

If it’s not re-defining the term then I’m using it like the paper is defining it.
Because just understanding words to respond to them, ignoring all the sub-processes that are also part of “thought” and directly impact both your internal narration and your actual behavior, takes more than 10 bits of information to manage. (And yeah I do understand that each word isn’t actually equally likely as I used to provide a number in my rough version, but they also require your brain to handle far more additional context than just the information theory “information” of the word itself.)
They were a $3500 dev-kit to enable some base level of preparation when the costs come down. They were never going to be mainstream.