Complete.Org: Mailing Lists: Archives: freeciv-dev: April 2001:
[Freeciv-Dev] Re: Client AI development
Home

[Freeciv-Dev] Re: Client AI development

[Top] [All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index] [Thread Index]
To: Glenn Taylor <glenn@xxxxxxxxxxxx>
Cc: <cameron@xxxxxxxxxx>, <rf13@xxxxxxxxxxxxxxxxxxxxxx>, freeciv development list <freeciv-dev@xxxxxxxxxxx>
Subject: [Freeciv-Dev] Re: Client AI development
From: Cameron Morland <cjmorlan@xxxxxxxxxxxx>
Date: Thu, 19 Apr 2001 08:05:05 -0500 (EST)
Reply-to: <cameron@xxxxxxxxxx>

La 2001-04-19, Glenn Taylor skribis:

> I'd argue against this micro approach-- Brooks' subsumption architecture,
> at least originally, suffered from a lack of goal-directed behavior, and
> it's never been clear that the subsumption approach can scale well
> higher than the insect level Brooks originally demonstrated.

But ants have only insect level behaviour, and yet they are quite
successful. Under certain circumstances they will even co-operate with
other tribes of ants, essentially having an alliance. (This was reported
IIRC in New Scientist a few years ago.)

You can even get the idea, discussed by Hofstadter, of the ant hill being
sentient; "Aunt Hillary" has discussions in _Godel, Escher, Bach_ with the
anteater. We try to make intelligence emerge from chaos.

I also disagree that BB hasn't been demonstrated to scale; check out their
more recent work on Cog, a humanoid robot which they suggest displays
human-infant-like intelligence.

<http://www.ai.mit.edu/projects/humanoid-robotics-group/>

> Why is it AI discussions always sound like religious discussions? :)

Because religion and AI are both very poorly understood; many ways are
successful, each believes his way to be best, but the choice is unclear
(to an impartial observer).

I thoroughly support other tactics for AI. I like BB because it's what we
do at work, and it's quite successful for the intelligent robots we build.
Ultimately the choice of the best AI is simple: let them battle it out. If
we all examine the problem from our own unique perspectives, we'll get
interesting results.

(Actually it's not quite as simple as battling it out; requiring a Beowulf
cluster to run "the best" AI wouldn't really be an acceptable solution.)

I would actually not like to make the only AI for Freeciv, because then I
would have no way of verifying it, except for playing against humans, and
that takes a long time.

[Agh, I mustn't keep writing this, I have to study.]

+----------------------------------------------------------
| PGP http://www.eng.uwaterloo.ca/student/cjmorlan/pgp.txt
| cjmorland@xxxxxxxxxxxx
|
| Thought is not a management function.
|     --John Ralston Saul
+----------------------------------------------------------




[Prev in Thread] Current Thread [Next in Thread]