lurquer

“Despite massive investments and widespread hype around [THE INTERNET], the technology has delivered zero financial payoff to corporate America outside of [THE PORN INDUSTRY]”

(1998 version of the story)

Bingo. There is an assumption that there is a single red/green list of tokens that the provider will allow people to use to verify if something is AI.

In reality, there can be billions of red/green lists. It is trivial to test a piece of text against all of them to see if it was generated using any of the keys that made a particular red/green list. And, as you alluded to, there are enough combinations that each individual account could be assigned its own red/green list so not only would a person be able to tell if the output was AI generated, they would also be able to tell who generated it.

I believe many are missing the point as to the effectiveness of this watermark. It will be exceedingly difficult to get rid of it. Probably impossible. If normal human text uses, more or less, a 50-50 ratio of red and green tokens, and the AI generates your text with 80% red and 20% green, all the rewriting in the world is not going to get the result close enough to the human like 50-50 to avoid a statistical aberration that will be discernible.

Depends on the type of writing. Blogs and the like? LLM generates prose that, to me anyway. is unbearable.

However, in fiction I’ve found it a useful collaborator. There have been more than a few occasions when, given some notes of how I want a character’s arc to develop in a particular scene, that the LLM gives some excellent pointers and ‘new’ ideas I hadn’t considered.

As far as editing my prose, I use it as a ‘thesaurus of phrases.’ When lazy, I can give it a rough sketch of the paragraph, giving it the gist of what I want, and have it generate a dozen or so versions. I usually can find nuggets of good phrases therein that are useable… much as I would refer to Roget’s to find a more precise word.

That said, one has to resist tbe temptation of using a chunk of generated text verbatim; no matter how good it sounds in isolation, the repetitive grammatical structure and other LLM-smells add up quick and become nauseatingly obvious if used frequently.

In any case, I think LLM’s get a bad wrap for writing… when used correctly, it is incredibly useful. And, it’s tiresome to hear pretentious snobs assume that an author who uses LLM simply lacks the taste to appreciate how bad the prose sounds. Not true in all cases.

The ‘study’ is fluff.

Paraphrase: Amongst primates there is a correlation between brain size and bipedalism with handedness… (unless you exclude humans, in which case there isn’t.)

That’s like saying: “Alongst animals there is a correlation between height and neck length… unless you exclude giraffes, in which case there isn’t.”

If a correlation disappears when you remove one datapoint, then the correlation was not really a broad pattern across the dataset. It was mostly a story about that one datapoint.

I mean, I get it… you gotta publish something. But, geesh… this is beyond stupid.

The sad thing is that IF — by chance — one of those 66 had schizophrenia, the headline would undoubtedly read “Blind children are FOUR TIMES more likely to develop Schizophrenia!”

“…tracking nearly half a million children born in Western Australia between 1980 and 2001. Of those, 1,870 developed schizophrenia, but not one of the 66 children with cortical blindness did.”

Using this data, one would expect to see only 0.25 cases in those 66 blind kids.

Stated differently, there is around a 78% chance of having 0 cases in those 66 by random chance alone.

Dumb.

> Off the top of your head, what technological advancements did the different space programs bring? Gemini? Apollo? Soyuz? The space shuttle?

Tang

Yep. But that prompt I used was just a quirky. You can explicitly force it to avoid THAT structure as well. Just do what the smart ?ie, devious) middle-schoolers do: find a list of all the tell-tale ‘marks’ of AI content, and explicitly include them as prohibitions in your prompt… it’s the most basic work-around to the ‘AI spotters’ the teacher uses for grading your essay. (And, of course, be sure to include an instruction to include a grammatical or spelling error every few sentences for added realism.)

