AI model market concentration barely affects collapse speed or outcome
The Oligarch Barely Steers Model Collapse in Multi-Model Ecosystems
Artificial IntelligenceComputation and LanguageMachine Learning
Summary
When many AI language models train using text generated by other models, they risk becoming less diverse and collapsing to similar outputs. The paper investigates if having a few dominant AI models (an oligopoly) speeds up this collapse or steers the results. The authors find that even when one model dominates the data pool by 90%, the collapse speed and final outcome hardly change. Instead, the quality and type of text contributing to training have a bigger effect on collapse than market share concentration.
What this means in practice
- •For machine learning engineers: Assess risks of training new language models on mixed AI-generated data without overestimating the impact of dominance by few models.
- •For ai platform operators: Optimize training data composition by controlling text source diversity rather than focusing on reducing market share concentration among providers.
Authors
Yangze Liu, Zhongyi Han
Abstract
AI-generated text is flowing back into the training corpora of the next generation of models. Recursive training on it drives model collapse, and recent work extends the setting to many models feeding one another -- but almost always with the market split evenly, while real generative AI is an oligopoly. Concentration raises two worries: fewer, more uniform sources may make collapse faster, and later models may be dragged toward the oligarch's output. We test both in controlled ecosystems: 13 open 1--4B models form natural ecosystems of 3 to 13 players, plus an injected probe that pushes the top share to 90%; each generation, every model's output is mixed into a shared pool by market share and every model is retrained on that pool from clean base weights, for five generations. Yet within the range we test, neither worry materializes; what emerges instead is an invariance. Making the split more unequal barely changes the speed of collapse. Destinations move even less: the share and identity knobs shift five-generation endpoints by only a few percent of the drift common to all arms -- the ecosystems collapse to nearly the same place. An extreme share paired with the strongest injected bias still does not guarantee steering, and the topic shifts it does produce leave only a faint trace on the ruler that measures collapse. What sets the speed is who supplies the pool and how readily those suppliers are carried along: with every share held fixed, swapping the members of a K=3 ecosystem changes five-generation drift by 2.8x; a share-weighted index of each member's susceptibility explains the speed differences across nineteen arms with R^2 = 0.68; and replacing half the pool with human text roughly halves drift without changing its course. Within the tested range, concentration sets neither the destination nor the pace of collapse; the pace follows whose text fills the pool.