Arbitrary Intelligence: Imperceptible Forms and Extended Musical Coherence in Computer Music (researcher: Guy-Shawn Wong)
How can new musical forms be proposed by algorithmic systems of intelligence informed by contemporary machine-learning architectures, and how can these forms be presented to enable perceptual, interpretive, and aesthetic engagement by listeners?
While extensive research in music cognition attempts to address issues of music perceptibility, vast diversity of musical practices problematises precise problem definition. Therefore, experiments necessarily focus on narrow musical contexts where comprehensive conclusions are yet to be established. Rather than seeking universality, this research approaches perceptibility as a problem of interpretability. From the compositional perspective, this concerns how intent and structure are articulated; from the listening perspective, how meaning emerges through context, conditions, and modes of engagement.
Results
The written component documents compositional investigations of selected machine-learning architectures,
foregrounding personal interpretation and experimental processes rather than technical exposition; articulates a
theoretical framework for computational temporality in music; presents the design of interactive music systems;
and incorporates qualitative research, including interviews and comparative analyses of relevant artistic and
scholarly work.
Studio-based compositions realise the project’s theoretical ideas. Current work-in-progress explores divergent
learning errors between multiple neural networks trained on identical dataset, repurposing stochastic variation in
knowledge generalisation as musical feature.
Performances and interdisciplinary collaborations develop engagement strategies through direct interaction
and discussion with collaborators and audiences.
Beyond artistic outputs, the research produces open-source softwares, public talks, workshops, and other
educational media such as videos or podcasts to further elaborate the research.
