Mostly not. I am a fan of the em-dash, which a lot of people now see as absolute evidence of LLM usage. I occasionally think about swapping it out if I’m desperate not to be mixed up with an LLM, but in principle, I don’t really care if someone thinks I used an LLM to help me write something or not.
I work with a lot of non-native English speakers (with English as a lingua franca) and I’m more than happy for them to use LLMs to help them phrase their thoughts in a way that I can understand more easily.
I also sometimes use LLMs myself for low-stakes stuff, tidying up sloppy notes, etc.
I think it’s a bit Ludditical to want people to always write every word themselves. Should they also hand write it using a quill pen and ink they made themselves from oak galls?
There are some types of writing (creative writing, writing to persuade, etc) where the writing itself benefits from being hand crafted, but most writing is just an imperfect way of sharing thoughts.
But is it their thoughts? Another problem with LLMs is that while it sometimes can produce a better english phrasing of what they were thinking, other times it could be something conceptually different that they didn't catch because of thier english isn't proficient enough to see it, or worse case, they are just lazy and write broad vague prompts and accept whatever blob of text comes out.
I use it sometimes where I just write a stream of conciouness for describing how we can solve a Jira ticket. Then I let LLM edit it. It doesnt add anything or remove any pount. It just moves things around and make the text more clear by rephrasing it.
Not a day goes by where I don't mourn the desecration of the em-dash, but its frustrating for people that write “authentcally” (referring to academic work) to get accused of AI writing, when something is well phrased and carefully thought out. I do agree on the fact that its quite Luddite to expect people to slave away at a document outside of an academic context, using AI does make you much more efficient
I use em-dash all the time — and wish now I had not read that it is popular with AI. (I don't suppose I'll change though, for some dumb reason I like parens as well. Mimic me there, AI.)
Em-dash fan here as well. I've been watering down my use of the em-dash, because I don't want folk to think that my (always self-written) business emails, etc are a chatGPT job.
One potential future I could see with generative AI and fiction is a sort of infinite interactive fiction where the reader plays a more active role in exploring the space. Maybe they have some control, or maybe they can just ask the AI where to look next. And at least in that case, it would clearly be AI doing the work. Not that this would stop people from producing AI novels, or anything.
In my writing, I just go for it. If someone thinks it sounds like an AI, what can I do? (I don't use AI for any of my writing, for the record. What fun would that be?)
I also love using em dashes properly, so I guess that's points against. Vim: ^K-M gets you one.
But when I write something like a letter of recommendation or, heaven forbid, a cover letter, I do actually try to write it so that it sounds human. If someone suspects that the letter was AI generated, it becomes worthless. And I don't like to write worthless letters of recommendation.
> Could there be anything more insulting for a writer than someone assuming that their writing is an output of generative artificial intelligence?
Not sure if this is a bad thing. AI uses very professional and correct language (unless otherwise instructed) with well-structured paragraphs. If a system can't distinguish my writing from AI, it means that I'm doing a great job as humans sometimes make typos and structural/grammatical errors and AI generally excels against those.
Yes, it uses very professional and correct language, but the times I've asked it to help me improve something I've written, most of the suggestions just suck the work dry. It is ignorant of those subtleties of meaning that are so important for proper emotional impact.
Human goal: be a better writer than AI. It's been trained on some good stuff, but also a lot of crap. I'd expect its output to be fairly average.
Sometimes I'll read comment online that does the equivalent of breaking my "willing suspension of diabelief" by sounding just like a standard gpt reply. I even slipped up and commented as much, which is not in itself an interesting comment , just because it was so jarring to read. But idk what it is about reading something lexically equivalent to injection molded plaatic that feels so annoying, plastic is great! But maybe it's similar to the feeling of buying an upcharged "branded" Alibaba product on Amazon. I can't trust the reseller on the basis of their effort because I don't know them and it's just as likely they put 0 effort into quality control (thoughtful selection, validation).
But in the end, like the article implies, any comment could be "manufactured."
All a question of "authenticity", which is always in demand in conversation, travel, food, music etc. Resulting in a complicated dance of trying to mass produce authenticity without bursting the illusion.
AI cannot be authentic, but it can look close enough to authentic for people who aren't paying deep attention.
Neither can the human slop writers, the internet turned to a stinking swamp full of slop content long before AI, what AI did was to actually improved the baseline slop quality by a huge margin(but an ad for weight loss is going to be slop even with high production values, just like the latest capeshit movie)
a subject that is close to my heart as I've been accused recently of being AI by someone who evidently didn't like what I wrote.
Admittedly it did make me chuckle that they thought AI sounds like a depressed character in science fiction, which was what I was writing. Poor AI.
