The Elephant in the Room: Why No Persistent Conversational Memory in LLMs?

Would be good to have a solid mechanism to access that memory with a level of abstraction that allows efficient retrieval of connectable concepts like human brain does.