LiviaAI

LiviaAI investigates the potential of artificial intelligence (AI) to identify patterns, connections, and associations between digitized objects in three Viennese art collections: the Belvedere, the Wien Museum, and the Museum für Angewandte Kunst (MAK). The project has three goals. First, to explore how AI can serve as a tool for scholars and museum curators to study art collections at scale, to understand how different museums have applied different collection practices and classification schemes over time. Second, to establish connections between online collections, so that the associations between them and their individual objects become more visible. Such connections can be based on different notions of similarity - common styles, motifs, materials, or composition, related artists, etc. The project will train AI to learn to identify such similarities in a process called “representation learning”. Third, the project will build an online showcase, allowing the public to navigate the connections uncovered through AI. Through a so-called “generous interface”, a visualization that emphasizes serendipitous browsing, aims to enable playful exploration of the collection, and supports an understanding of context and relationships, LiviaAI also wants to demonstrate how AI can support new online exhibition formats that bring Vienna's cultural heritage to life in new ways.

Building on cutting-edge deep learning techniques and natural language processing methods - most notably contrastive learning, triplet loss networks and word embeddings - LiviaAI will develop a method and toolset that enables museum professionals to design their own AI processes, without the need for the prohibitive manual data annotation work normally required to train AI models. The project will work in close cooperation with Viennese museum professionals and educators in Lower Austria to build expertise, and lay the groundwork for embedding AI skills, methods, and tools more deeply in future digital humanities curricula and, ultimately, everyday professional scholarly and curatorial practice.