l六级阅读资料.pdfVIP

1. Can the Computer Learn from Experience 计算机会总结经验吗 1 Computers have been taught to play not only checkers, but also championship chess, which is a fairly accurate yardstick for measuring the computer s progress in the ability to learn from experience. 2 Because the game requires logical reasoning, chess would seem to be perfectly suited to the computer .all a programmer has to do is give the computer a program evaluating the consequences of every possible response to every possible move, and the computer will win every time. In theory this is a sensible approach; in practice it is impossible. Today, a powerful computer can analyze 40 000 moves a second. That is an impressive speed. But there are an astronomical number of possible moves in chess—literally trillions. Even if such a program were written (and in theory it could be ,given enough people and enough time), there is no computer capable of holding that much data. 3 Therefore, if the computer is to compete at championship levels, it must be programmed to function with less than complete data. It must be able to learn from experience, to modify its own programm, to deal with a relatively unstructured situation—in a word, to “think” for itself . In fact, this can be done. Chess-playing computers have yet to defeat world champion chess players, but several have beaten human players of only slightly lower ranks. The computers have had programs to carry them through the early, mechanical stages of their chess games. But they have gone on from there to reason and learn, and sometimes to win the game. 4 There are other proofs that computers can be programmed to learn, but this example is sufficient to demonstrate the point. Granted , winning a