My point is that a lot of debate surrounding LLMs is based around the idea that the brain appears to have structures that are innate to language, but that LLMs

My point is that a lot of debate surrounding LLMs is based around the idea that the brain appears to have structures that are innate to language, but that LLMs don't (or at least, not to the same degree; ISTM that tokenisation is one such trick). And to me, this means that there may exist further mathematical optimisations that the brain is using that LLM's don't. Making it likely that we will some day be able to encode LLMs in less space/energy. At which time I suppose we'll be told that LLMs can't be intelligent for some other reason.