Team
Eva Cetinić
Principal investigator
Laura Juliane Wagner
PhD student
Maria-Teresa De Rosa Palmini
PhD student
“The Canon of Latent Spaces: How Large AI Models Encode Art and Culture” is a 4-year long project funded by Swiss National Science Foundation (Ambizione Grant 216104). The project started in November of 2023. It is hosted by the Institute of Art History and is supported by the Digital Society Initiative of the University of Zurich.
The project’s aim is to explore various aspects of multimodal AI technologies in the context of art and culture. Multimodal foundation and generative AI models are currently garnering enormous attention not only within the AI community, but also among a broad scope of scholars, artists and practitioners fascinated by this technology. The range of potential applications of these technologies is wide and manifold, but so are their various cultural, societal, ethical and political implications.
Trained on millions of image-text pairs sampled from the internet, multimodal foundation models integrate and propagate various biases, often including dominant societal perspectives and selective cultural memories. These models encode numerous and complex associations which exist between data items collected at a certain point in time. However, how those associations are integrated within the numerical hyperdimensional feature spaces of such models is yet to be fully understood. In other words, the question of what constitutes the canon of the latent spaces of multimodal foundation and generative AI models and how to systematically explore them, emerges as a new and exciting research challenge.
The goal of this project is to conduct a comprehensive analysis of multimodal AI technologies, particularly focusing on questions relating to explainability and the critical exploration of their development and use, as well as their impact on artistic and cultural production. The project aims to integrate diverse outlooks and incite interdisciplinary debates on the globally relevant topic of AI-based content generation and its far-reaching impact on the future of art, culture and society.
Eva Cetinić
Principal investigator
Laura Juliane Wagner
PhD student
Maria-Teresa De Rosa Palmini
PhD student
Eva Cetinić is a researcher at the University of Zurich (UZH). Before starting this project, she was a postdoctoral fellow at the Digital Society Initiative at UZH and a postdoctoral fellow at the Center for Digital Visual Studies at UZH. Prior to joining the University of Zurich, she was a postdoctoral researcher in Digital Humanities and Machine Learning at the Department of Computer Science, Durham University, UK and as a postdoctoral researcher and professional associate at the Ruđer Boškovic Institute in Zagreb. She received her Ph.D. degree in Computer science from the Faculty of Electrical Engineering and Computing, University of Zagreb in 2019, with the thesis “Computational detection of stylistic properties of paintings based on high-level image feature analysis”.
She has been involved in the intersection of AI and art for more than a decade, mainly exploring how computer vision and deep learning methods can be used for computational analysis of digitized artwork collections. Her current research interests are focused on multimodal foundation models and generative AI in context of visual art and culture. She is interested in studying the various cultural, ethical and societal implications of such models, with a particular focus on how text-to-image models encode various socio-cultural patterns, as well as impact artistic and cultural production and appreciation.
Contact: eva.cetinic@uzh.ch
Laura is an interaction designer and joined the project in February 2024 after working as a lecturer and research associate at Cologne International School of Design. Her current research focuses on the implications of creating and sharing highly personalized, open-source text-to-image/video generative AI models. She aims to understand how the appropriation of unlicensed and/or violent material in these systems affects the communities that adopt, modify, and share them, exploring broader systemic issues within open-source multimodal AI environments.
Contact: laura.wagner@khist.uzh.ch
Since February 2024, Maria-Teresa has been a PhD Student at the Digital Society Initiative of the University of Zurich. She holds a Bachelor's degree in English Language and Literature from the National and Kapodistrian University of Athens and a Master's degree in Natural Language Processing from the University of Konstanz. Her current research, under the guidance of Dr. Eva Cetinić, seeks to unravel the diverse cultural, ethical, and societal implications of advancing multimodal deep learning models. Specifically, her work focuses on examining the explainability and potential biases of these models, alongside their capabilities in risk mitigation, cultural analysis, and artistic exploration
Contact: maria-teresa.derosa-palmini@khist.uzh.ch