In that world, high quality pre-AI texts might be come really valuable, much like low-background steel.
But we also know that kid who learned everything from books, pronounces the words wrong and uses definitions for them that nobody has used in decades (work with one of those now. I thought I could talk people to death, and he wears even me out.) those AIs will sound like out of touch nerds too.
Someone once ranted about people using big words, “having a crush on their high school English teacher they never got over.” I knew exactly what he meant. I spent my childhood hiding how smart I was and part of my 20’s reveling in it. It’s off putting. Wisdom comes from everywhere, and the smartest often have the least. Now I’m solidly in the Feynman camp: if you can’t explain your domain to college freshmen then you don’t know what you’re talking about (yet).
When I’m refactoring code and introduce a new concept that’s like another one but different rules, I jump straight to a thesaurus. The word I pick out of the air might be good enough, but I guarantee you there’s a better one out there. Not fanciest or longest word, the most concise one. (Example: kind vs type in some circles of Type Theory). Some people act like that’s a crutch, but I’ve reviewed or refactored their code so I know that opinion and $5 isn’t worth a cup of coffee.
Do you really think it is just a coincidence that you happen to exist at the very last point of high value pre-AI data? ;-)