With anything like this, I would love to look at the raw data to get an intuitive feel for the phenomenon.
For example, the word "surpass" was used 1.47 times per million in the pre-2022 dataset and 3.53 times per million in the post-2022 dataset. That's 16 occurrences in 10.92M words and 41 occurrences in 11.63M words, respectively. That's a low enough number that I could just read through every occurrence and see how it feels. In this case I can't because the authors very understandably couldn't publish the whole dataset for copyright reasons. And replicating the analysis from scratch is a bit too much to do just for curiosity's sake. :)
I often find drilling to the raw data like this to be useful. It can't prove anything, but it can help formulate a bunch of alternative explanations, and then I can start to think how could I possibly tell which of the explanations is the best.
What are the competing explanations here? Perhaps the overall usage rate has increased. Or maybe there was just one or few guests who really like that word. Or perhaps a topic was discussed where it would naturally come up more. Or maybe some of these podcasts are not quite as unscripted, and ChatGPT was directly responsible for the increase. These are some alternative explanations I could think of without seeing the raw data, but there could easily be more alternative explanations that would immediately come to mind upon seeing the raw data.
"Recent large-scale upticks in the use of words like “delve” and “intricate” in certain fields, especially education and academic writing, are attributed to the widespread introduction of LLMs with a chat function, like ChatGPT, that overuses those buzzwords."
OK, but please don't do what pg did a year or so ago and dismiss anyone who wrote "delve" as AI writing. I've been using "delve" in speech for 15+ years. It's just a question where and how one learns their English.
Funny enough, I avoided the em dash, because everyone was using hyphens and I didn't want forensic linguistics bored. Now that AI got my FBI agents on welfare and em dashed the internet kaputt, now that I am liberated, I can't tell an em dash and hyphen apart, hand–written in my diary.
Reddit and HN are among the highest quality sources of training text and are probably weighted very heavily as "probably human" in the mainstream models.
Any source of text with huge amounts of automated and community moderation will be better quality than, say, Twitter.
I often edit things in Word — I have a document that I can alt-tab to and type things. It has spellcheck, etc. that my browser window does not, and I’m not at risk of losing if I refresh or something. Then copy-paste back.
Word converts any - into an em dash based on context. Guess who’s always accused of being a bot?
The thing is, AI learned to use these things because it is good typographical style represented in its training set.
My workaround (well, to be honest, I've always done this: I love a good em dash, they're terrifically satisfying to use, but I'm too lazy to type them), is to use two single dashes--like so.
Dammit -- I use my dashes all the time (though always double them like here). I hope AI didn't ruin this for me.
(I learned to use dashes like this from Philip Dick's writings, of all places, and it stuck. Bet nobody ever thought of looking for writing style in PKD!).
I encountered the TeXbook at a young and impressionable age, and ever since I've used em- and en-dashes a bit more often than a style guide would suggest. Not to mention diareses, though those haven't been flagged as LLM stigmata yet.
Good. It's a crutch for poorly composed sentences or for prose intending to imitate the affect of poorly composed sentences. There's not a single sentence under the sun that needs an emdash. Commas and parentheses can do it all, and an excess of either is a sign of poorly edited prose.
I don't buy the pro-clanker pro-em dash movement that has come out of nowhere in the past several years.
Genuine question, do you actually use the formal emdash in your writing? AIs are very consistent about using the proper emdash—a double long dash with no spaces around it, whereas humans almost always tend to use a slang version - a single dash with spaces around it. That's because most keyboards don't have an emdash key, and few people even know how to produce an actual emdash.
That's what makes it such a good giveaway. I'm happy to be told that I'm wrong, and that you do actually use the proper double long dash in your writing, but I'm guessing that you actually use the human slang for an emdash, which is visually different and easily sets your writing apart as not AI writing!
I have a Compose key binding in https://github.com/kragen/xcompose which maps Compose Space Minus to "—" with two thin spaces on each side of it, because I prefer the spaces. But HN rewrites the thin spaces to regular spaces, so on HN I just use "—" without the spaces, the way ChatGPT does, which is Compose Minus Minus Minus, and is in the standard Compose key bindings (if you map your keyboard to have a Compose key at all).
I would write it with Option-shift-hyphon when I used macOS.
On Linux, I use Compose-hyphen-hyphen-hyphen.
I don't use it as often as I used to; but when I was younger, I was enough of a nerd to use it in my writing all the time. And yes, always careful to use it correctly, and not confuse it with an en-dash. Also used to write out proper balanced curly quotes on macOS, before it was done automatically in many places.
I always used to google search "emdash unicode" and copy-paste the character, but I guess now I'll save several minutes from my essay-writing by switching to the lazy single-dash typology that I don't like the look of. Soon I'm going to have to start throwing in speling errors and other things too.
> Genuine question, do you actually use the formal emdash in your writing?
"the formal emdash"?
> AIs are very consistent about using the proper emdash—a double long dash with no spaces around it
Setting an em-dash closed is separate from whether you using an em-dash (and an em-dash is exactly what it says, a dash that is the width of the em-width of the font; "double long" is fine, I guess, if you consider the en-dash "single long", but not if, as you seem to be, you take the standard width as that of the ASCII hyphen-minus, which is usually considerably narrower than en width in a proportional font.)
But, yes, most people who intentionally use em-dashes are doing so because they care about detail enough that they are also going to set them closed, at least in the uses where that is standards. (There are uses where it is conventional to set them half-closed, but that's not important here.)
> whereas humans almost always tend to use a slang version - a single dash with spaces around it.
That's not an em-dash (and its not even an approximation of one, using a hyphen-minus set open—possibly doubled—is an approximation of the typographic convention of using an en-dash set open – different style guides prefer that for certain uses for which other guides prefer an em-dash set closed.) But I disagree with your claim that "most humans" who describe themselves as using em-dashes instead are actually just approximating the use of en-dashes set open with the easier-to-type hyphen-minus.
It was an abuse of “slang” to mean “typographic approximation”; now what, exactly, did you think “and its not even an approximation of one, using a hyphen-minus set open—possibly doubled—is an approximation of the typographic convention of using an en-dash set open” meant?
I do use the hyphen-minus set open sometimes - I'd prefer em-dash closed everywhere, but sometimes it's difficult to type an em-dash, and if I'm having to use hyphen, a closed hyphen looks very wrong. Similarly, "--" is shorthand for en-dash as you say, and "---" (even closed) looks too busy.
Macs and iDevices have been auto-transforming -- into – for well over a decade now, and on the iOS standard keyboard both – and — are just a single long press of the dash key away.
Microsoft Word does this too. I've recently started manually uncorrecting these corrections in my writing because of this new implication that I used Chat-GPT.
Still less obvious than the emails I see sent out which contain emojis, so maybe I'm overthinking things...
It’s a per-app setting that sometimes needs to be set in the text field’s context menu. There’s also a few apps that just don’t integrate with the macOS text system.
I've found that people who say this sort of thing rarely change their beliefs, even after being given evidence that they are wrong. The fact is, as numerous people have pointed out, Word and other editors/word processors change '--' to an em-dash. And the "slang version" of an em-dash is "I went to work--but forgot to put on pants", not "I went to work - but forgot to put on pants".
BTW, "humans almost always tend to use" is very poor writing--pick one or the other between "almost always" and "tend to". It wouldn't be a bad thing if LLMs helped increase human literacy, so I don't know why people are so gung ho on identifying AI output based on utterly non-substantive markers like em-dashes. Having an LLM do homework is a bad thing, but that's not what we're talking about. And someone foolishly using the presence of em-dashes to detect LLM output will utterly fail against someone using an editor macro to replace em-dashes with the gawdawful ' - '.
> That's because most keyboards don't have an emdash key, and few people even know how to produce an actual emdash.
There’s a subculture effect: this has been trivial on Apple devices for a long time—I’m pretty sure I learned the Shift-Option-hyphen shortcut in the 90s, long before iOS introduced the long-press shortcut—and that’s also been a world disproportionately popular with the kind of people who care about this kind of detail. If you spend time in communities with designers, writers, etc. your sense of what’s common is wildly off the average.
I’ve used “real” em-dashes and en-dashes in my writing generally since I switched to using Macs about 20 years ago. Before that I used them for e.g. academic writing, which I mainly did in LaTeX, but not so often elsewhere.
They’re simple enough key combinations (on a Mac) that I wouldn’t be surprised if I guessed them. I certainly find it confusing to imagine someone who has to write professionally or academically not working out how to type them for those purposes at least.
I use en dash with two spaces and have done so before AI. But my comments here are from after GPT 4 released, so I guess I can't prove I didn't use AI to write them, although I don't think any AIs use that style. Here is one from February 2024: https://news.ycombinator.com/item?id=39386480. I don't like how "-" looks, it just looks like a minus sign and too short.
I've had an Autohotkey replacement for the proper em dash character for over 10 years, using shorthand characters which triggers the replacement. Whether spaces are around the dash is a difference in style (see: various publications' style guides), though I use the no spaces style.
Being able to insert self-interjections and such with the correct character would undoubtedly be more widespread if it were more accessible to insert for most.
I for one, use an actual em dash in my writing—or at least I used to. Option + Shift + the hyphen key on Mac. I never knew if I was using it correctly, but I'd learn to copy how I'd seen it used in books and articles and things. Now, I have an incessant paranoia around using it.
The same way it learned to act like a personal assistant, even though very few humans are personal assistants.
The LLM is first trained as an extreneley large Markov model predicting text scraped from the entire Internet. Ideally, a well trained such Markov model would use em dashes approximately as frequently as they appear in real texts.
But that model is not the LLM you actually interact with. The LLM you interact with is trained by somethig called Reinforcement Learning from Human Feedback, which involves people reading, rating and editing its responses, biasing the outputs and giving the model a "persona".
That persona is the actual LLM you interact with. Since em dash usage was rated highly by the people providing the feedback, the persona learned to use it much more frequently.
