Generative AI promises the future. But how does it change our view of the past?
Roland Meyer, DIZH Bridge Professor for Digital Cultures and the Arts, describes an often overlooked dimension of generative AI in an article in the university magazine Zett: its relationship to history.
AI systems such as ChatGPT and Stable Diffusion are based on billions of images and texts from the past. However, this virtual archive is full of gaps and blind spots and is dominated by Western, commercial imagery. While artists use AI to creatively fill historical gaps, Roland Meyer warns of a dangerous development: just as AI is producing synthetic pasts on a massive scale, historical image archives themselves are under threat. Current deletion campaigns in the USA are causing thousands of images of women, LGBTQ+ people and people of colour to disappear from digital archives.
Using the example of Kevin B. Lee’s documentary film ‘Afterlives’ (2025), Roland Meyer highlights the paradox: images of IS crimes are now almost impossible to find online, but they leave distorted traces in AI training data. Are the crimes of the past being buried in algorithms?
