Computer generated Art and Story

["Joe Andrieu" <joe@andrieu.net>]

Thom@indiana.edu says
> Machines will never 'create' art or story or music of any significance
because humans create significant art and story and music for human needs.

Wow.  That's an arbitrarily limiting claim.  But so juicy I have to delurk
for a moment.

I agree with your main point... that computer generated art of any kind is a
harder problem than computers as a medium for a new type of experiential
(interactive) story. That computers can never do it is  just silly.  That's
like that early computer era assertion that computers will never need more
than 64k of memory.  Just because it is beyond the bounds of current systems
or thinking doesn't mean it can't be done.

1. Get a computer to model human needs, fears, desires.
2. Then get it to incorporate current news/media into that model.
3. Have it systematically translate the "classics" by reconnecting the
needs, fears, and desires of the original with contemporary needs, fears,
and desires, maintaining the structure (timing & presentation).

Input: Romeo & Juliet, Life in New York, Love, death, pain, parents, feuds,
friends, youth.

Output: Westside Story (or similar)

Now... we can get into all sorts of arguments about ART and whether or not
Westside Story or other similar "interpretations" or "translations" are
art... and by some definitions they aren't, especially mediated by a
computer (and of course, as a /musical/ there is definitely new art in the
real Westside Story). We could also talk about whether or not the real
artist is the creator of the program that does the translation.  

But I think there is a valid definition that does includes this type of
output as Art. 

Definition I: Art as a human connection between the artist and the audience.

Ok. By that definition, the artist must be human. But try definition II.

Definition II: Art connects the audience with their own humanity.

That I think you can do, using modeling and structuralism (my own
bastardized meaning of that term) to reinterpret classic (or newly authored)
forms in contemporarily relevant contexts (or interactive contexts).

The challenge is how do you model the humanity and define the structure in a
way that both produces quality output and allows the player/audience/hero
some form of engagement/interaction/control.

To me, that's the crux, defining the structure of story in terms of human
and emotional transitions.  

Frankly, I've seen very little deep understanding of story and drama in the
interactive story efforts out there.  Story is a subtle, slippery topic and
few master it even in fields built on it, which is why we like to complain
about TV shows and movies that flop or fail to maintain their initial power.
After fifteen years thinking about story in an interactive context, I feel I
am just barely beginning to get a wisp of an aroma of the real matter
underlying story.  And what we need is to explicitly model that underlying
story matter so the computer can play with it in some fashion.

As an experiment, I like Façade's use of the classic dramatic impact of
human relationships going awry, but I think there is a lot of room for
improving the structure to be more than just a loose, partially ordered
bundle of beats. It is certainly the best effort I've seen so far (by the
light of my own goals) and I know Michael & Andrew made great headway at
that sort of deep understanding of story I'm talking about.  But there is so
much more to be done.

I'll have to look at the other recent examples posted here, although it
seems like they are still predominantly relying on a human in the RPG to
make the artistic manifestation of the narrative turn of events.  But
perhaps there is some insight into how they are managing narrative
structure. 

Speaking of Façade, did anyone here go to the AAIDE conference this week?

-j

--
Joe Andrieu
joe@andrieu.net
+1(805) 898-9389

-----Original Message-----
From: owner-idrama@mail.flutterby.com
[mailto:owner-idrama@mail.flutterby.com] On Behalf Of thom@indiana.edu
Sent: Saturday, June 04, 2005 12:52 AM
To: idrama@flutterby.com
Subject: Re: Interactive storytelling and me; and a challenge

Quoting "Brandon J. Van Every" <vanevery@indiegamedesign.com>:

> I think I'd rather have my AIs generate art assets, ala procedural 
> modeling.  If an AI could do it decently, then the results are "known to 
> be saleable."  Seems like visual art generation would be a lot easier 
> than story generation.  Crap just has to look interesting.

This is a pretty silly statement since the quest for automated art is about
as 
long and successful as the quest for automated story or the quest for
automated 
music.

Glassner in his Interactive Storytelling book points out the
responsibilities 
of the author and reader and points out that interactive storytelling
violates 
these responsibilities by asking people without real training as writers and

storytellers to take on the roles of writers and storytellers. You can 
substitute artist and viewer above and apply this thought to visual art or 
musician and listener for music as well.

Art of any sort is difficult and often takes decades to master before one
can 
actually create 'art' of any sort. Machines will never 'create' art or story
or 
music of any significance because humans create significant art and story
and 
music for human needs. Heck, the visual aspect of visual art barely exists 
these days simply because of the function of 'art' these days. Art isn't 
something which matches the couch any more.

-Thom