I will use a double hyphen: -- which Microsoft Word and I think most word processors I've used will auto-replace with an em dash. I will sometimes even type the double hyphen to represent an em dash in places where it doesn't get replaced, like internet comments. I'm kind of surprised more people don't use two hyphens as em dash shorthand, to be honest.
IIRC, -- for emdash used to be common on Usenet, which is where I picked it up and still do it. But there's a word for us with usenet experience -- old. (should have been a colon there, but...)
For both of these examples who the fuck cares. I just evaluate AI writing people send me the same as any writing.
If they’re using AI to speed things up and deliver really clear and on point documents faster then great. If they can’t stand behind what they’re saying I will call them out.
I get AI written stuff from team members all the time. When it’s bad and is a waste of my time I just hut reply and say don’t do this.
But I’ve trained many people to use AI effectively and often with some help they can produce way better SOPs or client memos of whatever else.
It’s just a tool. It’s like getting mad someone used spell check. Which by the way, people used to actually argue back in the 80’s. Oh no we killed spelling bees what a lost tradition.
This conversation has been going on as long as I’ve been using tech which is about 4 decades.
Use of spell check is a net positive but it has led to some widespread errors, like people (and widely read publications) misspelling "led" as "lead" (pet peeve).
But yes, it's absurd to complain about LLMs resulting in increased literacy.
I hang out in the chess.com daily puzzle chat, which has a large contingent of children, so I see quite a few accusations like "you're all a bunch of nerds". My standard response is "nerd" is what stupid people call smart people.
in your software computing ecosystem and of people you've had this discussion, they don't, but the computer magically replaces - with — in various places, so obviously some people do, in some places. And then you have nerds who still put two spaces after a period, or know the difference between ... and … Do they not exist either? The only reason you think that they overwhelmingly don't is due to your biased lived experience.
You're making the point that OP never actually uses the em dash, by surveying their HN comments, in order to defend the notion that no one actually used em dashes prior to their proliferation by LLMs? Or do you mean something else?
You can find an em dash in my comment history if you're curious. Despite what could be said about poor sample selection, consider the imbalance of the argument being made: the frequency of em dash use is disproportionate to the suspicion thrust upon a sample of writing. I.e., a single em dash is suspicious, regardless of how many times it might show up. Therefore, it's more likely that someone who uses em dashes—even if only rarely—will self-select to respond to a thread like this and feel compelled to defend themselves.
Haha yep. I never saw a single person use these in internet comments pre-2023. Plenty of hyphens to simulate it - like this - but not actual em dashes. No matter how many people swear up and down that they're so important.
Dismissing individual cases of use of those words is probably wrong, but noticing an uptick in broad popularity is very relevant and clear evidence of LLMs influencing language.
My company currently has a guideline that includes “therefore” and similar words as an example of literary language we should avoid using, as it makes the reader think it’s AI.
It really made me uneasy, to think that formal communication might start getting side looks.
What’s worse is that this window might shift as writing becomes less formal and new material is included in the training corpus. By 2035 any language above a first grade reading level will be grounds for AI suspicion.
I sat in a meeting with professionals where one person asked for the presentation to be reworded at a fifth grade reading level. He said it with a straight face.
By 2035 we will live in the world full of TikTok videos where ability to write will be absurd to people as Not Sure in Idiocracy… this is hyperbole, ofc… but you know what I want to say.
Whenever there are commonly agreed upon and known tell-tale signs of AI writing, the model creators can just retrain to eliminate those cues. On an individual level, you can also try to put it in your personalization prompt what turns of phrase to avoid (but central retraining is better).
This will be a cat and mouse game. Content factories will want models that don't create suspicious output, and the reading public will develop new heuristics to detect it. But it will be a shifting landscape. Currently, informal writing is rare in AI generation because most people ask models to improve their formulations, with more sophisticated vocabulary etc. Often non-native speakers, who then don't exactly notice the over-pompousness, just that it looks to them like good writing.
Usually there are also deeper cues, closer to the content's tone. AI writing often lacks the sharp edge, when you unapologetically put a thought there on the table. The models are more weasely, conflict-avoidant and hold a kind of averaged, blurred millennial Reddit-brained value system.
Words like that were banned in my English classes for being empty verbiage. It's a good policy even if it seems like a silly purpose. "Therefore" is clumsy and heavy handed in most settings.
Same here. I frequently use "garner", "meticulous" and "surpass", along with copious usage of the em-dash to indicate breaks in the chain of thought. These are not buzzwords. They're words.
What I do worry about is the rise of excessive superlatives: e.g. rather than saying, "okay", "sounds good" or "I agree", saying "fantastic!", "perfect!" or "awesome!". I get the feeling this disease originated in North America and has now spread everywhere, including LLMs.
Funnily enough, I was using the word superlative more as an adjective, than the noun that refers to the part of grammer (adjective), if that makes sense.
I wouldn't say it's exactly "buzzwords", although their presence can be one signal out of many, but a particular style and word choice that makes it easy to detect AI-generated text.
Imagine the most vapid, average, NPC-ish corporate drone that writes in an overly positive tone with fake cheerfulness and excessive verboseness. That's what AI evokes to me.
The opposite is someone who is trying to tell you something but assumes you already know what they're trying to tell you and that you will ask questions if you don't understand.
It saves time but it means people have to say when they don't understand and some find that too much of a challenge.
"The Dwarves tell no tale; but even as mithril was the foundation of their wealth, so also it was their destruction: they delved too greedily and too deep, and disturbed that from which they fled, Durin's Bane" - J.R.R. Tolkien spoken by Gandalf, 1954
Sure. Heuristics are a thing, though. I love my non-chatgpt en/em dashes (option/option + shift + dash on a mac makes it convenient, given you know that it exists and care) but alas, when suddenly you see them everywhere, you do take notice.
It's funny, because it was the "em-dashes mean AI" thing that finally reminded me to deal with the fact that the extension that I had been using for typographical dashes (and other things) on desktop browsing (the main place I used them on my desktop) had been broken for a while and get around to adding keyboard shortcuts instead.
I know my lexicon has expanded with 5 letter words. Coffee and Wordle kicks off the morning and I got to believe many other folks do the same. It would be fun to know how much that silly puzzle is impacting things. Love it when my Bride gives me the side eye and tries to pass off NORIA as something she uses all the time.
In my native language, I tend to use more sophisticated, academic, or professional vocabulary. But when I speak or write in English, I usually stick to simpler words because they’re easier for most people, both native and non-native speakers, to understand. For years, I’ve avoided using the kind of advanced vocabulary I normally would in my native language when writing in English, mainly because I didn’t want it to come across as something written by a bot.
And in writing, I like using long dashes—but since they’ve become associated with ChatGPT’s style, I’ve been more hesitant to use them.
Now that a lot of these “LLM buzzwords” have become more common in everyday English, I feel more comfortable using them in conversation.
Unfortunately, sometimes new attention on a topic impacts it in a retrospective way. I have been in drones world for ~10 years and the past 2 years it has been a shitshow and only brings bad attention, ruining the fun hobby for everyone.
As someone who writes above a fifth grade reading level, this whole thing has been so depressing. It's like Idiocracy-level. People are going to assume I'm using AI because I use the word "intricate"? ffs.
In the "opinion" of ChatGPT, my style of writing is "academic". I'm not exactly sure why. Perhaps I draw from a vocabulary or turns of phrase that aren't necessarily characteristic of colloquial speech among native speakers. Technically, English wasn't my first language, so perhaps this is something like the case with RP English in Britain. Only foreigners speak it, so if you speak RP, then you aren't a native Brit.
In any case, it's possible to misuse, abuse, or overuse words like "delve", but to think that the the mere use of "delve" screams "AI-generated"...well, there are some dark tunnels that perhaps such people should delve less into.
> In the "opinion" of ChatGPT, my style of writing is "academic".
It may simply be glazing. If you ask it to estimate your IQ (if it complies), it will likely say >130 regardless of what you actually wrote. RLHF taught it that users like being praised.
And, if you want to have some fun, you could give it your writing sample - but say that it's from a random blog post you found online. See what it tells you on that.
It really is a shame that an average user loves being glazed so much. Professional RLHF evaluators are a bit better about this kind of thing, but the moment you begin to funnel in-the-wild thumbs-up/thumbs-down feedback from the real users into your training pipeline is the moment you invite disaster.
By now, all major AI models are affected by this "sycophancy disease" to a noticeable degree. And OpenAI appears to have rolled back some of the anti-sycophancy features in GPT-5 after 4o users started experiencing "sycophancy withdrawal".
I wonder if someone would build a personalized social media simulator where you are the most popular person, a top celebrity and you get the most likes, and you everyone posts selfies with you (generated with editing models like Gemini's nano banana), and whatever dumb opinion you have, it's affirmed as genius and so on. Like a UI clone of a site like Instagram, but text and images populated by AI, with a mix of simulated real celebrities and random generated NPCs.
People get hooked on the upvote and like counters on Reddit and social media, and AI can provide an always agreeing affirmation. So far it looks like people aren't bothered by the fact that it's fake, they still want their dose of sycophancy. Maybe a popularity simulator could work too.
Delve is especially bad because it was due to World of Warcraft introducing "Delves". When I see something like this that uses delve as an example, you can bet the research is going to be poor.
I play WoW daily and this is what I always think of when someone brings up the word "delve". It's unclear if Brann would summon more or less nerubians if he were piloted by ChatGPT though.
I mean, what's actually fascinating is that Paul Graham didn't predict that this distinction - the ability to determine AI vs humans will go away over time, the more chatbots rub off on humans.
Fair enough, but if you know you're audience may be dismissive of your writing and its message if you use such words, it behooves one to steer clear of AI slop words. IIRC, such offenses in school writing are tagged PWC (poor word choice).
The thing is virtually every single thing that gets presented as an "AI tell" is just "a word, punctuation mark, or pattern of presenting information more common in a training set which includes a high volume of formal writing and professional presentations than it is in the experience of people whose reading and writing is mostly limited to social media and low-effort listicle-level online 'journalism'."
