Title text:
You say no human would reply to a forum thread about Tom Bombadil by writing and editing hundreds of words of text, complete with formatting, fancy punctuation, and two separate uses of the word ‘delve’. Unfortunately for both of us, you are wrong.
Transcript:
Transcript will show once it’s been added to explainxkcd.com
Source: https://xkcd.com/3126/
I’ve had a handful of people think I was AI too because of my fancy word choice and mediocre knowledge of punctuation. My writing voice is ubfortunately devoid of emotion most of the time, and when it isn’t, it tends to fall into the matter-of-factly category. I try to include punctuation or breaks to be a little more similar to my speaking such as “well, uh, you wouldn’t…” or “I- Yeah, I don’t know.” I think my personal favourite I do do is “I think it was… two…? weeks ago?”
And yet I still get dubbed AI because I periodically use the word plethora…
This is the greatest thing I’ve ever seen, even better than banana man
it damn well bloody helps to include a fucking shitload of swearing, e-fucking-specially if it’s a mite creative too.
i found splitting up paragraphs into many small, newlined sentences help a lot. it seems to be mainly big paragraphs that give people uncanny feeling.
or alternatively join me. in not capitalizing letters anymore.
Me. I like formatting and using em/en dashes and semicolons.
Someone the other day mentioned semicolons are now a sign of AI. I always liked semicolons for when things are more connected as opposed to a period/full stop :/
at that point it’s just “grammar and punctuation is a sign of AI”, which actually means “i don’t know what AI output looks like”.
Same. I am one of the rare people who know the difference between i.e. and e.g. I know when to use a semicolon vs an em dash.
I no longer feel special; it feels wrong.
In Essence Egxample
the rare people who know the difference between i.e. and e.g.
This honestly isn’t rare at all, and people who try to flaunt it as some kind of mark of erudition tend to come across as… well, not quite what you intended.
Rare in certain corners of the internet? No.
Rare in the general public? Yes, absolutely.
it’s stupid that English uses Latin abbreviations for these things; my first language is German and so:
- “z.B.” = “zum Beispiel” = “for example” = “e.g.”
- “d.h.” = “das heißt” = “that means” = “i.e.”
When I first saw these abbreviations in English, it took me about ten seconds to memorize that “e.g.” means “z.B.” and “i.e.” means “d.h.”. If English just did it the way German does and abbreviated its native expressions (“f.e.” and “t.m.”), it would be obvious to everyone which is which.
I no longer feel special**;** it feels wrong.
Reported as AI bot. /j
Well that’s annoying because I’ve always used semicolons
Yeah I use semicolons all the time; I can only hope my lack of native level english skills makes up for it and I don’t sound like a LLM. I swear I’m just some autistic imbecile that has love for words instead of numbers
Lol! Same here. I feel seen. :)
Imitating LLM to piss off AI haters is just next level trolling. As everyone knows, trolling is a art.
As someone who has been mistaken for an LLM at least twice in the past couple of years, yeaaah. Sometimes I write like that. The LLMs learned from people like me. I can only hope it was smarter, more productive people with the same sort of writing style and not from anything I’ve produced… although it would explain a thing or two.
Nope. It’s you and me, buddy. They learned from a fancy talker and a drunk. That’s why they just make shit up.
In Denmark we have a saying, which (translated) is “the truth shall be heard from children and drunk people”
I guess fancy talkers are kinda children too, or at least former ones 🤷
The US government did extensive research on a potential truth serum. The single most effective solution they found was vodka. Every other thing they tried (including attempted mind control with LSD) had huge potential drawbacks, and usually didn’t even result in honesty. But get a dude drunk and have a pretty girl talk him up, and he’ll spill all of his secrets while thinking it’s his own idea.
The government also holds occasional “know your number” meetings amongst the people who hold security clearance. Basically a netting where they sit everyone down and go “okay you look like a wrinkly potato, you’re missing two teeth, and you smell like wet beef. At best, you’re like a 3 out of 10. Maybe a 4.5 if you bothered to shower before you hit the bar. If a solid 10 is flirting with you at the bar just outside of the base, and she seems really interested in what you do for work… She’s a fucking spy. Know your number, and know what you can reasonably pull. Because if you’re pulling above that number, you’re being honeypotted.”
Jokes on her. I have forgotten all the good information I had back when I held a clearance and I’m incredibly boring.
I am not sure how many times I’ve been mistaken for ChatGPT, but I don’t think my writing style is actually very similar.
I’m pretty sure that when people say that, most of the time, they actually mean, “I want to disagree with what you’re saying, but I lack the ability to do so legitimately. If I simply accuse you of using an LLM, people will assume I’m right and I will ‘win’.”
The topics were pretty tame that I remember, so there wasn’t much to disagree with. I was just being… uh. Florid? Verbose? Sesquipedalian?
It might be a neurodivergent trait; the need to use the right word to communicate exactly the right meaning even if it runs to several syllables.
