Eventually there will be so much ChatGPT spam that the only way to cope with it is to have an LLM respond to incoming requests as a first line of defense, gener

Eventually there will be so much ChatGPT spam that the only way to cope with it is to have an LLM respond to incoming requests as a first line of defense, generating _even more_ machine content in response, and a thread that someday a crawler will consume and use to train another generation of models. And every step of the way companies get their cut of the generation costs.