What is the pro-AI argument against "Model Collapse"?

A user on reddit.com is seeking pro-AI arguments against "model collapse," a phenomenon they learned about after searching "AI trained on AI problem." They express skepticism about AI's future, particularly 10-20 years from now, when a significant portion of online content might be AI-generated. The user questions how AI models can avoid being trained on other AI-generated content, leading to an inevitable model collapse, given the widespread push for AI usage across all aspects of life.