Autonomous language-image generation loops converge to generic visual motifs
Patterns · 2026-01 · 7 (1) · 101451
DOI: 10.1016/j.patter.2025.101451 ↗Kurz gefasst
Fachartikel in der Zeitschrift Patterns, 2026-01. Verfasst von Arend Hintze und 2 weiteren Autor:innen. Beteiligte Einrichtung: Department of Data Analytics, Dalarna University, Falun, Sweden. Die Arbeit wurde bisher 12-mal zitiert (Crossref) und nennt 45 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Michigan State University.
Abstract (Originalsprache)
Autonomous AI-to-AI creative systems promise new frontiers in machine creativity, yet we show that they systematically converge toward generic outputs. We built iterative feedback loops between Stable Diffusion XL (SDXL; image generation) and Large Language and Vision Assistant (LLaVA; image description), forming autonomous text → image → text → image cycles. Across 700 trajectories with diverse prompts and 7 temperature settings over 100 iterations, all runs converged to nearly identical visuals-what we term "visual elevator music." Quantitative analysis revealed just 12 dominant motifs with commercially safe aesthetics, such as stormy lighthouses and palatial interiors. This convergence persisted across model pairs, indicating structural limits in cross-modal AI creativity. The effect mirrors human cultural transmission, where iterated learning amplifies cognitive biases, but here, diversity collapses entirely as AI loops gravitate to high-probability attractors in training data. Our findings expose hidden homogenizing tendencies in current architectures and underscore the need for anti-convergence mechanisms and sustained human-AI interplay to preserve creative diversity.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Data Analytics, Dalarna University, Falun, Sweden.
- BEACON Center for the Study of Evolution in Action, Michigan State University, East Lansing, MI, USA.
Förderung
- Michigan State University
Zitieren
APA
Hintze, A., Proschinger Åström, F., & Schossau, J. (2026). Autonomous language-image generation loops converge to generic visual motifs. Patterns, 7(1), 101451. https://doi.org/10.1016/j.patter.2025.101451
Vancouver
Hintze A, Proschinger Åström F, Schossau J. Autonomous language-image generation loops converge to generic visual motifs. Patterns. 2026;7(1):101451. doi:10.1016/j.patter.2025.101451
BibTeX
@article{Hintze2026,
author = {Hintze, Arend and Proschinger Åström, Frida and Schossau, Jory},
title = {Autonomous language-image generation loops converge to generic visual motifs},
journal = {Patterns},
year = {2026},
volume = {7},
number = {1},
pages = {101451},
doi = {10.1016/j.patter.2025.101451}
}
EUVIMED