Latent Space Symposium

In the organizer's words:

Latent Space Symposium

13:00 - 18:00 | Free admission
Ludovica Schaerf: Do Latent Spaces Still Exist?
Ting-Chun Liu & Leon-Etienne Kühr: Stereotype Encoding
Hedda Roman: Artists as "Rogue Agents"

20:00 | Admission 15 € | Tickets
Wolfgang M. Schmitt & Stefan Schulz: The New Twenties - live

The Latent Space Symposium brings together researchers, artists and theorists to discuss current perspectives on artificial intelligence, culture, art and society based on the concept of latent space as it can be found in generative AI models.
In addition to the functionality and a historical and technical-historical localization, the focus will be on questions of social and power-political dynamics as well as the artistic potential of the appropriation and creative use of AI.

The symposium is part of Latent Space. Latent Spaces of Art - a project by Filmwerkstatt Düsseldorf.

Lectures 1 - 6 pm, Free admission

Ludovica Schaerf: Do Latent Spaces Still Exist?
[Lecture in English]

This talk explores the changing concept of latent space in the context of AI-generated image synthesis and traces its transformation across different model architectures - from GANs to modern diffusion and autoregressive models. Latent space is located as a central, yet increasingly fragmented, space of meaning within contemporary image synthesis. The presentation examines how image generation has evolved from a single, representative vector to a dynamic interplay of noise, conditioning and spatial embedding.

Ting-Chun Liu & Leon-Etienne Kühr: Stereotype Encoding
[Lecture in English]

Through collaborative artistic research, we investigate how machine vision AI systems translate cultural biases into mathematical form. In hands-on experiments with image models and their training data, we trace how machines learn to "see" through stereotypical associations. From the "happiest face" to correlations between food and ethnicity, our experiments show how these systems replace the material indexicality of photography with statistical biases and translate biases from statistical to geometric relationships.

Hedda Roman: Artists as "Rogue Agents"
[Lecture in German language]

The duo Hedda Roman presents their working method: Generative models function as new sample matter. Self-trained partial models are patched like modules of a synthesizer - weighted, routed, looped; coincidences and glitches are integrated as material and transferred into digital/analog collages. The lecture positions artistic practice as a subversive act against the "autoregressive scaling laws" of dominant AI development - and against its social reflections, such as the uninhibited belief in "unsupervised" market mechanisms. Generative AI does not serve as an effects machine, but as a critical tool against the logics of power that it itself created; as a sampling technique, it is operated in productive latency. The models appear as "alien savants" whose epistemic potential is thus exposed. In the dialog of algorithmic emergence and subversive intention, the emancipatory potential of the artist-in-the-loop unfolds: a negotiation with the uncanny that breaks up rigid patterns of man and machine and opens up new spaces of experience.

-

8 pm | Admission 15€ | Tickets
Wolfgang M. Schmitt & Stefan Schulz: The New Twenties - live
[Live podcast in German]

AI is not a trivial machine, it produces a double contingency that opens up spaces of possibility, not least for discussion. Stefan Schulz and Wolfgang M. Schmitt, the hosts of the Zeitgeist podcast "Die Neuen Zwanziger", talk about their view of AI, communication and the latent spaces. The focus will also be on the political: whether Alex C. Karp, Peter Thiel or Sam Altman - a new form of rule is emerging in which the state and private capital are merging in an uncanny way. The citizen becomes a user or prosumer.

Price information:

Free admission to the lecture program from 1-6 pm. Tickets must be purchased for "Die Neuen Zwanziger - live" at 8 pm. See ticket link in the event text.