Interesting read. I wonder what the authors would think of the recent Google "Inner Monologue" AI composite system? The paper mentions often that perhaps giving

Interesting read. I wonder what the authors would think of the recent Google "Inner Monologue" AI composite system? The paper mentions often that perhaps giving an LLM concrete references, a perceptual understanding of the environment, and agency in which its decisions affect that environment, could all bring LLMs very close to human cognition. Those are precisely what Inner Monologue provide to PaLM, and the results are certainly worth pondering.