This chapter offers a brief history of, and new directions for, artificial intelligence (AI) in design. Drawing together examples from architecture, music, and the visual arts, it shows how an experimental tradition of computational aesthetics and design underpins present-day approaches to architecture and AI. In exploring this history, the chapter calls attention to two aspects that have frequently been overlooked: first, the materially specific acts of codification that structure these experiments, and second, the reconfigurations of author and user roles that they entail. With these as points of reference, the chapter identifies new directions for creative research on AI and design that eschew simplistic approaches to style transfer or machine autonomy. As illustrations, three recent projects developed at the Computational Design Laboratory at Carnegie Mellon University are discussed. These explore new drawing interfaces that enable users to manually control the weights of a neural network; new approaches to computational urban analysis that reveal morphological gradients at the urban scale; and “distant readings” of architectural data that open up new approaches to design description, analysis, and synthesis.