This paper examines the role of artificial intelligence(AI) as an artistic agent through its creative logic, and looks into the AI-facilitated paradigm shift in the relation between medium and artist from the perspective of art history. It analyzes the basic structure of neural networks such as Convolutional Neural Network (CNN) and Generative Adversarial Network (GAN), which have often been used to generate creative images. It also notes that the training and evaluation of neural networks with proper sample datasets is a crucial step in selecting the aesthetic features that will compose the generated images. The comparison of deep neural networks with video synthesizers reveals that the creative logic of AI runs on digitally encoded linguistic instructions rather than functioning as a physical apparatus. This means that the paradigm of postmedium art, which has been supported by technological advances, would have to shift further to cover artificial agents. Through an attempt to locate AI in Rosalind Krauss’ diagram of the expanded medium, this paper shed light on the potential linkage of AI to the concepts of medium and installation, and brings AI into the realm of the discourse on art medium.