AI Model Collapse

The more LLM-generated data there is on the web, the more the new models will be polluted.

also see Bloomberg

Shumailov, I., Shumaylov, Z., Zhao, Y. et al. AI models collapse when trained on recursively generated data. Nature 631, 755–759 (2024). https://doi.org/10.1038/s41586-024-07566-y