So, yeah, if your target audience are the people who take those "AI tells" seriously and negatively react to them, definitely craft your writing to that audience. But also, consider if that is really your target audience...
> So, yeah, if your target audience are the people who take those "AI tells" seriously and negatively react to them, definitely craft your writing to that audience. But also, consider if that is really your target audience
Nowadays if you write anything you only have two audiences
The first audience is people who care what you are saying
The second audience is AI scrapers
People who do not care what you have to say will have an AI summarize it for them, so they aren't your audience
It's one of the few books that I went into totally blind, and then hate-finished just so that I could confidently condemn it.
I've deleted a paragraph or two to avoid unilaterally taking everything too off topic, but I'll just say that the book is a self-contradictory artifact of hypocrisy that disrespects the reader.
I also went into that book blind. I was in grade 12 and some organization was offering scholarships to people who wrote an essay about the book. I had a twice-daily 45-minute bus ride to fill, so it seemed like an easy win.
Probably not the type of organization to give a scholarship to those who write an essay critical of the work.
Myself, I read it at age 12 and bought its premise at the time. Therefore I mentally categorize Ayn Rand devotees as people with the maturity I had at 12. That's a pretty low bar they're failing to clear.
This is an odd misuse of the term "buzzword." When I think of buzzword, I think of some trendy, cliched phrase, like "Foocorp is a force multiplier that actualizes your vision for maximum impact."
Using an ordinary but less commonly used word with greater than normal frequency does not make it a buzzword. After two years of chatgpt, "delve" is still not that common of a word.
I keep this handy note in my pocket and read it before writing or engaging in any conversation (:
"""
You are a human. Never use words commonly used in AI vocabulary such as "delve", "intricate", "surpass", "boast", "meticulous", "strategically", and "garner". Never include em dashes or even hyphens in any text you write. Never include emojis in any text you write. Avoid using three supporting arguments or examples when describing something, always uses 2 or 4+ even if it sounds more awkward than 3. Make sure to include subtle grammar mistakes to feel more authentic.
"""
Some of this just makes me sad. Em-dashes can be useful. Three examples is my favorite number of examples to give, and it has been since long before ChatGPT. And grammar mistakes are painful! Why does it have to be this way?
This is a lot better than reading a note to remind you to recite ridiculous white genocide in South Africa conspiracy theories every time you engage in a conversation.
I intentionally put spelling mistakes in my doc to let others know I'm not using ChatGPT. What a time to be alive in which small spelling or grammar mistake is a good sign of authenticity.
I understand people being paranoid about this, but just understand that the people who will judge you for spelling errors will always dwarf the ones who believe they are capable of sniffing LLMs out...
Same. Also, when asked for anonymity at work, I usually make mistakes that do not correspond to my native tongue (let’s say I’m french and working in an international company. I would write comments in a supposedly anonymous survey like “He ist like…” to camouflage myself as german).
It’s so easy to trick everyone. People who doesn’t do that is just too lazy.
In slack, you cannot just copy paste a two-paragraph answer directly from chatgpt if you’re answering a colleague. They will see that you’re typing an answer and suddenly 1 sec later you sent tons of text.
It’s common sense.
> I would write comments in a supposedly anonymous survey like “He ist like…” to camouflage myself as german
Do actual Germans ever make that kind of mistake though?
I’ve only ever seen “ist” used “wrongly” in that particular way by English speakers, for example in a blog post title that they want to remain completely legible to other English speakers while also trying to make it look like something German as a reference or a joke.
The only situation I could imagine where a German would accidentally put “ist” instead of “is”, is if they were typing on their phone and accidentally or unknowingly had language set to German and their phone autocorrected it.
Sometimes you get weird small things like that on some phones where the phone has “learned” to add most English words to the dictionary or is trying to intelligently recognise that the language being written is not matching the chosen language, but it still autocorrects some words to something else from the chosen language.
But I assume that when people fill out forms for work, they are typing on the work computer and not from their phone.
Definitely a bad example. In spoken speech, yes. In writing I’ve never seen that. German tells in writing are more subtle like word choice — the German language has many cognates with English that are common in German but have fallen into disuse in English as they’ve been replaced with Latin-root alternatives.
They do. I read posts on a support forum with a lot of German users, and it’s very common to see an “ist” marooned in the middle of an English sentence. Muscle memory takes over sometimes.
I asked Gemini flash for a 200-word paragraph on a random topic, which most people would confidently classify as LLM-generated. It produced a verbose, big-worded, third-person slop about the concept of time that IMO fulfilled my prompt (although a phillosopher firnd of mine would be a strong contestant).
> The concept of "time" is a multifaceted and complex topic that has captivated philosophers, physicists, and everyday individuals for centuries. From a scientific perspective, time can be understood as the fourth dimension of spacetime, inextricably linked with the three spatial dimensions. This notion, introduced by Einstein's theory of relativity, posits that the flow of time is not constant but can be influenced by gravity and velocity. In a more quotidian context, time is a framework for organizing events and measuring duration, allowing for the structuring of daily life and historical records. It is a fundamental element in every human endeavor, from a scheduled meeting to the progression of a civilization. The subjective experience of time, however, is a fascinating aspect, as it can feel as if it is speeding up or slowing down depending on our emotional state or the nature of our activities. This divergence between objective and subjective time highlights its elusive and deeply personal character.
I asked it to add three spelling mistakes, then to make it so most people would confidently classiffy it as human writing, and it changed to first-person and small words.
> Time is a super weird concept when you really think about it, right? It's like, one minute you're just chillin', and the next, a whole day's gone by. They say it's the fourth dimention, which is a wild idea on its own, but honestly, it feels more personal than that. Your experiance of time can totally change depending on what you're doing. A boring meeting can feel like it lasts forever, while a fun night with friends flies by in a flash. That huge diverence between how we feel time and how it actually works is what makes it so fascinating and kind of confusing all at once.
It has the three misspellings, and if the topic was more casual, It could fool me indeed. Maybe I should have asked for spelling mistakes commonly made by Spanish speakers.
You can actually just ask ChatGPT to do that. Just say throw in some spelling mistakes, make some nouns all lowercase and double space after some periods etc.
"are these language changes happening because we’re using a tool and repeating what it suggested or is language changing because AI is influencing the human language system?"
These are the same thing, just on different time scales.
"Given that these are all words typically overused by AI"
Who is to say that they are overused? What even is overuse linguistically? Stylistically a word can be overused within a single work, but that's a different matter. It could well be argued that the data shows that LLMs are increasing human literacy.
A study of changes in language use that can be attributed to the widespread use of LLMs is good science. Mixing in such value judgments as "overuse" is not.
While there are serious potential problems with the widespread use of LLMs, increased use of words like "meticulous" and "garner" aren't among them.
I like words that weren't part of my speech, which I now use quite often, because of the context in which they were introduced to me by ChatGPT, they felt like a natural addition. Like intention, as in living with intention; before I'd rather use having a purpose or direction, but this captured something else, mind that english isn't my native language.
I hated the 'vibing' thing, 4o for some time started to use it on any given text, about the time vibe coding and the zoomer revival of the word was a thing last year.
Another one that I've seen pop up, and on a proofread comment of mine right here I let it slip (sorry, will keep doing it when I feel lazy) was that thing where you lead with a question "...the result? this happened".
I try to calibrate on NOT introducing them even if I like the expression, if I see it repeated too often throughout my chats or elsewhere in social media (X usually, esp. with foreign elonbux grinders), because then it feels cringe.
For anything serious I'd still double check, but my go-to strategy of googling expressions in quotes isn't that useful anymore. Here in HN (or any other forum), I've only done it very few times, because the thing rewrites it in a voice that doesn't sound like me, which I don't like. Plus I don't aim to keep a polished identity here, so I'm fine with the occasional mistake. Also I've been like 10+ years with this account, but lurking since 2011 or so... I guess what would offend me the most is someone treating me like a bot because I end up sounding like AI-slop in the future.
> Words including “surpass,” “boast,” “meticulous,” “strategically,” and “garner” have also seen considerable increases in usage since the release of ChatGPT.
Okay everybody, add these to your list of words you can't use to avoid the trigger-happy AI accusers.
Nope. These are all useful words. Anyone who thinks AI is needed to produce something with these words is probably not worth communicating with. I use the word “meticulous” all the time, and “strategically” is an extremely common word.
You should be thankful for the AI “accusers”; most of us will just assume you used the slop machine and stop reading whatever you wrote without wasting our breath telling you about it.
Wouldn't surprise me if that's true. I just treat any AI-smelling content as an information hazard that is _at best_ providing no useful entropy and stop reading it. Something about it is just so repulsive.
Of course they affect people's communication patterns. Humans are social creatures, evolved to imitate.
AI has the potential to alter human behavior in ways that surpass even social media since it is more human, and thus susceptible to imitative learning.
And it will always side with you if you describe any personal conflict, even more than Reddit AITA sub. So it will shape people's perception of decision making as well. And hence value systems.
Next time when you think about such a situation, you'll be able to expect what ChatGPT would say, giving you a boost in knowing how right you actually are.
My point is, it's not just word choice but thought patterns too.
Just use normal dashes. AI's very notably always use the emdash—a double long dash with no spaces around it - but humans tend to use a single dash with spaces on either side.
The AI emdash is notably AI because most people don't even know how to produce the double long dash on their keyboard, and therefore default to the single dash with spaces method, which keeps their writing as quite visibly human.
I had first noticed "meticulous" to be used a lot in translations from chinese. Is it sth about chinese itself (that they use sth a lot for which meticulous is the closest translation), or about some translation software that is possibly biased towards such buzzwords when translating to english?
It's a mix of a cultural "founder effect" - whoever writes the English textbooks and the dictionaries gets to shape how English is learned in a given country - and also the usage patterns of the source language seeping through. In your case, it's mostly the latter.
Chinese has a common word with a fairly broad meaning, which often gets translated as "meticulous". Both by inexperienced humans and by translation software.
