In online discussions many people do not really make arguments from knowledge. People will arbitrarily choose a "side" on some topic, and then formulate arguments by searching for "proof for why my random opinion is totally right" or "proof for what this guy is saying is totally wrong."
There's no exchange of ideas or views, the entire point of discussion, because the person you're discussing the issue with doesn't know what they're talking about. So it's a complete waste of time. Yet this can be hard to see at first because, thanks to the internet, even a blind man could describe the art of the Sistine Chapel with absolute clarity.
ChatGPT will make this orders of magnitude worse. Because now that "proof my biases are right" will become more eloquent and more dynamic. And the lack of knowledge even more extreme, because now you need not even be able to paraphrase what you're reading. Just dump it straight from the bot.
And then there will be active abuse. Now spammers and governments alike can have a million bots spamming either for viagra or war. And it will be done using dynamically generated text that can be tied into any arbitrary topic. And populating those bots with "real" histories will be trivial by having them formulate random, but relevant, comments in topics outside their "real" purpose.
It feels hyperbolic, yet it seems very possible that text-generation software may kill off meaningful discourse on the internet, at least outside of small vetted communities. And while internet dialogue has always been a bit of a bubble, the inevitable "dynamic propaganda" now means you'll never be able to even get a remote sense of the views of the public at large.