The answer is not to "disconnect". The answer is to read critically, to read for argument, and to try to connect with the intent of what the author was saying. All the while realizing that most of the content out there is bullshit(https://en.wikipedia.org/wiki/On_Bullshit), whether written by a human or not. And yes, the amount of bullshit is about to increase exponentially. But it was already a majority.
Ironically, this post itself doesn't pass that test. I'm not necessarily claiming that it was written by GPT-3 (although it could have been and I have a track record of pointing that out: https://news.ycombinator.com/item?id=23894742) There are a lot of humans that write like this too.
But it has the hallmarks a language model text; it lacks an argument (beyond a simple sentence that could have been given in the prompt), its explanation of how generative models work isn't particularly accurate, its recommended course of action is anodyne and one has the sense that most of the sentences are just there to take up space.
Same as we do now. Nobody currently goes around reading random books or articles. We use link aggregators from communities we trust, reviews from reputable establishments, or recommendations from our friends.
These mechanisms are certainly going to have to adjust (I estimate at least 10% of links posted on HN are already GPT-generated, for example, and who knows how many comments -- it's much harder to tell when thinks are shorter.) StackOverflow is already in the midst of a ChatGPT-induced moderation crisis.
Being judicious about the communities you take recommendations from, the content you spend time consuming and thinking about things critically is already an intellectual survival trait. It's just going to become that much more important.
I skim checked some paragraphs using https://huggingface.co/openai-detector/ and the article looks to be written by a human. However, the article from the link you gave I quickly found a “fake” (generated) paragraph.
That said, ChatGPT can produce text that looks 99.98% real on the GPT2 measurement tool. Also some people write GPT-looking text naturally (shallow mimicked shit without deeper semantic meaning!).
Ironically, this post itself doesn't pass that test. I'm not necessarily claiming that it was written by GPT-3 (although it could have been and I have a track record of pointing that out: https://news.ycombinator.com/item?id=23894742) There are a lot of humans that write like this too.
But it has the hallmarks a language model text; it lacks an argument (beyond a simple sentence that could have been given in the prompt), its explanation of how generative models work isn't particularly accurate, its recommended course of action is anodyne and one has the sense that most of the sentences are just there to take up space.