Information Technology Reference
In-Depth Information
acceptance of novel media such as software is generally rather slow. For instance,
only in 2009 did the Royal Academy in London first accept video installations for
its Summer Exhibition.
The visual art software we produce in Computational Creativity circles largely
fits into the mould of art-generating programs. However, there are two important
differences which set our programs aside from others in this mould. Firstly, there is
the underlying assumption that our software has some creative input to the process.
Sometimes, this creative input is in terms of the automatic assessment (and rejec-
tion or selection) of artefacts. Alternatively, the input may be in terms of searching
a space of artworks which can lead to surprising and interesting results. As a loose
rule of thumb, and without wanting to be too exclusive, if the software is not making
some kind of decision (whether about assessment and/or avenues of exploration), it
is unlikely to be considered to be within the realm of Computational Creativity. Sec-
ondly, Computational Creativity software can itself produce new programs, hence
it can act at a meta-level. Sometimes, this meta-level is not obvious, for instance,
the majority of evolutionary art systems produce programs (genotypes) which are
compiled or interpreted and executed to produce artworks (phenotypes). However,
the user is normally only ever shown the phenotypes, and in this sense the evolution-
ary software can be seen as an interactive art installation which enables the user to
produce aesthetically pleasing artworks. Occasionally, the meta-level is more obvi-
ous, for instance in Colton and Browne (
2009
) we evolved simple art-based games,
where the user could click on a spirograph being drawn in order to affect the draw-
ing process. If the user clicked correctly, the spirograph would be drawn to look like
a given one, which provided the game playing challenge. In this instance, therefore,
our evolutionary software was employed to produce new interactive programs for
artistic and playful purposes.
The Painting Fool is a generative art program with decision making abilities that
place it in the realm of Computational Creativity. It is definitively not a tool for
artists to use, and hence we do not make it available as such. Rather, we see it as
a fledgling artist that is being trained to act increasingly more creatively. In this
sense, our automated painter most closely resembles the
AARON
program written
by Harold Cohen and described in McCorduck (
1991
). This is a very well known
system that has been developed over 40 years to produce distinctive figurative art,
according to Cohen's unique guidance at both generative and aesthetic levels. The
software has been through a number of stages of development: early versions pro-
duced child-like simple line drawings, and in a recent novel development, AARON
has started producing abstract images.
It is over-simplistic to say that AARON has been developed to paint in the style
of Cohen, as he has been influenced himself by feedback from the software, so the
process has been somewhat circular. However, it is fair to say that AARON has not
been developed to be independent of Cohen. Taken together as a package, Cohen
and AARON represent one of the biggest success stories of AI art, both in terms
of critical appraisal, acceptance (to a certain level) by the art world and sales to
collectors and galleries. Part of this success can be attributed to AARON being
seen as creative by various groups of people (although only in a guarded way by