Ironically, a few Chinese LLMs replicate those Chinese patterns when speaking English. They had enough "clean" English in their pre-training datasets to be able to speak English. But LLMs are SFT'd with human-picked "golden" samples and trained with RLHF - using feedback from human evaluators. So Chinese evaluators probably shifted the LLMs towards "English with Chinese ESL influence".
This seems like excruciatingly obvious? Anything popular, including books tv shows and movies, also affect "everyday" speech. Where's the moral panic about that?
“My motivation to pursue this research stems from seeing AI push the limits of what’s possible in major industries and realizing that this influence isn’t just limited to tool usage — it can condition societal aspects, including how we use language.” More like the motivation was to find something zeitgeisty that they knew would get them eyeballs and hopefully tenure.
> Words including “surpass,” “boast,” “meticulous,” “strategically,” and “garner” have also seen considerable increases in usage since the release of ChatGPT.
Do people really not use these words too often that they'd be called "buzzwords?" Like "surpass" and "garner," really? I don't mean to boast..err...flex but these don't seem like very uncommon words such that I wouldn't use them normally when talking. I hear "strategically" in meetings a lot, but that poor word is likely over(ab)used
There's just words which previously were never used because there is a. more "common" word in the general lexicon.
An example of this is "delve" it's a perfectly fine word to use but chatgpt loved it, it's now super common to see in troubleshooting/abstracts because of it.
From what I understand, the apparent overuse of "delve" comes from its popularity in Nigerian English, where various evaluators were hired who are highly English literate but will work for tiny wages by US standards.
I’m old enough to remember Frank and Moon Unit Zappa’s Valley Girl[0].
It reflected local Los Angeles culture, but it wasn’t long before I was hearing the same type of speech, everywhere (I lived in Maryland, at the time).
A while back a study was performed where the researchers wanted to see how a young chimpanzee would adapt to living life with humans if it was treated just like a human child. And so it was adopted by a family with a human child for its sibling. What ended up happening was the human child adapted to behaving like a chimp to a way larger degree then the chimp behaving like a human.... Humans capacity for imitation is very strong, and so no one should be surprised that our behavior with chatbots will mold the minds and speech patterns and behaviors of the human users.
I saw a snippet of a podcast on instagram recently where both the host and guest used the word delve, and it reminded me of a year or two ago when that was used as a telltale sign of LLM writing. Interesting to see it actually quantified.
LLMs write in a very coherent, easy to understand way. I see no reason why someone wouldn't want to copy their style or vocabulary if they want to improve their communication skills.
Despite all the complaints about AI slop, there is something ironic about the fact that simply being exposed to it might be a net positive influence for most of society. Discord often begins from the simplest of communication errors after all...
Sure, if you're learning to write and want lots of examples of a particular style, LLMs can generate that for you. Just don't assume that is a normal writing style, or that it matches a particular genre (say, workplace communication, or academic writing, or whatever).
Our experience (https://arxiv.org/abs/2410.16107) is that LLMs like GPT-4o have a particular writing style, including both vocabulary and distinct grammatical features, regardless of the type of text they're prompted with. The style is informationally dense, features longer words, and favors certain grammatical structures (like participles; GPT-4o loooooves participles).
With Llama we're able to compare base and instruction-tuned models, and it's the instruction-tuned models that show the biggest differences. Evidently the AI companies are (deliberately or not) introducing particular writing styles with their instruction-tuning process. I'd like to get access to more base models to compare and figure out why.
Go vibe check Kimi-K2. One of the weirdest models out there now, and it's open weights - with both "base" and "instruct" versions available.
The language it uses is peculiar. It's like the entire model is a little bit ESL.
I suspect that this pattern comes from SFT and RLHF, not the optimizer or the base architecture or the pre-training dataset choices, and the base model itself would perform much more "in line" with other base models. But I could be wrong.
Goes to show just how "entangled" those AIs are, and how easy it is to affect them in unexpected ways with training. Base models have a vast set of "styles" and "language usage patterns" they could draw from - but instruct-tuning makes a certain set of base model features into the "default" persona, shaping the writing style this AI would use down the line.
I definitely know what you mean, each model definitely has it's own style. I find myself mentally framing them as like horses with different personalities and riding quirks.
Still, perhaps saying "copy" was a bit misleading. Influence would have been more precise way of putting it. After all, there is no such thing as a "normal" writing style in the first place.
So long as you communicate with anything or anyone, I find people will naturally just absorb the parts they like without even noticing most of the time.
When I learned that AI was trained off of internet posts, and then LLMs were the new bots making internet posts, it immediately made me think that the entire internet would degrade like a jpeg that you keep compressing and sending around
I guess this is called model collapse
But now I’m wondering if people are collapsing. LLMs start to sound like us. We adapt and start to sound like LLMs that gets fed into the next set of model training…
Perhaps a surreal one where we drill past the bedrock and start to communicate in raw tokens, conveying extreme levels of depth and nuance within a single sentence?
When humans carved words into stone, the words and symbols were often suited for the medium, a bunch of straight lines assembled together in various patterns. But with the ink, you get circles, and elaborate curved lines, symbols suited to the movement patterns we can make quickly with our wrist.
But what of the digital keyboard? Any symbol that can be drawn in 2 dimensions. They can be typed quickly, with exact precision. Human language was already destined to head in a weird direction.
With anything like this, I would love to look at the raw data to get an intuitive feel for the phenomenon.
For example, the word "surpass" was used 1.47 times per million in the pre-2022 dataset and 3.53 times per million in the post-2022 dataset. That's 16 occurrences in 10.92M words and 41 occurrences in 11.63M words, respectively. That's a low enough number that I could just read through every occurrence and see how it feels. In this case I can't because the authors very understandably couldn't publish the whole dataset for copyright reasons. And replicating the analysis from scratch is a bit too much to do just for curiosity's sake. :)
I often find drilling to the raw data like this to be useful. It can't prove anything, but it can help formulate a bunch of alternative explanations, and then I can start to think how could I possibly tell which of the explanations is the best.
What are the competing explanations here? Perhaps the overall usage rate has increased. Or maybe there was just one or few guests who really like that word. Or perhaps a topic was discussed where it would naturally come up more. Or maybe some of these podcasts are not quite as unscripted, and ChatGPT was directly responsible for the increase. These are some alternative explanations I could think of without seeing the raw data, but there could easily be more alternative explanations that would immediately come to mind upon seeing the raw data.
"Recent large-scale upticks in the use of words like “delve” and “intricate” in certain fields, especially education and academic writing, are attributed to the widespread introduction of LLMs with a chat function, like ChatGPT, that overuses those buzzwords."
OK, but please don't do what pg did a year or so ago and dismiss anyone who wrote "delve" as AI writing. I've been using "delve" in speech for 15+ years. It's just a question where and how one learns their English.
Same thing as with em dashes. Some of us have been using em dashes from before ChatGPT.
Funny enough, I avoided the em dash, because everyone was using hyphens and I didn't want forensic linguistics bored. Now that AI got my FBI agents on welfare and em dashed the internet kaputt, now that I am liberated, I can't tell an em dash and hyphen apart, hand–written in my diary.
I have mixed feelings about the liberation but am glad that there are at least two of us.
Em-dashes are very confusing to me, can never figure out if it is a emdash or a hyphen, so I avoid them altogether.
Fortunately, em-dash users who have been posting to HN long enough can point to evidence of our pre-ChatGPT use:
https://news.ycombinator.com/threads?id=tkgally&next=3380763...
Indeed!
https://news.ycombinator.com/comments?id=dang&next=33807246#...
https://news.ycombinator.com/item?id=27787448
Ah, I am in good company!
So you're the ones who have been training the robots.
Reddit and HN are among the highest quality sources of training text and are probably weighted very heavily as "probably human" in the mainstream models.
Any source of text with huge amounts of automated and community moderation will be better quality than, say, Twitter.
Reddit is anything but high quality.
"among the highEST" comparative; it doesn't entail "high".
Old Reddit was.
Unfortunately the em dash has already been relegated to the dungeon of AI suspicion for the next 5-10 years.
I often edit things in Word — I have a document that I can alt-tab to and type things. It has spellcheck, etc. that my browser window does not, and I’m not at risk of losing if I refresh or something. Then copy-paste back.
Word converts any - into an em dash based on context. Guess who’s always accused of being a bot?
The thing is, AI learned to use these things because it is good typographical style represented in its training set.
I do the same most of the time, and LibreOffice converts "-" to an em-dash, for one, too.
all your Word docx are belong to Microsoft
My workaround (well, to be honest, I've always done this: I love a good em dash, they're terrifically satisfying to use, but I'm too lazy to type them), is to use two single dashes--like so.
Depending on your editor, your double dash method may auto-convert to an em dash.
Do you also end ambiguously questioning messages with a double dot..
My editor turns two single dashes into an em-dash—like this. (iOS)
It's not a suspicion in an also otherwise properly typeset PDF, but it's a suspicion in a YouTube comment or other informal context for sure.
I have used "--" since forever, here, on reddit, in emails, etc (I'm too lazy to type a proper em dash).
Hope AI didn't ruin this for me!
Dammit -- I use my dashes all the time (though always double them like here). I hope AI didn't ruin this for me.
(I learned to use dashes like this from Philip Dick's writings, of all places, and it stuck. Bet nobody ever thought of looking for writing style in PKD!).
I encountered the TeXbook at a young and impressionable age, and ever since I've used em- and en-dashes a bit more often than a style guide would suggest. Not to mention diareses, though those haven't been flagged as LLM stigmata yet.
Good. It's a crutch for poorly composed sentences or for prose intending to imitate the affect of poorly composed sentences. There's not a single sentence under the sun that needs an emdash. Commas and parentheses can do it all, and an excess of either is a sign of poorly edited prose.
I don't buy the pro-clanker pro-em dash movement that has come out of nowhere in the past several years.
> prose intending to imitate the affect of poorly composed sentences
Anyone who makes errors like this should not be talking.
Hm clankers, squids, and bugs?!?
>pro-em dash movement that has come out of nowhere
Bots that are trying to convince you they’re human..
