Paterson, Justin ORCID: https://orcid.org/0000-0001-7822-319X (2026) AI2AI – AI pair programming towards policy-space control: a harmony-aware MIDI-channel multiplexer with captured and bred XY trajectories. In: Innovation in music: TBC. Innovation in music. Routledge, Focal Press, Abingdon, New York. (Submitted)

Full text not available from this repository.

Abstract

Contemporary software development is increasingly shaped by AI pair programming, in which complex systems are co-authored through human intent and machine inference. This chapter extends an ongoing case study by examining MorphMux, a harmony-aware, real-time MIDI-channel multiplexer that forms a new module within the author’s wider MIDI Channel Switcher (MCS) system. Where the host previously redistributed a single MIDI stream across multiple timbres using stochastic, modulation-driven routing, MorphMux reframes that redistribution as a musical decision rather than an abstract algorithm. Each incoming note is assigned to a destination channel by a morphable, heterogeneous policy space whose four corners embody complementary, non-neural inference strategies: a conservative rule set, a Bayesian estimator, an intent-transformative mapping and a Markov/state model. A front-panel XY pad continuously blends these strategies through smoothed convex weighting, letting a performer navigate a continuum of interpretive behaviours and ‘conduct’ a hybrid arrangement in real time – modulating the policies’ own decision parameters rather than crossfading finished audio. The system records smoothed XY gestures as reusable trajectory presets, and a non-real-time ‘gesture lab’ breeds new trajectories from existing ones. A ‘harmony guard’, built from a progression-agnostic pitch-class profile captured during ingestion, moderates controller-induced policy swings without ever altering pitch. Situated within a wide musical, cybernetic and contemporary AI-music literature, the chapter argues that policy-space control offers expressive, interpretable and serendipitous routing, while remaining difficult to fully anticipate – a tension that is itself musically generative. AI is used to develop AI: AI2AI.

Item Type:	Book Chapter or Section
Subjects:	Music
Date Deposited:	16 Oct 2026
Dates:	Date Publication status 30 July 2026 Submitted
School, department or research centre:	London College of Music
URI:	https://repository.uwl.ac.uk/id/eprint/15497
Sustainable Development Goals:	Goal 9: Industry, Innovation, and Infrastructure

Actions (admin access)