Computational Creativity and Music - Computational Creativity Research: Towards Creative Machines - page 309

Fig. 15.4 A data-driven creativity program
more sensitive and creative possibilities that likelywill end in reasonable time lengths
and on reasonable pitches.
The resulting program from the description just given looks something like that
shown in Fig. 15.4 .
Figure 15.4 begins with as much data as possible presented in a form that allows
logical segmentation in the form previously described. The Analysis portion then
stores these segments in their appropriate groups. The association process follows the
forward and backward chaining Markov reorganization processes producing original
output. As should be clear, this design is simple, data-driven, and highly contextual
in both its forward and backward directions.
Figure 15.5 presents eight examples representing potential creativity, four of
which by humans and four by a version of the program just described.
Note that each melody begins identically and none contain chromatic pitches. The
durations of the computer-composed examples were similarly derived fromdurations
provided by the extended human-composed ones. I won't indicate which are human
composed and which by computer program, but leave that up to readers to decide.
Note that the database, and thus the models for each, consisted of Mozart melodies
in Classical style simply because the rules are so apparent and the style obvious. The
ones completed by humans (four different people not including myself) may seem
obvious to some readers. However, I caution those making serious guesses here, that
what may seem obvious may not be so at all.
For example, note that melodies (c), (d), (g), and (h) each contain two imitative
phrases of equal length, that (a), (f), and (h) contain leaps of a sixth or more, all
but (d) and (g) end with downward motion, most contain only half, quarter, and
eighth notes, and that (e) and (h) have the widest ranges. Thus, the examples present
significant diversity that makes it difficult to determine which demonstrate human
and which computer origins.
I would agree with those that insist that such short and simple examples do not
provide a perfect test upon which to base judgment as to human versus computational
creativity. At the same time, short and simple melodies such as those presented here
can focus our attention on the nitty-gritty of the problem. While creativity can and
does exist at all levels of music composition, the smallest decisions may likely be
the most telling.