Comment by mananaysiempre

OTOH I’ve had blog posts I wrote two decades ago vehemently called out as AI generated. AI generated style unfortunately means writing that tested positively in human A/B testing, now over represented in a style used largely by AI.

So if you write in a way that engages the reader, you’re going to struggle not to use em dashes and the occasional a/b contrast, because those are challenging the reader to engage… but when overused, they not only don’t have the intended effect ( to break the reader out of passivity) , they also constitute a new kind of sin.

So no, don’t “trust your gut”. Trust the math. Is it too much? Or is it just trying to jar you out of not engaging with the prose?

But yeah, I’d say this article is likely written primarily with AI. Which doesn’t mean it’s not guided with intention and potentially important, it just means the article was probably commissioned and edited by a human, not written by one.

Indeed, consider these two posts linked below also from this blog. They look the same, they maintain the same impersonal writing style. There's no humanity to it at all.

They maintain such a consistent paragraph length that they're either a professional copyeditor or, as is clearly the case, are an LLM.

Humans deviate a lot more than this, they use run on sentences or lose the thread in their writing.

This blog however reads like every-other post on LinkedIn. Semi-professional tone, with a strong "You, Me" hook to most posts.

I encourage everyone to make an LLM-generated blog, don't post the articles anywhere, but generate one, to get a feeling for how these things write.

Because this is unmistakably LLM. I'd even go so far as to identify the model of these particular posts as ChatGPT.

Yet when we point this out, we're told it is "unmistakably human" and that we're rude for pointing it out.

https://adele.pages.casa/md/blog/the-joy-of-a-simple-life-wi...

https://adele.pages.casa/md/blog/finding_flow_in_code.md

I started off hedging but by the end of the comment came to think that AI use—or lack thereof—was actually beside the point. I have feelings with regards to the situation where “the situation” includes some largely irrelevant-to-writing things like the mainframization and the “feelings” are not nearly coherent enough to graduate to thoughts. Thus (unlike some others) I don’t think that calling out writers or warning readers about AI is all that useful (or for that matter courageous). With respect to writers who use AI due to a lack of confidence, it’s probably even harmful. (Saying that as a person who manages to absolutely suck in embarrassing ways in multiple foreign languages. And also in English but less obviously. And likely in my native language too due to lack of use.) Meanwhile, TFA makes a decent point, and I am in no position to criticize people for being wordy.

The thing is, by now it doesn’t actually matter if AI or not AI or partly AI or whatever, because the record scratch is still there and still breaks my immersion. I could be oversensitive (I definitely am to some other English-language things, and also feel that others are to yet other things like em dashes), but it feels like there’s a new language/social-signalling thing now, and you may have to avoid it even if you’re not an LLM.

Any style can cross the border into bad and get in the way of itself when it's turned up to 11, no matter who wrote it.

There've been stylistic fads before LLMs where a thing, with results just as chalkboard-screech-inducing as the current one. That this one is just a button-push away does make it worse, though, because it proliferates so greedily.

Bad writing is bad writing, and writing like an LLM is writing like an LLM. We should be able to call this out. In fact, calling out the human responsibility in it is the very opposite of dehumanizing to me.

Only to a limited extent, the fine tuning of these models uses a much smaller more curated set to generate tone and defaults.

The whole corpus is in there, but the standard style is tuned for.

I wonder how much marketing copy has poisoned the "default" writing style of LLMs, it surely has those undertones of pitching a sale in an uncanny valley way.

So I will say that things I read were not written in this style.

And people I read had better ability to not put in unneceasary random completely made up facts or illogical implications.

LLMs don’t own these expressions in the same sense that McDonald’s doesn’t own salt: they are undoubtedly making use of a strong reaction that humans have had—have been having—long before; but they did develop a way to mash that button on an industrial scale like few before them. (With of course a great deal of help from humans, be it via customer surveys or RLHF; or you could call it help from Moloch[1] in that the humans unwittingly or negligently assembled themselves into a runaway optimizer.) So I think it’s fair to say that LLMs do own this style, as in the balance of ingredients, even if they do not own the ingredients themselves. And anyway nothing in the social perception of language cares about fairness: low-class English speakers did not invent negative agreement (“double negatives”), yet it will still sound low-class to you and even me (and my native language requires negative agreement).

As for being dehumanizing, perhaps I did commit the sin of psychoanalysis at a distance here, but I’ve felt enough loose wires sticking out of my brain’s own language production apparatus that I don’t think pointing out the mechanistic aspects reduces anyone’s humanity.

For instance, nobody can edit their own writing until they forget what’s in it—that’s why any publishing pipeline needs editors, and preferably two layers of them, because the first one, who edits for style and grammar, consequently becomes incapable of spotting their own mechanical mistakes like typos, transposed or merged words, etc. Ever spotted a bug in a code-review tool that you’ve read and overlooked a dozen times in your editor? Why does a change in font or UI cause a presumably rational human being to become capable of drawing logical inferences they were not before? In either case, there seems to be a conclusion cache of sorts that we can’t flush and can’t disable, requiring these sorts of actually quite expensive hacks. I don’t think this makes us any less human, and it pays to be aware of your own imperfections. (Don’t merge your copy- and line editors into a single position, please?..)

As for syntactic patterns, I’ve quite often thought of a slick way to phrase things and then realized that I’d used it three times in as many sentences. On some occasions I’ve needed to literally grep every linking word in my writing to make sure I haven’t used a single specific one five times in a row. If you pay attention during meetings or presentations, you’ll notice that speakers (including me!) will very often reuse the question’s phrasing word for word regardless of how well it fits, without being aware of it in the slightest. (I’m now wondering if lawyers and witnesses train to avoid this.) Language production is stupidly taxing on the brain (or so I’ve heard), so the brain will absolutely take every possible shortcut whether we want it to or not.

Thus I expect that the priming effect I’m alleging can be very real even before getting into equally real intangibles like “taste”. I don’t think it dehumanizes anyone; you could say it dehumanizes everyone equally instead, but my point of view is that being aware of these mechanical realities of the mind is essential to competent writing (or thinking, or problem solving) in the same way that being aware of mechanical realities of the body is essential to competent dancing (or fighting, or doing sports). A bit of innocence lost is a fair trade for the wisdom gained.

(Not that I claim to be a particularly good writer.)

[1] https://slatestarcodex.com/2014/07/30/meditations-on-moloch/

[dead]

[dead]