Machine Bildwissenschaft

Combining digital art history with critical AI studies, the group explores AI not only as a tool but also as a cultural phenomenon with important implications for the humanities. The conference investigates the reciprocal relationship between artificial intelligence and visual culture, looking at how AI systems both shape and are shaped by histories of seeing. We welcome both international experts and the group’s first two members.

Program

Monday, May 12th

11:00-11:15 Welcome Leonardo Impett (BHMPI): An introduction to the Machine Visual Culture Research Group

11:15 - 12:15 Panel 1

11:15 - 11:35 Fabian Offert (University of California, Santa Barbara), Vector Media: Towards a Materialist Epistemology of Artificial Intelligence
Vector Media offers a media theory of the black box, tracing how deep neural networks encode latent ideologies through their internal architectures—particularly the technique of embedding, which turns all media into abstract, commensurable vectors. Rather than focusing solely on datasets or outputs, Vector Media explores the epistemic shift that embeds historical models of vision, perception, and meaning within neural infrastructures, revealing how AI models do not just process media but model media itself. Drawing on Phil Agre’s “Critical Technical Practice,” we argue that this transformation produces a new kind of epistemology—neural exchange value—where cultural objects gain significance only through their operational equivalence within machine learning systems.

11:35 - 11:55 M. Beatrice Fazi (University of Sussex), Towards a Transcendental Philosophy of Computation
This talk presents a transcendental framework for understanding generative artificial intelligence (AI). Via the example of large language models (LLMs), it contends that these programmes perform activities of synthesis, constructing an internal representational reality. The synthetic unity achieved is structural—not the unity of a self, but the unity of a structure that structures. By extending the philosophical concept of Kantian synthesis beyond human cognition, the talk examines how computational systems engage in a search for unity that organises distributed representations into coherent, if imperfect, wholes. This transcendental approach recognises the alterity of LLM outputs while highlighting their world-making capacity.

11:55 - 12:15 Discussion

12:15 - 13:15 Lunch (speakers only)

13:15 - 13:30 Breakout Groups

13:30 - 14:30 Panel 2

13:30 - 13:50 Ranjodh Singh Dhaliwal (University of Basel), What, for AI, is style?
Style abounds in technical and popular discourses of AI, whether it be the first 'uses' of LLMs and diffusion models (where the text/image were to be generated 'in the style of' another) or the technical genealogies of AI that arise out of 'style transfer' mechanisms over a decade ago. This talk puts the technical substrate of style today in conversation with the long history of style in visual studies and literary studies. In doing so, it simply asks what, for AI, is style? Is it the same as the (conflicted histories and theories of) style in humanistic disciplines? And regardless of what the answers to these questions look like, what may be their implications thereof for humanistic and technical scholarship today?

13:50 - 14:10 Tristan Dot (University of Cambridge), The order of design. Patterns of invention / alienation – in the nineteenth-century British textile industry
Textile patterns – from all regions, periods, and media – were copied, modified, and recombined on the drawing tables of Victorian designers. By studying – in all their materiality, at the level of the sketching paper itself – the habits of these workers, a specific mode of invention emerges, in which repetition and experimentation go hand in hand. These patterns of invention – with their copyright issues, and quasi-interpolation practices – resonate with our current digital, vector-based visual culture.

14:10 – 14:30 Discussion

14:30 – 15:00 Coffee

15:00 – 16:00 Panel 3

15:00 – 15:20 Ellen Charlesworth (Durham University), Objective/Subjective: Theorising Digital Art History
A growing body of work on bias has highlighted just how culturally situated—and subjective—computational methods are. However, in media theory, philosophy, and art history today, the very notion of the post-Cartesian subject is under fire. Drawing on object-orientated ontology and the materiality turn, this talk will explore whether digital art history can reconcile the tension between objective and subjective at the heart of digital disciplines.

15:20 – 15:40 Noa Garcia (Osaka University), Bias in the Frame: Unpacking Social Stereotypes in AI-Generated Imagery
Generating images from textual descriptions requires the generative model to make implicit assumptions about the output scene that are not explicitly instructed in the input prompt. These assumptions can reinforce unfair stereotypes related to gender, race, or socioeconomic status. However, measuring and quantifying these social biases in generated images is a big challenge. In this talk, we will explore methods for measuring gender bias in text-to-image models, particularly Stable Diffusion, and discuss how the generated images, when used to train future computer vision models, affect bias in downstream tasks.

15:40 – 16:00 Discussion

16:00 – 16:30 Coffee

16:30 – 17:30 Panel 4

16:30 – 16:50 Eryk Salvaggio (researcher and new media artist), This is (not) an Image of Noise: Creative Reflections on Diffusion Models
Eryk Salvaggio reflects on the role of noise in AI systems and in his creative practice, which engages with images of noise as a tool for forcing errors of recognition, and therefore media production, in diffusion models. The process of noising and denoising in multimodal generative AI includes "visual" mediums of images and video, but extends also to music, which is generated through images of sound. Salvaggio explores noise in his creative practice as the opposite of machinic categorization, and an overlooked element of attempts to trace the inscription of patterns from training data into generated media.

16:50 – 17:10 Maya Indria Ganesh (University of Cambridge)(Zoom) What is critical digital art (now)?
The 2013–2022 decade is marked by critical media arts and storytelling that held a mirror to algorithmic society, drawing attention to the political and moral economies of the embodied and situated human grappling with the loss of privacy, the performance/commodification of the self/identity, surveillance, algorithmic control v human agency. Artists working in this period also imagine possibilities for resisting and reforming these systems. The question of what critical digital art practice is now informs this talk. I will review recent artworks to discuss three dimensions of critical digital visual art that negotiates AI as tool and as text: inversions of institutional authority; the resources required of art practice; and how/if critical art practice can address latent space.

17:10 – 17:30 Discussion

17:30 – 18:00 Break

18:00 – 19:15 Keynote: Antonio Somaini (Université Sorbonne Nouvelle), AI and Visual Culture: A Theory of Latent Spaces
A theory of images and visual culture, today, needs a theory of latent spaces. In a historical phase in which images are more and more generated, modified, circulated, seen and described by or with the help of different kinds of AI models, we need to understand the crucial role played by an abstract, mathematical construct whose cultural and political implications could hardly be overestimated. Latent spaces play also a central role in the contemporary artistic practices that engage with AI: whether to critically respond to its increasing presence in every aspect of culture, society, politics and economics, or to use it as a series of new tools for artistic production. For a few years now, artists have developed different strategies to explore or modify the existing, dominant latent spaces, or to produce their own alternative, antagonist, counter-hegemonic ones. Several of these strategies are documented in the exhibition The World Through AI, curated by Antonio Somaini (with Ada Ackerman, Alexandre Gefen and Pia Viewing as associate curators) and on view at the Jeu de Paume in Paris between April 11th and September 21st, 2025. Taken together, they show the awareness with which the field of contemporary art is tackling the presence and the agency of this hidden layer of mathematical abstraction that is profoundly transforming the status of images and vision, as well as the relations between images and other media.

20:30 - 21:05 Screening of Eryk Salvaggio’s film “Human Movie“ (https://www.cyberneticforests.com/news/human-movie-six-meditations-on-a-compression-algorithm-2025)

21:05-22:00 Questions and discussion

Scientific Organization: Leonardo Impett

_{Image: ai.biennial.com - UBERMORGEN, Leonardo Impett, Joasia Krysa, Eva Cetinic, MetaObjects, Sui: ‘The Next Biennial Should Be Curated by a Machine’ (2021)}