Twentieth International Conference on the Arts in Society

Abstract

AI-generated images feature an unprecedented event in the history of image creation: images made with words. From the early days of cave painting to the rise of digital photography, image making has always involved material devices and media. Given the material aspects of image creation, this process entailed considerable experimentation, particularly at the early stages of development, when technology is “new”. For example, early photography was characterized by frenzied experimentations with different substances and devices. AI image generators, however, bypass this traditional process by inputting text prompts. Image creation through natural language presumably takes out an important element of materiality and craftsmanship from the creative process. We argue, however, that materiality in fact shifts to language, and experimentations with language take center stage. We refer to prompt culture as the discourse, practices, actors, institutions, and devices which are geared towards understanding, defining and refining, training and experimenting with crafting language for Image generators. We focus on three case-studies: 1. Promotional videos of Dall-E (showing a historical trend by which the app encourages users to move from commending to communicating); 2. Social media groups of AI-professionals (following linguistic experimentations in creative industries; and 3. Dedicated AI devices (focusing on Paragraphica, “a context-to-image camera that uses location data and artificial intelligence to visualize a ‘photo’ of a specific place and moment.” Analyzing these sites, we argue that prompt-culture not only creates new images but also a new language through which humans and machines can communicate.