2023 | 37

Now showing 1 - 3 of 3

The AI Image, the Dream, and the Statistical Unconscious
Schröter, Jens (2023) , S. 112-120
As has been remarked several times in the recent past, the images generated by AI systems like DALL·E, Stable Diffusion, or Midjourney have a certain surrealist quality. In the present essay I want to analyze the dreamlike quality of (at least some) AI-generated images. This dreaminess is related to Freud’s com- parison of the mechanism of condensation in dreams with Galton’s composite photography, which he reflected explicitly with regard to statistics – which are also a basis of today’s AI images. The superimposition of images results at the same time in generalized images of an uncanny sameness and in a certain blurriness. Does the fascination of (at least some) AI-generated images result in their relation to a kind of statistical unconscious?
Generative AI and the Collective Imaginary: The Technology-Guided Social Imagination in AI-Imagenesis
Ervik, Andreas (2023) , S. 42-57
This paper explores generative AI images as new media through the central questions: What do AI-generated images show, how does image generation (imagenesis) occur, and how might AI influence notions of the imaginary? The questions are approached with theoretical reflections on other forms of image production. AI images are identified here as radically new, distinct from earlier forms of image production as they do not register light or brushstrokes. The images are, however, formed from the stylistic and media technological remains of other forms of image production, from the training material to the act of prompting – the process depends on a connection between images and words. AI image generators take the form of search engines in which users enter prompts to probe into the latent space with its virtual potential. Agency in AI imagenesis is shared between the program, the platform holder, and the users’ prompting. Generative AI is argued here as creating a uniquely social form of images, as the images are formed from training datasets comprised of human created and/ or tagged images as well as shared on social networks. AI image generation is further conceptualized as giving rise to a near-infinite variability, termed a ‘machinic imaginary’. Rather than comparable to an individualized human imagination, this is a social imaginary characterized by the techniques, styles, and fantasies of earlier forms of media production. AI-generative images add themselves to and become an acquisition of the reservoirs of this already existing collective media imaginary. Since the discourse on AI images is so preoccupied with what the technology might become capable of, the AI imaginary would seem to also be filled with dreams of technological progress.
How to Read an AI Image: Toward a Media Studies Methodology for the Analysis of Synthetic Images
Salvaggio, Eryk (2023) , S. 83-99
Image-generating approaches in machine learning, such as GANs and Diffusion, are actually not generative but predictive. AI images are data patterns inscribed into pictures, and they reveal aspects of these image-text datasets and the human decisions behind them. Examining AI-generated images as ‘info- graphics’ informs a methodology, as described in this paper, for the analysis of these images within a media studies framework of discourse analysis. This paper proposes a methodological framework for analyzing the content of these images, applying tools from media theory to machine learning. Using two case studies, the paper applies an analytical methodology to determine how information patterns manifest through visual representations. This methodology consists of generating a series of images of interest, following Roland Barthes’ advice that “what is noted is by definition notable” (BArThEs 1977: 89). It then examines this sample of images as a non-linear sequence. The paper offers examples of certain patterns, gaps, absences, strengths, and weaknesses and what they might suggest about the underlying dataset. The methodology considers two frames of intervention for explaining these gaps and distortions: Either the model imposes a restriction (content policies), or else the training data has included or excluded certain images, through conscious or unconscious bias. The hypothesis is then extended to a more randomized sample of images. The method is illustrated by two examples. First, it is applied to images of faces produced by the StyleGAN2 model. Second, it is applied to images of humans kissing created with DALL·E 2. This allows us to compare GAN and Diffusion models, and to test whether the method might be generalizable. The paper draws some conclusions to the hypotheses generated by the method and presents a final comparison to an actu- al training dataset for StyleGAN2, finding that the hypotheses were accurate.

Browsing 2023 | 37 by Subject "AI-generated images"

Results Per Page

Sort Options