Our current AI models have massive problems. They do not plan. They gravitate towards normalcy. Their world model is NOT a set of functional components. This does not mean we will not solve these issues. Current LLMs are likely over hyped. Future AI systems perhaps not.

LLMs are a massive step in AI. But lets not kid ourselves… Intelligence, especially general intelligence is very difficult. Scaling LLMs is looking less and less likely to lead us to this general intelligence.