Comment by evanelias
1 month ago
I didn't mention em dashes anywhere in my comment!
If this isn't AI writing, why say "The “New Account” Trap" with then further sub-headers "The Legal Catch", "The Technical Trap", "The Developer Risk"... I have done a lot of copyreading in my life and humans simply didn't write this way prior to recent years.
> humans simply didn’t write this way prior to recent years.
Aren’t LLMs evidence that humans did write this way? They’re literally trained to copy humans on vast swaths of human written content. What evidence do you have to back up your claim?
Decades of reading experience of blog posts and newspaper articles. They simply never contained this many section headers or bolded phrases after bullet points, and especially not of the "The [awkward noun phrase]" format heavily favored by LLMs.
So what would explain why AI writes a certain way, when there is no mechanism for it, and when the way AI works is to favor what humans do? LLM training includes many more writing samples than you’ve ever seen. Maybe you have a biased sample, or maybe you’re misremembering? The article’s style is called an outline, we were taught in school to write the way the author did.
2 replies →
You’re pointlessly derailing a conversation with a claim you can’t support that isn’t relevant even if true.
Regardless of whether AI wrote that line he published it and we can safely assume it is what he thinks.
[flagged]
I don’t think you even know what you’re arguing about anymore. You claimed that what the author wrote wasn’t what the author thinks. As evidence you provided weak arguments about other parts of it being AI written and made an appeal to your own authority. It doesn’t matter if AI wrote that line, he wrote it, a ghost writer wrote it or a billion monkeys wrote it. He published it as his own work and you can act as if he thinks it even if you don’t otherwise trust him or the article.
1 reply →
LLMs learned from human writing. They might amplify the frequency of some particular affectations, but they didn't come up with those affectations themselves. They write like that because some people write like that.
[flagged]
Those are different levels of abstraction. LLMs can say false things, but the overall structure and style is, at this point, generally correct (if repetitive/boring at times). Same with image gen. They can get the general structure and vibe pretty well, but inspecting the individual "facts" like number of fingers may reveal problems.
That seems like straw man. Image generation matches style quite well. LLM hallucination conjures untrue statements while still matching the training data style and word choices.
2 replies →
This is a bad faith argument and you know it.
1 reply →
> I didn't mention em dashes anywhere in my comment!
I know. I just mentioned them as another silly but common reason why people unjustly accuse professional writers of being AI.
> I have done a lot of copyreading in my life and humans simply didn't write this way prior to recent years.
What would you have written instead?
Most of those section headers and bolded bullet-point summary phrases should simply be removed. That's why I described them as superfluous.
In cases where it makes sense to divide an article into sections, the phrasing should be varied so that they aren't mostly of the same format ("The Blahbity Blah", in the case of what AI commonly spews out).
This is fairly basic writing advice!
To be clear, I'm not accusing his books as being written like this or using AI. I'm simply responding to the writing style of this article. For me, it reduces the trustworthiness of the claims in the article, especially combined with the key missing detail of why/how exactly such a large gift card was being purchased.
> To be clear, I'm not accusing his books as being written like this or using AI. I'm simply responding to the writing style of this article.
It's unlikely that the article had the benefit of professional, external editing, unlike the books. Moreover, it's likely that this article was written in a relatively short amount of time, so maybe give the author a break that it's not formatted the way you would prefer if you were copyediting? I think you're just nitpicking here. It's a blog post, not a book.
Look at the last line of the article: "No permission granted to any AI/LLM/ML-powered system (or similar)." The author has also written several previous articles that appear to be anti-AI: https://hey.paris/posts/govai/ https://hey.paris/posts/cba/ https://hey.paris/posts/genai/
So again, I think it's ridiculous to claim that the article was written by AI.
9 replies →
> I know. I just mentioned them as another silly but common reason why people unjustly accuse professional writers of being AI.
The difference is that using em dashes is good, whereas the cringe headings should die in a fire whether they’re written by an LLM or a human.
Heuristics are nice but must be reviewed when confronted with actual counterexamples.
If this is a published author known to write books before LLMs, why automatically decide "humans don't write like this". He's human and he does write like this!
The author is reputable, just look at the rest of their website.
Your accusation on the other hand is based on far-fetched speculation.