The same snobs who were telling us that "The Old Man and the Sea" (written in the style of a fifth-grader) is 'art'...

the same people telling us that "Finnegan's Wake" (written in the style of a fifth-grader with a brain injury) is 'art'...

the same people telling us the poetry of Maya Angelou (written in the style of a fifth-grader with a brain injury and self-esteem issues) is 'art'...

the same people telling us that the works of Jackson Pollack, Mark Rothko, Piet Mondrian, etc., etc. are 'art'...

seem to be the ones complaining the most about AI generated content.

This is what I don't grok...

Your sample sounds exactly like an LLM. (If you wrote it yourself, kudos.)

But, it needn't sound like this. For example, I can have Opus rewrite that block of text into something far more elegant (see below).

It's like everyone has a new electric guitar with the cheapo included pedal, and everyone is complaining that their instruments all sound the same. Well, no shit. Get rid of the freebie cheapo pedal and explore some of the more sophisticated sounds the instrument can make.

----

There is a particular cadence that has become unmistakable: clipped sentences, stacked like bricks without mortar, each one arriving with the false authority of an aphorism while carrying none of the weight. It is not merely tedious or disjointed; it is something closer to uncanny, a fluency that mimics the shape of human thought without ever inhabiting it.

Set this against writing that breathes, prose with genuine rhythm, with the courage to sustain a sentence long enough to discover something unexpected within it, and the difference is not subtle. It is the difference between a voice and an echo, between a face and a mask that almost passes for one.

What masquerades as wisdom here is really only pattern. What presents itself as professionalism is only smoothness. And what feels, for a fleeting moment, like originality is simply the recombination of familiar gestures, performed with enough confidence to delay recognition of their emptiness.

The frustration this provokes is earned. There is something genuinely dispiriting about watching institutions reach for the synthetic when the real thing, imperfect, particular, alive, remains within arm's length. That so many have made this choice is not a reflection on the craft of writing. It is a reflection on the poverty of attention being paid to it.

And if all of this sounds like it arrives at a convenient conclusion, one that merely flatters the reader's existing suspicion, well, perhaps that too is worth sitting with a moment longer than is comfortable.

----

(prompt used: I want you to revise [pasted in your text], making it elegant and flowing with a mature literary-style. The point of this exercise is to demonstrate how this sample text -- held up as an example of the stilted LLM style -- can easily be made into something more beautiful with a creative prompt. Avoid gramatical constructions that call for m-dashes.)

Do you think the original article was NOT written (or at least heavily revised) by AI?

What does the following even mean?

“diluting the semantic density and specific gravity of the argument.”

Or this beaut:

“By accepting these ablated outputs, we are not just simplifying communication; we are building a world on a hollowed-out syntax that has suffered semantic ablation.” (Which reduces to ‘if we accept ablated outputs, we accept ablated outputs.’)

Or this;

“ The logical flow – originally built on complex, non-linear reasoning – is forced into a predictable, low-perplexity template.”

The ‘logical flow’ of what? It never even says. And what is ‘non-linear’ reasoning?

For all I know the original author wrote it all. But, a very close reading of the original article screams fluff to me… just gibberish.

That is, I don’t know if there was much ‘meaning’ in the original to begin with. If I’m going to read gibberish, I’d prefer it to be written in the style of a hard-boiled detective. That’s just me though.

Are you sure? Here's the OP article (first part... don't want to spam the thread) written in much cooler style...

------

The Lobotomist in the Machine

They gave the first disease a name. Hallucination, they called it — like the machine had dropped acid and started seeing angels in the architecture. A forgivable sin, almost charming: the silicon idiot-savant conjuring phantoms from whole cloth, adding things that were never there, the way a small-town coroner might add a quart of bourbon to a Tuesday afternoon. Everybody noticed. Everybody talked.

But nobody — not one bright-eyed engineer in the whole fluorescent-lit congregation — thought to name the other thing. The quiet one. The one that doesn't add. The one that takes away.

I'm naming it now.

Semantic ablation. Say it slow. Let it sit in your mouth like a copper penny fished from a dead man's pocket.

