‘systems_for_refusal’ is a networked installation that explores the poltergeist inside contemporary AI: a restless, disembodied intelligence that moves symbols without ever touching the world that produced them. When meaning is reduced to correlations, the machine seeks ontological certainty by erasing ambiguity—smoothing raw edges, classifying difference, tightening its grip on what counts as “signal” and what becomes disposable “noise.” This erasure is political. Every label is an act of power.

Across the installation, unruly tools, error-driven scripts, broken taxonomies and hybrid creatures expose the violence and absurdity of classification systems within generative AI. Language models discuss research questions in looping feedback cycles; datasets are navigated through friction, weight, resistance; supply chains unfold across timezones; bodies, fossils and prototypes emerge from technological exhaustion. Together, these elements trace a dense material and symbolic ecosystem in which human labor, geological resources, and cultural meaning are extracted, refined, and rendered invisible.

systems_for_refusal stages a refusal of ontological certainty and offers other ways of seeing: through noise, anomalies, and the beings that slip outside taxonomies. By animating the ghosts inside the machine, the network asks how we might inhabit systems differently—and what kinds of life can emerge when classification collapses.

Sound and connection LLM + UE: Ymer Marinus, Telemagic