Through February 2024, Nasher Museum of Art at Duke University is exhibiting works from the museum’s collection curated by artificial intelligence. The Nasher website notes:

…[w]hile museum professionals are far from relinquishing control of exhibition making and interpretation, this exercise is a powerful way to explore the applications of AI in the creative realm as related to curatorial authorship and expertise, the subjectivity of the selection process, and the future impact of technology on museums.

Students and faculty created a tool to extract publicly accessible information from the museum’s database of nearly 14,000 objects. This dataset was then converted into machine-readable data that is understandable by OpenAI’s ChatGPT. They then developed prompts and instructions for ChatGPT to select the artworks and write the accompanying descriptions and introductory text.

According to Chief Curator Marshall Price, “[w]e hope this experiment will shed some light on [AI’s] usefulness and limitations within museums, higher education, and the creative industries more broadly.” While ChatGPT produced a number of quirks and inaccuracies, “it was a new lens through which we could see and understand our collections,” Price noted in the New York Times. Specifically, the experiment confirmed that the human curators need to be mindful of bias and how they use keywords as well as describe artwork in a machine-readable database.

In the face of tight arts budgets, the technology expenses for this version of ChatGPT were only $10.75, which makes it easier to forgive its mistakes. As underscored during the lengthy SAG-AFTRA strike, the ways in which AI can complement rather than replace human creativity in the arts will be debated for the foreseeable future.