‘Deep Thought’ in chess

Press, 23 February 1989, Page 31

‘Deep Thought’ in chess

By

MICHAEL ROCKS

As readers of Douglas Adams’s four-book trilogy “The Hitch-hiker’s Guide to the Galaxy” will recall, Deep Thought was a computer designed to discover the meaning of “Life, the Universe and •Everything.” After some seven and a half million years contemplation the ultimate answer was the puzzling ”42.” In honour of this fictional machine the latest computer chess creation from CarnegieMellon University (Pittsburgh, U.S.A.) hrs been named Deep Thought. Late last year it placed first equal with expatriate English GM Tony Miles in a Californian tournament defeating en route GM Bent Larsen, a former world championship candidate. The defeat of Larsen was the first time a machine had beaten a leading grand-mas-ter under tournament conditions. (Larsen is rated in the world’s top 50 players.) A former world correspondence chess champion, Professor Hans Berliner, who heads the Deep Thought project, confidently predicted that within three years the son of Deep Thought would be stronger than the world champion, Garri Kasparov. While the rest of the world’s leading players were probably secretly relieved that it was Larsen and not themselves that became the computer’s first GM scalp, I doubt very much that they were all out looking for alternative employment over Christmas. On past results the predictions by the computer chess programming fraternity have been wildly optimistic, just as each chess writer who has pontificated on the advance of computer chess and cried “never” has eventually discovered that never is a very, very long time. Deep Thought, like most chess computers, operates by

brute force — attempting to calculate the outcome of all possible moves, although only to an average depth of 10 moves per variation subject to the time; limit the game is • played unden Without the restrictions on thinking time it would take Deep Thought several million years to'make its first moye, despite its ability to consider 700,000 positions per second, due to - the almost limitless number of move permutations. By comparison, the top programmes of the late seventies 1 ' only considered 4000 positions per second, still a number several orders of magnitude faster than the strongest human players. An increase in the number of positions examined each second is usually reflected in an increase in the programme’s tactical ability, but as any strong human player discovers the stronger one becomes the smaller part tactics play in determining the result of the game. Hence improved tactical ability on its own will only take a player so far. . -

Larsen lost because he became impatient and tried to beat Deep Thought at its own game, that is, outplaying it tactically. He could have saved his reputation by playing to his strengths and not the machine’s. The machine is programmed for precise calculation whereas the human thought process involves intuition, the setting of goals and the ability to break complex problems into smaller pieces. By followiing the maxim “Do nothing, but do it well” even moderately strong human players can give strong computers a difficult game as they strive for complex tactical attacks. It is interesting to note that despite the resources behind the Deep Thought project it could only tie for first place in the 1988 North American

Computer Championship with a commerical Fidelity model..

On the Christchurch scene an estimated 2500 chess computers were sold in New Zealand in 1987. Surprisingly enough it is the casual player rather than the club player who is buying them. Possibly the club player needs the human element in order to enjoy the game.

For readers without a» specialised knowledge of - computers, a good introductory text on computer chess is “The Computer • Chess Handbook" by David Levy. Levy, president of the International Computer Chess Association and an international master, is the author of the maxim quoted above. Readers interested in the reasons behind the academic interest -in computer chess might like to read the article by Dominic Lawson in “The Spectator,” January 28, 1989, available from the Canterbury Public Library. The game detailed below played a few years ago by the then world’s number two player was played at a time limit that suits computers: each player had five minutes to complete the game. White: CHESS 4.8 Black: V. Korchnoi Bird’s Opening 1 f 4 Nf6 2 e 3 g 6 3 b 3 Bg7 4 Bb2 0-0 5 Nf3 c 5 6 Bbs d 5 7 0-0 a 6 8 Be2 d 4 9 exd4 Nds 10 Nes cxd4 11 Bxd4 Nxf4! 12 Rxf4 g 5 13 Re 4 f 5 White’s Rook can no longer protect his Bishop. Korchnoi had to see this when he sacrificed his Knight on move 11. 14 Re 3 QXO4 15 c 3 Qd6 16 Nc4 Qc7

17 Rg3 h 6 18 Nca3? e 5 19 Bc4+ Kh7 20 Qhs Qe7 21 b 4 f 4 22 Rf 3 e 4 23 Rf2 Nc6 Although material is level, White’s Queenside pieces are not working. 24 Nc2 Nes 25 Be2 f 3 26 gxf3 exf3; 27 Rxf3 Bg4f White’s Queen is netted. No better was 27 Bxf3 Rxf3 28 Rxf3 Bg4 ' 28 RxfB Bxhs 29 RxaB Bxe2 30 Nd 4 Nf3+ 31 Kf2 Nxd4 32 Na 3 Qf6+ 33 Kgl Qf4 34 Ra 7 Nf3+ 35 Kf2 Nxh2+ 36 Kg2 Qf3+ 37 Kgl Qg3+ 38 Khl Bf3++

The study above is a joint composition by H. Mattison and J. Behting. White is to move and draw. Solution to No. 32: (4Q3/4p3/ 4k3/8/IP2NK2/24) Unfortunately the diagram for the last problem contained an error. The White Bishop that appeared on b 4 should have been a White pawn. Key: 1 Ke3 1 ... Kds 2 QcB Kes 3 Qd7 e 6 4Qd6++ 1 ... Kfs 2 QgB Kes 3 Qf7 e 6 4 Qf6++ 1 ... Kes 2 Qd7 e 6 3 Qd6++ 2 Qf7 e 6 3 Qf6++