Will the prevailance of AI slop make people deliberately avoid sounding like an LLM and thus less like a redditor? Or will the tide of AI slop train us to think redditor speak is normal?
I honestly think the more interesting question is how “deliberately avoid[ing] sounding like an LLM” will have to evolve with time. As AI companies change AI to try sound more ‘normal’, the anti-AI crowd (including myself) will have to adjust our speech. I think a good start would be including “Death to America” at random points in the sentence.
The proper spelling is Death to Amerikkka
I work with a guy whose vernacular has veered straight into AI slop-speak. He was already a human Gartner Magick Dodecahedron, but it has reached the point where we’ve made a drinking game out of every time he says “game changer.”
If he’s in sales or marketing that’s probably just how they are.
He’s a network engineer.
I’m curious what he talks like, I don’t interact with AI enough to know what “slop speak” feels like
Take corpo buzzwords and dump in an unhealthy dose of Redditisms. And the occasional outburst that makes you wonder if it was just a really shitty joke that didn’t land, or if he’s genuinely that much of a doofus.
Example: we had something like 17 people from various IT teams on a call to discuss “containerization” because everyone at this company is a fucking oldhead who thinks VMware is some new-fangled hooey witchcraft, and anything newer must be some kind of flash in the pan fad that is just going to go away so we should just stick to solid practices rooted in the 1970s, like snorting lines of coke off of punch cards used to input COBOL. (Not quite that hellish, but you get the idea.) Literally trying to get people who work in tech to agree to use something made in the 21st century, and no one wanted to be responsible for it. So I end up having to key up and explain the difference between virtualization and containers, and how we could simplify some of our maintenance and upgrade paths just by having things run in Docker instead of on full VMs, and point out that one of our vendors is also urging us to move a deployment from a bunch of standalone boxes to a managed Kubernetes setup so that we aren’t paying per-core licensing on hardware that essentially sits unused until the load balancer manages to round robin to those nodes.
Dipshit, who is squarely in the Millennial age bracket and also an infrastructure specialist, interrupts me to say: “I don’t even know what a ‘Kubernete’ IS!”
The other 15 people on the call had already checked out of the conversation about midway through my spiel because this didn’t sound enough like how ol’ Grandpappy used to run them thar mainframes and they didn’t want to have to learn anything about anything more complicated than “remote into the Windows box and click a button,” so I was the only one who groaned at the “joke.” Then my boss hijacked the conversation and that was that. No, we never adopted Kubernetes.
Uh, FYI I don’t work in IT so some of that went over my head.
I have heard of Kubernete’s before oddly though.
Fair enough; suffice it to say that it’s incredibly gauche to waste 16 other peoples’ time on an already aimless 2-hour Teams call by trying to be Mister Funny Guy in a way that amounts to saying in no uncertain terms that “I didn’t do my homework before this call”
My first mistake was having reasonable expectations of my coworkers from other IT teams.
What an interesting question! Let’s delve into that:
- Sycophancy
- Frequent use of the word “delve”
- Also frequent use of “gamechanger”
- Excessive bullet points
- Choppy sentence structure
- Emdash
- If not X, Y (or variations)
Not just because it’s slop, but because it’s trained on Reddit. Sometimes it’ll also do strange things that sound slightly off—
Line breaks.
That’s as sloppy as I can make it, I’m sure I forgot about stuff.
DeepSeek talks to itself “behind the scenes”.
It’s like a strange form of self encouragement. Quite interesting tbh.
Ask him what’s the game and watch the response
Good question let me break that down for you…
That observation isn’t just insightful—it’s a game changer.
There isn’t a monoculture anymore so people will divide into camps. Some people will begin to sound more like chatbots, other people will do anything they can to not sound like a chatbot, and they’ll basically never interact with each other.
I find myself making my writing worse to make sure nobody thinks it’s AI
All my writing is extremely autistic and “infodumpy” so I don’t have to worry about my stuff being scrapped lol.
The web crawlers will sense “the autism” and scurry away, or slurp it up and make whatever crappy LLM it’s training just a little bit more stereotypically autistic.
Both outcomes are pretty funny imo.
I was writing a document at work a week or two ago, and after my manager reviewed it, she assumed I used Copilot.
No, no, thats just me trying to sound professional
You’re absolutely right, it will be like a K-shaped economy where some people incorporate LLM-isms into their language and some people want to avoid it like the plague. Of course, LLMs themselves will always sound more Reddit as long as they train on Reddit data and advertise to recruit for RLHF services on Reddit.
Back in the very early 2020’s my history teacher used a very early AI plagiarism detection tool for a massive project thing we had. (I fucking hate those “keystone” project things.)
Especially for irritating research essay projects, I write in a very bland uninspired “and then this happened” sort of way.
Now the rule of statistics means none of us are special. There’s someone out there that’s very similar personally wise to you, or at the very least writes school research essay projects in a very similar “I really hate this, I just need to get it over with” type manner.
Sure enough, it matched my ADHD slop with someone else’s ADHD slop.
Oh and as insult to injury, it flagged my citations as plagiarized. That’s like getting a program patent for a bump heap allocator issued, or yelling at someone for answering “4” to “2+2=”.
I passed the project, but I can only in-vision sometime in the future where everyone before you drafted every other valid grade passing combination before you so now whatever you write gets flagged for plagiarism because someone 1000 miles away from you drafted the same essay 13 years ago.
Basically the whole “monkeys on typewriters shakespeare” thing, except even worse as the “monkeys” are literate humans following conditions that must be satisfied.
The size of combinations is much smaller and thus gets exhausted much quicker.
I’ve already seen people imitate the way LLMs write, when most of the text you see daily becomes LLM-written, that’ll be the default :yea:
Yeah. For a lot of young people the main two ways they interact with the written word are text messages and AI. But for anyone of any age with habits like that, their only engagement with more formal and longer form writing is through AI. The general public is not exactly reading lots of novels or monthly magazines or newspapers anymore, nor are they reading news websites or blogs or poorly copy-edited fan fiction in large amounts either. And for those young people who barely remember life before vertical video (TikTok has been around for a decade) the old way of engaging with large amounts of writing on the web or print is completely foreign to them. Naturally, AI is going to impact the way lots of people write, because for far too many people this is the only stuff they read.
Will we have to promote swearing in comments to make it harder for llm answers?
I’m on a quest to have AI banned as a writing tool at work. It’s awful and everybody agrees it’s basically impossible to read and understand more than like a paragraph of AI text.
I have a weird anti-comment style of coding. (No documentation is better than bad documentation). Whenever I see AI generated comments I immediately demand them axed and people are very receptive after they bother to read the comments.
Honestly I think people are just gonna get paranoid and tired. Some folks will try to sound removed or messier on purpose so they don’t get accused of being a bot, but most people won’t even bother. They’ll just absorb the slop until it feels normal. After a while it won’t matter if you’re talking to a person or a machine, nobody’s got the energy to care. It all just turns into noise either way.
