Newsgroups: comp.ai.games
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!gatech!news.mathworks.com!nntp.primenet.com!netcom.com!virtuall
From: virtuall@netcom.com
Subject: Re: Programable AI, Why?
Message-ID: <virtuallDw1DC3.1I4@netcom.com>
Organization: NETCOM On-line Communication Services (408 261-4700 guest)
X-Newsreader: TIN [version 1.2 PL1]
References: <Dvz486.A37@mv.mv.com>
Date: Mon, 12 Aug 1996 17:45:39 GMT
Lines: 29
Sender: virtuall@netcom23.netcom.com

Steve Hodsdon (hodsdon@scoot.netis.com) wrote:
: Does John & Jane Q. Public really want to have programmable AI?  Or
: are we creating something that only a few of us (AI types) are
: interested in?

As far as I am concerned, the target audience for writing programmable AI
is ME and the rest of my development team.  If I hard-code some big state
machine, and in final QA it turns out to be dumb as a brick, I've got a real
problem.  If, however, it is a programmable system, I (or even the testers)
can easily change the programming in the final mix.  

: I can't see too many people even having the ability to write a new AI
: from scratch.  Does this mean that the AI portion of the program is
: supplied in source form?  Or are we, as developers, expect to sell an
: update disk with a handful of "new & improved" opponents?

Well, programs don't have to be written at such a low level.  If my programming
language is simply a way to define transitions between several non-modifiable
behaviors and set goals, then it should be easy enough for a non-programmer
to fiddle with my AI behaviors.  This 'AI language' could be as simple as a
text file or as complex as something parsed by LEX/YACC modules in my game.

After having done it both ways, I would say some level of programmability
(preferrably fairly high/abstract) is a required part of the design.
----
Robert Huebner
LEC