I do use my share of em-dashes but hardly ever use the word delve, unless discussing something that Adam and Eve did, or if I were to write about Dwarves, Kobolds, Lovecraftian horrors and their worshipers because in those cases I would expect delving to occur.
I've mostly escaped it by speaking Australian English, which scatters text with little archaisms like "reckon" and prefers an en-dash with spaces around it over an em-dash.
I also have faced this accusation. I believe it's indicative of younger people who haven't been exposed to older people who have pompous writing styles, like I unquestionably do. Like using idoneous words such as "unquestionably", or "idoneous"
Em dashes are just an affectation.
Patrick O'Brien (aubrey/maturin) used "idoneous" which is how I learned it, and he was both pompous, and erudite. And plain when it suited him: he translated "papillon" as well as writing lit and biography.
The distinction between an em dash, an en dash and a hyphen is quite distinct from the general use of punctuation. It's more akin to '' vs " and `' in some respects. I have some sense 《 and ¿ are at least functional in their native languages.
But for lots of people being unemotional is normal. I kind of get it, but I think this is putting the cart before the horse. I mean Hemingway.. all those short manly sentences..
This is unequivocally NOT a Turing test moment. Simply parroting forms of writing which leads some people to accuse other people of being machines does not mean the thinking component is present.
> I believe its indicative of younger people who haven't been exposed to older people who have pompous writing styles.
That, or reading comprehension has dropped below a point in the young where they can no longer recognize or perceive intelligence, and so conflate AI with intelligence much like what is described in the Allegory of the Cave.
Lots of people are calling intelligent well read people AI because the accusers worldview can't distinguish those that are intelligent from those that are AI, or as an imposition of cost the same as any invective. Its just a new form of gaslighting.
You can't possibly be that intelligent so you must be AI. /s
We live in a wide world, which they don't realize because they've been coddled, and often lied to, where the lie traversed more than a generation. Brittle people break.
I assure you I am not that intelligent. Erudite sentences from me are very much repetition of things seen. In that narrow sense, they may be closer to an llm than I like.
"I'm sorry that writing like this is so beyond your reach that you can't imagine anyone doing it without a program" is one of those retorts which, regretfully, only works well in my head.
I used to use a grammar and punctuation checking browser plugin that worked in any text box. Among all the normal suggestions, it would always insist that instead of using a - dash, that it was proper to use an — emdash. I would accept the suggestion until people on Reddit accused me of being chatgpt because I wrote "too properly" with emdashes. So now you can't do that anymore.
if they cannot make critiques of your content/comment without resorting to appeal to authority or ad hominem attacks (which i consider being accused of using chatgpt as one), then their "critique" is worthless and should just be ignored (with a comment saying so as a response).
I literally do not care that someone is using chatgpt, or anything generated, as a response or comment, or content. The content _itself_ speaks for itself, and the worthiness stands alone regardless of how it was made. Slop can be easily written by an author or ai, and pedigree has nothing to do with it.
I think the fear of being confused for AI is a generational thing. Wait until the first "AI native" generation takes over and AI is the primary means by which people communicate with one another and interact with the world, the very concept will seem alien and absurd.
Of course "writing" itself will likely become passe as literacy drops with attention spans, and most AI interaction is done through dynamically generated video and audio. That process is already underway.
AI writing will be impossible to distinguish from human writing in a year or two. The weird crappy way it writes now will be adversarially driven out of it in the next LLM releases, and probably gone in one or two more.
No more lists of 3 things, no more emdash, no more vacant live laugh love level vapid niceties.
I don't think or at least hope so. The corporate desire to sanitise LLMs for "safety" (for PR/liability) and making it agreeable (for engagement) will hopefully leave AI with a recognisable non-human style.
I wish I knew how to (ask an AI to) characterize this style. Is it always three examples? Nice formatting and bullet points? Evenly issued, lengthy and didactic explanations?
When I write, I now go out of my way not to use lists or em dashes. I learned before LLMs, that using lists in writing was lazy. Even when I do use an LLM for writing, I tell it to explain everything in paragraph form. I start off with my own outline and then “discuss” what I want to emphasize.
The second thing about LLM generates text especially when doing technical writing, is that anytime it explains something, even with a lot of context of what I’m working on, it always adds lines about “benefits”.
In my use case something like “this provides a secure, adaptable environment…”. I have had to remove wording I had like that on my resume before LLMs and definitely don’t put it in my own writing now.
Ironically enough, out of the four examples ChatGPT generated about “what makes a good leader” in the link above, the example of “AI generated”, is the one I would lean toward in my own writing.
Also, putting AI assisted writing back through a new session with ChatGPT and asking it does it sound AI generated multiple times, and taking its suggestions will make it sound less like it’s AI generated.
When I do write something “thought leader”ish which unfortunately is part of my job now as a staff architect, I give a lot of real world examples that couldn’t easily be fabricated by AI and lean on those examples as I make my points.
AI writing is good, only people who cry about AI are people afraid it will take their jobs, or people who think AI slop is the default output as opposed to the product of a human with shitty taste using AI.
All these tears and tantrums over AI are so tiresome.
Mostly not. I am a fan of the em-dash, which a lot of people now see as absolute evidence of LLM usage. I occasionally think about swapping it out if I’m desperate not to be mixed up with an LLM, but in principle, I don’t really care if someone thinks I used an LLM to help me write something or not.
I work with a lot of non-native English speakers (with English as a lingua franca) and I’m more than happy for them to use LLMs to help them phrase their thoughts in a way that I can understand more easily.
I also sometimes use LLMs myself for low-stakes stuff, tidying up sloppy notes, etc.
I think it’s a bit Ludditical to want people to always write every word themselves. Should they also hand write it using a quill pen and ink they made themselves from oak galls?
There are some types of writing (creative writing, writing to persuade, etc) where the writing itself benefits from being hand crafted, but most writing is just an imperfect way of sharing thoughts.
> use LLMs to help them phrase their thoughts
But is it their thoughts? Another problem with LLMs is that while it sometimes can produce a better english phrasing of what they were thinking, other times it could be something conceptually different that they didn't catch because of thier english isn't proficient enough to see it, or worse case, they are just lazy and write broad vague prompts and accept whatever blob of text comes out.
I use it sometimes where I just write a stream of conciouness for describing how we can solve a Jira ticket. Then I let LLM edit it. It doesnt add anything or remove any pount. It just moves things around and make the text more clear by rephrasing it.
Not a day goes by where I don't mourn the desecration of the em-dash, but its frustrating for people that write “authentcally” (referring to academic work) to get accused of AI writing, when something is well phrased and carefully thought out. I do agree on the fact that its quite Luddite to expect people to slave away at a document outside of an academic context, using AI does make you much more efficient
I use em-dash all the time — and wish now I had not read that it is popular with AI. (I don't suppose I'll change though, for some dumb reason I like parens as well. Mimic me there, AI.)
Em-dash fan here as well. I've been watering down my use of the em-dash, because I don't want folk to think that my (always self-written) business emails, etc are a chatGPT job.
One potential future I could see with generative AI and fiction is a sort of infinite interactive fiction where the reader plays a more active role in exploring the space. Maybe they have some control, or maybe they can just ask the AI where to look next. And at least in that case, it would clearly be AI doing the work. Not that this would stop people from producing AI novels, or anything.
In my writing, I just go for it. If someone thinks it sounds like an AI, what can I do? (I don't use AI for any of my writing, for the record. What fun would that be?)
I also love using em dashes properly, so I guess that's points against. Vim: ^K-M gets you one.
But when I write something like a letter of recommendation or, heaven forbid, a cover letter, I do actually try to write it so that it sounds human. If someone suspects that the letter was AI generated, it becomes worthless. And I don't like to write worthless letters of recommendation.
> Could there be anything more insulting for a writer than someone assuming that their writing is an output of generative artificial intelligence?
Not sure if this is a bad thing. AI uses very professional and correct language (unless otherwise instructed) with well-structured paragraphs. If a system can't distinguish my writing from AI, it means that I'm doing a great job as humans sometimes make typos and structural/grammatical errors and AI generally excels against those.
The thing is, I think it's possible to outdo AI.
Yes, it uses very professional and correct language, but the times I've asked it to help me improve something I've written, most of the suggestions just suck the work dry. It is ignorant of those subtleties of meaning that are so important for proper emotional impact.
Human goal: be a better writer than AI. It's been trained on some good stuff, but also a lot of crap. I'd expect its output to be fairly average.
And I have seen already people dismissing a thing as "AI slop" without any evidence.
Writer's have always needed thick skins in order to deal with criticism, this is just another insult they need to shrug off.
Sometimes I'll read comment online that does the equivalent of breaking my "willing suspension of diabelief" by sounding just like a standard gpt reply. I even slipped up and commented as much, which is not in itself an interesting comment , just because it was so jarring to read. But idk what it is about reading something lexically equivalent to injection molded plaatic that feels so annoying, plastic is great! But maybe it's similar to the feeling of buying an upcharged "branded" Alibaba product on Amazon. I can't trust the reseller on the basis of their effort because I don't know them and it's just as likely they put 0 effort into quality control (thoughtful selection, validation).
