Information Technology Reference
In-Depth Information
Maybe the most successful approach so far (according to Boden 2004 ) has been
the use of evolutionary algorithms, i.e. simplified emulations of Darwinian evolu-
tion applied to data representations, as search techniques in open-ended conceptual
spaces, inspired by nature's creativity. The numerous examples include works by
Sims ( 1991 ), Todd and Werner ( 1999 ), Jacob ( 1996 ) and myself (Dahlstedt 2004 ;
2007 ; 2009b ).
Well implemented evolutionary systems are capable of generating interesting
novelty; they can be creative in a sense. But there are several problems with this
approach. Firstly, while evolution is good at searching spaces, it has been difficult
to design really open ended systems. Secondly, the kind of creativity it exhibits is
not very similar to human artistic creativity. It uses blind variation directly on the
genetic representation, which corresponds to the conceptual representation in my
model. In artistic creativity, the variation is instead inferred by extracting a new
conceptual representation from the current material form in whatever way this came
to be. To understand human creativity, I think we need to base our implementations
on a model of human creativity, and not on natural evolution. Evolution is one ex-
ample of how new things or ideas can be created, but maybe not how we create. See
Gabora ( 2005 ) for further discussion about this distinction.
In this context it might be interesting to consider two completely different types
of creative processes, both existing in nature, but in different domains. The first is the
reiteration of a particular generative process until it is “just right”, with evaluation
only of quasi-complete results. This is analogous to natural evolution, where each
new individual is developed all the way from the blueprint, in each generation. From
this perspective every living thing is a generative artwork. The other alternative is
the accumulated treatment and processing of a temporary form, exemplified by nat-
ural structures such as mountains, rocks and terrain. They record their own history
of coming into being, through generative and erosive processes. We may call these
generative and accumulative creative processes. So, one is typical of living things,
the other of dead matter exposed to additive, transformative, and destructive pro-
cesses. Both can be accounted for by the proposed model, with different frequency
of re-conceptualisation, and both types of process exist in art. I would say that the
accumulative process is a crucial part of human artistic creativity, with the excep-
tion of explicitly generative art. Evolutionary algorithms, as powerful as the may
be, are limited to generative creative processes, which may indicate that they are not
entirely suitable for emulation of artistic creativity.
8.4.1 Implementation of the Model
Implementing the proposed model involves several difficult and challenging prob-
lems. They are discussed below, with some preliminary speculation about possible
initial approaches.
To fully model human creativity, we would need to successfully model most es-
sential features of the human mind, which is of course impractical. However, there
Search WWH ::




Custom Search