IBM's Watson Trounces Puny Humans At Jeopardy

Inside Watson, IBM's Jeopardy Computer

"I for one welcome our new computer overlords." With that, Jeopardy record holder Ken Jennings conceded defeat to an IBM supercomputer that handily defeated Jennings and Brad Rutter, another past champ from the game show, in a two-match contest that aired over three nights this week.

In the final match, shown Wednesday, IBM's Watson machine racked up a winning score of $41,413. Jennings was second, with $19,200, while Rutter placed third with $11,200. Watson's cumulative score over the two matches was $77,147, compared to $24,000 for Jennings and $21,600 for Rutter.

In vanquishing its human opponents, Watson won a grand prize of $1 million on behalf of IBM, which the company has said will be donated to charity.

"Watson is fast, knows a lot of stuff, and can really dominate a match," said Jeopardy host Alex Trebek, stating what became obvious over the three-episode exhibition of the state-of-the-art in artificial intelligence.

But the computer also showed that even the most advanced AI program can be strikingly fallible at times. On Tuesday's episode, it drew guffaws from the audience when it answered "Toronto" to a question that asked which city was home to airports named after a World War II hero and a famous WWII battle.

Watson, in his soft mechanical voice, named the Canadian city despite the fact the question was under the category "U.S. Cities."

On its Smarter Planet blog, IBM said Watson's training may have led to the error. "Watson, in his training phase, learned that categories only weakly suggest the kind of answer that is expected, and therefore the machine downgrades their significance," IBM said.

IBM also noted that there are seven U.S. towns or cities that bear the same name as the Ontario provincial capital. Possibly adding to Watson's confusion was the fact that Toronto is home to baseball's Blue Jays, which play in MLB's American League.

IBM says Watson, named after company founder Thomas J. Watson, is much more than a science experiment. Much of the program is built on technology that IBM has already commercialized for applications such as economic modeling, weather forecasting, the prediction of disease vectors, and the tracking of trends in financial markets.

"Beyond our excitement for the match itself, our team is very motivated by the possibilities that Watson's breakthrough computing capabilities hold for building a smarter planet and helping people in their business tasks and personal lives," said David Ferrucci, who leads IBM's Watson team.

Watson simultaneously runs natural language processing, information retrieval, knowledge representation and reasoning algorithms to fathom the intent of questions and yield what it thinks is the best answer—all in a matter of seconds or less. The Watson program runs on IBM's new massively parallel POWER7 processors, which the company rolled out last year.

The shows were taped last month at IBM's research center in Yorktown Heights, NY.