The Image as the Living Counterpart: How Art and Virtual Ephemera can Contribute to Beneficial AI Modelling

This project proposes that visual features found in images in general, like agency and persuasion, can be utilized to promote a provably beneficial development of artificial intelligence (AI) systems. For centuries, artists, designers, and their audiences have negotiated for image conventions that effectively package human preferences, objectives, and behaviour, for good or for bad. Images can therefore, after analysis through an AI system trained by a vast visual databank, be used as qualified sources for the development of altruistic machines that work for human well-being. We attempt to bring the machine-analysis of images to the next level: Not only should the AI be able to recognize for example objects, animals, and simple facial features in images. The aim is to classify a large number of images not only defining what the images are or depict, but what they show to and what they are about. The AI program should thus be able to recognize human behaviour, as well as contextual knowledge. The AI system should also be taught how to target inclusion, diversity, and intersectionality. Constructing such complex categories of images will be an important task for the project. We want to provide food for AI thought.

Recent developments of AI have already led to conditions of reflexive empowerment, given it is provided advanced knowledge not only of the technical side but of its cultural aspects. We take a close look at the current state of the AI-as-artist and its impact on human artistic agency. Instead of eliminating the “human” from culture, AI might aid in discovering what is essentially human in art and visual ephemera. If a machine can learn to discern between images of for example love and hate, it could move on to other media and could eventually be able to make moral, and even autonomous choices in aid of the well-being of humanity. At such a stage, some steps have been taken on the way of making “beneficial AI” a reality. We would like to push the development along that path.

Project group: Jan von Bonsdorff, Britt-Inger Johansson, Johan Eriksson, Department of Art History, Anna Foka, Department of ALM and Digital Humaniora Uppsala Universitet, Lars Oestreicher, at the Visual Information and Interaction unit at the Department of Information Technology, all Uppsala University.