Oliver Bown

Executive Summary

Oliver Bown reframes AI's role in creative practice as subject and medium rather than tool or partner. This is not semantic quibbling; it represents a fundamental philosophical repositioning. Tools imply instrumentality, missing how technology shapes creative practice from the inside. Partners carry political baggage given corporate platform ownership, masking extraction dynamics. Treating AI as subject means technology becomes the focus of creative investigation itself.

His flagship Sydney Opera House commission exemplifies this philosophy. Tasked with expressing the building's agency and daily dynamics as a cybernetic system for its 50th anniversary, he created algorithmic composition driven by building telemetry. The LLM served as routing interface, not composer. Humans composed the musical space; AI guided which pre-authored elements surfaced when.

His platform capitalism analysis is equally unflinching: he expects at least one AI-triggered catastrophe due to human implementation error, not evil AI intent, and believes individual practitioner choices are insufficient against systemic platform dynamics. Change requires policy and regulation.

AI as Subject, Not Tool or Partner

Oliver explicitly rejects both dominant framings of AI in creative practice. The tool metaphor implies instrumentality, AI as neutral means toward human ends. This misses how technology shapes creative practice itself. Using a synthesiser versus writing code versus training neural networks does not just change efficiency; it changes what becomes thinkable, what aesthetic possibilities emerge.

The partner metaphor carries different problems. Partnership and collaborator are status attributions rather than functional descriptions. When your partner is owned by a tech company, learns from your interactions at scale, and feeds that learning back into systems serving millions, partnership language masks extraction dynamics.

His alternative: treat AI as subject. Not asking what AI can do for you, but examining how it operates, how it bends and shapes artists, how forms and aesthetics emerge through systemic engagement.

The Sydney Opera House: Building as Instrument

The brief asked him to express the building's agency and daily dynamics as a cybernetic system for its 50th anniversary. He structured the work around four interacting elements: water, air, power, people, the core systems maintaining the Opera House as a functioning organism. The centrepiece was the harbour-water thermal regulation system, seawater pumping through building infrastructure to control temperature.

His materials palette combined field-recorded cooling system sonics with pipe organ recordings. The organ session proved transformative.

That late-night session inside the Opera House recording the professional organist was one of the best experiences of my life.

The recordings became the keystone, the foundational element everything else built from.

The technical architecture reveals his approach to AI integration. Real-time telemetry from building systems fed into the LLM as a preprocessed stream. The LLM issued selection and routing decisions to his custom generative remix engine, which functioned more like a game-music stem remix machine than a generator. It did not create new musical material; it selected, arranged, and processed pre-composed elements. Humans composed the musical space; AI guided which pre-authored elements surfaced when, responding to building state.

Amoeba-Level Agency and Space-Shaping Methodology

Oliver's central concept for machine musicianship is amoeba-level agency: not human-equivalent intelligence but simple lifelike systems that can still be musically beguiling. More complex than synthesisers, far less than humans, occupying a productive middle ground where machine behaviour remains interesting without requiring anthropomorphisation.

His litmus test: can clarinetist François Houle sustain ten or more minutes playing with the system without boredom? When musicians engage naturally with machine systems through extended improvisation, that validates the design.

His methodology centres on space-shaping: preconfigure system spaces so that in-the-moment creative play becomes fluid. LLMs serve as flexible control surfaces for last-minute aesthetic adjustments. The programming work happens in advance; performance explores a pre-configured possibility space where the system constantly varies but never goes bad.

Client-Producer Realism and Platform Capitalism

Oliver proposes a realistic alternative to partnership metaphors: most human-AI making today is a client-producer relation, not peer collaboration. One party instructs; the other delivers. This describes functional reality more accurately than partnership talk, and reveals actual power asymmetries rather than projecting fantasies about mutual creative growth.

His platform capitalism analysis is unflinching. AI represents the endgame of platform capture, from Netflix recommendation algorithms controlling what people watch to personalised movie generation at scale. Creative labour gets displaced, authorship becomes opaque, individual choice shrinks despite the appearance of customisation.

He expects at least one AI-triggered catastrophe, a market shock or runaway agent scenario, due to human implementation error rather than malicious AI intent. Individual practitioner choices prove insufficient against systemic platform dynamics. Change requires policy and regulation: data rights, model accountability, provenance standards, labour protections.

Technical Curiosity Alongside Political Concern

Despite his political scepticism, Oliver maintains genuine technical curiosity. He is intrigued by vibe coding, AI-assisted coding that yields what he describes as weird, mutant, evolving software that humans would not write themselves. He finds this more compelling than static image or audio generators because the outputs are themselves systems capable of surprising behaviour.

His commissioned research reinforces a consistent finding: the tools that work best for artists offer both surprise and a sense of control.

What I learned from commissioning artists is that AI systems need both opportunity for surprise and a sense of control. Without both, it's not creative partnership.

This shapes his broader research agenda. The question is not how to make AI more powerful but how to design interaction models that keep the artist genuinely in the loop, able to understand, interrogate, and redirect what the system is doing. Explainability is not a technical nicety; it is a precondition for meaningful creative agency.

His optimism is measured but real. The creative arts stand to benefit from AI, but only if the tools are built around the artist's process rather than around the technology's capabilities, and only if the political and regulatory conditions exist to prevent platform capture from hollowing out the creative economy entirely.