Can an art exhibition be curated by algorithms? What advantages and limitations do algorithms bring when applied to curating?
This exhibition displays the work of 20 young Colombian artists, selected through the use of algorithmic and machine learning techniques applied to the open-access collection of the Metropolitan Museum of Art in New York.
This vast collection was used to train machine learning models (MLM), which were then used to assign vectorial values to images and text. Thousands of artworks and hundreds of publications were turned into vectorial representations, through which calculations of proximity and relevance were applied. "Curating" here was understood through proximity, and selection took into account visual and conceptual resonance with each artwork. Exhibition design and curatorial concept was also dictated by these MLMs.
While the application of machine learning algorithms for image and text processing proved useful in sorting through a vast number of artworks and artists, final selection still relied on human criteria, in pursuit of unexpected or serendipitous connections. While automation proved effective in sorting large datasets, human sensibility remained the most valuable component in creating compelling narratives.
This exhibition was supported by the China Academy of Art in Hangzhou, China, and the Museo de Arte de Pereira in Pereira, Colombia.