Other Images
Other Images
Research Project
Introducing the aesthetic dimensions of image synthesis and contextualizing generative imagery in (visual) culture.
Artistic Research, Generative AI
The work "Other Images" introduces the aesthetic dimensions of image synthesis as a discipline and provides both an overview and contextualization of this technology within (visual) culture and touches on the questions of what it means to be in the midst of a paradigm shift and how to navigate through an ever-changing media environment?
By now, being the summer of 2023, artificial intelligence technologies are an integral part of our society, recognized both as opportunities for social progress & economic growth, as well as threats that will automate human labor and are already having a significant impact on many areas, including the creative industries.
Media creation, using generative artificial intelligence algorithms, has established a new category of media assets known as synthetic media, which appears to have arrived in most areas of cultural production; from AI-powered software features, such as numerous style transfer or face filters, to generative AI platforms that enable media production through text input for mass audiences.
The work "Other Images" deals specifically with AI-generated images, also commonly referred to as generative imagery or (generative) AI art, which represent a radical shift in image production and whose pace of technological development has surprised most observers.
No Thoughts, Just Vibes, 2023
The essay No Thoughts, Just Vibes explores the role of vibes as a concept in the context of our algorithmic culture and machine learning processes. In a long-read format, it introduces the notion itself, distinguishes it from other similarly used terms such as style or mood, and elaborates on the relevance of vibes in the field of artificial intelligence.
Download PDF
A Short History of (Generative) AI Art, 2023
The longform essay A Short History of (Generative) AI Art provides an introduction to the field of generative imagery by offering an overview of the major technological developments in the field. It also contextualizes AI-generated imagery as a distinct image category and highlights the aesthetic characteristics of this new medium.
Download PDF
Algorithmic Culture Content, 2023
In a condensed form Algorithmic Culture Content contributes to the discussion around the phenomenon of “cultural flattening” in our media landscape through both recommender and generative AI systems. It addresses the emergence of content as a media form and discusses how algorithms influence the way we produce and consume media.
Download PDF
This project explores the phenomenon through a collection of (scientific) essays, an academic course syllabus, suggested readings, as well as artistic experiments and commentaries. Intended to be treated as a dossier that allows for different levels of engagement with (generative) AI, each contribution can either be viewed as an independent work, considered in relation to other parts, or experienced together as a whole body of work.
Hidden Spheres, 2023
Video, 00:22:15, 1080 x 1080 resolution
The work Hidden Spheres provides a deep dive into the inner workings of a generative machine learning model by revealing the otherwise invisible structures of a deep neural network. For this work, a Generative Adversarial Network (GAN) was trained on a dataset of planetary imagery using the StyleGan3 architecture, and then inspected using its interactive model visualization tool.
An astronaut riding a horse in photorealistic style, 2023
Scratch ink on pigment print, framed, 200 x 200 mm
The work An astronaut riding a horse in photorealistic style focuses on both the imaginative and arbitrary aspects of AI-generated imagery by presenting an essential technical element of the image production process as the artwork itself, thus offering a glimpse into the inner workings of generative AI.
Arbitrary Imitations, 2023
Pigment prints, framed, 210 x 300 mm / Artist book, edition of 10, 105 x 148 mm
The work Arbitrary Imitations addresses the commodification of images through artificial intelligence and the resulting perception, that the emergence of algorithmic social media feeds as well as generative AI systems, have led to a devaluation of cultural production, creating a media landscape of aesthetic relativism.
Computational Colors, 2023
Video, 00:09:30, 320 x 1480 resolution
Bypassing the distinctive aesthetics of AI-generated imagery, the work Computational Colors offers a monochromatic view on the capabilities of generative AI, while also addressing the transition from one era of (generative) AI art to another.
By “zooming in” on the trained model, this work brings the transformational qualities to the forefront, revealing the process of how such models learn simple features at the beginning of the training and learn to recognize more complex and abstract features as the training progresses.
Scratching the framed artwork reveals the image underneath, which consists entirely of Gaussian noise, an essential technical element in image synthesis. Noise serves several purposes in the image generation process, such as allowing the exploration of the model’s latent space, as random noise is used at the beginning of each generation process, thus representing the seemingly infinite number of possible images that can be generated by multimodal AI systems.
Arbitrary Imitations shows a portrait of two lilies, one being a freshly cut flower and the other an artificial silk reproduction, on a gray background, as well as various AI-generated imitations of this photograph.
This work involves a combination of both value-added and value-decreasing modes of production and is presented in two ways: first, as a conventionally printed limited edition artist book containing 390 alternative versions, with the original photograph added to the book cover as a high-quality pigment print; and second, as a framed selection of prints featuring eight imitations along with the original photograph.
By recreating the DALL·E 2 watermark, which consists of five solid color squares placed side by side, and offering alternative versions of the mark through the medium’s interpolation feature, the video installation mediates the statistical capabilities of pre-multimodal AI systems, which have often been humanized and referred to as dreaming or hallucinating.
Web Safe color palette dataset, consisting of 216 color squares.
