Information Technology Reference
In-Depth Information
Cohen). This is because it invents scenes from imagination, i.e., each scene that
it paints is different, and doesn't rely on digital images, etc. Moreover, the scenes
are figurative rather than abstract, hence the software uses information about the
way the world works, which is not often the case with generative computer art.
Cohen has used AARON to raise issues about the nature of software in art, which
has further increased the interest in the artworks it produces. For instance, he ends
(Cohen
1995
) by asking:
If what AARON is making is not art, what is it exactly, and in what ways, other than its
origin, does it differ from the “real thing?” If it is not thinking, what exactly is it doing?
The main difference between The Painting Fool and AARON is in terms of the
range of artistic abilities in the two pieces of software. For instance, the range of
scene types that can be painted by AARON have differed somewhat over the years,
but are largely limited to figurative scenes involving multiple people, pot plants and
tables in a room. We discuss later how, when properly trained, The Painting Fool
can produce pieces which depict a wide variety of scenes, including ones similar
to those produced by AARON. The notion of training highlights another difference
between the two systems. To the best of our knowledge, AARON has only ever
been programmed/trained by Cohen, and that is not likely to change. In contrast,
again as described below, we have built a teaching interface to The Painting Fool
which enables artists, designers and anyone else to train the software in all aspects
of its processing, from the way in which it analyses digital photographs to the way
in which it constructs and paints scenes. We hope that allowing the software to be
trained by artists will ultimately enable it to produce more varied and culturally
valuable pieces. In particular, while The Painting Fool will be able to draw on and
refer directly to some of the training it has been given, with knowledge of the styles
of those who have trained it, the software will also be able to find its own path, its
own style. In addition to this, we have enabled the software to interact with online
information sources, such as Google and Flickr and social networking sites such as
Facebook and Twitter, as described below and in (Krzeczkowska et al.
2010
). Again,
the hope is that the software can be trained to harness this information to produce
more culturally interesting paintings.
Future versions of The Painting Fool will be further distinguished from AARON
by their ability to critically appraise their own work, and that of others. Cohen pro-
vides aesthetic guidance to AARON by programming it to generate pieces in a cer-
tain style. However, he has not supplied it with any critical ability to judge the value
of the pieces it produces—and ultimately, Cohen acts as curator/collaborator by ac-
cepting and rejecting pieces produced by the system. In contrast, not only do we
plan for The Painting Fool to use critical judgement to guide its processing, we also
plan for it to invent and defend its own aesthetic criteria to use within these judge-
ments. For instance, it will be difficult, but not impossible, to use machine vision
techniques to put its own work into art-historical context, and appraise its pieces in
terms of references (or lack thereof) to existing works of art. In addition, we plan
a
committee splitting
exercise, whereby we use crowd sourcing technologies such
as Facebook apps to enable members of the public to score pieces produced by The