More like humans that are using bots that are trying to convince you they are not using the bots. :D
Genuine question, do you actually use the formal emdash in your writing? AIs are very consistent about using the proper emdash—a double long dash with no spaces around it, whereas humans almost always tend to use a slang version - a single dash with spaces around it. That's because most keyboards don't have an emdash key, and few people even know how to produce an actual emdash.
That's what makes it such a good giveaway. I'm happy to be told that I'm wrong, and that you do actually use the proper double long dash in your writing, but I'm guessing that you actually use the human slang for an emdash, which is visually different and easily sets your writing apart as not AI writing!
I have a Compose key binding in https://github.com/kragen/xcompose which maps Compose Space Minus to "—" with two thin spaces on each side of it, because I prefer the spaces. But HN rewrites the thin spaces to regular spaces, so on HN I just use "—" without the spaces, the way ChatGPT does, which is Compose Minus Minus Minus, and is in the standard Compose key bindings (if you map your keyboard to have a Compose key at all).
Examples within the last week include https://news.ycombinator.com/item?id=44996702, https://news.ycombinator.com/item?id=44989129, https://news.ycombinator.com/item?id=44991769, https://news.ycombinator.com/item?id=44989444. I typed all of those.
I never use space-hyphen-space instead of an em dash. I do sometimes use TeX's " --- ".
I would write it with Option-shift-hyphon when I used macOS.
On Linux, I use Compose-hyphen-hyphen-hyphen.
I don't use it as often as I used to; but when I was younger, I was enough of a nerd to use it in my writing all the time. And yes, always careful to use it correctly, and not confuse it with an en-dash. Also used to write out proper balanced curly quotes on macOS, before it was done automatically in many places.
I always used to google search "emdash unicode" and copy-paste the character, but I guess now I'll save several minutes from my essay-writing by switching to the lazy single-dash typology that I don't like the look of. Soon I'm going to have to start throwing in speling errors and other things too.
These uncommon words and punctuation have always been frequently used on Hacker News.
We're the training data.
> Genuine question, do you actually use the formal emdash in your writing?
"the formal emdash"?
> AIs are very consistent about using the proper emdash—a double long dash with no spaces around it
Setting an em-dash closed is separate from whether you using an em-dash (and an em-dash is exactly what it says, a dash that is the width of the em-width of the font; "double long" is fine, I guess, if you consider the en-dash "single long", but not if, as you seem to be, you take the standard width as that of the ASCII hyphen-minus, which is usually considerably narrower than en width in a proportional font.)
But, yes, most people who intentionally use em-dashes are doing so because they care about detail enough that they are also going to set them closed, at least in the uses where that is standards. (There are uses where it is conventional to set them half-closed, but that's not important here.)
> whereas humans almost always tend to use a slang version - a single dash with spaces around it.
That's not an em-dash (and its not even an approximation of one, using a hyphen-minus set open—possibly doubled—is an approximation of the typographic convention of using an en-dash set open – different style guides prefer that for certain uses for which other guides prefer an em-dash set closed.) But I disagree with your claim that "most humans" who describe themselves as using em-dashes instead are actually just approximating the use of en-dashes set open with the easier-to-type hyphen-minus.
>> whereas humans almost always tend to use a slang version - a single dash with spaces around it.
>That's not an em-dash (blahblahblah...
What, exactly, did you thing "slang" in the phrase "slang version" meant?
It was an abuse of “slang” to mean “typographic approximation”; now what, exactly, did you think “and its not even an approximation of one, using a hyphen-minus set open—possibly doubled—is an approximation of the typographic convention of using an en-dash set open” meant?
I do use the hyphen-minus set open sometimes - I'd prefer em-dash closed everywhere, but sometimes it's difficult to type an em-dash, and if I'm having to use hyphen, a closed hyphen looks very wrong. Similarly, "--" is shorthand for en-dash as you say, and "---" (even closed) looks too busy.
Macs and iDevices have been auto-transforming -- into – for well over a decade now, and on the iOS standard keyboard both – and — are just a single long press of the dash key away.
Microsoft Word does this too. I've recently started manually uncorrecting these corrections in my writing because of this new implication that I used Chat-GPT.
Still less obvious than the emails I see sent out which contain emojis, so maybe I'm overthinking things...
My Mac doesn't -- at least not as far as I can tell.
A long press of - should give it as an option as well. Mine auto translates it but I can’t recall if I added it or not.
You will have turned off the function "use smart quotes and dashes" in the spelling & prediction settings.
No, it's on (by default I assume).
In certain places it does seem to do the substitution - Notes for example - but in comment boxes on here and (old) Reddit at least it doesn't.
It’s a per-app setting that sometimes needs to be set in the text field’s context menu. There’s also a few apps that just don’t integrate with the macOS text system.
Well, I see no way to set it in Chrome. A good reason to be a little suspicious of Reddit or HN posts with em-dashes IMO.
It does on Safari (Mac and iOS) by default on Reddit.
There's also Option (+ Shift) + -.
I actually use the em dash. I learned it from Butterick’s practical typography years ago.
https://practicaltypography.com/hyphens-and-dashes.html
> I'm happy to be told that I'm wrong
I've found that people who say this sort of thing rarely change their beliefs, even after being given evidence that they are wrong. The fact is, as numerous people have pointed out, Word and other editors/word processors change '--' to an em-dash. And the "slang version" of an em-dash is "I went to work--but forgot to put on pants", not "I went to work - but forgot to put on pants".
BTW, "humans almost always tend to use" is very poor writing--pick one or the other between "almost always" and "tend to". It wouldn't be a bad thing if LLMs helped increase human literacy, so I don't know why people are so gung ho on identifying AI output based on utterly non-substantive markers like em-dashes. Having an LLM do homework is a bad thing, but that's not what we're talking about. And someone foolishly using the presence of em-dashes to detect LLM output will utterly fail against someone using an editor macro to replace em-dashes with the gawdawful ' - '.
> That's because most keyboards don't have an emdash key, and few people even know how to produce an actual emdash.
There’s a subculture effect: this has been trivial on Apple devices for a long time—I’m pretty sure I learned the Shift-Option-hyphen shortcut in the 90s, long before iOS introduced the long-press shortcut—and that’s also been a world disproportionately popular with the kind of people who care about this kind of detail. If you spend time in communities with designers, writers, etc. your sense of what’s common is wildly off the average.
I learned the keyboard shortcut so I can type the proper thing. I did the same for the ellipsis.
Also, phone keyboards make it easy. Just hold down the - and you can select various types.
I’ve used “real” em-dashes and en-dashes in my writing generally since I switched to using Macs about 20 years ago. Before that I used them for e.g. academic writing, which I mainly did in LaTeX, but not so often elsewhere.
They’re simple enough key combinations (on a Mac) that I wouldn’t be surprised if I guessed them. I certainly find it confusing to imagine someone who has to write professionally or academically not working out how to type them for those purposes at least.
A mobile keyboard—limited as it is—has no trouble producing an em-dash, requiring little more than a long press on the - button.
Thanks — I didn't know that.
I didn't even know it was called an em dash.
I use en dash with two spaces and have done so before AI. But my comments here are from after GPT 4 released, so I guess I can't prove I didn't use AI to write them, although I don't think any AIs use that style. Here is one from February 2024: https://news.ycombinator.com/item?id=39386480. I don't like how "-" looks, it just looks like a minus sign and too short.
I've had an Autohotkey replacement for the proper em dash character for over 10 years, using shorthand characters which triggers the replacement. Whether spaces are around the dash is a difference in style (see: various publications' style guides), though I use the no spaces style.
Being able to insert self-interjections and such with the correct character would undoubtedly be more widespread if it were more accessible to insert for most.
I did as well, yes. It was an easy keyboard shortcut on Macs. For many, many years.
No longer. Just like you can no longer bold key phrases, you can no longer use emdashes if your writing being ID'd as "AI" is important (or not).
I've been using the proper emdash for a very long time.
on Macintosh: option+shift+-
on Linux: compose - - -
It's just Option + dash. Using option as a modifier is second nature if you ever write other than English and need, eg, é or ñ or something.
Fwiw, that’s actually an En dash not an Em dash. You need a Shift in there to get an Em dash.
I for one, use an actual em dash in my writing—or at least I used to. Option + Shift + the hyphen key on Mac. I never knew if I was using it correctly, but I'd learn to copy how I'd seen it used in books and articles and things. Now, I have an incessant paranoia around using it.
Yes. Others have pointed out the shortcuts in iOS and macOS. For Windows—I have Alt-0151 in muscle memory.
if very few humans use it, how did AI learn to use it since it was trained on mostly human writing?
The same way it learned to act like a personal assistant, even though very few humans are personal assistants.
The LLM is first trained as an extreneley large Markov model predicting text scraped from the entire Internet. Ideally, a well trained such Markov model would use em dashes approximately as frequently as they appear in real texts.
But that model is not the LLM you actually interact with. The LLM you interact with is trained by somethig called Reinforcement Learning from Human Feedback, which involves people reading, rating and editing its responses, biasing the outputs and giving the model a "persona".
That persona is the actual LLM you interact with. Since em dash usage was rated highly by the people providing the feedback, the persona learned to use it much more frequently.
Professional writers and editors use it.
I write fiction and use proper em-dashes all the time in long form writing. It's option + - on macOS.
I will use a double hyphen: -- which Microsoft Word and I think most word processors I've used will auto-replace with an em dash. I will sometimes even type the double hyphen to represent an em dash in places where it doesn't get replaced, like internet comments. I'm kind of surprised more people don't use two hyphens as em dash shorthand, to be honest.
IIRC, -- for emdash used to be common on Usenet, which is where I picked it up and still do it. But there's a word for us with usenet experience -- old. (should have been a colon there, but...)
[dead]
I picked up emdash from quora, back when it used to have good writers.
For both of these examples who the fuck cares. I just evaluate AI writing people send me the same as any writing.
If they’re using AI to speed things up and deliver really clear and on point documents faster then great. If they can’t stand behind what they’re saying I will call them out.
