Newsgroups: comp.ai.games
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!udel!gatech!howland.reston.ans.net!ix.netcom.com!netcom.com!activis
From: activis@netcom.com (ActiVision)
Subject: Re: What is "real" AI???  (Was: AI routines for RPG game?)
Message-ID: <activisD5nrr4.qB@netcom.com>
Organization: NETCOM On-line Communication Services (408 261-4700 guest)
X-Newsreader: TIN [version 1.2 PL1]
References: <D5E5oo.GDw@emr1.emr.ca> <SMISHRA.95Mar14151317@kiwi.acns.nwu.edu> <1995Mar15.092901.29098@news.unige.ch> <SMISHRA.95Mar15123306@kiwi.acns.nwu.edu> <3k9sl5$avb@Venus.mcs.com>
Date: Sat, 18 Mar 1995 22:57:04 GMT
Lines: 86
Sender: activis@netcom22.netcom.com

Jorn Barger (jorn@MCS.COM) wrote:
: Sunil Mishra <smishra@nwu.edu> wrote:
: >It also does have a lot to do with the language. Could you write a routine
: >to have an arbitrary composable list like lisp does as it's basic data
: >structure in C? It would be hard.

: You only need that during the r&d phase.  If you know what your code
: neds to do, your data structures should be simple and straightforward.

If you know what your code needs to do, you're probably not doing anything that
most programmers would classify as AI.

Consider the example of writing a program that plays a killer game of chess.
10 years ago this was considered `hard;' 20 years ago it was considered an
exercise in more-or-less `pure' AI.  A programmer 20 years ago might very well
have chosen to write in Lisp in order to take advantage of Lisp's very
orthogonal set of operators over very flexible data structures, just because
they weren't sure how best to represent the game's state space.

Now, of course, chess programs are written in C/C++, and the programmer worries
more about whether to use MinMax or Alpha-Beta, or perhaps A*, and how good
their dictionary of openings and endgames is, and how much lookahead to do.

No one considers chess an AI problem anymore.  It's a relatively
straightforward evaluation and search problem.

: > Besides, most C programmers have a
: >(mis)conception that lisp crawls. It does need RAM, and an executable image
: >is hard to put together.

: And it has to pause frequently for garbage collection.

Wrong.  Lisp implementations with archaic garbage collectors have to pause,
perhaps too frequently for some users, for garbage collection.  Lisps with
modern garbage collectors don't introduce noticable pauses at all, at a cost of
approximately 5% more total CPU time than the collectors that do, and truly
macha garbage collectors can even satisfy hard real-time constraints.

: >And you are right, the problem is also in complexity. I would hate to have
: >to deal with that using C.

: Translation?-- "C sourcecode has a more complex *appearance* than LISP"?

I doubt it.  I interpret the above to mean simply that `there is some set of
algorithms/heuristics that Lisp lends itself to a more parismonious
representation of than C,' which to my way of thinking is a truism.

: Sunil, I'm afraid you're introducing into this group a sort of
: low-self-esteem *meme*, that carries a (totally false) implication that
: academic AI has accomplished these wonderful things that are just-
: around-the-corner but so-complex-no-one-in-the-game-community-can-
: understand-them...

: ***There aren't.***

How do you know?

Military, commercial, and academic AI has, in fact, accomplished things that
are amazing (SNePS, CLASSIC, and LOOM come immediately to mind), and a few
things that are bordering on the scary (Eurisko and Cyc come immediately to
mind).

I (and you) can FTP SNePS right now.  If you would be willing to sign a
license, you could FTP CLASSIC or LOOM.  I happen to be a LOOM licensee.  It's
amazing.  Too bad the license says that I can't use it commercially...

: If there ever *is*, we can talk about it when we experience it.  Until
: then, we should have some self-respect and ***assume*** that that
: talk is all hot air and ego, designed to channel the big grant
: money into the pockets of the big egomaniacs!!!

That's not an expression of self-respect; that's an expression of your own ego
and anti-academic prejudice.

: If you disagree with this, ***what are your "real" apps???***

That depends upon how you define `real app.'  If you're referring to things
that are mass-marketed and shrink-wrapped, that's an awfully small playing
field that doesn't tend to get the benefit even of the AI technology that is
transitioned out of the military/private research/academic communities.

: j
: jorn@mcs.com

Paul Snively
activis!paul_snively@netcom.com
