Newsgroups: comp.ai.games
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!gatech!news.mathworks.com!news.kei.com!wang!news
From: bruck@actcom.co.il (Uri Bruck)
Subject: Re: What's really being done
Organization: ACTCOM - Internet Services in Israel
Date: Wed, 3 May 1995 18:49:52 GMT
Message-ID: <D80Mz5.HHx@actcom.co.il>
References: <mrizzo.13.00126DCC@mpd.tandem.com> <3niphh$cpe@nntp.interaccess.com> <3nn4jr$ap9@case.cyberspace.com>
Sender: news@wang.com
Lines: 35

Dan Thies (rtfm@cyberspace.com) wrote:
: denryan@flowbee.interaccess.com wrote:
: : Err, I have a degree in AI, Mike, and I had the same questions as you, until
: : I eventually realized that "AI", in the context of games, means "the
: : code that handles the opponant, the adversary, etc.". It doesn't
: : necessarily indicate canonical AI. It could, maybe it should, and
: : certainly, it will, one day.

: One of the great benefits of this newsgroup should be the opportunity it 
: provides for people like you to share what you know with people who want 
: a better computerized opponent.  In my experience, applying real AI 
: methods to computer opponents makes a better game, with less fudging 
: required.  One of the projects I'm involved with is a 3-d space dogfight 
: type of thing, and using neural nets has made the computer opponents much 
: tougher and more importantly, unpredictable.  We're still basically 
: emulating the brain of a jellyfish, but that's better than no brain at all.

: Dan (a mind is a terrible thing)
: -- 
: rtfm@cyberspace.com

Using AI concepts certainly makes for better computer oponentes. I have been
working on a strategy war-game for a while, actually for several whiles,
I tried implementing a hierarchical structure, the higher you go on the 
command hierarchy, the more information you get on the game state, but
in less detail. This also means that while field units shuld basically follow
the instructions of their commanders, and react to what ever happens "now"
(i.e. real time), the command-level routines whoud be able to plan
ahead a little in order to use their forces effectively. There was no
escaping using some AI concepts, specifically some goal-oriented searching
with mid-path evaluations in order to evaluate the execution of plans,
and react to changing conditions.
Uri
bruck@actcom.co.il