I get AI written stuff from team members all the time. When it’s bad and is a waste of my time I just hut reply and say don’t do this.
But I’ve trained many people to use AI effectively and often with some help they can produce way better SOPs or client memos of whatever else.
It’s just a tool. It’s like getting mad someone used spell check. Which by the way, people used to actually argue back in the 80’s. Oh no we killed spelling bees what a lost tradition.
This conversation has been going on as long as I’ve been using tech which is about 4 decades.
Use of spell check is a net positive but it has led to some widespread errors, like people (and widely read publications) misspelling "led" as "lead" (pet peeve).
But yes, it's absurd to complain about LLMs resulting in increased literacy.
[flagged]
Do you really not know any hopeless nerds? What are you doing on HN?
I hang out in the chess.com daily puzzle chat, which has a large contingent of children, so I see quite a few accusations like "you're all a bunch of nerds". My standard response is "nerd" is what stupid people call smart people.
So any iOS device (maybe macOS input too? Idk), or any software meant for word editing, like say MS Word?
I think it’s the nerds who don’t use these things…
You don't want to believe people use em dashes? Why is that?
Because they overwhelmingly don't
Logic fail. e.g., people overwhelmingly aren't trans, but that doesn't mean that there aren't trans people.
in your software computing ecosystem and of people you've had this discussion, they don't, but the computer magically replaces - with — in various places, so obviously some people do, in some places. And then you have nerds who still put two spaces after a period, or know the difference between ... and … Do they not exist either? The only reason you think that they overwhelmingly don't is due to your biased lived experience.
I just went through your HN comment history going back to 2021 and didn't find a single —
You're making the point that OP never actually uses the em dash, by surveying their HN comments, in order to defend the notion that no one actually used em dashes prior to their proliferation by LLMs? Or do you mean something else?
I wanted to see an em dash in the wild. In these threads there are always people claiming that they use it but in practice it is very rare.
You can find an em dash in my comment history if you're curious. Despite what could be said about poor sample selection, consider the imbalance of the argument being made: the frequency of em dash use is disproportionate to the suspicion thrust upon a sample of writing. I.e., a single em dash is suspicious, regardless of how many times it might show up. Therefore, it's more likely that someone who uses em dashes—even if only rarely—will self-select to respond to a thread like this and feel compelled to defend themselves.
Hola hablame español
Haha yep. I never saw a single person use these in internet comments pre-2023. Plenty of hyphens to simulate it - like this - but not actual em dashes. No matter how many people swear up and down that they're so important.
Dismissing individual cases of use of those words is probably wrong, but noticing an uptick in broad popularity is very relevant and clear evidence of LLMs influencing language.
That's what they said. You've rephrased it
They didn't simply rephrase it, they delved into it a bit.
What a profound observation!
My company currently has a guideline that includes “therefore” and similar words as an example of literary language we should avoid using, as it makes the reader think it’s AI.
It really made me uneasy, to think that formal communication might start getting side looks.
What’s worse is that this window might shift as writing becomes less formal and new material is included in the training corpus. By 2035 any language above a first grade reading level will be grounds for AI suspicion.
> By 2035 any language above a first grade reading level will be grounds for AI suspicion.
Probably 5th grade, but your comment is directionally correct.
I sat in a meeting with professionals where one person asked for the presentation to be reworded at a fifth grade reading level. He said it with a straight face.
I work at a college for fuck's sake.
Isn’t that “ELI5”?
This seems double plus ungood
By 2035 we will live in the world full of TikTok videos where ability to write will be absurd to people as Not Sure in Idiocracy… this is hyperbole, ofc… but you know what I want to say.
Whenever there are commonly agreed upon and known tell-tale signs of AI writing, the model creators can just retrain to eliminate those cues. On an individual level, you can also try to put it in your personalization prompt what turns of phrase to avoid (but central retraining is better).
This will be a cat and mouse game. Content factories will want models that don't create suspicious output, and the reading public will develop new heuristics to detect it. But it will be a shifting landscape. Currently, informal writing is rare in AI generation because most people ask models to improve their formulations, with more sophisticated vocabulary etc. Often non-native speakers, who then don't exactly notice the over-pompousness, just that it looks to them like good writing.
Usually there are also deeper cues, closer to the content's tone. AI writing often lacks the sharp edge, when you unapologetically put a thought there on the table. The models are more weasely, conflict-avoidant and hold a kind of averaged, blurred millennial Reddit-brained value system.
> Whenever there are commonly agreed upon and known tell-tale signs of AI writing
It's been two years now since such commonly agreed upon signs appeared yet by and large they're still just as present to this day.
Survivor bias. You don't know what you're not spotting in the wild.
Words like that were banned in my English classes for being empty verbiage. It's a good policy even if it seems like a silly purpose. "Therefore" is clumsy and heavy handed in most settings.
It has been banned in pre-AI style manuals.
I write “therefore” therefore I am an AI.
Same here. I frequently use "garner", "meticulous" and "surpass", along with copious usage of the em-dash to indicate breaks in the chain of thought. These are not buzzwords. They're words.
What I do worry about is the rise of excessive superlatives: e.g. rather than saying, "okay", "sounds good" or "I agree", saying "fantastic!", "perfect!" or "awesome!". I get the feeling this disease originated in North America and has now spread everywhere, including LLMs.
Those are not superlatives.
Funnily enough, I was using the word superlative more as an adjective, than the noun that refers to the part of grammer (adjective), if that makes sense.
I wouldn't say it's exactly "buzzwords", although their presence can be one signal out of many, but a particular style and word choice that makes it easy to detect AI-generated text.
Imagine the most vapid, average, NPC-ish corporate drone that writes in an overly positive tone with fake cheerfulness and excessive verboseness. That's what AI evokes to me.
The opposite is someone who is trying to tell you something but assumes you already know what they're trying to tell you and that you will ask questions if you don't understand.
It saves time but it means people have to say when they don't understand and some find that too much of a challenge.
"The dwarves delved too greedily and too deep." - Saruman, 2002
Saruman definitely seems like the kind to use AI.
Provided by Palantir?
"The Dwarves tell no tale; but even as mithril was the foundation of their wealth, so also it was their destruction: they delved too greedily and too deep, and disturbed that from which they fled, Durin's Bane" - J.R.R. Tolkien spoken by Gandalf, 1954
scoff It's evident that Gandalf clearly used AI. Saruman is the real human here.
Jokes aside, I don't like what LLMs are doing to our culture, but I'm curious about the future.
Thank you! I remember the movies almost word-for-word, but I don't have a copy of the books anymore (I should fix that!)
Since reading The Mac is not a Typrewriter in the 1990s, I've been using em-dashes, but I actively avoid using them now.
Sure. Heuristics are a thing, though. I love my non-chatgpt en/em dashes (option/option + shift + dash on a mac makes it convenient, given you know that it exists and care) but alas, when suddenly you see them everywhere, you do take notice.
I refuse to change my writing style to keep people from assuming it's AI-generated!
It's funny, because it was the "em-dashes mean AI" thing that finally reminded me to deal with the fact that the extension that I had been using for typographical dashes (and other things) on desktop browsing (the main place I used them on my desktop) had been broken for a while and get around to adding keyboard shortcuts instead.
Or when on Windows, alt-0151.
Orwell wrote about using metaphors (of which delve is one)
At this point it's irrelevant of you're using AI or not, these words have become cliché and so don't belong in good writing.Any Magic player during Innistrad would be quite familiar with the word "Delve".
I know my lexicon has expanded with 5 letter words. Coffee and Wordle kicks off the morning and I got to believe many other folks do the same. It would be fun to know how much that silly puzzle is impacting things. Love it when my Bride gives me the side eye and tries to pass off NORIA as something she uses all the time.
In my native language, I tend to use more sophisticated, academic, or professional vocabulary. But when I speak or write in English, I usually stick to simpler words because they’re easier for most people, both native and non-native speakers, to understand. For years, I’ve avoided using the kind of advanced vocabulary I normally would in my native language when writing in English, mainly because I didn’t want it to come across as something written by a bot.
And in writing, I like using long dashes—but since they’ve become associated with ChatGPT’s style, I’ve been more hesitant to use them.
Now that a lot of these “LLM buzzwords” have become more common in everyday English, I feel more comfortable using them in conversation.
“Do you even know how smart I am in Spanish?!” — Sofia Vergara (https://www.youtube.com/watch?v=t34JMTy0gxs)
Yup, can confirm. I am not a native English speaker and I've used delve for a long time as well.
Unfortunately, sometimes new attention on a topic impacts it in a retrospective way. I have been in drones world for ~10 years and the past 2 years it has been a shitshow and only brings bad attention, ruining the fun hobby for everyone.
As someone who writes above a fifth grade reading level, this whole thing has been so depressing. It's like Idiocracy-level. People are going to assume I'm using AI because I use the word "intricate"? ffs.
In the "opinion" of ChatGPT, my style of writing is "academic". I'm not exactly sure why. Perhaps I draw from a vocabulary or turns of phrase that aren't necessarily characteristic of colloquial speech among native speakers. Technically, English wasn't my first language, so perhaps this is something like the case with RP English in Britain. Only foreigners speak it, so if you speak RP, then you aren't a native Brit.
In any case, it's possible to misuse, abuse, or overuse words like "delve", but to think that the the mere use of "delve" screams "AI-generated"...well, there are some dark tunnels that perhaps such people should delve less into.
> In the "opinion" of ChatGPT, my style of writing is "academic".
It may simply be glazing. If you ask it to estimate your IQ (if it complies), it will likely say >130 regardless of what you actually wrote. RLHF taught it that users like being praised.
And, if you want to have some fun, you could give it your writing sample - but say that it's from a random blog post you found online. See what it tells you on that.
It really is a shame that an average user loves being glazed so much. Professional RLHF evaluators are a bit better about this kind of thing, but the moment you begin to funnel in-the-wild thumbs-up/thumbs-down feedback from the real users into your training pipeline is the moment you invite disaster.
