Newsgroups: comp.ai.games
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!gatech!news.mathworks.com!news.kei.com!nntp.coast.net!dispatch.news.demon.net!demon!usenet2.news.uk.psi.net!uknet!usenet1.news.uk.psi.net!uknet!uknet!newsfeed.ed.ac.uk!hwcee!meszaros
From: meszaros@cee.hw.ac.uk (Tamas Meszaros)
Subject: Re: AI MUD agents
Message-ID: <DuqJI8.BAs@cee.hw.ac.uk>
Sender: news@cee.hw.ac.uk (News Administrator)
Organization: Dept of Computing and Electrical Engineering, Heriot-Watt University
X-Newsreader: TIN [version 1.2 PL2]
References: <Pine.SGI.3.94.960717135507.15480A-100000@golgi>
Date: Thu, 18 Jul 1996 10:51:44 GMT
Lines: 59

Joseph Strout (jstrout@ucsd.edu) wrote:
: It would be great to see a MUD (Multi-User Dungeon game) dedicated (or at
: least open to) "players" which are really AI programs.  It would work just
: like a normal MUD, though you might make it a little easier for newbies to
: be able to eat... but all the players would be programs rather than
: people.

I've been thinking about this, too.

It's quite easy to build (let's call it) a bot which do the basic things
like moving, speaking, eating, fighting (self defense or attacking others),
etc. Most of the "smart monsters" in MUDs do these things.

More difficult problems are for example the cooperation and coordination
(as Joseph mentioned), or "talking that makes sense" (like Eliza, Julia),
or solving a non-precoded problem, and other things that are part of the AI
and agent research.

In fact, there are such "AI bots", with limited goals (mostly for talking,
like Julia), see for example

Michael L. Mauldin: Chatterbots, Tinymuds, And The Turing Test ...
(Presented at AAAI-94), or

http://www.cs.umbc.edu/agents/muds/

The MUDs are open to such bots because they are the very same as human
players from their point of view.

: You could host periodic contests where everyone starts from zero, then
: have at it, and the highest-level agent at the end of a month (or
: whatever) wins.

This would be a very challenging competition for agent builders.
I was thinking about something similar. Recently, my research was
partly related to the social issues of agency and I imagined a
(restricted) MUD-like world as a framework for simulating societies.

: Anybody else think this is a neat idea?  I welcome your comments.

I think it wouldn't be difficult to open (or find) a MUD for such a
competition. However, it might be necessary to build a restricted world
with limited scope (i.e. to focus on restricted problem solving,
or on communication, or on other issues). (My imagined world concentrated
on the social issues, especially on modeling the emotional states and
relations.)

I agree with Joseph: this would generate very useful lessons for many
real-world applications. Let's see if anybody else is also interested
in this. :)

Tamas

---
Tamas Meszaros
Dept. of Measurement and Instrument Eng.
Technical University of Budapest, Hungary
email: meszaros@mmt.bme.hu

