Earlier this month, a team of researchers at Santa Clara University in California launched an AI chatbot called AI Provenance Research. The tool is intended to enable anyone to ask questions about Nazi-looted artworks and receive answers based on information culled from a searchable, illustrated online database with over 40,000 entries. The news was recently reported by the Jewish News of Northern California.
The group is made up of three business school colleagues: management professor Michael Santoro and analytics professors Haibing Lu and Michele Samorani. The database use AI to corral the vast array of existing provenance data that are spread across different archives, databases, and catalogues, in many different languages into a more easily accessible tool.
Santoro was inspired to launch the project after following his friend Timothy Reif’s decade-long effort to locate artworks that had been stolen from his ancestor, a Jewish Austrian cabaret star. (Reif, a judge in the US Court of International Trade, has since recovered 11 artworks by Egon Schiele.) Santoro also thinks the AI bot could be used to aid in the repatriation of Indigenous art and artifacts in American and European museums to their communities of origin.
For phase one of their project, Santoro, Lu, and Samorani zeroed in on a particular database managed by the Jeu de Paume in Paris. The database, whose records have been accessible since 2010, collects information on the more than 40,000 art objects taken from Jews in German-occupied France and Belgium between 1940 and 1944. During that time, plundered artworks were processed at the Jeu de Paume, which temporarily served as a sorting facility for the Nazi party.
“The AI tool makes it so that anyone can write a question in natural, human language, and it will automatically translate it to the code,” said Lu in a release about the project.
At press time, the AI Provenance Assistant seems a bit glitchy. A question, suggested in the release—How many pieces of art were shipped to Nazi leaders?”—returned an error (“The research service is temporarily unavailable. Please confirm that the backend is running.”) Another question—“Which artist appears more often in this database of Nazi looted artworks?”—returned a link to a .csv file that didn’t answer the question.
The prompt “List artists in the David Weill collection,” referring to one of the famous Nazi-looted collections belonging French banker David David-Weill, returned a table that include the names of artists like Edgar Degas, Eugéne Delacroix, Antoine-Jean Gros, Jean-Auguste-Dominique Ingres, Honoré Daumier, and Théodore Rousseau, though formatted oddly, as well as nonsensical entries like “Turkestan” and “17 Jh., Französisch.”
In addition to refining the functionality of the tool, an obvious next step is to access datasets from additional archives and databases beyond just the one created by the Jeu de Paume. To that end, Santoro and his colleagues hope to publish a paper about their project and establish a foundation that could hire lawyers and provenance specialists.
