Ed Toutant is the fourth highest game show money-winner in the United States, having successfully competed on both Jeopardy! and Who Wants to Be a Millionaire. He's also a retired IBM employee, giving him a unique perspective on the Jeopardy! IBM Challenge.
Toutant has played several practice games against Watson, and was in attendance during the taping of the two matches featured in the Jeopardy! tournament. While he certainly can't give away any secrets, he was kind enough to answer a few questions about the event.
About.com: Hi Ed! Thanks so much for taking some time out of your schedule to speak with us.
Ed Toutant: Thanks, Carrie. I've been a fan of your site for several years and I'm glad to join you here.
Q: How long have you been involved in the Watson project, and in what capacity?
Toutant: IBM and Jeopardy first announced this this event in April of 2009, without providing many specifics. I've been thinking about Watson off and on ever since. In May of 2009 I spent some time with Ken Jennings at the GSN Game Show Awards, and we speculated about how the competition might work and who might be involved.
I had several contacts with the IBM Watson group over the next year, but my first direct involvement was in September of 2010, when I was invited to watch Watson play six matches against past Tournament of Champions contestants, and to meet with IBM's strategy team. Based on those observations and discussions, I suggested several ideas for new and improved game playing strategy. At least a few of my ideas were implemented, and some of the others had already been included in Watson's algorithms.
I met again with the development team the first week of January 2011, and played in eight games against Watson. These were the final practice matches before Watson played against Ken Jennings and Brad Rutter. A week later I returned and watched Watson play the televised match against Ken and Brad. My role was small compared to the many talented and dedicated IBMers who have been developing Watson for the past four years. I was honored to contribute to a great team effort.
Q: Since you've been active in getting Watson ready for the Jeopardy tournament, how have you seen "him" change and improve?
Toutant: Probably 90% of the work was already done before I had a chance to get involved, so I was blown away by Watson's abilities the first time I saw him. The most obvious improvement I saw in recent months was the appearance and design of Watson's avatar. The first time I saw it, I thought it was uninteresting, unrecognizable and unmemorable. Maybe that was just an early prototype or maybe I saw it on a day when it was having technical problems. My opinion completely changed when I saw the updated version in January, and now I love it. I think it's everything IBM could hope for and more.I met the artist who created the avatar, Joshua Davis, and told him I look forward to wearing his design on a t-shirt soon. The best way to appreciate the four years of steady improvement is to watch the excellent videos IBM has posted on its website and on YouTube. What I observed in my visits was a consistent focus on making small changes to produce incremental improvements. The one change I was most excited to see was the implementation of an unusual wagering strategy that I had suggested for Daily Doubles. I'm probably not supposed to talk about it until after the broadcasts, so I'll just tell you it will be obvious when you see it.
Q: Before you saw the IBM Challenge taping, who did you feel had the advantage in the game of Jeopardy? Why?
Toutant: I felt that Watson had the best chance of winning, for two main reasons. First, there are two humans in the game and only one computer. That means Watson should do well on the computer friendly questions, but Ken and Brad would have to split the human friendly questions. If the match had two computers and one human, it would be a different story. Watson's other big advantage is his lightning fast buzzer "thumb." In a typical Jeopardy game, all three contestants know most of the answers, so the game is usually won by the person with the best buzzer timing. Still, Watson showed some weaknesses in the practice matches. And Ken is widely regarded as the greatest player in Jeopardy history, but Brad dominated Ken in the Ultimate Tournament of Champions. So any result seemed quite possible.
Q: When playing against Watson, is there an urge to interact with him like you would with a human competitor?
Toutant: Yes, although I tried not to. The main emotion that comes out during a game is frustration, especially when your opponent outbuzzes you on a clue you want to answer. My human opponent in the games I played was David Sampugnaro, who won five games on Jeopardy in the era when five wins was the limit. It didn't really matter whether I was beaten on the buzzer by David or by Watson. I'd briefly scowl and repress evil urges toward either one of them, then try to refocus on the next clue. It is counterproductive to be emotional while playing Jeopardy.
Q: Those of us with just a basic understanding of computers and artificial intelligence might have a hard time understanding how Watson functions. Can you give a brief explanation of how his "thought process" actually works? How does Watson know to both decipher a tricky Jeopardy question and also come up with the correct answer?
Toutant: I'm far more qualified to talk about game show strategy than about artificial intelligence. The IBM people were always careful not to tell me too much about how Watson works. The basic idea is that Watson doesn't "think" in the same way that a human does. He has an enormous collection of documents in his memory. These are mostly standard English language texts, including encyclopedias, web sites, novels, almanacs, plays, lyrics, etc. Watson is programmed to look for relevant connections and relationships between key words in each clue and the raw data stored in his memory. He uses many different schemes simultaneously to generate an internal list of hundreds of possible answers, then runs verification tests on those answers to determine the most likely correct responses. If the question is especially tricky, Watson may decide that he is not confident enough to buzz in with an answer.
Q: While Jeopardy is a high-profile and easily understandable way to display Watson's potential, what are some of the other fields in which this type of technology will become beneficial?
Toutant: Possible uses for Watson's technology are almost limitless, but IBM executives are especially focused on applying it to the field of medicine and health care. If Watson is loaded with a huge amount of medical information, he may be able to help diagnose illnesses and suggest treatments. Of course, IBM wants to move beyond fun and games and into more profitable areas, so medicine is an area with a lot of financial rewards for boosting productivity.
Personally, I hope IBM will find another "Grand Challenge" in the years ahead to capture the imagination of the public. It has been fourteen years since IBM's Deep Blue computer defeated Garry Kasparov in chess (in a 1997 rematch after Kasparov won the first match in 1996). Now Watson has rekindled interest in man-versus-machine competitions that are much more challenging and potentially more important to society. Whenever this phase is over, I hope IBM will set its sights on another totally unexpected goal that seems crazy or impossible. My suggestion is for IBM to buy an NFL football team and build the next generation of Watson to be its head coach. It took Watson four years to be competitive on Jeopardy. In four more years, I'd love to see Watson2 win the Super Bowl. I just hope he's designed to survive a Gatorade shower.
Q: You and I have both been reading extensively about Watson and the tournament - in your opinion, what's the biggest misconception people seem to have about Watson and this event?
Toutant: Gosh, only one? I've seen so many! Watson is going to take our jobs and destroy humanity. A Google search would be just as good as Watson's answer.Chess is harder than Jeopardy, so what's the big deal? Watson is a computer so his answers can be never be wrong. Jeopardy matches are won by the smartest contestant. IBM would never have allowed this match to happen if they weren't certain that they would win. IBM is a boring, ancient company and it would be much cooler if Ken played against an Apple computer. Ken's the greatest, so who is that other guy? Jeopardy wants a particular contestant to win so the clues will be biased in his favor, etc.
In reality, I think the biggest source of viewer misunderstanding about any Jeopardy match is the buzzer, and it will be even more so for this match. For top Jeopardy contestants, it is not accurate to say they have quick buzzer reaction times. The buzzers don't work until Alex Trebek speaks the last syllable of a clue, then a studio crew member hits a switch that tells the contestants their buzzers are active. Human contestants try to anticipate that precise moment so they can eliminate their reaction times, but Watson will actually react to the buzzer activation signal and he is often fast enough to be successful with that technique.
Thanks again to Ed Toutant for his insight into Watson and the Jeopardy! tournament. It's a fascinating subject and I can't imagine a better showcase for IBM to introduce Watson to the world.