vonunov

Yes, this is the kind of thing I've been hoping more people would notice.

I think another tell with a long lifespan is that it will try to write about a thing it experienced, but having no real knowledge of the qualia, it'll end up only referring to it (not uncommonly using "that") and leaving all the substance as an exercise for the reader

/* This function doesn't return an int. It doesn't return a float. It doesn't return a char. It doesn't ret-- */

Not always or inherently, I would say, but if you can't help consciously noticing how much a piece uses tricolons, or negative parallel constructions, or dashes offsetting punchy clauses at the ends of sentences, then the style has probably become what you might call overseasoned

In my experience, and as someone who has technically been a professional copy editor before (in the sense that I got paid by someone to perform line/copy editing of some material or other, but definitely not who or what you would think if I only said I used to be a copy editor), the uses of "not only, but also" that didn't have anything wrong with them (or didn't deviate from convention, if you like) were in the minority by far. Assuming people in the last few years haven't all become suddenly interested in that particular construction and skilled in its use, the presence of an improperly implemented NOBA should remain vaguely reliable as an indicator of authenticity :D

The robot can be made to write any such thing

In the interest of not occupying significant page/screen height with LLM output, example prompts+responses here: https://dpaste.com/H9DXKNYQH.txt

Surely it hasn't escaped your notice that the LLMs don't tend to stick to just the one set of idioms/devices. They're constantly running new batches of them into the ground

I don't want to make this about people's faith or whatever, but to put it frankly, I've heard a lot of contemporary Christian music, I don't care for it, and [I like to think] I can reliably recognize it in three notes or fewer,¹ which may or may not bear out in rigorous testing, but saves me a lot of time either way. This feels like it parallels strongly with the topic at hand

1. erring on the side of sounding cooler

(I would drop this somewhere more relevant but apparently replying to a thing from 17 days ago would be necroing here + PMs are hard or something)

>[1] With what prompt!? I like the terse output! Do share...

Not sure if it's complete or completely right, but the matter came up in a recent session, and when asked what gave, Jim 'n' I came up with some supposedly relevant factors, at least one of which is news to me (supposedly, steering LLMs using negative instructions isn't counterproductive anymore (not that I'd been resisting the temptation anyway)):

https://gemini.google.com/share/7af54a6861d7#:~:text=What%20...

With the caveat (or bonus) that it can go (?:too)? far when told to "be blunt" and not to "pull punches":

https://i.vgy.me/WHRZD7.png (from 2024-09) (in this case, in user-config persistent instructions in Kagi's multi-LLM thing)

Just off the top of my head, an offense of impersonation will have an element along the lines of "doing [a] thing[s] such that a reasonable person [does/would] believe you're a real cop", which [optimistically] would not be satisfied as there would be no actual person being led to believe anything, or the court would [optimistically] not find that its model of a reasonable person would be genuinely convinced by someone on the internet typing "I'm an FBI agent" or whatever.

I bet it could be some interesting caselaw actually, if it resulted in circuit court judges (or whoever) writing opinions about the essence of impersonation, fraud, etc. and what kind of actual or hypothetical agent is needed to make the crime a thing that could have happened. E.G., basically, if you sit alone in a room where nobody else can see or hear you, and you put on a realistic local police uniform and declare to the room that you're a licensed police/peace officer, is a crime being committed (i.e., is the nature of the crime "pretending/claiming to be a cop" or "making an actual person really believe it" or something else)

(could also be an intent element to satisfy, not sure)

the FCC doesn't seem to think so

>And, just as I can now recognize Christian music in three notes or less, I can spot ChatGPT output without necessarily even reading any one contiguous string of it. I can just tell by the shape or something.¹

1. https://v-n-n-v.github.io/chatgpt-voice.html (bit outdated now, from 2025-03)

Well, I mean, kind of https://i.vgy.me/UtZJt0.png

If it has any regularity to it, then LLMs will, so to speak, figure it out. Maybe if you do it like a game of Mao¹ you could make it a little bit harder for them.

1. https://everything2.com/node/e2node/How%20to%20play%20Mao

If you also want a thing to do typewriting on, later-model word processors (e.g., Smith Corona PWP series) tend to have a feature that auto-types text entered and edited earlier. It'll have the imprinting insofar as the type impacts the paper, but it's not going to have the off-center / over-inked / patchy manual typewriter look. For that you may just want to find a font face that replicates it.

