Low-life tactics of means to end

[WageSlave1]

Re: i think he's off

You'll probably find that half the people above you in the rankings are doing exactly what you were thinking of doing

And to think, such an idea entered the mind of a Russian. AtW, Botvinnik will be turning in his grave. Don't let down a century of Russian tradition.

Best way to beat a machine/program, is to sacrifice pieces; they find it incredibly difficult to calculate strategic advantage, let alone a tempo advantage.

[ScotsPine]

um

Chess grandmaster Micheal Adams has been comprehensively beaten 5 games
to 1 by IBM supercomputer Hydra, proving once and for all (until the
next high profile man vs. machine match-up) the dominance of silicon
over puny carbon-based life forms. Hydra can analyse an enormous 200
million moves per second, and plans its game up to 40 moves ahead, six
more than IBM's Deep Blue. The next challenge, it is suggested, is a
supercomputer able to compete at the game of Go. Currently even the best
example to come out of Microsoft's Cambridge research facility is easily
beaten by a competent human player.

[ScotsPine]

er

Michael Adams

[WageSlave1]

Re: um

Yes, yes, yes; I've posted my thoughts on this issue before, when Franco stated that Deep Blue II beat Kasparov.

Michael Adams has never been the greatest player in the world, so beating him proves nothing.
Machines don't really play chess; they access an enormous database of human games and select the best move based on historical outcome. If you play Fischer Randomised Chess, machines are very weak. Machines play tactically, not strategically.
Cars can travel faster than man, but does this mean they are better athletes than the Olympians?

[ScotsPine]

i thought they

used AI algorithms, developed in languages like prolog? are you saying they are more a rules/data based expert system with a modicum of heuristics thrown in for good luck? if so, no wonder everybody could beat my 1987 attempt in prolog.

[MilanB]

Re: i thought they

Alexey,

you sad little man !

Milan.

[WageSlave1]

Re: i thought they

are you saying they are more a rules/data based expert system with a modicum of heuristics thrown in for good luck

Pretty much. Deep Blue II beat Kasaparov a couple of years back, which everyone always refers to. But the important thing to remember is that Kasparov lost interest in the match after game 2.
Every so often they play a machine in a tournament. Usually the players treat the game as a curiosity, so results against humans don't really reflect the true reality of ability.

Offer a machine a sacrifice and it will nearly always throw them.

[ScotsPine]

Offer a machine a sacrifice and it will nearly always throw

huh! you didn't have to do that with my program. all you had to do was make a move. that usually threw it quite well.

[WageSlave1]

Re: Offer a machine a sacrifice and it will nearly always th

you didn't have to do that with my program. all you had to do was make a move. that usually threw it quite well.

SP, get the program to search a database of grandmaster games and build a 'possible moves' tree from the winning lines it finds.
For example, what is the best response for white against the poisoned pawn variant of the Scillian Najdorf. Build the line based on successful outcomes.

[ScotsPine]

but [shurely] if it is

just a search against a database on the best option from the current position, you are likely to paint yourself into an indefensible corner somewhere down the line, bearing in mind that your opponent will be operating with a different strategy/gameplan?