Reclaiming Data - Skill Sharing 1: Michael von zur Mühlen

Memory Machines LLMs archival silence, and the politics of historical voice

This workshop introduces Will the Revolution Not Be AI Scripted?, an interactive art installation in development that stages an encounter with AI-mediated voices of resistance — drawn from labor movements, anti-colonial struggles, and subaltern histories. The project is built on archival material from European and South Asian collections, including sources whose gaps and silences are themselves the product of colonial archival practice.

AI systems are predominantly developed to optimize workflows, drive economic growth, and enable targeted marketing, military operations, and surveillance. This workshop asks what an emancipatory — perhaps even subversive — use of these same technologies might look like. Not AI as a self-optimization tool, but AI as a space for politicizing individual experience and embedding it within a heterogeneous memory of past and present social struggles.

Working with the actual technical stack of the project — large language models, retrieval-augmented generation pipelines, and self-curated archival datasets — participants explore how institutional knowledge gaps, training data politics, and LLM hallucination become artistic material rather than engineering problems. A central practical question is how to work with large, heterogeneous, self-assembled collections: how to structure, embed, and query material that was never designed to be machine-readable, and how the shape of that data — its inconsistencies, its silences, its uneven coverage — becomes legible in the behavior of the system.

The workshop takes Saidiya Hartman's Critical Fabulation and Michel-Rolph Trouillot's Silencing the Past as methodological anchors: archival absence is treated not as a failure to be corrected but as a political statement about whose voices were excluded from the record. The central question is not how to make AI speak more accurately about the past, but how to design systems in which the limits of that speech remain visible — and what it means to build memory infrastructure on top of technologies produced through their own forms of invisible labor.

Key info:

— Date: Saturday, 13 June 2026, 10:00–11:30

— Venue: Flutgraben e.V., Am Flutgraben 3, 12435 Berlin

— Capacity: 15 participants

— Admission: free, registration required

— Language: English

— Prerequisites: none, but curiosity about AI, archives, and political memory is welcome

Reclaiming Data is a two-day symposium and exhibition on 12 and 13 June 2026 at Flutgraben e.V. Berlin. It brings together artists, researchers, journalists, and cultural institutions around the question of how digital archives, data sets, and generative systems shape what is remembered. Three Skill Sharing Sessions accompany the symposium, each held in a small group (max. 15 participants). Please choose one session per registration.

More: newpractice.net/funded-projects/reclaiming-data