“Large Language Models (LLMs) function as large interpolative memories.”