Study of historical material in generative artificial intelligence

The researchers in the project aim to examine the role that historicity and history play in AI-generated text and images.

The ability of artificial intelligence to autonomously produce convincing synthetic images and text challenges conventional notions of human creativity, aesthetics, ethics and law.

The fact that generative AI – such as ChatGPT or Stable Diffusion – is trained on massive amounts of online data that is necessarily historical raises questions about the role historicity and history can play in synthetic creativity and AI models. What are the implications for historiography when AI generates synthetic images of an historical nature? How is historical material expressed in synthetic media?

“Historicity” means historically true or real, often in contrast to fictionality. Although parts of a history may stem from myth-making, propaganda or fiction, not everything need be untrue.

The main purpose of the project is to study the role of historicity in generative artificial intelligences by examining and reflecting on synthetic image and text production, based on scientific and artistic expressions from the 19th and 20th centuries. The researchers will do this by training and interacting with generative AI models while posing critical and historical questions about “synthetic pasts” articulated by these technologies.

The project will be carried out in an interdisciplinary environment involving experts in media history, literary studies, media theory, and artistic research. The project team will

• Investigate how history is incorporated into the framework of generative AI.
• Situate contemporary AI technology in media history.
• Integrate media theory with generative AI technologies in an educational framework to examine machine-generated historiography.
• Develop critical perspectives that provide an understanding of how AI relates to history, historicity and human creativity.