Has AI research hit a plateau in the language model (LLM) route, and do we require a new breakthrough?