[MUD-Dev] Neural Networks as the AI system for a MUD?

Lo Lo
Tue Dec 7 10:06:57 New Zealand Daylight Time 1999


> On Thu, Dec 2, 1999 19:15 +0000 PartyG2816 at aol.com wrote:
> 
> > The subject says it all.  I'm out researching various AI tools and i
> > hit upon  NN's. From the bit I have read, they seem rather
> > interesting, and very well  suited to a MUD's needs for AI.. But I
> > still haven't read far enough to  really /know/ if you know what I
> > mean... I thought maybe some of you more  experienced people on here
> > might have used em or know ab it about them.. And  if that fails, if
> > you can also provide a few other ideas for me to look up  after I
> > get done digging into NN's :)
> > 
> 
The short answer is no. NN's 'can' be used to provide an AI but in terms
of bang for your bucks, they're atrocious.  Short of finding a rat brain and
connecting it to your computer, you aren't going to get particularly good
results.  Most of the uses for NN's have been in pattern recognition, I
think you may be using the wrong tool for the job.  NN's have been used
successfully, to emulate the equivalent of a gerbil.

Unfortunately, I've lost the link to a handy page that had brief
explanations
of the different types of NN's and their typical uses.  Different layouts
are
used for different tasks.  Try <URL:http://www.creaturelabs.com> then click
on science for a rather interesting take on a neural network.  It is limited
though.

Ask what you want from your AI.  It is very likely to be artificial
humanity.
You don't want the AI to be dumb yet you also don't want it to be
unrealistic perfect.

Here are a few more bits and bobs you may be interested in:

Finite State Machine: Looks like a state transition diagram, I hope you
know what this one is.  If you admire games like Half-Life, Valve did it
with one of these, although they hyped it a bit by implementing superstates
to emulate emotions.

Genetic Algorithms: A technique to optimise solutions using the theory of
evolution.  Consider combining this technique with the above.  I reckon it
is possible to have a constantly adjusting combat engine which can
calibrate itself.  Evolve the FSM.

Fuzzy Logic: A model that allows variables to be analogue rather than
binary.  Actions are carried out depending upon the magnitude of the
variables through rulesets.  Difficult to define rulesets.  Perhaps this
would be better than using an FSM since the ruleset would be robust
to small changes and it is possible to tweak through random bit flipping.

Artificial Life: Complex behaviour emerging from the interactions of many
simple entities, eg: ants and bees.  Mainly been used for simulation, such
as simulating the actions of a crowd in a fire.  Entities in artificial life
are
often called agents.  Try: <URL:http://alife.santafe.edu/> (which also
happens
to have a zillion pointers to all sorts of things)

Avoid agent-based software engineering, it has nothing to do with AI.
It is just a fancy name for reversing the role of data and code, instead of
static code and mobile data, it is the other way round.

If you are that serious about AI, you would use several, maybe all, the
methods available.  Oh, finally: <URL:http://www.gameai.com>.  Incredibly
informative yet practical (hurrah).
--
Ling Lo

Opinions are mine and not those of my employer but can be yours for a
nominal fee.



_______________________________________________
MUD-Dev maillist  -  MUD-Dev at kanga.nu
http://www.kanga.nu/lists/listinfo/mud-dev



More information about the MUD-Dev mailing list