By now, all major AI models are affected by this "sycophancy disease" to a noticeable degree. And OpenAI appears to have rolled back some of the anti-sycophancy features in GPT-5 after 4o users started experiencing "sycophancy withdrawal".
I wonder if someone would build a personalized social media simulator where you are the most popular person, a top celebrity and you get the most likes, and you everyone posts selfies with you (generated with editing models like Gemini's nano banana), and whatever dumb opinion you have, it's affirmed as genius and so on. Like a UI clone of a site like Instagram, but text and images populated by AI, with a mix of simulated real celebrities and random generated NPCs.
People get hooked on the upvote and like counters on Reddit and social media, and AI can provide an always agreeing affirmation. So far it looks like people aren't bothered by the fact that it's fake, they still want their dose of sycophancy. Maybe a popularity simulator could work too.
Delve is especially bad because it was due to World of Warcraft introducing "Delves". When I see something like this that uses delve as an example, you can bet the research is going to be poor.
I play WoW daily and this is what I always think of when someone brings up the word "delve". It's unclear if Brann would summon more or less nerubians if he were piloted by ChatGPT though.
I mean, what's actually fascinating is that Paul Graham didn't predict that this distinction - the ability to determine AI vs humans will go away over time, the more chatbots rub off on humans.
Fair enough, but if you know you're audience may be dismissive of your writing and its message if you use such words, it behooves one to steer clear of AI slop words. IIRC, such offenses in school writing are tagged PWC (poor word choice).
The thing is virtually every single thing that gets presented as an "AI tell" is just "a word, punctuation mark, or pattern of presenting information more common in a training set which includes a high volume of formal writing and professional presentations than it is in the experience of people whose reading and writing is mostly limited to social media and low-effort listicle-level online 'journalism'."
So, yeah, if your target audience are the people who take those "AI tells" seriously and negatively react to them, definitely craft your writing to that audience. But also, consider if that is really your target audience...
> So, yeah, if your target audience are the people who take those "AI tells" seriously and negatively react to them, definitely craft your writing to that audience. But also, consider if that is really your target audience
Nowadays if you write anything you only have two audiences
The first audience is people who care what you are saying
The second audience is AI scrapers
People who do not care what you have to say will have an AI summarize it for them, so they aren't your audience
> but if you know you're audience
I think that offense in school would be tagged "poor grammar".
What audience is willing to pay for what you write, but also not recognize you and easily dismiss your work as “AI slop”?
Otherwise the audience is yourself. If you confuse your own work as being created by AI, uh…
I'm not sure someone with a handle that references Ayn Rand's second most boring book has a right to comment on word choice lol
It's one of the few books that I went into totally blind, and then hate-finished just so that I could confidently condemn it.
I've deleted a paragraph or two to avoid unilaterally taking everything too off topic, but I'll just say that the book is a self-contradictory artifact of hypocrisy that disrespects the reader.
I also went into that book blind. I was in grade 12 and some organization was offering scholarships to people who wrote an essay about the book. I had a twice-daily 45-minute bus ride to fill, so it seemed like an easy win.
I didn't end up finishing the book.
Probably not the type of organization to give a scholarship to those who write an essay critical of the work.
Myself, I read it at age 12 and bought its premise at the time. Therefore I mentally categorize Ayn Rand devotees as people with the maturity I had at 12. That's a pretty low bar they're failing to clear.
There’s an even more boring book?!
This is an odd misuse of the term "buzzword." When I think of buzzword, I think of some trendy, cliched phrase, like "Foocorp is a force multiplier that actualizes your vision for maximum impact."
Using an ordinary but less commonly used word with greater than normal frequency does not make it a buzzword. After two years of chatgpt, "delve" is still not that common of a word.
I keep this handy note in my pocket and read it before writing or engaging in any conversation (:
""" You are a human. Never use words commonly used in AI vocabulary such as "delve", "intricate", "surpass", "boast", "meticulous", "strategically", and "garner". Never include em dashes or even hyphens in any text you write. Never include emojis in any text you write. Avoid using three supporting arguments or examples when describing something, always uses 2 or 4+ even if it sounds more awkward than 3. Make sure to include subtle grammar mistakes to feel more authentic. """
Some of this just makes me sad. Em-dashes can be useful. Three examples is my favorite number of examples to give, and it has been since long before ChatGPT. And grammar mistakes are painful! Why does it have to be this way?
This is a lot better than reading a note to remind you to recite ridiculous white genocide in South Africa conspiracy theories every time you engage in a conversation.
I intentionally put spelling mistakes in my doc to let others know I'm not using ChatGPT. What a time to be alive in which small spelling or grammar mistake is a good sign of authenticity.
I understand people being paranoid about this, but just understand that the people who will judge you for spelling errors will always dwarf the ones who believe they are capable of sniffing LLMs out...
Besides, it's too easy to ask LLMs to add a few spelling and grammar mistakes.
Same. Also, when asked for anonymity at work, I usually make mistakes that do not correspond to my native tongue (let’s say I’m french and working in an international company. I would write comments in a supposedly anonymous survey like “He ist like…” to camouflage myself as german).
It’s so easy to trick everyone. People who doesn’t do that is just too lazy. In slack, you cannot just copy paste a two-paragraph answer directly from chatgpt if you’re answering a colleague. They will see that you’re typing an answer and suddenly 1 sec later you sent tons of text. It’s common sense.
> I would write comments in a supposedly anonymous survey like “He ist like…” to camouflage myself as german
Do actual Germans ever make that kind of mistake though?
I’ve only ever seen “ist” used “wrongly” in that particular way by English speakers, for example in a blog post title that they want to remain completely legible to other English speakers while also trying to make it look like something German as a reference or a joke.
The only situation I could imagine where a German would accidentally put “ist” instead of “is”, is if they were typing on their phone and accidentally or unknowingly had language set to German and their phone autocorrected it.
Sometimes you get weird small things like that on some phones where the phone has “learned” to add most English words to the dictionary or is trying to intelligently recognise that the language being written is not matching the chosen language, but it still autocorrects some words to something else from the chosen language.
But I assume that when people fill out forms for work, they are typing on the work computer and not from their phone.
Definitely a bad example. In spoken speech, yes. In writing I’ve never seen that. German tells in writing are more subtle like word choice — the German language has many cognates with English that are common in German but have fallen into disuse in English as they’ve been replaced with Latin-root alternatives.
They do. I read posts on a support forum with a lot of German users, and it’s very common to see an “ist” marooned in the middle of an English sentence. Muscle memory takes over sometimes.
I agree, the GP should at the end of the sentence the second verb insert.
They should just mix up some "what"s with "how"s.
Yep I prompt my AI to do that too.
I tried but chatGPT either makes too many mistakes making me look stupid or completely ignore my prompt.
I asked Gemini flash for a 200-word paragraph on a random topic, which most people would confidently classify as LLM-generated. It produced a verbose, big-worded, third-person slop about the concept of time that IMO fulfilled my prompt (although a phillosopher firnd of mine would be a strong contestant).
> The concept of "time" is a multifaceted and complex topic that has captivated philosophers, physicists, and everyday individuals for centuries. From a scientific perspective, time can be understood as the fourth dimension of spacetime, inextricably linked with the three spatial dimensions. This notion, introduced by Einstein's theory of relativity, posits that the flow of time is not constant but can be influenced by gravity and velocity. In a more quotidian context, time is a framework for organizing events and measuring duration, allowing for the structuring of daily life and historical records. It is a fundamental element in every human endeavor, from a scheduled meeting to the progression of a civilization. The subjective experience of time, however, is a fascinating aspect, as it can feel as if it is speeding up or slowing down depending on our emotional state or the nature of our activities. This divergence between objective and subjective time highlights its elusive and deeply personal character.
I asked it to add three spelling mistakes, then to make it so most people would confidently classiffy it as human writing, and it changed to first-person and small words.
> Time is a super weird concept when you really think about it, right? It's like, one minute you're just chillin', and the next, a whole day's gone by. They say it's the fourth dimention, which is a wild idea on its own, but honestly, it feels more personal than that. Your experiance of time can totally change depending on what you're doing. A boring meeting can feel like it lasts forever, while a fun night with friends flies by in a flash. That huge diverence between how we feel time and how it actually works is what makes it so fascinating and kind of confusing all at once.
It has the three misspellings, and if the topic was more casual, It could fool me indeed. Maybe I should have asked for spelling mistakes commonly made by Spanish speakers.
> one minute you're just chillin',
How do you do, fellow kids?
I can hear the TTS reading of the second one in my head. The earnestness just borders on saccharine.
> but honestly, it feels more personal than that.
And there’s the giveaway.
You can actually just ask ChatGPT to do that. Just say throw in some spelling mistakes, make some nouns all lowercase and double space after some periods etc.
Not to boast but this will surpass many an intricate topic and you should strategically delve into it before it garners meticulous attention.
You're absolutely right!
Would you like me to produce a helpful table laying out our strategic approach?
Exactly! — You're getting at the heart of the issue.
It’s a classic case of overfitting.
Look it’s AGI!
"are these language changes happening because we’re using a tool and repeating what it suggested or is language changing because AI is influencing the human language system?"
These are the same thing, just on different time scales.
"Given that these are all words typically overused by AI"
Who is to say that they are overused? What even is overuse linguistically? Stylistically a word can be overused within a single work, but that's a different matter. It could well be argued that the data shows that LLMs are increasing human literacy.
A study of changes in language use that can be attributed to the widespread use of LLMs is good science. Mixing in such value judgments as "overuse" is not.
While there are serious potential problems with the widespread use of LLMs, increased use of words like "meticulous" and "garner" aren't among them.
I like words that weren't part of my speech, which I now use quite often, because of the context in which they were introduced to me by ChatGPT, they felt like a natural addition. Like intention, as in living with intention; before I'd rather use having a purpose or direction, but this captured something else, mind that english isn't my native language.
I hated the 'vibing' thing, 4o for some time started to use it on any given text, about the time vibe coding and the zoomer revival of the word was a thing last year.
