Abstract
First opened to the public in 1915, The New York Public Library’s Picture Collection functions like a quirky, local Google Image Search—but one that is entirely analog. Visitors leaf through folders stuffed with physical cutouts of images from publications ranging from the 1700s to present day. A giant binder acts as the finding aid, listing categories ranging from “Abacus” to “Hands” and from “Sorrow” to “Zoology.” The contents are in no particular order, can be checked out for several weeks at a time, and are reorganized and maintained by a small staff. In an age when an online search can pull up reference photographs of almost any imaginable subject within seconds—and in which image generators driven by artificial intelligence can produce new, composite images trained from those sources just as quickly—why maintain a physical collection like the NYPL’s? For this paper, I revisit the original intentions as well as the current functionality and operations of the Picture Collection in light of evolving technologies and digital taxonomies that seek to rapidly advance computer vision. Comparing the collection to the neural network and training sets behind image generator DALL-E, I demonstrate how the design of the NYPL’s Picture Collection—one of the few archives left of its kind—allows for a level of chance, subjectivity and free association that runs counter to the predictive qualities of the algorithm, offering a user experience that highlights some of the biggest limitations of AI-driven generators and digital search engines.
Details
Presentation Type
Paper Presentation in a Themed Session
Theme
2024 Special Focus—Images and Imaginaries from Artificial Intelligence
KEYWORDS
Artificial Intelligence, Image Search, Photography, Image Generation, Local Archives, Analog
