On Fri, 1 Apr 2005 12:17:57 +0100, Tom Betts wrote
> > Writing programs on the fly, I got the impression that the relation
> > between statistical and deteministic properties of the algorithm gets
> > quite relevant. If there is too much randomness, modifying the program
> > does not yield any new insight in the process. But this is the case for a
> > certain durational range - very fast changes are overlooked easily enough.
> > I would guess that gp + ip is really good for these ranges.
> >
>
> Exactly, again you return to the issues faced by anyone using
> generative processes in live context. It can become a struggle
> between signal and noise (or rather the users signal and the
> computers evolving signal). Of course it becomes more about what
> controls and limitations you put in place to begin with. As the
> designer of the basic performance environment you can set yourself
> up for an edgey battle, where the GP aspects can modify high profile
> elements of the performance, global tempo, global key etc. Or you
> can limit them to evolving lower profile themes such as subtle fx
> processing or note masking etc. However if setup properly having GP
> strands could allow the performer to demote some
> experminetation/development to the computer and increase their
> supernatural multitasking skills!
Coming back to what I first blurted out, I have to admit that there is
something about the pure GP approach with fitness functions and so on, that
doesn't sit well for me.
I think the ability to be able to write some code, then have variations of it
generated to choose from is really powerful though, and doesn't remove you
from the basic process. You can edit the code if you want, and the decisions
aren't made for you.
I do think it would be fun (and marginally possible) to set up a GP system to
evolve from the classical "device driver" level, to a complex musical piece
though! :)
cheers,
dave
Received on Fri Apr 01 2005 - 12:42:05 BST