Comment by armchairhacker | Hacker News Reader

Comment by armchairhacker | Hacker News Reader

How much is ChatGPT replacing human writing?

It still seems to me like most of the articles on HN are human-written, and I heard CNBC tried replacing writers with ChatGPT and it went terribly. But I've also heard anecdotes of freelance writers getting replaced [1] [2] [3], and every once in a while I spot or hear about something ChatGPT-written in the wild. I'm sure ChatGPT replaces boilerplate writing (e.g. legal documents, mandatory code documentation, technical specifications, formal communication) but what about "real" writing (articles, papers, slogans, and books)?

ChatGPT (including GPT4) has a distinctive writing style (neutral-helpful tone, verbose), and in my own experiments and observations I can still recognize it even when it's prompted to write differently. But this is selection bias: when ChatGPT writes something I can't spot, I just don't notice and assume it's human-written. And I know for a fact I can't recognize ChatGPT's writing in isolated paragraphs (I took the quizzes), so it's clear some writings slip past my radar.

This is an important thing to track. Idk how we'd track it in a way that's even remotely accurate; perhaps by taking small samples where you know for sure whether each article was written by GPT or human (if your subjects disclose, or if they have edit history turned on). But we can't say how much ChatGPT is replacing human writing, and how it's affecting human thinking, until we have actual evidence this is happening at a large scale.

[1] https://www.reddit.com/r/ChatGPT/comments/139o1q6/lost_all_m...

[2] https://www.reddit.com/r/ChatGPT/comments/138clv9/spent_5_ye...

[3] https://www.reddit.com/r/freelanceWriters/comments/12ff5mw/i...)

ChatGPT can't replace writing intended to convey information, but that's only a small fraction of the writing that's getting created. There's a whole industry where people are given prompts like "five paragraphs about the history of spaghetti" and get paid $0.01 per word, then that text gets put into an Italy-themed travel blog full of ads.

You and I wouldn't notice if that entire industry was replaced by ChatGPT because nobody reads that text, it doesn't exist to be read, it exists to tickle some ranking model deep in the bowels of Google. But you can be sure that the people selling coherent text for $0.01/word will notice when ChatGPT can generate the same thing for $0.000001/word.

I don't remember what I search yesterday, but the whole first Google page result was useless websites like that, endup using site :-/

I think you're right, and it's so depressing.

What I need is not an AI that generates that bullshit, but an AI that detects it and ensures I never have to encounter it.

Those people are not going to be hired in the future. The companies will use ChatGPT directly.

I don't disagree, by the way. I think ChatGPT may be a net negative for us, but the issue you describe has more to do with the way Google rewards the web

While reading your post I wondered if it was written (in part) by chatgpt. I hope you don't take offense, I'm not saying that it's written badly, it's more about of paranoia activating when there is long-form content talking about chatgpt writing posts in forums and such.

> there is long-form content

The post you're replying to has had a lot more effort put in than the usual forum post but... long form? Anything longer than 160 characters is long form now?

I don't think it sounded particularly ChatGPTish but yeah for me any longish comment split into three even paragraphs sets off alarms in my head now.

Later on, gpt or other spun off gpts can be trained to write with a specific voice. The future is limitless with this tech, for good or bad

Reply on news.ycombinator.com