If you're plan to keep improving LLM's is based on more training, then we have an AI hype problem.