I finally got all the nits picked out, except it seems to erroneously point out supposed section reference errors:

----

Reference to 'Section 13.1.3' but there is no Section 13 in the document - this appears to be a copy-paste error from another lease form

Cross-reference to 'Section 2' is incorrect - the non-real property items are actually listed in paragraph 12, not in any 'Section 2'

----

It seems to have trouble with the technical-writing numbering (e.g., 3.1.23) and insists on referring to sequential paragraph numbers even though the paragraphs aren't labeled and referenced that way. This numbering is done using the usual numbered list object but I don't remember if I had to configure a variant of it to do this. If you need I'll anonymize the document and send it, as long as you promise not to make too much fun of it (I was a party to the contract so I can get away with having drafted or substantively edited it despite not being an attorney, according to some guy on the internet; govern yourself accordingly or whatever it is you people say these days)

> disingenuous pretending at human experience; alluding vaguely to "that [relatable experience]" and hoping we won't notice that there is clearly no actual knowledge of the qualia it tries to invoke

> subtly detached or glossed-over metaphors

> feels empty

> alluding to things without actually talking about them

> using many words to say nothing

https://v-n-n-v.github.io/chatgpt-voice.html

See also: https://i.vgy.me/ZIxnSh.png

You're really on to something there (-:

Interesting, thanks. Notwithstanding all I've said, I've also struggled with a nebulous sense that there's something to handwriting, possibly bordering on "sacred", to be a bit dramatic about it. This has resulted in phases of annoying obsession with handwriting where all its technical/practical shortcomings are overwhelmed by this unknown property for a time.

The sub-process thing sounds vaguely plausible, but at the same time, I'm under the impression that I'm already constantly evaluating whether what I'm writing makes sense, its implications and assumptions, etc. At least I feel that I'm attentive to the impression of it not making sense, though I also try not to confuse the feeling that something makes sense with the fact of it being so.

Anyway, what I'm going for here is, did you consciously apply any technique or process to evaluating this or do you come by this insight naturally?

https://i.vgy.me/JuGzDb.png (user input marked)

I suspect it's to do with fluency, at least partially. Have you looked into how it goes for very fluent typists? Problem is, I suspect you might get enough fluency that typing isn't distracting as compared to handwriting only at significantly faster speeds than most people consider the beginning of "fast".

Chicago says to format dashes like this—and ellipses . . . like this. . . .

AP says to format dashes like this — and ellipses ... like this. ...

Who's "correct"?

Embrace the double hyphen -- it's still attested in Garner's ;)

Yes, the LLMs have made great progress in that regard. It wasn't too long ago that the majority of dashes seen in LLM material could have been commas, periods, or nothing at all with no loss of tone or meaning, and almost none were used to offset parenthetical phrases. It was nearly exclusively an overdramatic flourish. And now look at them. They're growing up. Just makes you want to squeeze them until they pop.

Trivia Madness!:1 4 lyf

Now I'm just burning to know how the debate about which way to round it went down

Are we talking about with OTR? You're meant to verify fingerprints out of band as usual. Without, I guess you check if they've authenticated to nickserv if there are services. Or do your own checks or heuristics.

or what

∙ Paragraphs: 32; ∙ Sentence-level rule of three: 9; ∙ Paragraph-level rule of three: 10; ∙ Parallel contrast: 11; ∙ Dramatic sentence-initial conjunction (polysyndeton): 6; ∙ Literary conjunctionless list (asyndeton): 9; ∙ Foo x, foo y, [foo z] (anaphora): 14; ∙ Escalation of sentiment/gravitas (auxesis): 12

One of the most glaring and insulting parts, of course, is the "too . . . yourself" when we know good and well that there is nobody hesitating and questioning whether to be vulnerable, only something pretending. Like it thinks we're stupid or something. That's the bottom-line or nutshell or whatever regarding why it's irritating to receive lazy LLM output from people. It's like they think everyone but them is stupid and won't notice that they don't care.

I was wondering. Maybe "refined", as a derivative of the verb?