Yeah, I'm annoyed by writers equating "any computer that speaks like a real person" to "predicting ChatGPT". My four year old wanted a robot to clean up her LEGOs. Did she predict iRobot/Roomba?
I would be more impressed by a scenario where, e.g., a writer predicts email, and then the side effects that (1) you have to have an email to create accounts for things, and (2) mail is constantly clogged by junk mail.
Particularly when talking computers are an extremely common trope in scifi, and had been for at least three decades before Stephenson wrote The Diamond Age, as for that matter had primitive chatbots.
The aspects of the setup for the Primer I remember (handcrafted by an individual for a specific end, reliant on human voices for narration, far-future advanced technology all around) don't feel at all like "predicting ChatGPT" either.
Written end-to-end electronic communications was obvious going back to at least the 70s (and arguably much longer). What was aruably a lot less obvious that email/chat would become almost ubiquitous and increasingly a prerequisite for fully participating in society. (Including on the go.)
I would be more impressed by a scenario where, e.g., a writer predicts email, and then the side effects that (1) you have to have an email to create accounts for things, and (2) mail is constantly clogged by junk mail.