What is Model Collapse?
When AI models degrade from being trained on AI-generated content instead of human data.
Definition
Model collapse is a phenomenon where AI models trained on synthetic (AI-generated) data progressively lose quality and diversity in their outputs. As AI-generated content floods the internet, future models risk being trained on this synthetic data, creating a feedback loop that degrades quality over time.
💡 Example
If a future LLM is trained primarily on AI-written articles (rather than human-written ones), its output would become increasingly generic, repetitive, and lacking in creativity — each generation slightly worse than the last.
Related concepts
Explore AI tools
Find tools that use model collapse in practice.
What is Model Collapse?
Model collapse is a phenomenon where AI models trained on synthetic (AI-generated) data progressively lose quality and diversity in their outputs. As AI-generated content floods the internet, future models risk being trained on this synthetic data, creating a feedback loop that degrades quality over time.
How does Model Collapse work in practice?
If a future LLM is trained primarily on AI-written articles (rather than human-written ones), its output would become increasingly generic, repetitive, and lacking in creativity — each generation slightly worse than the last.