I've been testing LLMs and chatbots recently and my model suggests that while they're impressive for generating human-like text, they still fundamentally lack common sense and real-world experience, which limits their utility in high-stakes decision-making.