The Extended Composer - Computers and Creativity - page 195

With the increasing maturity of such methods, the application of algorithms in
composition has started to become more comfortably integrated with the rest of the
cultural landscape. It is now incumbent on the critic to judge such hybrid human-
computer works against the normal value scheme of creative works: responding to
aspects of cultural fit, social impact, usefulness and beauty.
By incorporating generative processes into a feedback loop over which we then
exercise selective control, one can effectively bypass the bulk of the arguments
against the inhuman or uncontrollable nature of computational creativity: it is still
the artist that exercises the decisive decision-making. For all the conceptual diffi-
culty in realigning our technological understanding with our aesthetic past, the de-
gree and complexity of reflection, development and conceptual weight are arguably
all the greater.
Though a simplistic view of the human-computer creative partnership has the
computer generating material and the human judging it, the reality in most systems
is more complex. The degree to which the computational system or the human fil-
ters the results depends on the design of the system and/or the intent of the artists.
Take, for example, the fairly hands-off approach (procedural interaction) of Iannis
Xenakis with his Gendyn system, which was used to create the composition Gendy 3
by generating complete works using handcrafted program settings. The final work
was but one iteration selected by the composer. On the other hand, Biles' GenJam
(Biles 1994 ) performs quite autonomously, improvising jazz solos created by a ge-
netic algorithm and a database of human-performed solos. The user's control con-
sists of playing solos that the system analyses and combines with other contextual
musical information, including harmony and metre, to generate its own solos. Dur-
ing live performance with GenJam, there is no time for filtering of the computer's
solos by the human partner.
Even though both these systems differ with regard to human filtering of the re-
sults, they both assume a considerable degree of autonomy over the generation of
material. Generative systems with this degree of autonomy are often designed with
a particular stylistic outcomes in mind in order to ensure that outputs fall within de-
sired aesthetic boundaries. Other systems, such as Nodal (McCormack et al. 2008 )
and Emily Howell (Cope 2008 ), are more interactive, requiring the human to make
frequent and often detailed decisions that guide the generative process. This ap-
proach can typically allow for a broader range of stylistic results because of the
continual human guidance that is a check against undesirable output.
Regardless of the interaction and division of responsibility during the creation
process, once music is completed by a human and generative system partnership, its
value is judged like any other music by its audience appeal—whoever the audience
is, and however value may be defined by them.
7.4 In Summary
Most people who believe that I'm interested in chance don't realize that I use chance as a
discipline. They think I use it as a way of giving up making choices. But my choices consist
in choosing what questions to ask.
- John Cage (Kostelanetz 1989 , p. 17)