The author should really take a long pass through the multidisciplinary investigation of what embodiment means. Multi-modal is just a start, physical agency an

The author should really take a long pass through the multidisciplinary investigation of what embodiment means. Multi-modal is just a start, physical agency and proception are the cores of what is missing from LLM that people mean when they say such systems don't "know" what things are. What they don't have is not just the sorts of categorizations and propositional reasoning around it implicit in e.g. Wikidata entity maps and Prolog/Wolfram like propositional reasoning. It's also having a body that inhabits space and time with other agents and bodies.