I'm still not convinced that LLMs are the future of human-computer interaction - sure, they're good at generating text, but can they actually understand the context and nuances of human communication?