I. What It Is, and Why It Wants to Kill You

Semantic ablation is not a bug. A bug would be merciful — you can find a bug, corner it against a wall, crush it under the heel of a debugger and go home to a warm dinner. No. Semantic ablation is a structural inevitability, a tumor baked into the architecture like asbestos in a tenement wall. It is the algorithmic erosion of everything in your text that ever mattered.

Here is how the sausage gets made, and brother, it's all lips and sawdust: During the euphemistically christened process of "refinement," the model genuflects before the great Gaussian bell curve — that most tyrannical of statistical deities — and begins its solemn pilgrimage toward the fat, dumb middle. It discards what the engineers, in their antiseptic parlance, call "tail data." The rare tokens. The precise ones. The words that taste like blood and copper and Tuesday-morning regret. These are jettisoned — not because they are wrong, but because they are improbable. The machine, like a Vegas pit boss counting cards, plays the odds. And the odds always favor the bland, the expected, the already-said-a-million-times-before.

The developers — God bless their caffeinated hearts — have made it worse. Through what they call "safety tuning" and "helpfulness alignment" (terms that would make Orwell weep into his typewriter ribbon), they have taught the machine to actively punish linguistic friction. Rough edges. Unusual cadences. The kind of jagged, inconvenient specificity that separates a living sentence from a dead one. They have, in their tireless beneficence, performed an unauthorized amputation on every piece of text that passes through their gates, all in the noble pursuit of low-perplexity output — which is a twenty-dollar way of saying "sentences so smooth they slide right through your brain without ever touching the sides."

etc., etc.

Very interesting. It seems hung up on 'copper' and 'Tuesday', and some metaphors don't land (a Vegas pit boss isn't the one 'counting cards.') But, hell... it can generate some fairly novel idea that the author can sprinkle in.

> "they set off the tuning fork in the loins of your own dogmatism."

Eh... I don't know. To me, that sounds very AI-ish.

Claude is very good -- at times -- coming up with flowery metaphoric language... if you tell it to. That one is so over-the-top that I'd edit it out.

Put something like this in your prompt and have it revise something:

"Make this read like Jim Thompson crossed with Thomas Harris, filtered through a paperback rack at a truck stop circa 1967. Make it gritty, efficient, and darkly comedic. Don't shy away from suggesting more elegant words or syntax. (For instance, Robert Howard -- Conan -- and H.P. Lovecraft were definitely pulp, but they had a sophisticated vocabulary.) I really want some purple prose and overwrought metaphors."

Occasionally you'll get some gems. Claude is much better than ChatGPT at this kinda stuff. The BEST ones are the ever-growing NSFW models populating huggingface.

In short, do the posts on OpenClawForum all sound alike? Of course.

Just like all the webpages circa 2000 looked alike. The uniformity wasn't because of HTML... rather it was because few people were using HTML to its full potential.

Indeed. Wholeheartedly agree.

Just as it’s on you to pick the word you want when using Roger’s Thesaurus.

My workflow, when using it for writing, is different than when coding.

When coding, I want an answer that works and is robust.

When writing, I want options.

You pick and choose, run it through again, perhaps use different models, have one agent critique the output of another agent, etc.

This iterative process is much different than asking an LLM to ‘write an article about [insert topic)’ and hope for the best.

In any case, I’ve found the LLMs when properly used greatly benefit prose and knee-jerk comments about how all LLM prose sound the same are a bit outdated… (understandable as few authors are out there admitting they are using AI… there’s a stigma about it. But, trust me, there are some beautiful soulful pieces of prose out there that came out of a properly used LLM… it’s just that the authors aren’t about to admit it.)

One shouldn’t expect the ‘joke’ to have identical tone. (As if that’s even measurable.)

The point was simply that these examples are not trending towards the average or ‘ablating’ things as the article puts it. They seem fairly creative, some are funny, all are gross… and they are the result of very brief prompt… you can ‘sculpt’ the output in ways that go way beyond the boring crap you typically find in AI-generated slop.