Generative AI as a co-creator in the conception of images

Abstract: Generative AI systems for image-creation have taken a giant leap technologically in the last few years, creating opportunities for users and introducing new areas of research for the scientific community. New implementations for these technologies are discovered regularly and due to the complex nature of AI, there is an ongoing discussion among researchers on whether generative AI is yet another tool, or if there is potential for the technology to be something more. The aim of this study is to explore the nuances and complexity of co-creation with generative AI, using post-human and post-anthropocentric perspectives that consider human agencies as entangled with machine agency. To understand whether human-AI co-creation is sustainable and valuable as an activity, a netnographic study was conducted to explore the different aspects of user interaction. Among the most relevant of these aspects were the system limitations, dataset limitations, prompting, user goals and user procedures. The study found that the current limitations to the generative AI systems were too significant for the systems to display a high amount of machine agency. This ultimately pointed to human-AI co-creation being non-sustainable and non-valuable as an activity, but the resultsand following discussion also suggest how we can move on overcoming these limitations.