RENEE MONTAGNE, host:
Back in 1997, IBM's Deep Blue supercomputer defeated world chess champion Garry Kasparov. This week brings a new contest between another IBM computing system, and the two most celebrated contestants from the TV quiz show "Jeopardy!"
(Soundbite of promotion)
Unidentified Announcer #1: These "Jeopardy!" champions are so smart, people say their brains are like computers. This new IBM computer is so smart, they say its brain is almost human.
Mr. ALEX TREBEK (Host, "Jeopardy!"): At "Jeopardy!" we say: Let the games begin.
MONTAGNE: Those heavily promoted games begin tonight. NPR's Nell Greenfieldboyce reports why this machine is a formidable competitor even though it doesn't think exactly like a human.
NELL GREENFIELDBOYCE: When Deep Blue demolished humans at chess, people could dismiss its victories, saying the game really is just math, perfect for a computer. Oren Etzioni is a computer science professor at the University of Washington in Seattle.
Professor OREN ETZIONI (Computer Science, University of Washington): People said, OK, that's amazing. But ultimately, look, this is chess. This is something that's very precise. It's very constrained - right? Black and white, if you will.
GREENFIELDBOYCE: Chess didn't involve anything really quirky or fuzzy or ambiguous, nothing really human.
Unidentified Announcer #2: This is "Jeopardy!"
GREENFIELDBOYCE: Now, "Jeopardy!" on the other hand, is a game full of tricky wordplay. The clues contain little jokes, puns, hidden meanings; the kind of human language that has traditionally baffled literal-minded computers. That's what made "Jeopardy!" such an enticing challenge for IBM.
About two dozen researchers spent four years building a computer system called Watson. Last month, it squared off in a short practice round against the two top "Jeopardy!" champions, Ken Jennings and Brad Rutter. The two men stood behind podiums in front of a big "Jeopardy!" board. Between them was a black, rectangular computer screen. Watson hit the buzzer again and again.
Unidentified Man: Watson?
WATSON: Who is Mary Leakey?
Unidentified Man: You're right.
WATSON: Eight hundred, same category.
Unidentified Man: Harriet Boyd Hawes was the first woman to discover and excavate a Minoan settlement on this island. Watson.
WATSON: What is Crete?
Unidentified Man: Yes.
WATSON: Let's finish: Chicks Dig Me.
GREENFIELDBOYCE: Chicks Dig Me.
(Soundbite of laughter)
GREENFIELDBOYCE: That was the category name. The questions were about women and archaeology.
David Ferrucci leads the team that made Watson. He says never mind figuring out the answer; the game's first hurdle is figuring out the question.
Dr. DAVID FERRUCCI (Senior Manager, Semantic Analysis and Integration Department, T.J. Watson's Research Center, IBM): I mean, the computer has to find out, you know, where are the individual words and then how do the words group together? You know, what's the verb, what's the subject, what's the object, what's the preposition, what's the object of the preposition?
GREENFIELDBOYCE: He says even after doing that, Watson still doesn't understand a word's meaning, like we do. Humans have real-world experience. We know chicks. We've been to islands.
Dr. FERRUCCI: When we hear language, we bring so much context to interpreting the question that we come up with some sensible and reasonable answers. The computer struggles with that.
GREENFIELDBOYCE: To aid in this struggle, Watson's creators gave it awesome speed and memory. It searches through some 200 million pages of reference material - everything from the Bible to encyclopedias to novels - looking for the words or phrases in the "Jeopardy!" clue, and seeing what other words are frequently associated with them on these pages. This produces lots of potential answers. But which one is likely to be correct?
Dr. FERRUCCI: And so it reads other things and says, does this passage support this as the answer?
GREENFIELDBOYCE: Watson knows how to weigh all this evidence because it knows the game. It's studied thousands of "Jeopardy!" clues plus their correct solutions. In mere seconds, Watson comes up with a list of possible answers, and ranks them. The top choice is the one it has the most confidence in.
Dr. FERRUCCI: And if that confidence is above a threshold, it says: Oh, I want to buzz in.
Unidentified Man: Watson.
WATSON: Who was Tennyson?
Unidentified Man: Yes, Alfred Lord Tennyson.
GREENFIELDBOYCE: Watson can make mistakes. Ferrucci says the errors can be funny and revealing.
Dr. FERRUCCI: You know, of my favorites is, you know, what do grasshoppers eat? And it came back and said: kosher.
GREENFIELDBOYCE: How do you think it got that answer?
Dr. FERRUCCI: It turns out, grasshoppers is an actual kosher food.
GREENFIELDBOYCE: Experts say that even with the occasional mistake, Watson has what it takes to perform very well against humanity's best.
Oren Etzioni expects to see a Watson win.
Prof. ETZIONI: Does that mean that it's game over for humans, that robots will keep us as pets? Absolutely not. But it does mean it's a demonstration that we've significantly expanded the envelope of what computers can really achieve.
GREENFIELDBOYCE: The contest will be broadcast over three nights, starting this evening.
Nell Greenfieldboyce, NPR News.
NPR transcripts are created on a rush deadline by Verb8tm, Inc., an NPR contractor, and produced using a proprietary transcription process developed with NPR. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.