. IBM is holding a private screening of the episode in New York City, where company researchers will further expound on what Waston means for the future of computing.

--Written by Maggie Overfelt in New York.

>To follow the writer on Twitter, go to http://twitter.com/maggieoverfelt.

>To submit a news tip, send an email to: tips@thestreet.com.

If you liked this article you might like

Warren Buffett Is Sticking to His Strengths By Selling IBM and Buying More Apple

Warren Buffett Is Sticking to His Strengths By Selling IBM and Buying More Apple

Five-Day Winning Streak in Stocks Expected to Continue

Five-Day Winning Streak in Stocks Expected to Continue

Bitcoin Today: Prices Burst Through $10,000 as Traders Mitigate Stark Warnings

Bitcoin Today: Prices Burst Through $10,000 as Traders Mitigate Stark Warnings

Apple Stock Is Being Bought by Warren Buffett; Should You Buy It?

Apple Stock Is Being Bought by Warren Buffett; Should You Buy It?