The technological foundation of the project lies in the emergence of new AI computer models capable of automatically generating synthetic images that resemble real photographs (e.g., GAN models like [thispersondoesnotexist.com]) or translating text input into images with increasing levels of detail and syntax (e.g., GPT-3 models, as demonstrated by openai.com/dall-e2/).

This development raises a number of pressing questions about how these ‘artificial’ image practices—shaped by training on vast collections of big (image) data—are influenced by specific, existing image cultures, and how synthetic images, in turn, shape our visual citizenship in the future.

A key democratic challenge is that while everyone potentially contributes to the creation of these images (e.g., through their data), the conditions for active participation remain opaque. These new AI image technologies not only expand citizens' capacity to produce images but also establish new boundaries for what images can be created. These boundaries are shaped by biases inherent in the computer models, which favor certain visual expressions and motifs. Moreover, the meanings assigned to these images depend on how we use and interpret them—whether as art, superficial entertainment, “fake news,” visual dreams and imaginations, narrow stereotypes, or something else entirely.

The project is conducted by the Centre for Aesthetics of AI Images (AIIM).