Can A Computer Become A Jeopardy! Champ? Coming soon: a first-ever human vs. machine Jeopardy! competition. The show will air in February, with two matches being played over three consecutive days. Watson, an IBM computing system, will challenge top Jeopardy! winners Ken Jennings and Brad Rutter. Host Guy Raz speaks with Dave Ferrucci, the scientist leading the IBM Research team that created Watson. The computer is programmed to rival the human ability to answer spoken questions with speed and accuracy.
NPR logo

Can A Computer Become A Jeopardy! Champ?

  • Download
  • <iframe src="" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript
Can A Computer Become A Jeopardy! Champ?

Can A Computer Become A Jeopardy! Champ?

  • Download
  • <iframe src="" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript

(Soundbite of music)

GUY RAZ, host:

A few years ago, an IBM computer beat Garry Kasparov on a chess match. The question is, can a computer become a "Jeopardy!" champion? IBM seems pretty confident it can.

Four years ago, the company started to work on a computer they called Watson. And programmers have been fine-tuning it ever since. And in about a month, Watson will make its debut appearance on "Jeopardy!," playing against two of the most formidable players of all time.

David Ferrucci is IBM's chief scientist behind Watson.

Dr. DAVID FERRUCCI (Researcher; Senior Manager, IBM T.J. Watson's Research Center): Watson will appear like to the audience as essentially a screen, a screen of some form where there's a design on that screen and it reacts to Watson's different states, whether it's sort of thinking or speaking. And that's all you'll see at the podium.

Off stage are the actual computers, the hardware that is trying to understand and answer the questions.

RAZ: How does Watson know the answers? I mean, there are an infinite number of topics, questions that could be asked. I mean, is he just loaded with information? Does he have access to infinite data?

Dr. FERRUCCI: No, it doesn't. In fact, Watson, when he plays "Jeopardy!," is completely self-contained.

RAZ: Not connected to the Internet?

Dr. FERRUCCI: Not connected to the Internet. Moreover, I mean, think about it, even if you were connected to the Internet and you got - looked at the top 10, 20 documents, let's say the answer was in there somewhere, how do you know what the right answer is?

RAZ: Right.

Dr. FERRUCCI: You have to have a deeper understanding. And that's where all sort of the analytics come in, these algorithms that run in his hardware to analyze all that data from lots of different dimensions to decide I think this answer is more right than this other one, and I'm going to compute for the probability.

RAZ: David, let me play a clip of Watson. This is a sort of a trial run on "Jeopardy!" answering the question correctly.

Unidentified Man: You got to know when to hold them, know when to fold them. On July 28, 1994, this Texas ranger with a familiar name did. Watson?

WATSON: What is Kenny Rogers?

Unidentified Man: That is correct. Once again, Watson...

RAZ: He's talking about Kenny Rogers, a baseball player, not Kenny Rogers the singer who sang that song. So that was a trick question. How did Watson figure that out?

Dr. FERRUCCI: Watson looks at the different parts of the clue and looks at the date, looks at the names, looks at the times that were referred to and pieces it all together. So it may find many, many documents that refers - use various elements in the clue, so it's got to knit all those pieces of information together.

RAZ: But there are cases when it gets a wrong. For example, there is the question to identify the people with the initials L.B. in the R.E.M. song "It's the End of the World As We Know It." And here's how Watson replied.

Unidentified Man: In R.E.M.'s "It's the End of the World As We Know It," two of the men with the initials L.B. Watson?

WATSON: What is I feel fine?

Unidentified Man: Ooh, no.

RAZ: I feel fine is a lyric in the song, Dave.

Dr. FERRUCCI: Right.

RAZ: And as an R.E.M. fan, I should say the answer was it could have been Lester Bangs, Lenny Bruce, Leonid Brezhnev, Leonard Bernstein. Have you guys worked out the kinks?

Dr. FERRUCCI: So I guess Watson is not an R.E.M. fan.

RAZ: No, he's not an R.E.M. fan.

(Soundbite of laughter)

RAZ: But they listen to this show, so they're not going to be happy to hear that.

Dr. FERRUCCI: I think that's one worry. It's really not understanding the clue. It's not interpreting the need for those initials, or it's not weighing that enough. And this is an example of Watson looking, again, sort of many different dimensions of the clue and what it's reading and analyzing. Putting it all together in the end, then that isn't favoring the right answer.

RAZ: So now you guys are clearly confident in all the refining you've done over the years that you're taking Watson to "Jeopardy!" You're going to put it up against two of the best "Jeopardy!" players of all time: Ken Jennings, he had the longest winning streak; Brad Rutter, he won the most money. Who should I put my money on?

Dr. FERRUCCI: I couldn't tell. That wouldn't be fair if I told you that. What I could tell you is that Watson wins and loses. We've played a series of 55 sparring matches here at IBM against former Tournament of Champion players.

And while we can't release that record yet, I can tell you that, you know, Watson did well enough to give it his television debut. I could also tell you it won and it lost. So it's going to be an interesting, and I think, at least for me, edge of my seat contest.

RAZ: That's David Ferrucci, the scientist who led the team that created the IBM computing system Watson, which is set to appear on "Jeopardy!" next month.

David, thank you.

Dr. FERRUCCI: Thank you.

Copyright © 2011 NPR. All rights reserved. Visit our website terms of use and permissions pages at for further information.

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.