“Close Reading Distant Viewing”

Machine learning has not only changed how computers “read”, but also how they process the visual world. In this talk, I will investigate some of the implications of the machine learning revolution for the digital humanities. I will present an analysis of popular distant viewing practices that leverage machine learning to analyze large corpora of images and show how these practices fail by failing to take into account the radical differences between human and machine intelligence. Specifically, I will argue that it is exactly a “close reading” of machine learning that is required to uncover such epistemological limitations, and that machine learning thus needs to be understood as a natural domain of humanist inquiry.