Hacker News Reader

Comment by Garlef | Hacker News Reader

It seems reading comprehension is also declining:

Furniture is established as an image for memories just a few lines earlier. And the quote directly afterwards is framed precisely in this image.

View on HN

It's confirmed that Claude was involved: https://news.ycombinator.com/item?id=48247127#48248070

True. And the article is not great anyways.

But the claim was that it "makes zero sense".

Convince me that "Sir Terry Pratchett, who knew more about furniture than most" makes sense.

The fact that he used the word "furniture" in a quoted sentence from one of his books does not convince me.

I don't really want to defend the article.

My initial reading was that the author intended to imply that philisophizing about memories was a repeated thing in Pratchetts writings. Which - given my limited exposure - seemed plausible to me.

Good article? Definitely not. But I've also read similar try-hard / pseudo literary blog posts pre AI.

That's the exact problem: you wasted mental energy thinking about whether the author was implying that Pratchett was known for philosophizing on the role or memories and their similarity to furniture, when actually it was Claude spitting out a sentence that looked like the kind of thing Pratchett might have written.

Wasting time and mental energy is why this kind of thing upsets me so much.

I think we would all save a lot of time and avoid a lot of stress if it was the standard thing to do to declare at the start of a piece of writing how much of it was written by hand, and how much with or by AI and in what sense (i.e. was it fully vibe-written or just proofread to correct grammar etc).

I mean I really, really wish people did that. I don't want to fulminate since it's against the guidelines but presenting a piece of text that you haven't written entirely yourself and claiming its authorship as entirely your own is the definition of plagiarism. I think that's what bothers me the most. If you use AI and it contributes to the content of your text at all, why not just let us know? It's the fair thing to do. Don't make it look like you wrote something, or wrote it entirely on your own, when you haven't.

I think that would suffice to cool down the debate quite a bit, because then we wouldn't be constantly trying to double-think each other all the time. We'd be free to concentrate on the merits and demerits of a piece of writing for its own worth.

Not saying I, for one, wouldn't still be prejudiced, but if AI writing really gets better than human writing there will come a time that even hard-headed curmudgeons like me will have to admit it. Or just go and fume alone in the corner.

This is certainly not the case now and since almost nobody ever says "this was written with/by AI" we end up in the kind of double-guessing spiral we're in here.

It's more that:

1. The previous paragraph establishes a metaphor between furniture and memories. So you can take that sentence to also be metaphor, not literally about furniture.

2. A sentient animated luggage is a main character in the first two Discworld novels.

But neither of those things come close to justifying the claim that Pratchett knew "more about furniture than most".

Is this (ironically) the new ad hominem?

I guess it would be something like an “ad inhominem,” but that something was AI generated is a legitimate gripe so I hope we don’t start using that phrase.

definitively, no. it’s not against the person for their characteristics, but the composition of the content they’re sharing.

i think it’s more like accurately pointing out a logical fallacy, i.e., an assertion that thought was not properly applied which results in nonsense.

The reason why ad hominems are a fallacy is because it fails to engage with the actual argument being made. The same seems likely to happen when arguments are dismissed because of their AI origin. While that dismissal is somewhat reasonable now, at least as a heuristic, I expect it to become increasingly less valid.

An ad-LLM-em?

Reply on news.ycombinator.com