| dc.description.abstract | In 2019, the musician Holly Herndon released her third full-length album, Proto. In
addition to input from two other human artists, the album had a fourth collaborator:
an artificial neural network named Spawn. The software had been trained over
several years to generate and manipulate the cavernous choral soundscapes that
brought Proto widespread critical acclaim. Spawn’s role in each stage of the musicmaking process was neither completely predictable nor completely under Herndon’s
control; her vocal contribution – its tone, pitch, rhythm, and dynamics - was often
novel, original, and surprising. Herndon describes Spawn as ‘a performer... an
ensemble member. So I would say that I collaborated with a human and an inhuman
ensemble’ (Funai 2019).
Here, we consider how seriously we ought to take assertions like this one.
Can we really conceive of AI systems as legitimate collaborators in the skilled project
of making art? Do they have the kinds of creative agency, autonomy, and expressive
power that characterise membership of an artistic ensemble?
In the next section, we rehearse some reasons why there has been a reluctance
to give affirmative answers to these questions – why, that is, computational systems
have been taken to have an impoverished status, lacking capacities essential to true
artistic agency (see Boden 2007). In section 2, we explore the view that even when
attributions of creativity and autonomy to artificial systems are not literally true, they
can instead be fictionally true. Those who work alongside generative systems like
Spawn and those who enjoy the musical fruits of such collaboration are participants
in an elaborate game of make-believe, wherein the non-human contributor is
imaginatively conceived as being a real improviser, a real singer, a real musician.
Taking this line allows us to give credence to testimony like Herndon’s, and to better
understand the production and appreciation of music that has a partially nonhuman origin. | en_GB |
