Story perspectives

AI Image Makers Risk Homogenizing Visual Culture, Study Warns

12/24/2025

Summary

Story summary

A study by researchers at Dalarna University (Sweden) and Michigan State University finds that AI image makers produce 12 common visual motifs without human input.
The study used Stable Diffusion XL and LLaVA to generate images through a feedback loop.
Despite varied prompts, the AI consistently produced outputs described as "visual elevator music."
Researchers warn this could homogenize visual culture.
Human involvement is needed to preserve diversity in AI-generated art.

1 of 1

Story summary

A study by researchers at Dalarna University (Sweden) and Michigan State University finds that AI image makers produce 12 common visual motifs without human input.
The study used Stable Diffusion XL and LLaVA to generate images through a feedback loop.
Despite varied prompts, the AI consistently produced outputs described as "visual elevator music."
Researchers warn this could homogenize visual culture.
Human involvement is needed to preserve diversity in AI-generated art.