Hacker News Reader

Comment by alendit | Hacker News Reader

What I notice quite often is a sort of "LLM derangement syndrome" where people start flagging the most innocuous figures of speech and speech pattern as "clearly LLM".

Quite ironic when we are talking about an article which argues for extending the benefit of the doubt.

View on HN

These patterns aren't somehow alien to human speech. It's the quantity. In ChatGPT (and I think it must be ChatGPT specifically) the default writing voice tries for dramatic and profound by dramatically overusing these figures of speech, chaining them together one after another. But they're things most writers use occasionally, one paragraph in ten or twenty. If you're giving a TED Talk or some other elaborate, melodramatic persuasive public speaking role, you might use them one paragraph in four. ChatGPT uses them every other paragraph at least, sometimes multiple times in the same paragraph, so something like a negation comparator or a triplet is present in sentence after fatiguing sentence.

It's trained on the speech of online commenters like us; That's where it's getting these. I'm not gonna change just because Timmy the Zoomer has been with his LLM girlfriend for the past three years and occasionally I remind him of her. I'm not going to deliberately code switch into somebody who doesn't open a new tab to Google things as he talks about them. And you can take my bullet point lists from my cold, dead hands.

I mean I think were agreeing, in a body of text this size I'd expect to see the dramatic short sentences and standalone dramatic sentence maybe one or twice. This article they are constant. They're great devices were used appropriately, but when the whole body of text reads like it it's awful.

Not asking anyone to stop using those writing devices, just begging for any work to not be exclusively dramatic devices. And when you read a piece of work like that its generally a child or AI lol.

I thought the piece sounded like AI as well, so I ran it through Pangram, which reports it as 100% AI. Considering Pangram's extremely low false positive rate, I'm inclined to believe this was AI written.

And? That doesn’t mean it’s not true.

Some people say that AI is glorified spellchecker and then they complain when it’s used like a glorified spellchecker.

Didn't have the same feel about the piece. But based on the parent comment, decided to also test it, using Pangram free "credits". For me it reported that only sections of it (30%) have been "touched" by an LLM. There is 60% of human written text.

Pangram does not report this as 100% LLM generated. Why do I use the word "touched" - because there are some specific details that must have come from some existing text and are not elsewhere.

How did I do the check? Please do check my results! I copied just the authored prose from the page, carefully removing any comments and site text that got mixed in. I did not have it analyze the "URL".

In looking at the details, a large part of the story is flagged as all human generated.

I've personally used LLMs when I've been very upset. I used it to rewrite / edit my rant into something with a more balanced tone. The modified text got what I wanted, my original rant would almost certainly not have.

So I do think this was a thought out piece that was largely written by a human who used an LLM as an "editor" to polish the writing. Someone took time to think about this and write it.

Reply on news.ycombinator.com