I wonder… If humans were all super serious, direct, and not funny, would LLMs trained on their stolen data actually function as intended? Maybe. But such people do not use LLMs.