March 15, 2011

Watson and Kasparov


Around work, it's well-known that I'm a bit of a chess nut, always up for a game. I'm not a very strong player, but I understand the game well enough to appreciate the play of those who are. I love games: board games, puzzles, computer games, crosswords, you name it.

So I was quite interested in the recent Jeopardy contest in which IBM's Watson computer program competed, and won.

This, of course, is not the first time that IBM's computer programmers have taken on the best human game players and defeated them: I still vividly remember the Kasparov-versus-DeepBlue chess matches of the late 1990's. Interestingly, Kasparov recently wrote a fascinating and detailed memoir of the experience for the New York Review of Books: The Chess Master and the Computer. If you haven't seen his article, I recommend it highly.

As Kasparov notes, building a program which played better chess than a human turned out to take us in a different direction than originally anticipated:

The AI crowd, too, was pleased with the result and the attention, but dismayed by the fact that Deep Blue was hardly what their predecessors had imagined decades earlier when they dreamed of creating a machine to defeat the world chess champion. Instead of a computer that thought and played chess like a human, with human creativity and intuition, they got one that played like a machine, systematically evaluating 200 million possible moves on the chess board per second and winning with brute number-crunching force.

Not surprisingly, Kasparov's observations on Watson's victory are both insightful and rather similar to his analysis of DeepBlue:

Much like how computers play chess, reducing the algorithm into "crunchable" elements can simulate the way humans do things in the result even though the computer's method is entirely different. If the result—the chess move, the Jeopardy answer—is all that matters, it's a success. If how the result is achieved matters more, I'm not so sure. For example, Deep Blue had no real impact on chess or science despite the hype surrounding its sporting achievement in defeating me.

As Mig Greengard, one of Kasparov's long-time assistants, notes, widely available search engines such as Google already do a similarly fine job of searching for knowledge and displaying the results of that search.

Interestingly, as Kasparov notes, although computers never turned out to play like humans, increasingly humans are learning how to play like computers. The widespread availability of powerful cheap chess-playing computer software has dramatically improved and enhanced the world of chess. Where once only a few players such as Kasparov had access to these tools, now the use of a sophisticated computer chess assistant is routine, and the result is that the top players are more closely competitive than ever before.

Here in America, the best current player is Hikaru Nakamura, who was born in Japan but moved to the United States when he was two years old. Nakamura's use of computer chess-playing software for training and preparation was described in a wonderful article last year by Debra Littlejohn Shinder: The Role of Computers In Planning Chess Strategy. As she says, "you grow stronger in your game by playing against those who are equal or higher in rating than you," and your tireless computer is ideal for this practice and training.

So where does this leave us, as we consider the implications of the Watson result, and the future of so-called "Artificial Intelligence"? I find myself agreeing with Professor Sean Dorrance Kelly of Harvard and Professor Hubert Dreyfus of Berkeley, who write:

Watson’s ability to process natural language, to resolve linguistic ambiguities and to evaluate the puns, slang, nuance and oblique allusions characteristic of a typical “Jeopardy!” question or category is impressive indeed. Nevertheless, the arrival of our new computer overlords is not exactly around the corner.

As Kelly and Dreyfus observe, the real key is not knowing when you are right, but knowing when you are wrong:

Knowing when it’s relevant to pay attention to the mismatch and when it’s not is trivial for a human being. But Watson doesn’t understand relevance at all. It only measures statistical frequencies.

Computers are wonderful tools, and competitions such as the IBM DeepBlue and Watson events are great for focusing effort on advancing our knowledge of how to build more powerful and more useful tools. But let's not waste time fretting about the computer's "intelligence" or worrying about whether it cheated or not; instead, let's keep trying to improve our own abilities and skills using these wonderful computing tools as our tireless assistants.