By Ahmed Elgammal
Publication Date: 2026-01-22 13:36:00
Generative AI has been trained on centuries of art and writings created by humans.
But scientists and critics have wondered what would happen if AI became widespread and began training its results.
A new study provides some answers.
In January 2026, artificial intelligence researchers Arend Hintze, Frida Proschinger Åström and Jory Schossau published a study showing what happens when generative AI systems are allowed to run autonomously – that is, generate and interpret their own output without human intervention.
The researchers linked a text-to-image system to an image-to-text system and had them iterate over and over again – image, caption, image, caption.
No matter how diverse the initial stimuli were—and no matter how much randomness the systems allowed for—the results quickly converged on a narrow set of general, familiar visual themes: atmospheric cityscapes, grandiose buildings, and rural landscapes. Even more striking…
