If writing is thinking then what happens if AI is doing the writing and reading?

Writing isn't necessarily thinking. For example you can have a spreadsheet of 20 000 words and use an algorithm to randomly pick a sequence of words. No one would call that thinking, even if such a process made intelligible sentences sometimes.

"AI" is using an algorithm and statistics in the same way---it's just more accurate at making intelligible sentences than the example above. I wouldn't call either thinking, would you?

That’s not at all the point the author is making. The point is people don’t read. Now we don’t even write. At least what USED to be written had to cost a human time, something relatively precious.

Now we can churn out text at rates unprecedented and the original problem, no one reading, is left untouched.

The author wonders what happens when the weird lossy gap in-between these processes gets worse.

There’s lots of evidence that writing helps formulate good thinking. Interestingly, CoT reasoning mirrors this even if the underlying mechanisms differ. So while I wouldn’t call this thinking, I also don’t think reducing LLM output to mere algorithmic output exactly captures what’s happening either.

EDIT: previous != precious.