The feed of a Hackaday writer brings many surprises, some rubbish, and the occasional diamond which of course we bring to you. Sometimes a page catches the eye that’s not a Hackaday piece per se, but makes for interesting reading. Today there was an example which provides a catalyst for a story if not being one itself.
It’s an AI skill intended to make a slop-writing operation look less sloppy, and since it contains a list of slop behaviours, of course it’s irresistible to turn it on yourself. To delve into the archives, if you will. So here goes, how slop-like is your scribe?
It kicks off with the old favourite, “Delve“. Oh dear, that one’s in the List Hackaday corpus quite a bit. Earliest appearance: 2016. Then landscape terms, in which it’s easy to recognise “Sphere” again. Oops. 2016 again. How about “Testament“? Well, that one’s a 2017 word. “Pivotal“? Four years of the Raspberry Pi, back in 2016. How about “Showcase“? Yup, it’s there, in 2019. Meanwhile “Foster“, goes back to 2017.
Next up are “Leverage” and “Utilise“. Safe ground here. Editors have what are referred to a “Pet Peeves”, language tics that annoy them on sight, and those two regrettable pieces of business-speak never evade this red pencil. The word you are looking for is “Use”.
The page goes on to name a load more words as tells, and we could continue to dissect them for your amusement, but we think you get the message from long-pre-LLM Hackaday. It gets more interesting when it talks about paragraph structure, and there are the things we’d recognise as tells. Staccato paragraphs, one-line stacks for impact, and repeated sentence start words. Whether AI slop is sounding like a business influencer’s Linkedin posts, or the other way round, is anybody’s guess. And yes, Hackaday does use em-dashes — get over it.
It’s amusing to poke mild fun at your past writing style, but if there’s a serious point in there it’s that spotting slop is never as easy as having a check-box list of tells when in fact you’re just looking at professional-grade command of the language. After all, given that Hackaday is in its corpus, the LLM can even make a passable effort at writing like one of us. Perhaps the key lies in that article, when the fake Hackaday piece became obvious because it lacked the inference of the real one.
To be fair AI generated text is easier to read.
To be fair, blended nutrient slurry is easier to consume
Soylent Green is … AI?
Lol, amazing
That says more about you than it does about LLM text. Personally I find it too painfully annoying to bother parsing, like when someone uses kindergarten teacher voice.
That´s my opinion also. It is irritating and tiresome to read. Boring to the point of skipping large parts of it, or stop reading entirely. It always sound like those people that speak a lot, fast, to get you confused and distracted because they want to sell you something.
Agreed; something in my brain turns off when I start reading AI-generated text — no matter whether it’s in “brutally dense/terse GH commit” mode or “tediously verbose chatbot” mode or “Don Draper copywriting for the 21st century” mode. These are all different genres, but there’s something that breaks my ability to parse it. It may be a prejudice borne of aesthetic revulsion, but I feel like it goes deeper — like it’s good at form but forgets about content. Or something.
Interesting, none of the words identified are ones I even recognise as AI tells. Some of them are good solid usable words (delve) and some of them sound like slop that predates the AI era (utilise and leverage). No, the ones that tip me off are phrases like “doing real work” and “load-bearing”, and longer phrases like “And that’s the part that matters”.
Really ? I Found AI text has too low Singal-to-Noise ratio. They say a lot of words for just simple thing.
Like my 9 year old.
this is ai slop. its not keywords. its the way in which it is written
its a pattern of communication, not a pattern of vocabulary.
It’s the rhythm, the transitions, the unnecessary elaboration, the exaggerated enthusiasm, and the way everything gets packaged into a neat little narrative.
Unlike seemingly most humans, yourself included, AI at least gets grammar and punctuation correct.
Grammar Nazis have never been happier! They’d crowdfund Skynet if they could!
In the AI age I’ve come to cherish bad grammar and punctuation — even to the point of leaving it in if I notice it in my own writing. It’s like an “I’m human” dog-whistle… at least until people start getting AIs to emulate mistakes on purpose.
Yes, and like any tool in skilled hands that can be smoothed over. But it depends upon one recognizing such things.
yeah that’s what i came here to say. people focus on individual words or like the em dash…and i never understand what they’re talking about. what strikes me about slop is the overall structure, the way it uses repetition.
the frustrating part is that the structure is not dissimilar from good structure. you should sum up your essay with a repetition of the thesis, etc. and on top of that, for humans much like for LLMs, using language is the practice of producing output that emulates your input. so i can find traces of this LLM structure in my own prose and it is a very uncomfortable feeling.
