From newshub.ccs.yorku.ca!ists!torn!utcsri!rpi!usc!wupost!darwin.sura.net!uvaarpa!mcnc!aurs01!throop Thu Jul  9 16:19:34 EDT 1992
Article 6351 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!torn!utcsri!rpi!usc!wupost!darwin.sura.net!uvaarpa!mcnc!aurs01!throop
>From: throop@aurs01.UUCP (Wayne Throop)
Newsgroups: comp.ai.philosophy
Subject: Re: 5-step program to AI
Keywords: chess
Message-ID: <60855@aurs01.UUCP>
Date: 22 Jun 92 17:41:11 GMT
References: <60842@aurs01.UUCP> <1992Jun18.205639.3093@mp.cs.niu.edu> <60848@aurs01.UUCP> <1992Jun20.022757.31828@mp.cs.niu.edu>
Sender: news@aurs01.UUCP
Lines: 31

> rickert@mp.cs.niu.edu (Neil Rickert)
> I tend to be skeptical about these types of depths.  The human player
> may adopt a strategy which will not pay off for 18 moves, but I doubt that
> he has carefully gone through and analyzed those 18 moves and the likely
> responses to them all.  More likely he has recognized clear evidence
> in the first two or three moves that this strategy puts him in the
> offensive and will give the opponent difficulties.

Well, that's 18 half-moves, but the point remains.

I agree that the human player hasn't carefully analyzed the full
tree to 18 plys deep, nor even a very thick "surround" to the
path from the current position to the desired one.

But the human player *has* done three things: first, verified that
the position is indeed reachable efficiently, and second that the move
sequence is plausible, and finally that there are no obvious
monkey wrenches that the oponent can throw into the works in answer
to the move sequence the player has in mind.  Doing these things is
what takes the bulk of the time after the desired future position
is "seen" (or created or chosen or however best to look at it).
Settling on the position itself (or a small suite of positions to
compare) is done quickly in comparison.

But yes, with the above caveats, and with allowances for my memory of
year-old magaine articles, and remembering that this model is a
simplification, human players (at least the ones interviewed) really do
have in mind fairly detailed play for more moves into the "future" than
a computer can handle.  At least in the mid-game.

Wayne Throop       ...!mcnc!aurgate!throop


