Information Technology Reference
In-Depth Information
keep adding skillful, appreciative and imaginative behaviours so that the software is
perceived as increasingly creative.
1.3.6 Beauty Is in the Mind of the Beholder
By changing the sentence “Beauty is in the eye of the beholder” to the one above,
we want to emphasise that when people appreciate/buy artwork, the actual look
of the finished piece is only one thing they take into consideration. Other things
that occupy their mind may include details about the artist and their previous work,
other pieces of art owned by the art appreciator, or they have seen in museums,
whether the artwork will increase in value, etc. Most importantly, as argued previ-
ously (Sect.
1.3.5
), people tend to take into account how a piece of art was produced
when assessing the finished product. If no information pertaining to the production
of an artwork is available, then people can fall back on general knowledge about the
struggle artists have in taming paint on a canvas, and can try and reverse engineer the
specifics of this from the paint strokes exhibited. These fallbacks are not available
for software generated artefacts, as most people have little idea about how software
works. Turing-test style experiments may seem attractive because it shows some
level of success if the artefacts being generated by a creative system are vaguely
comparable to those produced by people. However, computers are not humans, and
this fact should be celebrated, rather than hidden through Turing tests. In the visual
arts in particular, Turing-style tests ignore process and promote pastiche, both of
which are done at great peril, as expanded on in Pease and Colton (
2011
).
We argue that Computational Creativity researchers should be loud and proud
about the fact that our software is generating artefacts that humans might be physi-
cally able to produce, but might not have thought to actually bring into being. Many
people have asked why The Painting Fool produces artworks that look like they
might have been hand drawn/painted. It does seem like we are missing an oppor-
tunity to produce pieces that humans can't produce, thus supplementing global art
production, rather than producing more of what people are already good at produc-
ing. This is a valid point, which we address to some extent in Sect.
1.4.5
below.
However, automatically producing images which can't be produced by people is
easy, but not necessarily enough to demonstrate creativity. We have largely cho-
sen instead to aim at automatically producing images which look like they could
have been produced by people (because they include figurative details, messages,
intriguing references, skillful flourishes, etc.), but—importantly—have not yet been
produced by people because no one has so far thought to do so. This has the advan-
tage that audiences have a frame of reference, namely human painting, in which to
appreciate the behaviour of the software. It is for this reason that The Painting Fool
continues to produce images that look hand drawn. No self-respecting art school
graduate wants to be mistaken for another artist, and should be horrified if they
were mixed up with Picasso or Monet in a blind test. We should write software that
similarly wants to produce uniquely interesting works of art, which are not confused
with anyone else's, whether human or computer.
