Abstract This paper addresses the question of the role that artificial intelligence (AI) image generators play in the reception of the ancient world, examining the assumptions on which they draw in the generation of images, and how the creation of such images influences perceptions about the classical past. After a brief outline of how AI image generators work, highlighting the inherent presumptions and biases of the visual productions, a small case study is then presented, in which the prompt ‘Roman Britain’ was submitted to eight different free image generators. The conclusion drawn from this experiment is that while the technology is impressive, none of the image generators have managed to produce pictures that effectively conjure up Roman Britain. Although the tools may be good at creating a general impression, individual details are often incorrect. Moreover, the output depends heavily on the training data available. In the case of the ancient world, no photographs exist; only archaeological remains, fragments, and later imaginative reconstructions survive. Consequently, these limitations inevitably shape the generated images. Despite these disadvantages, it is likely that AI-generated images will become part of cultural heritage, and it is, therefore, important to consider the role that such images might play in the reception of antiquity. In recognition of the problems, and the advantages, of this technology, some suggestions are made in the final section of the paper as to how generative artificial intelligence (GenAI) images may be used in a positive manner, particularly within the classroom.
Lisa Maurice (Tue,) studied this question.