Evans On Chess. May 10, 1996. Copyright Chesstours. All rights reserved.

                             MACHINES WINNING WAR

     "I think I did see some signs of intelligence in Deep Blue,
     a weird kind, an inefficient, inflexible kind that makes me
     think I have a few years left." -- World champ Gary Kasparov
     upon beating IBM's Deep Blue in a $500,000 match last February.

"Our program is so complex that we don't really understand how Deep Blue
arrives at a decision. Indeed, sometimes when faced with exactly the same
position it will make a different move than it made before!" marvelled
one of its designers.

"In three minutes it can evaluate about 20 billion moves -- enough to
consider every single possible move and countermove 12 sequences ahead,
and selected lines of attack as much as 30 moves beyond that," he told
Scientific American. "The fact that this ability is still not enough to
beat a mere human is amazing. The lesson is that chessmasters are doing
some mysterious computation that we can't figure out."

Last April an 11th massive man vs. machine challenge was sponsored by
Aegon, Holland's largest insurance firm. 50 players of varying strength
were pitted against 50 different programs in 300 games at a time limit of
1.5 hours for each side.

Nine grandmasters led the human team and won their battle. They played
six games apiece and scored 41.5 out of a possible 54 points, a neat 77%.
However, machines won the war by a margin of 25 points (162.5 - 137.5)
when humans faltered on the lower boards.

To err is human, but computers also have glitches. Take a look at this
strange encounter between Dutch grandmaster Hans Ree vs. Hitech, a
program created at Carnegie Mellon Institute by former postal world champ
Hans Berliner. White got a fierce attack in the opening but Hitech
wriggled out and could force a draw just when it stumbled into a stupid
mate-in-two at the end.

This game nicely illustrates the difference between computers and flesh
and blood. Humans wouldn't denude their kingside with 42...Qb6. Instead
they would seek a draw by repetition with 42...Qg6 43 Rf6 Qb1--but not
43...Qg5? (hoping for 44 Rxc6? Qc1 45 Kg2 Qf1 mate) 44 Ne6! Qh5 45 Rf8!
Rxf8 46 Qg7 mate.

Hitech went haywire on the next move. A draw was still expected on
42...Kh8! since White seems to have nothing better than 43 Qd4 Kg8 44 Qg4
Kh8. Now if 45 Qh5? Bb5 wards off all the immediate threats.

White: HANS REE Black: HITECH Queen's Gambit Accepted 1996

1 d4 d5 2 g3 Nf6 3 Bg2 Bf5 4 c4 Be4 5 Nf3 dxc4 6 Nbd2 Bd5 7 Qc2 b5 8 0-0
Nbd7 9 e4 Bb7 10 e5 Nd5 11 e6 fxe6 12 Bh3 g6 13 Ng5 Bh6 14 Nxe6 Qb8 15
Ne4 Nb4 16 Qe2 Bxc1 17 Raxc1 Nd3 18 Ng7 Kf8 19 Ng5 Nf6 20 Rfe1 e5 21 dxe5
Nd5 22 N7e6 Kg8 23 b3 h6 24 Ne4 Nxe1 25 Rxe1 cxb3 26 axb3 Rh7 27 N6c5 Kh8
28 e6 Bc6 29 Qb2 Rg7 30 Qe5 Qf8 31 f4 a5 32 f5 gxf5 33 Bxf5 a4 34 bxa4
bxa4 35 Bg6 Kg8 36 Bf7 Rxf7 37 exf7 Qxf7 38 Rf1 Re8 39 Qd4 Qg6 40 Nf6
Nxf6 41 Rxf6 Qb1 42 Rf1 Qb6 43 Qg4 Kh7?? 44 Rf7 Kh8 45 Qg7 mate
