Re: [-empyre-] transparency+digidos



Cristiano Bianchi wrote:
> 
> > Damien Everett
> 
> > The main reason I've been working in the area of generative
> > programming is to evolve my limited imagination, and make
> > original art/music. By creating agents to inhabit new worlds
> > of computational possibility, I allow them to suggest
> > interesting manifestations.
> 
> I must reject that. I mean the very idea and concept in its essence.
> Original art/music does not need a computer for its generation. Not even
> computer/digital art. I trust the unique judgement power of the artist
> as creator. Art, ultimately, is for the fruition of human being, not
> computers (yet!). Mozart, Arvo Part and the likes CREATE ORIGINAL art
> without the aid of a computer as their source of inspiration.
> While neural net are useful and good when it comes to washing machines
> programming and genetive algorithms have been successfully employed in
> managing airports, there are tasks where a limited number of final
> results, well defined constrains and set conditions are given, for the
> resolution of a practical problem in unexpected ways. That's fine.

Indeed, though much of the public's concept of neural nets and AI is
heavily influenced by the fantasies of hollywood, most AI is employed in
extremely mundane tasks, like converting a scanned document into
editable text. However we are beginning to see more concrete examples of
machine intelligence (ie creativity) from research labs and the gaming
industry.  Many characters in recent games have what is refered to as
"domain knowledge," a less than perfect (by design) view of their world
that they build on, and draw from, to make decisions in the course of
the game.  This deviates from a rigid rule set (expert system) in that
the rule set is modified during play, so the programmer doesn't
completely define ahead of time, all the behavior a character can have.

When the artist/programmer invents or employs generative algorithms,
what they are doing is establishing a structure and an initial set of
parameters they then let loose.  if they have a lick of sense, they
watch the outcome and make aesthetic decisions, modifying their
structure and parameters to achieve a better (what ever that might mean)
result. So yes, its not terribly interesting to create a generative
algorithm, let it loose, and accept what ever result it offers as "Art"
simply because it was generative and algorithmic. with a domain system,
the creator may spend hours, even days, tweaking the "training set"
until a desired or pleasing outcome is achieved. the beauty of it being
generative is that it never works out the same way twice. but the
computer is not making the art here, the artist is, or should be,
constantly making aesthetic decisions and modifying their structure
accordingly.  All an artist ever really does is make decisions.

j




This archive was generated by a fusion of Pipermail 0.09 (Mailman edition) and MHonArc 2.6.8.