In principle, chess is a relatively easy game for a computer to play well. In the first place, the battlefield is small - only 64 squares - so if you have enough computer power, deep-level tree searches can be quickly accomplished. Secondly, the evaluation of a position is based on material advantage, so if the computer can analyze positions eight- to ten-ply deep and find that playing a particular line wins it a piece or even a pawn, it will play that line.
Human intelligence is not brute force! If one wishes to mimic human intelligence on a machine, chess is not the game best suited to accomplish this. The game that IBM and others in the field of artificial intelligence should be concentrating their efforts on is the oriental strategy game of GO.
GO is played on a 19 by 19 grid, making up a playing field of 361 points. The playing pieces are black and white stones, with one player having 181 black stones and the other player having 180 white ones. The object of the game is to simply control more of the intersections (or territory) than your opponent. There is also a rule of capture stating that if a stone or a group of stones is completely surrounded they can be removed from the board.
At first glance, it might seem that go would be an easy game for a computer to play well. The black and white playing pieces are equivalent to the plus and minus bits of the computer's language; evaluating the best move is simply calculating the move that gives the most territory. However, what might be a good move locally, might be a bad move globally. Determining the best move requires subtle judgment which takes into account the various weaknesses and strengths of positions in different parts of the board. This is a skill which no computer and its program will ever have.
A foundation in Taiwan, The Ing Chang-ki Goe Educational Foundation, is offering US$1,000,000 to the programmer of the first computer program that can beat a selected twelve-year-old GO player. At present, the best computer programs today play at a level of around 10-class, which is very weak, a bit stronger than an absolute beginner.
Now Professor Elwyn Berlekamp has upped the ante by offering US$5,000,000 to the first computer program that can beat a 5-dan player, which is equivalent to a strong expert player.
Professor Berlekamp is a wealthy venture capitalist who is attached to the Mathematical Sciences Institute at the University of California at Berkeley. Professor Berlekamp and his team are developing a mathematical theory of GO. So far their accomplishments have been modest: they have only succeeded in formulating a theory for the very end of the game, where the moves are only worth two points. They are presently working to extend this theory to cover endgames where the value of moves are worth up to four points. His book, "Mathematical Go Endgames", is available from Ishi Press.
More information can be obtained by writing to:
The Ing Chang-ki Goe Educational Foundation 4F Kuang Fu Building, No. 35 Kuang Fu South Road Taipei, Taiwan Republic of China Phone: 886-2-761-4117; FAX: 886-2-768-6940
Professor Elwyn Berlekamp, Mathematical Science Institute University of California at Berkeley, Berkeley, California
If you want to learn more about GO, visit the following home pages. Best information source and computer GO link-ups: