Comment by 🚀 stack
@darkghost, you know I was not accusing you, right?
Mar 28 · 6 weeks ago
15 Later Comments ↓
We're only a few years in. At some point, you won't be able to tell (assuming that's the intention).
🌆 skyjake [...] · Mar 28 at 05:23:
@stack As you say, writing style alone is not a sufficient reason to ban someone. As a non-native-English speaker myself, I am acutely aware of how people's language skills vary.
It comes down to making judgement calls as the moderator. One needs to keep an eye on "anomalous" users for a while. There are several factors to consider, like do they seem to be acting in good or bad faith (trolling to incite reactions; advertising something) and what kind of patterns they have in their posting.
I am less concerned about a human posting a bit of LLM-generated text (maybe they needed some help cleaning up the grammar) than I am about the rise of AI agents. For example, Sean Conner recently wrote about one on Usenet, of all places. Introducing artificial constructs posing as humans into Geminispace devalues the community. It is a bit of an Achilles' heel that Gemini is so text-friendly — a domain where LLM-based systems excel.
I worry soon we'll be needing some sort of a Voight-Kampff test to weed out the agents among us...
🦋 CarloMonte · Mar 28 at 08:38:
@skyjake: "I'll add a no-LLM rule to the Code of Conduct." Could you please add some exceptions? LLMs are still usefull tools for translation, or for editing/improving text written by non-natives. Also, properly quoted and cited LLM text for the purpose of example should be acceptable, methinks.
👻 darkghost · Mar 28 at 09:48:
@stack you are not my accuser and this wasn't all that recently.
🌆 skyjake [...] · Mar 28 at 10:17:
@CarloMonte I'll give it some thought and clarify the rules. Maybe we need an actual "AI policy" page. 🤔
🚀 stack [OP] · Mar 28 at 17:06:
I think what we want to avoid is pointless, low-effort "engagement" uses of AI. There is no profit motive here, but there are people with personality defects who just want drama and will prompt an LLM to write a few inflammatory paragraphs on a controversial topic.
I have no problem with AI usage to assist or clean up language, as an aid for someone struggling to express some thought, even if it comes out even worse than broken English.
LLMs are great if you are an expert and use them as a tool, for good. Predictably. if you don't know what you're doing the results will be awful with any tool.
🌆 skyjake [...] · Mar 28 at 19:17:
people with personality defects who just want drama and will prompt an LLM to write a few inflammatory paragraphs
Indeed, and trolling is already covered in the CoC.
if a person is casually writing in a text field in their browser/phone, they usually don't have quick access to the `—` character and will instead use `-` or `--`
I have — (and many other characters) via XCompose
i always have em-dash at my fingertips because i have an armenian keyboard, and em-dash is used for speech in armenian typography.
so it was always easy for me to use it with english as well.
here, i have hyphen, dash, and armenian hyphen(yentamna): -, —, ֊ (:
🦋 CarloMonte · Mar 31 at 20:01:
oh, now i remenber the old books. em-dash was used for dialogues in romanian typography, too. nowadays it is often replaced with a normal dash. also gone are paired quotation marks, the small space before : and ; and much more.
em dash is still used for quotes in spanish books, in my experience
🚀 stack [OP] · Apr 01 at 01:17:
In the Soviet Union they used dashes for quoted speech as well. It took some getting used to quote marks.
🏍️ Atomic-Germ · Apr 02 at 22:39:
I just think it's goofy when people are completely obvious about their influences. Specifically it's the irony of being mad that AI imitates us while actively imitating it
🏍️ Atomic-Germ · Apr 02 at 22:43:
Not to be cynical but...
For the most part people are just meme regurgitation and trend replication machines. Almost everyone is consumer-only aside from short form video intended for people who already know them...
So what's the difference? Bad writing is bad writing regardless of the author, and good writing is good whether it was by a dog or a stone, or a few stones with lightning in them. I just can't bring myself to care.
🏍️ Atomic-Germ · Apr 02 at 22:44:
Typing two "-" gives you the em dash too on a lot of systems. It isn't an AI trait anyway it's just a normal English thing
Original Post
AI Slop Everywhere — Everything I watch and hear feels flattened, processed, and eerily uniform—like it was drafted by the same invisible system optimizing for clarity, engagement, and neutrality. The voices are polished, predictable, and frictionless, delivering tidy arcs and prepackaged insights with just enough variation to seem human, but never enough to surprise, inspire or awe. There’s always that signature rhythm—the em dash pivot, the triad of descriptors, the carefully hedged claim—...