What you recognize as AI slop, I recognize as corporatized slop or general garbage genericism.
I’m certain that there were people on YouTube following scripts like this before LLMs and before that it was all over commercials and other business trash.
Hell, part of why they talk like that might be due to garbage in-garbage out more than a true fault with AI.
There’s the word used to describe plays, movies and TV shows “formulaic”. It existed long before computers were used to generate much of anything. Although, frankly, in the era when artists fed themselves by painting their patrons, or bowls of fruit, or landscapes, there was a lot of it in art, too. And (full circle) now there are ads with people pretending to be artists hawking “rain drums” that are actually all pumped out of factories in China, but if they sell you the formulaic story of a blacksmith in Sedona who is selling the last of her stock because she can’t hand tune them anymore, they can make a LOT more money.
Over twelve?! A whole baker’s dozen! 😱
The slop itself is not new.
If you remember, back in 2010-2011 there was a famous website called Bullshit Generator, which could create corporate-slang-type text on a click of a button. After we found it, we had regular jokes inside our team that the leadership presentations were basically 1:1 with the text generated from this website. All the typical bullshit-representative words were there: leverage, excellence, empower, stakeholders, expectations… Damn, it hurts my brain. I’m happily far away these days from corporate Murica.
Scott Adams, may his crazy angry racist ass rest: “Proactively leverage your synergies.”
Top!
I came here to make much the same point, so I’m piggy-backing on your comment.
When I looked at the AI Slop Detector list, I found a huge overlap with what I’ve come to refer to as HR/PR speak. I scorned that style of communication long before the advent of LLMs.
Were the LLMs heavily trained on corporate press releases, or is it simply that “lack of humanity” has a style all its own? I suppose both could be true…
The common factor is probably “generate text that sounds good to investors and board members”…
Ah, youth.
This was the “automation” of what was originally a dead-tree list that was often shared by fax dating back before the 1980s (the first time I ran into it). These always had three columns and you’d simply choose any word from each column in sequence.
Ah, yes, now I remember seeing similar 3-column lists as running jokes but on the early web! Damn, we’ve come a long way :D
Next: see if you can get an AI to generate an article published on HaD without anyone in the comments section noticing! (Or was that it…)
I’m in the process of writing a novel, while at the same time learning how to write.
A lot of the errors listed in that site can be read and digested into patterns that will cripple your ability to write interesting and readable text. Don’t do that, think about what the rule says, and try not to overextend it.
Taking examples at random: a list dump can be useful if you want to paint the image of going on and on about a subject. Used judiciously, it can show a personality trait.
Ending on a mike drop? That’s a useful technique at the end of a chapter, it encourages the reader to keep reading.
Run on sentences nicely illustrate fear or anxiety, short 5-word sentences between paragraphs make a sort of back-and-forth rhythm that’s pleasant to read, and bolding can be used to emphasize a particular word.
I didn’t know what a comma splice was when I started, for a while I removed them when found, now I’m of the opinion that they provide a useful timing slur. (BTW: That last sentence was a triple comma splice.) “I came, I saw, I conquered” just doesn’t have the same immediacy when punctuated properly: “I came. I saw. I conquered.”
The patterns shown in the web page are fine examples of bad writing generally, but don’t scrupulously dump them because you saw them listed on a web page.
Think about them first, and consider whether maybe you could use them for effect.
The novice is learning the rules.
The journeyman knows how to follow the rules.
The master knows when to break the rules.
Who beats cliches to death?
Build AI to check for AI slop.
It’s a very sloppy slope.
Kinda makes you not want to trust the slop detectors.
As for the humans who claim they know when they’re reading AI slop: https://www.pnas.org/doi/10.1073/pnas.2208839120#sec-1
Thank you! This:
‘…The word you are looking for is “Use”.’
…sooooo needed to be said! “Leverage” and “like” are my top two, and I mention “like” only because it, no longer just an American speech tic, is now showing up in written transcriptions…which means it will be stochastically parroted any minute now.
The one that grinds my gears is “begs the question.”
In fact, the very next posted HAD blog article has this in it: “But that, of course, begs a question…”
It leads to the question! Begging the question is completely different and doesn’t mean what the author thinks it means.
But I suppose language is fluid and modern usage can change the meaning over time. Even so, it still grates to read it.
My other bugaboo is “one of the only” instead of “one of the few”. Just bothers the heck outta me.
I’ve been going to a lot of art museums of late, and… wow! take a look at some of the artspeak descriptions on the wall next to a famous painting. Incomprehensible fluff with not a shred of meaning.
