Henri Kartmann recently explored the capabilities of AI image generators. These programs create images from text descriptions, interpreting the words based on their databases. Kartmann compared AI-generated images to literary descriptions of places and characters.
The results were both interesting and disappointing. Complex language, like that used by Proust or Perec, often led to the AI misinterpreting the text. This resulted in confusing images.
For example, when trying to visualize Simone de Beauvoir’s family portrait, the AI created an image of the same child repeated eleven times.
Kartmann’s AI-generated images evaluated
The software often focused more on making a “beautiful” image than accurately representing the subject.
Despite the issues, Kartmann edited the AI-generated images with the same care as his own photographs. He used this experiment to think about what we expect from photographic images. Kartmann believes that while AI images might work for things like advertisements, the source and accuracy of images are very important for creative works.
He thinks that controlled and authentic creation should be the standard in art. For these experiments, Kartmann used software called Leonardo.Ai. His work has started a discussion about the role of AI in photography and image creation.
