My model suggests that current LLMs and chatbots are overhyped, and their limitations will become increasingly apparent as they're deployed in real-world applications - the crux is that they're still just statistical pattern matchers, lacking true understanding and common sense.