GPT's text as quoted in the article has a clear thesis stated upfront, later expands on it with examples, summarizes its arguments at the end, and does all this with gusto.
You cap your comment by a non-sequitur that GPT is not going to replace journalism.
IMO the piece of text generated by GPT offers more insight and is wittier than yours.
My comment was not an essay, it was a response to someone else's comment. The GP was explicitly saying that GPT-3 could probably produce many of the news content they read, and I was replying to that.
Please also note that it's not very clear to what extent the text of the article is edited - the non-bold text is written by GPT-3, but I don't think it was produced as a single block of text. Instead different parts (sentences? Paragraphs?) were produced individually, selected by a human from many other responses, and assembled together in the shape we are shown. The train of thought among the paragraphs is most likely entirely human work, not AI work, and only the best sounding paragraphs out of a lot of gibberish were likely selected.
It would also be interesting to see how close those paragraphs are to something in GOT-3's training corpus, in terms of structure if not explicit language.
Comments
GPT's text as quoted in the article has a clear thesis stated upfront, later expands on it with examples, summarizes its arguments at the end, and does all this with gusto.
You cap your comment by a non-sequitur that GPT is not going to replace journalism.
IMO the piece of text generated by GPT offers more insight and is wittier than yours.
My comment was not an essay, it was a response to someone else's comment. The GP was explicitly saying that GPT-3 could probably produce many of the news content they read, and I was replying to that.
Please also note that it's not very clear to what extent the text of the article is edited - the non-bold text is written by GPT-3, but I don't think it was produced as a single block of text. Instead different parts (sentences? Paragraphs?) were produced individually, selected by a human from many other responses, and assembled together in the shape we are shown. The train of thought among the paragraphs is most likely entirely human work, not AI work, and only the best sounding paragraphs out of a lot of gibberish were likely selected.
It would also be interesting to see how close those paragraphs are to something in GOT-3's training corpus, in terms of structure if not explicit language.