Recently we said the construction of a machine for playing perfect chess was demonstrably impossible, no matter how far mechanical science might progress. This is confirmed in a most interesting paper of about 8000 words in the "Philosophical Magazine" of March 1950, by Claude E. Shannon. It was first presented at the American I.R.E. Convention, New York, in March 1949. Shannon points out that, even assuming such a machine could be constructed, and operated at the rapid rate of one chess variation per micro-micro-second it would take a rather long time to work out its first move - the number of years being at least 10 to the 90th power. This means that if it had been constructed by some quite exceptionally intelligent brontausaurus and had been permitted to "think" uninterruptedly ever since, it would still be only beginning its calculations. Civilisations would arise and decay - not just a few such as the world has seen so far, but billions of them, if the earth lasted long enough - and the computer would still be plugging along. Perhaps at last, with all mankind extinct, some future brontausaurus would see the machine finally disgorge 1. PK4.
Fortunately chess is, from its nature, not amenable to mechanical computation.
However, the author develops a thesis that a modern general purpose computer could be made to play a "tolerably good" game of chess.
Tuesday, May 26, 2009
Chess World on Chess Robots
The 1 January, 1951 issue of Chess World contains an interesting article entitled, "Chess Robot". The first couple of paragraphs read:
Posted by The Closet Grandmaster at 9:01 pm