This book illustrates how Generative AI models have rapidly revolutionized creative industries, media, and technology while addressing growing concerns surrounding issues like machine hallucinations, labor exploitation, and social biases.

In this book, Jihoon Kim introduces new concepts to define generative AI as the unique blend of statistical precision and seemingly magical effects.

Mathemagical Media introduces new concepts to offer a nuanced understanding of generative AI models—such as GPT-4o for text, DALL·E and Midjourney for images, and SORA for video—in relation to their discriminative counterparts for sociotechnical mediation. Amongst them is “mathemagical,” which stems from replacing the “t” in the word “mathematical” with “g,” therefore pointing to the conjuncture of “mathematics” and “magic.” As much as it clarifies that GenAI is not a superintelligence, it also advances critical AI studies by arguing that GenAI’s operations of vectors in its latent space are so complicated that their outputs are more than “copying and pasting” of training data.

Drawing from the material-technical approach to AI models and from the hyperparameters of cinema and media studies, photography theory, art history, and media archaeology, Jihoon Kim assesses GenAI as an emergent medium that increasingly rehearses and restructures photography and cinema. With its wide-ranging analysis of synthetic still and moving images in contemporary media art and visual culture, this book guides readers towards GenAI’s ambivalent magical effects: their stereotypical and uncanny, familiar and unfamiliar expressions enchant GenAI’s aesthetic capacities while also disenchanting its underlying cultural and machinic biases originated from discriminative AI.