Tools of the Future for Research of the Past. The Use of Generative Image Models for Humanities

Over the past year, there have been notable advancements in the field of artificial intelligence (AI). These developments have had a significant impact on the art world, specifically through the emergence of generative diffusion models which enable the production of high-quality images from textual descriptions, independent of any particular artistic ability. Although much of the discourse surrounding generative image models has centered on their impact on contemporary artistic production and the challenges they present to traditional artistic creation, the proposed project seeks to investigate a previously overlooked facet of AI advancements: their potential to facilitate the analysis of older artworks in a historical context. This project endeavors to employ generative image models to produce visualisations of significant Slovenian artworks that have either been lost, damaged, destroyed, or never realized. 
Through a series of case studies focusing on architecture, sculpture and painting across various historical periods, the project team aims to develop a novel interdisciplinary methodology integrating the approaches of art history, computer science, and philosophy. Our main objective is to develop a comprehensive set of guidelines that support the integration of generative image models into art-based research workflows while taking into account the ethical implications of AI implementation in the humanities with the aim of responsible and equitable utilization. We believe that the proposed methodology has the potential to revolutionize day-to-day research in the humanities by enabling the regular verification of research hypotheses through the use of affordable and user-friendly generative image models. With the project, we seeks to foster a paradigm shift in the development of generative diffusion models to better support the humanities while providing AI tool developers with a user-centric perspective on new directions for technological advancement. 

Keywords: Generative image models, art history, artificial intelligence, visualisation of art and architecture, methodology