Talks, Workshops & Writing

Workshops I offer: using AI locally on your own computer, working hands-on with embeddings, and seeing the biases inside AI models. They work for anyone, from people who have never written code to artists, companies and scholars.

Upcoming

2026-11-13WorkshopConditions for Emotions, Alphabete der Automatisierung (Burg Hülshoff – Center for Literature), with University of Münster, TH Köln Code & Context and KHM Computational Thinking & Aesthetic Doing

2026-12-17WorkshopSpeaking the Token Language (workshop), Alphabete der Automatisierung (Burg Hülshoff – Center for Literature), with TH Köln Code & Context and KHM Computational Thinking & Aesthetic Doing, w/ Leon-Etienne Kühr

2027-01-13TalkLLM & Operational Image, guest talk in the seminar “A Different View of the World (What Are/Do/Change Operational Images?)”, Prof. Dr. Jan Distelmeyer, European Media Studies, University of Potsdam / FH Potsdam

Workshops

Conditions for EmotionsInstalling small language models on your own laptop. Through absurd questions and answers, the workshop discusses the gap between what a machine computes and the meaning we read into it.

51 Ways to Spell the Image Giraffewith Leon-Etienne KührWhat does an image generator actually read? We learn what a token dictionary consists of and how an image generator understands language, and use tokens to build images that no prompt can produce.

Stereotype Encodingwith Leon-Etienne KührUnderstanding the multimodal model CLIP, which judges whether an image matches a text. While learning how it scores images, we look at the biases of today’s AI models towards images and text.

The Given ImageAn image version of the telephone game: each group sees only the picture sent by the previous group, writes a description, and lets AI regenerate it from the text. In the gradually changing images, the conflicts and differences between computational thinking and images become visible.

Talk recordings

Watch on media.ccc.de ↗

51 Ways to Spell the Image Giraffe: The Hidden Politics of Token Languages in Generative AI

39C3, media.ccc.de · 2025

Talk with Leon-Etienne Kühr, in English with German translation. Before an image model reads a prompt, a tokenizer cuts the text into pieces from a fixed dictionary: “giraffe” can be spelled from tokens in 51 ways, and some images cannot be prompted at all because words were cleaned out of that dictionary. The talk reads tokenization as politics, decided before any training begins.

Watch on The Photographers’ Gallery ↗

Screen Walk with Leon-Etienne Kühr and Ting-Chun Liu

The Photographers’ Gallery, London, with Fotomuseum Winterthur · 2025

Online Screen Walk for The Photographers’ Gallery and Fotomuseum Winterthur, 26 March 2025. Live-coding on screen, Kühr and Liu feed generative models their own output and watch where the images drift: feedback as an artistic strategy, cybernetics as the frame. The recording is on the gallery’s page.

Watch on media.ccc.de ↗

arafed futures – An Artist Dialogue on Chip Storage and AI Accelerationism

38C3, media.ccc.de · 2024

Artist dialogue with Leon-Etienne Kühr. It starts from the chip shortage of the pandemic years and the load generative AI puts on hardware, then follows “arafed”, a meaningless word made by captioning models that has spread into image captions and product texts. These oddities of the embedding space are tied to the geopolitics of semiconductor supply chains, with NVIDIA A100 chips made by TSMC as the case.

Watch on media.ccc.de ↗

Self-cannibalizing AI

37C3, media.ccc.de · 2023

Talk with Leon-Etienne Kühr. What happens when machines learn from each other and feed on their own output? Can an image model pick the happiest person in a photo, or guess ethnicity from a random image? With feedback, misuse and hacking, the talk takes text-to-image pipelines apart: dataset filtering, labelling with CLIP, and recursive generation that turns images into purple patterns or mundane concepts.