Stephen Baker, author of "Final Jeopardy: Man vs. Machine and the Quest to Know Everything," spoke with Don Tennant about IBM's Watson project and the story behind the widely publicized Jeopardy tournament between Watson and two Jeopardy champions that will air this week.
Tennant: You wrote in the book about having almost beaten an early version of Watson. Have you since had any indication of exactly how elementary that version was?
Baker: No, I really don't know. They told me afterwards that the version I was so proud to have come so close to beating was actually an old version dragged off of somebody's laptop. For all I know, it was the cutting-edge version and they were just saying that to try to make me feel bad.
"When we think we know things, when we think we're sure about them, we close our minds to further inquiry."
Tennant: You wrote about IBM's concerted effort not to make Watson too human-like, given that the economy was tanking and people were already fearful of a lot of things, including being replaced by technology. In that context, you quoted an IBM executive who said, "We evaluate every job, and we calculate whether it could be handled more efficiently offshore or by a machine." Do you think there's enough damage control in place to prevent any backlash from overpowering the positive vibe that IBM has spent untold resources to create?
Baker: I think they've put a lot of thought into this. They've been dealing with these issues for decades, because IBM has been a leader in offshoring tech work -- they've built up enormous operations in India. And they've also been a leader in building machines that replace people. That's basically why we have machines-tractors replaced people. So IBM is used to this, and they've got entire operations that should be ready to deal with this. But that doesn't mean there aren't going to be loud and persistent fears and complaints about Watson and machines like it.
Tennant: To your knowledge, has IBM placed a dollar figure on what this whole thing will have cost, when all is said and done?
Baker: My knowledge is that they've refused to tell me what that dollar figure is. I think it's safe to say it's tens of millions of dollars, and it could be even more. A lot of that stuff is really kind of hard to quantify, because Watson is built of state-of-the-art Power7 servers, for example. How much does IBM Research "pay" for those servers that are part of Watson's body? So there are difficult accounting issues.
Tennant: IBM has laid off thousands of workers since the Watson project began. Have you heard any backlash from workers who argue that the money spent to build a machine to play Jeopardy could have been better spent investing in them so they could keep their jobs?
Baker: I have not heard that line of argument. I could see where people could make that claim, but I think if IBM is going to continue to employ people in 2020 and 2030, it's going to be because they've built some of the best machines around. And the only way to do that is through research and development. Now, you could say that the Jeopardy thing is a gimmick, and it is. But the machinery that they've built to play Jeopardy is question answering, which has some very powerful real-world applications.
Tennant: I found it interesting that one of the Watson team members, James Fan, came from China. Can you elaborate on his role in the project?
Baker: One of the things that was interesting about him was that he was new to IBM when this project started. One of [chief scientist and Watson team leader David] Ferrucci's greatest concerns going into this project was that IBM would spend tens of millions of dollars building this fabulous question-answering machine, and publicize it, and just before the Jeopardy match, some hacker in a basement all by himself would create a machine that was just as good, using Google and Wikipedia and what not. That was a nightmare scenario-it would be hugely embarrassing to IBM. So Ferrucci assigned James Fan to try to be that basement hacker, to come up with a Watson-like system.
Tennant: Do you have any sense of how close Fan came?
Baker: He came close enough to show that, No. 1, the legacy question-answering system that IBM had in place was woefully inadequate for the job. No.2, Fan's own system was inadequate for the job, so that left the only conclusion that they would have to build an entirely new system. But they would use some of the ideas that Fan developed.
Tennant: You've written that Watson can teach us humility. Can you briefly encapsulate your thoughts on that?
Baker: Well, Watson is never sure about anything. Watson does all its calculations based on how certain it is that its responses are correct. It can look at even the most basic sentence: "Pat Nixon, Richard Nixon's wife, was First Lady of the United States." It can look at that sentence and think that it understands, but it's not entirely sure that it understands. Because it's not sure that it understands, it can't be sure any of the answers are correct. I think this is a useful thing, because many of us are sure about many things. When we think we know things, when we think we're sure about them, we close our minds to further inquiry. This means that there are certain areas in our lives and in our society that are beyond debate. The fact is, if you open your mind and say, "You know, I think this is true, but I could be wrong," if you do that, then you can enter into dialogue with people, and perhaps it can change your mind. This is really important in democracy, it's really important in diplomacy. I wish more of it existed in our society-in places like the Middle East, where people are absolutely sure about so many things. And what they're sure about is at the root of so many disagreements.
Tennant: What conclusion would you draw about what Watson means to the rest of us and what our interaction with computers will be like in the future?
Baker: I think the development of Watson is sort of the next step in computers becoming an external lobe of our own shared human brain. We already have that, to a degree, with Google, and machines like Watson are going to accelerate this process, where we're going to have all kinds of information and analysis, ever more sophisticated, available to us 24/7, almost as if it were in our own heads. So we have to decide, as a society and as individuals, what we want to learn ourselves-what does it make sense for us to put in our own heads when we have machines like this surrounding us. I think that reading this book, people will understand what Watson's weaknesses are. And that should help steer them to figure out what they can learn, and put in their own heads, to remain Watson's masters instead of being replaced by Watson.
Tennant: Have you come away from the project more uplifted by or more wary of what the future holds with respect to the impact of thinking machines on humanity?
Baker: I'm definitely in the uplifted camp. I'm excited about this-I think it's a terrific thing. There are some things we have to watch out for, but for the most part I love the development of smart machinery. That's what humans do-we build ever better tools, and it's really inspiring that we're building tools that make us smarter.
Tennant: I was struck by the line you used to close the penultimate chapter of your book: "The machine, unlike everyone else, had no stake in the outcome." Can you share the thought process that inspired that last line?
Baker: I was ending that chapter knowing that that was the last chapter people would be reading before the match, before they got to the final chapter. So my instruction from my editor at Houghton Mifflin, Amanda Cook, was, "This has to be a cliffhanger. You have to have people on the edge of their seats." So I tried to look ahead to the match, and also I wanted to make it clear to people that a lot of humanity, including ourselves, IBM, [Jeopardy champions] Ken Jennings [and] Brad Rutter, the producers at Jeopardy, and millions of Jeopardy fans, have a lot riding on this-it's an important match in many ways, although it's also just a game. But the one player who really doesn't care one bit, and doesn't even know what it means to care, is the central protagonist, Watson.