Hacker News new | ask | show | jobs
by exmadscientist 112 days ago
I disagree. Articles written by AI are inherently less trustworthy (they're notorious for fabrications and hallucinations) and often have a very low content density. "Write me a 10 paragraph article about high hardware prices for my blog" is the sort of thing that expands to a lot of fluff with not a lot of content. I don't really want to read that article.

Even if AI might (might) be justifiable as an editor, it's still such a negative signal for "is this worth reading?" that in my opinion it is worthwhile to point out and discuss.
2 comments
bryanlarsen 112 days ago
This case is interesting, because it seems obvious that the AI accusation is just plain wrong. The article is riddled with the kind of grammatical and spelling mistakes that humans regularly make but that a modern AI would never make.
link
exmadscientist 112 days ago
It's also very easy to paste a paragraph in to a chatbot and ask it to revise it. Or ask it to write an introduction.

I don't really have a problem with that use of AI.

But one of the costs is reputational: potential readers are now going to assume AI wrote the whole article, fairly or unfairly. That's a consideration writers have to weigh before choosing to do this.

link
83 112 days ago
That could easily be part of their prompt. I just did a quick test telling Copilot to "add a few spelling and grammar mistakes to look like a human wrote it" and it does a reasonably convincing job.
link
bryanlarsen 112 days ago
The errors are more than superficial; few authors would be willing to have AI screw up their work enough to invert the meaning of some of their sentences. https://news.ycombinator.com/item?id=47169256
link
AnimalMuppet 112 days ago
I agree that AI-written text often has a low content density. I wonder if it's a matter of information theory.

Information theory defines the information of a symbol as being related to how often it occurs and how often it is expected to occur. Something that isn't expected carries more information. (Usually "symbol" is defined as one character or byte, but it could be a word or word part.)

Well, if you think about LLMs that way, they give you the most-probable next word (or word part). That means that they give you less information than normal writing. I suspect that's why it reads as bland, low-content - because it really is low content, in the information theory sense.

Now, it doesn't always give you the most probable next symbol. There is some randomness. And you can increase the randomness by turning up the temperature. But if you do, then I suspect it becomes incoherent more quickly. (Random gibberish may have high information from an information theory standpoint, but humans don't want to read that either.)

link