Reed NewsReed News

AI Model Produces Stereotypical and Inaccurate Images of 1936 in Swedish Research Study

Science & technologyScience
Key Points
  • AI-generated images of 1936 were stereotypical and sometimes inaccurate, including anachronisms like mobile phones.
  • Researchers used the Stable Diffusion model trained on 5.8 billion web images for the study.
  • The study focused on 1936 in Sweden to test AI's handling of everyday historical contexts rather than dramatic events.

Researchers at Malmö University have tested how an AI model visualizes the year 1936, finding the results to be stereotypical and sometimes outright inaccurate. 8 billion web-based images. When asked to generate images of 1936 in Sweden and Scandinavia, the AI produced black-and-white pictures of people in period clothing—but sometimes holding mobile phones.

The researchers deliberately chose 1936 as a seemingly uneventful year marked by emerging modernity to examine how AI handles everyday historical representation rather than dramatic events.

Transparency

How we verified this article

UnconfirmedBased on 1 sources
1 sources3 Involved