It seems like the two approaches compliment each other nicely. Human intelligence also relies on parallel information processing. LLMs are like a massive workin

It seems like the two approaches compliment each other nicely. Human intelligence also relies on parallel information processing. LLMs are like a massive working memory, incredibly effective but with a similar set of limitations. What they lack is a symbolic model of reality, something that they can build and refine.