Information Technology Reference
In-Depth Information
1.3 Guiding Principles
The building of creative systems requires overcoming numerous technical problems
both of a general nature and in the particular domain within which the system works.
Given that the results arising from such systems are ultimately for general consump-
tion, building AI systems to create culturally interesting artefacts also requires a
certain amount of framing and promotion. Over the years, we have developed the
following seven principles to which we try to adhere when building creative soft-
ware and which we hope may be useful frames of reference for other people building
similar systems. They stand as a paradigm within which to build, test, employ and
promote the output of creative software.
1.3.1 Ever-Decreasing Circles
We start with the observation that it is much easier to put together artificially intelli-
gent systems if we have something concrete to work towards, especially when there
is a general and workable theory of human intelligence to guide us. This has led to a
somewhat unspoken notion in Computational Creativity that we should be looking
towards research about human creativity for guidance on how to get computers to
behave creatively. While such natural creativity research influences Computational
Creativity research to some extent, our efforts in building creative software simi-
larly influences our understanding of creativity in general. So, we shouldn't wait for
philosophers, psychologists, cognitive scientists or anyone else to give us a workable
impression of what creativity is. We should embrace the fact that we are actually un-
dertaking research into creativity in general, not just computer creativity. Hence, we
should continue to build software which undertakes creative tasks, we should study
these systems, and we should help in the goal of understanding creativity in general.
In this way, there will be ever-decreasing circles of research where we influence the
understanding of natural creativity, then it influences our research, and so on until
we pinpoint and understand the main issues of creativity in both artificial and natural
forms.
1.3.2 Paradigms Lost
The problem solving paradigm in AI research is well established and dominant. It
dictates that when an intelligent task needs to be automated, we immediately ask the
same questions: Does it involve proving something?; Does it involve generalising
a pattern?; Does it involve putting together a plan? and so on. If it is possible to
answer yes to any of these questions, then the task is pigeonholed forever as a the-
orem proving problem, or a machine learning problem, or a planning problem, etc.
This often means that the original aim of the task is lost, because only researchers in
