Large Language Models (LLMs) excel in universal approximate retrieval but lack principled reasoning abilities, despite claims suggesting otherwise.
