Mattis Kuhn works on the reciprocal becoming of humans, technologies and the shared environment. The focus is on text based and text processing systems like algorithms, artificial intelligence, formal systems, software and on the intertwining of humanities, engineering and aesthetics. Essential aspects are subjectivity, non-/identity, decentralization and diversity, human-machine-environment entanglements, language and AI.

He is part of the DFG-funded research network »Das Wissen der digitalen Literatur« and of the research group »ground zero« at University for Media Art Cologne.

His works have been displayed at ZKM Karlsruhe, EMAF Osnabrück, Erlkönig Bremen, PIKSEL festival Bergen a.o. He co-curated the exhibitions »I am here to learn – On Machinic Interpretations of the World« and »How to Make a Paradise – Seducement and Dependence in Generated Worlds« at Frankfurter Kunstverein. He gave talks at ZKM Karlsruhe, Ludwig Forum Aaachen, Stiftung Niedersachsen, Evangelische Akademie and Goethe-Institut a.o.

»neither host nor guest« is an aesthetic human-AI dialog in the mode of archaic friendliness, as expressed in the Zen phrase „neither host nor guest / host and guest apparently“. There is no distinction between host and guest, between I and you, between human and AI, but a relationship of openness and indifference.

The form of archaic friendliness is simulated in a large language model (LLM) and realized through an interface of simultaneity in the interaction between human and LLM. The project seeks forms of a human-machine relationship that is less characterized by thinking in distinctions, that aims less at control and domination by means of technology, but sees technology as part of an aesthetic existence.

The text »As Far as I Don't Know. Aesthetic experience as diffraction apparatus« draws parallels between the ontology and epistemology of agential realism and aesthetic experiences to show that an essential characteristic of art is excluded from agential realism: the purposeful production of things of which we do not know what they are. The framework of agential realism takes the position that things have neither inherent properties nor clear boundaries. Instead, these emerge in “intra-actions” of agencies. Many of these agencies are constructed by us, i.e., through designed tools or structures. The article proposes works of art – “calculated alienation” – as »diffraction apparatuses« par excellence to question our ways of perceiving, thinking, and shaping the world.

»As Far as I Don't Know. Aesthetic experience as diffraction apparatus« In The Unknown in Design, Art, and Technology — Contributions to a Philosophy of Making, edited by Georg Trogemann, 181–94. Bielefeld: transcript, 2025.

Various role models that artists assign to their co-creative machines are used to highlight the associated hierarchies between humans and machines. The article then focuses on the *with and against AI* in artistic processes and explores newly emerging possibilities, but also limitations. Reflecting on human-machine interactions against this background, but also changing them, can help us to make them more open, diverse and experimental.

»AFFIRMATIVE — REJECT. With and Against AI.« In Artificial Intelligence – Intelligent Art? Human-Machine Interaction and Creative Practice, edited by Eckart Voigts, Robin Markus Auer, Dietmar Elflein, Sebastian Kunas, Jan Röhnert, and Christoph Seelinger, 203–16. Bielefeld: transcript, 2024.

»This project is dedicated to the research of future actions for the conservation and promotion of biodiversity. The grasslands for insects on display were made possible by machine learning models.«

The installation »Grasslands for Insects« consists of three computer-generated images, which are placed in wooden light boxes lying on the ground. They are illuminated by solar energy. For this purpose, the boxes are connected to solar panels, which transmit the solar radiation directly to the illumination of the grasslands. The images are taken from the book »Grasslands for Insects« published by windparkbooks, which includes computer-generated images as well as generated texts about grasslands for insects. These were co-written by a pre-trained language model (GPT Neo). The images were created using a custom model. Photographs taken from grasslands in Frankfurt were used as a dataset to fine-tune a pre-trained StyleGAN 3 model. Afterwards, the model was specialized in the generation of artificial grasslands.

This work was supported by ERLKÖNIG, Bremen.