The Work of Art in the Age of Neural Reproduction:

Art historians have used repro-photography as a tool since the mid-nineteenth century, often without critically addressing limits of the technical medium. Today, when photo-documentary material is used in automated analysis, image production errors produced by limits of the medium may be built into machine learning, generative adversarial networks (GANs), and convolutional neural networks (CNNs). This article focuses on the uses of photographic reproductions of the works of Rembrandt van Rijn (1606–1669), including visual analysis of photographs, technical analysis of the “Rembrandt portrait” style, and projects that employ reproductions in the architecture of neural networks—these all define the new cultural shift generative reproduction brings to visual documentation. The article advocates for a critical approach to photo-documentation in the arts.