It makes me feel like I’m from a different planet, and can’t speak the language…
One word that drives me crazy is when people use “decimate” when what they really mean is “devastate”. But I think nowadays “decimate” has been redefined from its original meaning because its misuse is so prevalent.
Language is vibes to most.
This is very true. When you learn only by listening and context, meaning is a fuzzy thing that can shift rather quickly. Combine that with a mind that is not analytical, and it’s vibes all the way down.
I passionately hate the “like” word, for the same reasons, including “ammmmm” (I call these “filler words, when you want to continue making sounds, but have nothing meaningful to say).
Strangely enough this is a Pet Peeve of mine too. There is always a better word for the application.
“Actually” and “basically”. Redundant words. Also see “Literally”.
To be pedantic I want to point out that each of these words has a very specific function in the language. It’s not their fault they are mostly misused to the point of being annoying. Take utilize: it’s not a fancy way to say “use” despite what LLMs and speakers of Corporate English might think.
The word “utilize” has a specific meaning: to make use of something for a purpose other than that for which it was intended. You can’t use a screwdriver to open a can of paint; you can only use it on screws. To open a can of paint you utilize your screwdriver as a prybar. I do not make use of this distinction writing for Hackaday because I respect that that the word is overused and disliked… but it does vex me, sometimes, avoiding the proper word out of convention.
The thesaurus was a crime against the English language because it made us think that synonyms mean the same thing. They do not. If I am angry, I am not yet furious. If I am despondent, that’s something other than merely being sad. No soldier in the US Civil War writing home to his wife would have mistaken the two, but then, he likely had no thesaurus in his ruck sack. As Mark Twain may have said, “’tis the difference between the lightening and the lightening bug.”
There are shades and layers of meanings that each of these words once had that have slowly been lost to us as our language is impoverished by generations of increasingly poor education.
I’m pretty sure he said “tis the difference between the lightning and the lightning bug.”
Lightening is not a synonym of
lightning. 😇
Off topic, but I’d argue that you’re supposed to open a paint can with a screwdriver, so you do use it in that particular case. Now, I once did it with a chisel instead, that I’d say warranted the word “utilize”. (don’t get me started on the stupid differences between British and American spelling)
They’re called hesitation devices in linguistics. They tend to provide moments to organize and collect thoughts while not interrupting the flow of speech. I agree it’s rather annoying, but they do serve a purpose. It’s really not just about wanting to hear the sound of your own voice.
My issue is that I don’t really want to uhmmmm a lot, but I often want to express myself with care, which means I need to stop talking and think about the right word for a second. But then my conversation partner will interpret the silence as a signal that I was either done talking or need help thinking of the next word, and they’ll interject. If people could be relied upon to let each other finish talking, I think there would be fewer filler words.
I love discussing lingual pet peeves! Another one of mine, and I hear this on YouTube a lot, is when people say “less” instead of “fewer”. It sometimes feels like the latter has been banned or something.
(In case anyone needs to and wants to learn: you use “fewer” when the things you’re describing are countable, and “less” when they are not. If you have less current flowing through a wire, there are fewer electrons moving.)
I tend to think of the lucid traits of my speech patterns as a word salad. Slop just sounds so feeding trough.
I find self obsessed personalities talk like AI. They tend to revalidate their own thoughts frequently and this tends to sound AI-esque.
I have AI do creative writing tasks when I am bored and I do agree, there is a cadence and structural quality that is ineffable. Keywords might be a giveaway, but sentence structure absolutely is.
Also, I see frequent grammatic and spelling issues in AI outputs, they follow the rules as a probability which is not how natural language is actually structured.
I’ll utilize that methodology. (Shoots self in foot.)
“Girl, when I get done utilizing you, you gonna pay ME money!”
“get over it”? So to those readers concerned with the integrity of this site, I say “go suck an egg”… Maybe the excessive editorializing is the problem…
“Game Changer” is an immediate give-away.
In general, LLMs write to sound like marketing which is a form of slop that existed long before they did… we weren’t really aware of how much of that crap there was because, of course, no one actually reads marketing stuff because it’s just junk (aka marketing). But it turns out if you train models on the internet, the internet is full of marketing nonsense.
I tend to discard stuff as soon as I hit a spammy sounding sentence and I don’t know if I’m bailing out on LLM crap or marketing and promotion crap – it’s all the same and none of it feels like it was written by a real person with real information and ideas to convey.
I understood this sentence, but I suspect that’s down to my professional-grade command of the language, not yours.