A Personal Perspective into the Future for Computational Creativity - Computational Creativity Research: Towards Creative Machines - page 385

One involves establishing some means for spotting apple-like candidates among
system output. This could be achieved by defining a set of desirable properties of
apple-candidates, and using those to filter system output. Another involves modify-
ing further the set of ingredients and/or the constructive procedure, constraining our
conceptual space to include only those artefacts that resemble apples.
Think of it as five separate processes. An initial process of selecting a construc-
tive procedure, followed by a process of selecting ingredients to feed the construc-
tive procedure. A third process would combine those—for instance, successively
applying the constructive procedure to the set of ingredients—to produce a set of
apple-candidates. A fourth process would establish the selection criteria for picking
desirable candidates. A final process would look at this set of candidates and pick
out the one most likely to be accepted as an apple.
It is important to note that the last two processes are conceptually similar—in
as much as they both involve selection among a set of candidates—but at the same
time different in that one rules out unacceptable candidates and the other picks out
top performers. Think of them as lower and upper bounds on the fitness of the
candidates: candidates below the lower threshold get rejected outright, candidates
above the upper threshold effectively determine when the generation process can be
stopped if only one result is desired.
The procedure described in our thought experiment is essentially cyclic in nature:
note the phrase “as we progressively refine” in the paragraph above. This progressive
refinement may be applied to any or all of the five progresses described. In truth, the
procedure implies additional process of reflection upon the results of each of the five
processes already described, with a view to refining them as described.
The schema resulting from this thought experiment involves 7 basic processes:
1. Procedure Selection : selection of constructive procedure
2. Ingredient selection : selection of ingredients to use
3. Lower bound definition : establishment of candidate rejection criteria
4. Upper bound definition : establishment of candidate success criteria
5. Construction : application of the procedure that resulted from 1 to the ingredients
that resulted from 2
6. Selection : application of results of 3 and 4 to results of 5
7. Reflection : reflection upon the accumulated results with a view to refining steps
1to4.
From a procedural point of view, this conceptual schema could be applied to most
processes of artefact generation, including those carried out by humans. It can be
applied to an engineer trying to put together a robot to carry out a particular task. Or
to an artist considering what his next creation might be. Or a programmer building
a poetry generator. Or a poetry generator at work.
I want to make two important observations at this stage.
First, there are significant differences between the case of the engineer and the
artist. The engineer is very unlikely to make any drastic changes to his constructive
procedure (process 1 of procedure selection)—which would correspond to the design
of the robot—or his set of ingredients (process 2 of ingredient selection) beyond