Re: CF: Long term experimental ideas

[David Andrew Michael Noelle <dave@Straylight.org>]

Hwei Sheng TEOH wrote:
[...]
> But I perceive the implementation of this to be extremely complicated... now,
> having human players for the "dungeon crew" would probably work fine, but with
> AI monsters, things get a little tough... you'll need an AI that is flexible
> enough to handle the ever-changing dungeon: suppose your dungeon has a kobold
> servant whom you assigned to carry some stuff from your HQ to an upper dungeon
> level. To account for the possibility that some adventurer might be wandering
> down your dungeons, you'd want some way of knowing whether that kobold
> actually made it to where he's supposed to go. You'd also want to know about
> it if an adventurer is seen at a dangerously deep dungeon level, etc.. Now
> generalize this scenario to a full-fledged dungeon with all kinds of workers,
> with many different tasks, and you have to re-educate them every time you
> change the structure of your dungeon (else they'd get lost on an errand),
> plus you'll have to deal with racial squabbles, etc.. Very complex, intricate
> web of communications and dynamic AI programming, etc.. I'm sure everyone has
> seen how "dumb" an AI games like C&C clones and dungeon keeper have... well,
> they are very powerful already, yet they require constant supervision by a
> DK who can see the bird's eye view of the entire dungeon just by scrolling
> with the mouse. The AI we need here is one fool-proof enough to be manageable
> by a player who is limited by normal gameplay rules, and who probably wants to
> go on some quest rather than spend the rest of his existence maintaining a
> dungeon.

    Having studied A.I. at SUNY Buffalo for three years, then at U.C.L.A.
for two more, I can tell you this:  Egad!  I think maybe you're expecting a
bit much from Open Source.  An AI that sophisticated, and "fool-proof" (???)
would take years, lots of years.  Even with several people collaborating on
it.
    It's definitely an interesting idea, but that's exactly the problem. 
It's *too* interesting.  Ever heard the phrase "Turing Tarpit"?  Back when
computers were new and people didn't really understand them, there were all
kinds of projects trying to build computers that could write their own
programs.  A mathematician named Alan Turing told them this was a waste of
time, but nobody believed him.  So he made up a mathematical model of a
computer, now known as a Turing Machine, and proved that his theoretical
computer could do anything a real computer could do, and vice versa.  Then
he proved that with both theoretical machines and real machines, everything
is possible, but nothing interesting is easy.  And as the problem gets more
interesting, it gets exponentially more difficult.  Self-programming
computers are so interesting that they fall into a special category of
problems that are infinitely difficult.

    By all means, keep pounding out ideas.  A lot of it sounds good, and any
individual component of this might be a good step.  Just keep in mind that
an AI capable of holding together all these dynamic features seamlessly, and
in real time, is real deep in the Turing Tarpit already.  Don't drown in it.

-- 
            -Dave Noelle,                 dave@Straylight.org
            -the Villa Straylight,  http://www.straylight.org
Coalition Against Unsolicited Commercial Email  ==  http://www.cauce.com

Disclaimer:#include <std_disclaim.sig>

Quote of the Day:
"Twas brillig, and the slithy toves
   Did gyre and gimble in the wabe.
 All mimsy were the borogoves,
   And the mome raths outgrabe."  - Lewis Caroll
-
[you can put yourself on the announcement list only or unsubscribe altogether
by sending an email stating your wishes to crossfire-request@ifi.uio.no]