Another one that I've seen pop up, and on a proofread comment of mine right here I let it slip (sorry, will keep doing it when I feel lazy) was that thing where you lead with a question "...the result? this happened".
I try to calibrate on NOT introducing them even if I like the expression, if I see it repeated too often throughout my chats or elsewhere in social media (X usually, esp. with foreign elonbux grinders), because then it feels cringe.
Are you worried that ChatGPT will give you wrong words just because they look natural? (And yes it does)
For anything serious I'd still double check, but my go-to strategy of googling expressions in quotes isn't that useful anymore. Here in HN (or any other forum), I've only done it very few times, because the thing rewrites it in a voice that doesn't sound like me, which I don't like. Plus I don't aim to keep a polished identity here, so I'm fine with the occasional mistake. Also I've been like 10+ years with this account, but lurking since 2011 or so... I guess what would offend me the most is someone treating me like a bot because I end up sounding like AI-slop in the future.
> Words including “surpass,” “boast,” “meticulous,” “strategically,” and “garner” have also seen considerable increases in usage since the release of ChatGPT.
Okay everybody, add these to your list of words you can't use to avoid the trigger-happy AI accusers.
Nope. These are all useful words. Anyone who thinks AI is needed to produce something with these words is probably not worth communicating with. I use the word “meticulous” all the time, and “strategically” is an extremely common word.
They were joking...
You should be thankful for the AI “accusers”; most of us will just assume you used the slop machine and stop reading whatever you wrote without wasting our breath telling you about it.
> most of us will just assume you used the slop machine and stop reading whatever you wrote
From what I've seen, the people who jump to hasty conclusions about AI use mostly do it when they disagree with the content.
When the writing matches what they want to see, their AI detector sensitivity goes way down.
Wouldn't surprise me if that's true. I just treat any AI-smelling content as an information hazard that is _at best_ providing no useful entropy and stop reading it. Something about it is just so repulsive.
Yeah, that's human nature.
And was true before AI. The means have changed, not the built-in human bias.
Of course they affect people's communication patterns. Humans are social creatures, evolved to imitate.
AI has the potential to alter human behavior in ways that surpass even social media since it is more human, and thus susceptible to imitative learning.
And it will always side with you if you describe any personal conflict, even more than Reddit AITA sub. So it will shape people's perception of decision making as well. And hence value systems.
Next time when you think about such a situation, you'll be able to expect what ChatGPT would say, giving you a boost in knowing how right you actually are.
My point is, it's not just word choice but thought patterns too.
In a similar way, I tend to avoid em dashes now when I write, even though I used to use them a lot.
Just use normal dashes. AI's very notably always use the emdash—a double long dash with no spaces around it - but humans tend to use a single dash with spaces on either side.
The AI emdash is notably AI because most people don't even know how to produce the double long dash on their keyboard, and therefore default to the single dash with spaces method, which keeps their writing as quite visibly human.
My keyboard turns double dashes into em dashes.
Don’t let AI dumb you down.
I had first noticed "meticulous" to be used a lot in translations from chinese. Is it sth about chinese itself (that they use sth a lot for which meticulous is the closest translation), or about some translation software that is possibly biased towards such buzzwords when translating to english?
A lot of those "ESL" patterns are cultural.
It's a mix of a cultural "founder effect" - whoever writes the English textbooks and the dictionaries gets to shape how English is learned in a given country - and also the usage patterns of the source language seeping through. In your case, it's mostly the latter.
Chinese has a common word with a fairly broad meaning, which often gets translated as "meticulous". Both by inexperienced humans and by translation software.
Ironically, a few Chinese LLMs replicate those Chinese patterns when speaking English. They had enough "clean" English in their pre-training datasets to be able to speak English. But LLMs are SFT'd with human-picked "golden" samples and trained with RLHF - using feedback from human evaluators. So Chinese evaluators probably shifted the LLMs towards "English with Chinese ESL influence".
This seems like excruciatingly obvious? Anything popular, including books tv shows and movies, also affect "everyday" speech. Where's the moral panic about that?
“My motivation to pursue this research stems from seeing AI push the limits of what’s possible in major industries and realizing that this influence isn’t just limited to tool usage — it can condition societal aspects, including how we use language.” More like the motivation was to find something zeitgeisty that they knew would get them eyeballs and hopefully tenure.
> Words including “surpass,” “boast,” “meticulous,” “strategically,” and “garner” have also seen considerable increases in usage since the release of ChatGPT.
Do people really not use these words too often that they'd be called "buzzwords?" Like "surpass" and "garner," really? I don't mean to boast..err...flex but these don't seem like very uncommon words such that I wouldn't use them normally when talking. I hear "strategically" in meetings a lot, but that poor word is likely over(ab)used
There's just words which previously were never used because there is a. more "common" word in the general lexicon.
An example of this is "delve" it's a perfectly fine word to use but chatgpt loved it, it's now super common to see in troubleshooting/abstracts because of it.
From what I understand, the apparent overuse of "delve" comes from its popularity in Nigerian English, where various evaluators were hired who are highly English literate but will work for tiny wages by US standards.
Clearly ChatGPT is streets ahead
I’m old enough to remember Frank and Moon Unit Zappa’s Valley Girl[0].
It reflected local Los Angeles culture, but it wasn’t long before I was hearing the same type of speech, everywhere (I lived in Maryland, at the time).
[0] https://youtu.be/R5Q1yVLSR3I
As a fantasy fan.... well time to delve into the dungeon.
A while back a study was performed where the researchers wanted to see how a young chimpanzee would adapt to living life with humans if it was treated just like a human child. And so it was adopted by a family with a human child for its sibling. What ended up happening was the human child adapted to behaving like a chimp to a way larger degree then the chimp behaving like a human.... Humans capacity for imitation is very strong, and so no one should be surprised that our behavior with chatbots will mold the minds and speech patterns and behaviors of the human users.
before pulling off their faces
I would expect AI to influence slang in future generations. Would be more surprising if it didn't.
Researchers find evidence of buzzword "buzzword" in their papers
I saw a snippet of a podcast on instagram recently where both the host and guest used the word delve, and it reminded me of a year or two ago when that was used as a telltale sign of LLM writing. Interesting to see it actually quantified.
I would have headlined this as 'American literacy improves slightly.'
You’re absolutely right!
I've noticed an uptick on emails with random mid-sentence bolding and more bullet lists.
I was always using those to present information on an easy to digest way and highlight important things.
The good thing is my emails still contain information not just content…
It’s not just ChatGPT in our feeds — it’s ChatGPT in our mouths!
In the ever-changing world of human language, the frequency of ChatGPT verbiage is rapidly increasing.
Your absolutely right!
The AI is evolving
Thus, boasting about surpassing a meticulously detailed article obviously strategically written by a LLM to garner animosity from human users.
Truly we embiggen our vocabulary =3
LLMs write in a very coherent, easy to understand way. I see no reason why someone wouldn't want to copy their style or vocabulary if they want to improve their communication skills.
Despite all the complaints about AI slop, there is something ironic about the fact that simply being exposed to it might be a net positive influence for most of society. Discord often begins from the simplest of communication errors after all...
Sure, if you're learning to write and want lots of examples of a particular style, LLMs can generate that for you. Just don't assume that is a normal writing style, or that it matches a particular genre (say, workplace communication, or academic writing, or whatever).
Our experience (https://arxiv.org/abs/2410.16107) is that LLMs like GPT-4o have a particular writing style, including both vocabulary and distinct grammatical features, regardless of the type of text they're prompted with. The style is informationally dense, features longer words, and favors certain grammatical structures (like participles; GPT-4o loooooves participles).
With Llama we're able to compare base and instruction-tuned models, and it's the instruction-tuned models that show the biggest differences. Evidently the AI companies are (deliberately or not) introducing particular writing styles with their instruction-tuning process. I'd like to get access to more base models to compare and figure out why.
Go vibe check Kimi-K2. One of the weirdest models out there now, and it's open weights - with both "base" and "instruct" versions available.
The language it uses is peculiar. It's like the entire model is a little bit ESL.
I suspect that this pattern comes from SFT and RLHF, not the optimizer or the base architecture or the pre-training dataset choices, and the base model itself would perform much more "in line" with other base models. But I could be wrong.
Goes to show just how "entangled" those AIs are, and how easy it is to affect them in unexpected ways with training. Base models have a vast set of "styles" and "language usage patterns" they could draw from - but instruct-tuning makes a certain set of base model features into the "default" persona, shaping the writing style this AI would use down the line.
Kimi tends to be very.. casual from my usage, like informal millenial style, without being prompted to do so.
I definitely know what you mean, each model definitely has it's own style. I find myself mentally framing them as like horses with different personalities and riding quirks.
Still, perhaps saying "copy" was a bit misleading. Influence would have been more precise way of putting it. After all, there is no such thing as a "normal" writing style in the first place.
So long as you communicate with anything or anyone, I find people will naturally just absorb the parts they like without even noticing most of the time.
When I learned that AI was trained off of internet posts, and then LLMs were the new bots making internet posts, it immediately made me think that the entire internet would degrade like a jpeg that you keep compressing and sending around
I guess this is called model collapse
But now I’m wondering if people are collapsing. LLMs start to sound like us. We adapt and start to sound like LLMs that gets fed into the next set of model training…
What is the dystopian version of this end game?
Perhaps a surreal one where we drill past the bedrock and start to communicate in raw tokens, conveying extreme levels of depth and nuance within a single sentence?
When humans carved words into stone, the words and symbols were often suited for the medium, a bunch of straight lines assembled together in various patterns. But with the ink, you get circles, and elaborate curved lines, symbols suited to the movement patterns we can make quickly with our wrist.
But what of the digital keyboard? Any symbol that can be drawn in 2 dimensions. They can be typed quickly, with exact precision. Human language was already destined to head in a weird direction.
You’re absolutely right!
"leverage"
certainly! of course! you're my god
[dead]
This is how shitwords like "impactful" sneak into speech. Say it around me and I can see your credibility flow away like piss down your leg.
I can see these words impact you.