On the Fault Lines of Our Visual Culture: The Dynamics of Personalized Text-to-Image Model Creation and Sharing

The research is embedded within the project «The Canon of Latent Spaces: How Large AI Models Encode Art and Culture» and specifically focuses on investigating the implications of creating and sharing highly personalized, open-source text-to-image/video generative AI models. This research effort aims to identify how broader systemic issues, such as the practice of appropriating unlicensed and/or violent material in open-source multimodal AI systems, might cascade into the communities that adopt, modify and share them.