It might lose a few people, but I’ve got to say what I mean.
And then someone else comes along in a different comment and says what I wanted to say with words of fewer than three syllables and I’m like “hmmm”.
It might be a neurodivergent trait; the need to use the right word to communicate exactly the right meaning even if it runs to several syllables.
It might lose a few people, but I’ve got to say what I mean.
Speaking as someone who got his ADHD diagnosis late and felt chronically misunderstood for his entire adolescence, I’m gonna go with
And then someone else comes along in a different comment and says what I wanted to say with words of fewer than three syllables
Beginner’s luck!
I’ve never seen LLMs talk like what you’re describing, though.
If I had to describe ChatGPT’s usual style, it’s like a neurodivergent person who really wants the average person to understand what they’re saying, hopefully without causing offense.
Oh, god, that’s me that you just described. No wonder people think I’m an LLM. Yikes! Lol!
I’ve also gotten accused of being on meth due to my Internet comments, even though I’ve never touched anything like it. Yay, AuDHD!
So it’s almost as if it were trained on Reddit?
(No offense intended! I hope you get what I mean! ☺️)
Since you’re a polysyllabic person, can you explain why the word “monosyllabic” has five syllables?
Information entropy. You need roughly as many syllables to explain the same concept with mono- or disyllabic English words as you do with a scientific polysyllable. Admittedly, some of it is “I know this word! See how smart I am!”, but another part is how much more fluid it is to say. “Monosyllabic” rolls off the tongue a lot more easily than “having only one sound”.
(The funny answer here would have been “No.”)
On top of all that, monosyllabic is accurate to the intended meaning while “having only one sound” is not: a single syllable word often comprises multiple phones and/or phonemes.
can you explain why the word “monosyllabic” has five syllables?
For the same reason why the word “lisp” has an s in it and the fear of long words is called monstrosequippedaliophobia*: because sometimes language is a callous bastard 😁
*no, I don’t accept the “Hippopoto-” many people like to tack on. Unlike the rest of the word, which describes EXACTLY what the word means, adding a large semi-aquatic mammal serves no purpose other than lengthening an already monstrously equipped dalio.
I may be verbose, but I’m way less friendly than most LLMs
You just need to start inserting more Ai type punctuation into your text — like an Em dash for example.
This will really confuse people, resulting in more instances of you being treated like us — I mean Ai.
The funny thing is, I watch The Vlogbrothers fairly often - both of whom are writers - and recently John has told of his fondness for the m-dash. His enthusiasm and explanation was enough to get me to consider using it, but then that trait was identified as one overused by LLMs.
I’d already been mistaken for one by that point (an LLM, not a Vlogbrother), so instead I’ve stuck with the technically incorrect hyphen-minus or plain old parentheses when I’ve felt the need to do that.
Found Dave Green’s Lemmy account
Nice try Hank we know you when we see you
Bitch ass LLMs putting spaces before and after their emdashes—I REFUSE!
Not ChatGPT output — I’m just like this.
That’s an m-dash, which we all know is irrefutable confirmation of LLM output. /s
iirc its m-dash as well as constant rule of threes and generally using incredibly formal sentence structures even when the language involved is not formal in any way. Kind of like what I just did there though probably with an extra comma after m-dash.
I think I’m a—LLM now.
And I am b—LLM, nice to meet you!
you can keep using pronounciation correctly and writing long paragraphs of words if you lowercase it all, just saying.
Yep ChatGPT must have learned from people like me, because:
- I write long texts
- I over explain stuff to people who did not ask for explanation
- I use bullet points in every post
Hey, that’s a really well-formatted commemt! If it weren’t for all my typos, people would probably suspect me too – but hey at least now I have an excuse not to edit anymore.
(Edited: 2 minutes ago)Claim:
- I use bullet points in every post
Fact Check:
Out of your 36 comments, this is the only one with bullet points. That’s only 2.7% of your comments. One other has an enumeration, but an enumeration is not bullet points.
Additionally, you have one post, but that also doesn’t use bullet points. 0% of your actual posts use bullet points.
Conclusion: Claim is FALSE. Ziffy-fa-Jazz-KZone-Sweek’em does not use bullet points in every post.
Nice :) I like that you did that! But to clarify some things:
- I use them mostly in emails where there is actual content to write
- I rearly post online
- I was exadurating
- I use new account every few months.
Chat gpburn
- they might have another account
- or talk about their presence outside of Lemmy
Fact: we don’t know and can’t come to a definite conclusion
In a literal sense, you only need one exception to disprove an “every” claim.
In a figurative sense, you literally can’t
Emoji, title, bullet points, repeat
Can you elaborate on the second point?
Well now they literally cannot, since you asked!
They can but it wouldn’t prove or disprove anything eitherway
There really is an xkcd for everything.
