I'd argue that they don't even have speech in the same sense as we do. We choose words because they're connected to semantics we wish to convey, while LLMs choo

I'd argue that they don't even have speech in the same sense as we do. We choose words because they're connected to semantics we wish to convey, while LLMs choose words because they fit word patterns that exist in their training datasets.