Opinion Man vs. Machine By SCOTT T. HARDS	フレームで見る  英文をプリントする  脚注をプリントする
人間対機械の戦い ニューヨークで開かれたチェスの世界チャンピオン、ガルリ・カスパロフ氏とIBMのコンピューター「ディープブルー」との戦いは、5月11日、2勝1敗3引き分けでコンピューターの勝利に終わった。現役の世界チャンピオンにコンピューターが勝利したのは初めてで、機械が人間の知能を超えたかのように報道された。だが、筆者は…。

When IBM's chess computer known as Deep Blue beat the longtime world champion Garry Kasparov a few weeks ago, newspapers and magazines all over the world were filled with editorials and columns commenting on how, in some ways, this showed that machines were now better than man. Even in Japan, there was talk that it would only be a matter of time before a shogi or igo program would be beating the best human players. That's probably true, but statements suggesting this is some kind of revolution in machines cause me to chuckle, because they are really just stating the obvious.

Of course machines are better than people at the jobs they are designed to do! That's why we build them in the first place! We use calculators because they do math faster than we can. We use cranes because they can lift heavier things than we can. If machines were not better, we wouldn't bother with them! After all, nobody is writing a column about how sad it is that a steam shovel can dig a hole faster than a person with a spade.

Kasparov's loss to Deep Blue generated conversation because it made it seem as if a machine could think better than a man. But it actually only proved that chess is a giant math problem that can be solved through calculations. Think about it: all through the match we heard time and again about how powerful Deep Blue was, how it could evaluate 200 million positions every second

I suppose some people may consider that powerful, but inefficient is another way to interpret it. After all, Garry Kasparov certainly cannot study 200 million positions a second; he probably only studies a handful every minute, but he still plays chess just as well as Deep Blue. Why?

Because Deep Blue is stupid. It has no intuition that tells it if a particular approach is completely worthless or not, so it simply evaluates every possible one and finds the best. It's just as if you were trying to decide with what to build your house and actively considered using water, string or chocolate before settling on wood. This isn't intelligence, it's just a process of elimination.

True intelligence involves learning from one's mistakes, and Deep Blue cannot even do that. Given a certain position, Deep Blue will make the same move every time, even if it leads to checkmate. Intelligence also involves creativity. When they come up with computer programs that can write novels that are more interesting than those that people write, then I'll start to worry!

Shukan ST: June 6, 1997

(C) All rights reserved