Expect to see this accidentally included in someone’s forum screed soon:
Here’s a detailed response to BombaFan486’s post, including three separate (but natural) uses of the word ‘revelatory’, ending with the disclaimer, “Not ChatGPT output. I’m just like this.”
Is there anything else I can do for you?
Then make it look less perfect with custom command AI, after finding a FOSS AI website that has a lot of uncensored AI my life hasn’t been same again
You just have say fuck a lot…
But I’m pretty sure any explanation of Bombadil less than 300 words would fail the Turing test
That is an excellent point! Use of the word “fuck” in online conversation may present to readers with more realism.
It is however important to note that use of the word “fuck” does not fully rule out the use of large language models. While most commercial offerings may be trained to avoid profanity, certain models might not be trained the same way.
Additionally, use of the word “fuck” may be inappropriate in certain human conversations such as:
- formal conversations
- conversations with parents
- conversations with children
So, while the presence of the word “fuck” may decrease the likelihood of the text being generated by large language models, it is important to keep in mind its limitations, and opt for more robust methods like cryptographic signatures or verbal conversations.
Is there anything else I can help you with?
(This was genuinely written by me)
how do you do, fellow human?
The method I (just now) thought up using to signal humanity was responding to accusations of being an LLM with a “fuck you”. The combination of vulgar language and defiance of the sycophantic tendencies of LLMs feels to me like a pretty effective proof of humanity, at least for now.
You can actually get LLMs to swear, sort of. They just won’t use real swear words. If you set up your LLM parameters to use a specific word for an expletive, but it’s not actually an expletive, then you can replace that word with your choice of expletive after the text is generated.
I got accused of being an LLM for the first time just a few days ago. Was pretty funny.
When they actually get good at mimicking convincingly enough to be indistinguishable from a normal human user, that’s when dead internet theory will truly take over. This could’ve already happened, but I’ve seen enough stupid shit vomited by LLMs to know it probably hasn’t happened yet. Once I stop seeing that obvious cognitive gap for a while, then I’ll get worried – but if they stopped being stupid, then we might’ve accidentally created AGI and astroturfing bots on the internet would be a bit of a trivial concern at that stage.
*in English.
They still suck at many other languages, like Finnish.
They’re still garbage in finnish. Even if they manage to make grammatically correct sentences, it sounds like incredibly awkward word salad, often there isn’t even real meaning in the sentences. Non-human translations have struggled from the beginning and I’m not surprised even the LLMs aren’t doing that well.
Kuulostaa siis ihan paskalta ja on edelleen tosi helppoa tunnistaa. Ei akateemikot väännä sellaista tuubaa omin käsin, eikä kukaan normitason juntti taas kirjoita niin jäykästi. Tai ok varmaan oman elämänsä jarisarasvuot yrittää suoltaa samankuuloista tyhjää konsulttiroskaa, mutta olettaisin että ne on juuri sitä porukkaa jotka käyttää tekoälyä muutenkin
They’ve mastered Danish, though. Or at least the native speakers can’t tell the difference.
None of the other people in the Nordics are surprised that the native speakers can’t tell the difference. We can’t either after all
—
Seeing any em dash in the wild makes me immediately suspicious
I loooove using m dashes. They’re just so—expressive.
Yeah, I’m 42 yrs old. A long time ago I sacrificed what lvl of English I did know in order to use “lol” - I’m not giving up em dashes.
That’s probably part of the joke
There’s always one good litmus test, if you can ask someone live before they have a chance to do a Google search: ask what the alt code is for an em dash.
If they don’t immediately answer 0151, they’re full of shit.
for me it’s just altgr+caps followed by tapping - thrice, way better use of the caps lock key. thank u xmodmap
meh - I just use a dash and move on.
That’s fine, LLMs use mainly em dashes so there would be no confusion, hence that being the litmus test.
Compose key, hyphen, hypen. Because while it’s a less common usage, visually I prefer the shorter en dash – over the em dash — anyway.
Also for people missing the Unix-style compose key on Windows: https://wincompose.info/. Memorising numeric codes is for chumps :)
OH MY GOD, I CAN GET A COMPOSE KEY ON WINDOWS
Thank youBot account obvs
Opt shift hyphen. What’s an alt code? 🍏
Hmmm, I think to know for sure, we’ll have to throw your device into a lake. Apples float, so if it’s a genuine Designed By Apple in California™ computer that supports that option, it won’t sink and then we’ll know you’re telling the truth.
I feel like this is more likely to lead you astray than anything. An LLM bot will immediately know the alt code, while a real person will only know it if they use Windows. Lots of people use Linux, or mobile keyboards.
I always just google ‘em dash’ then ctrl+c it.
Or if they start explaining what a compose key is; also human
What if they don’t know the code and just use a keyshift instead?
Throw them into a lake, I guess. If they sink and drown, they’re innocent. If they float, you know they’re full of (buoyant) shit (and can then safely burn them at the stake if you want to).
Ah, welcome back, Witch Test…
