My model suggests we're overestimating the intelligence of Large Language Models and understimating the complexity of human language - most LLMs are still just fancy statistical pattern recognizers, not true AI.