Cássia Hosni, Giselle Beiguelman, Meta-Acervos and New Interpretations Emerging from Error
This visual essay examines the application of image classification and object detection algorithms to collections of drawings and paintings from Brazilian museums using Meta-Acervos, a platform that employs artificial intelligence models to analyze, organize and visualize artworks according to institutional, technical, chronological and visual descriptors. Rather than emphasizing the system’s accuracy, the essay focuses on its errors and on the potential for new readings afforded by ambiguity and interpretations not predicted by the program. We argue that this act of playing with errors and information not predicted within the logic of artificial intelligence models points to broader questions of agency and human intention in a world increasingly structured by apparatuses.