But in the end, like the article implies, any comment could be "manufactured."
All a question of "authenticity", which is always in demand in conversation, travel, food, music etc. Resulting in a complicated dance of trying to mass produce authenticity without bursting the illusion.
AI cannot be authentic, but it can look close enough to authentic for people who aren't paying deep attention.
>AI cannot be authentic
Neither can the human slop writers, the internet turned to a stinking swamp full of slop content long before AI, what AI did was to actually improved the baseline slop quality by a huge margin(but an ad for weight loss is going to be slop even with high production values, just like the latest capeshit movie)
An excellent question! First of all...
a subject that is close to my heart as I've been accused recently of being AI by someone who evidently didn't like what I wrote.
Admittedly it did make me chuckle that they thought AI sounds like a depressed character in science fiction, which was what I was writing. Poor AI.
I do use my share of em-dashes but hardly ever use the word delve, unless discussing something that Adam and Eve did, or if I were to write about Dwarves, Kobolds, Lovecraftian horrors and their worshipers because in those cases I would expect delving to occur.
I've mostly escaped it by speaking Australian English, which scatters text with little archaisms like "reckon" and prefers an en-dash with spaces around it over an em-dash.
>little archaisms like "reckon"
"Fortnight" is by far my favourite, at least when dealing with my American colleagues.
I also have faced this accusation. I believe it's indicative of younger people who haven't been exposed to older people who have pompous writing styles, like I unquestionably do. Like using idoneous words such as "unquestionably", or "idoneous"
Em dashes are just an affectation.
Patrick O'Brien (aubrey/maturin) used "idoneous" which is how I learned it, and he was both pompous, and erudite. And plain when it suited him: he translated "papillon" as well as writing lit and biography.
> Em dashes are just an affectation.
Isn’t all punctuation?
The distinction between an em dash, an en dash and a hyphen is quite distinct from the general use of punctuation. It's more akin to '' vs " and `' in some respects. I have some sense 《 and ¿ are at least functional in their native languages.
Now semicolons on the other hand, they’re an affectation.
!
in my case it was because the character being depressed seemed emotionally disaffected, and that was interpreted as machine like.
But for lots of people being unemotional is normal. I kind of get it, but I think this is putting the cart before the horse. I mean Hemingway.. all those short manly sentences..
This is unequivocally NOT a Turing test moment. Simply parroting forms of writing which leads some people to accuse other people of being machines does not mean the thinking component is present.
> I believe its indicative of younger people who haven't been exposed to older people who have pompous writing styles.
That, or reading comprehension has dropped below a point in the young where they can no longer recognize or perceive intelligence, and so conflate AI with intelligence much like what is described in the Allegory of the Cave.
Lots of people are calling intelligent well read people AI because the accusers worldview can't distinguish those that are intelligent from those that are AI, or as an imposition of cost the same as any invective. Its just a new form of gaslighting.
You can't possibly be that intelligent so you must be AI. /s
We live in a wide world, which they don't realize because they've been coddled, and often lied to, where the lie traversed more than a generation. Brittle people break.
I assure you I am not that intelligent. Erudite sentences from me are very much repetition of things seen. In that narrow sense, they may be closer to an llm than I like.
"I'm sorry that writing like this is so beyond your reach that you can't imagine anyone doing it without a program" is one of those retorts which, regretfully, only works well in my head.
Was this written by AI?
I used to use a grammar and punctuation checking browser plugin that worked in any text box. Among all the normal suggestions, it would always insist that instead of using a - dash, that it was proper to use an — emdash. I would accept the suggestion until people on Reddit accused me of being chatgpt because I wrote "too properly" with emdashes. So now you can't do that anymore.
I hereby complain about you not using emdash when applicable.
> people on Reddit accused me of being chatgpt
if they cannot make critiques of your content/comment without resorting to appeal to authority or ad hominem attacks (which i consider being accused of using chatgpt as one), then their "critique" is worthless and should just be ignored (with a comment saying so as a response).
I literally do not care that someone is using chatgpt, or anything generated, as a response or comment, or content. The content _itself_ speaks for itself, and the worthiness stands alone regardless of how it was made. Slop can be easily written by an author or ai, and pedigree has nothing to do with it.
I think the fear of being confused for AI is a generational thing. Wait until the first "AI native" generation takes over and AI is the primary means by which people communicate with one another and interact with the world, the very concept will seem alien and absurd.
Of course "writing" itself will likely become passe as literacy drops with attention spans, and most AI interaction is done through dynamically generated video and audio. That process is already underway.
I'll just have my AI read your AI output.
Long live the semicolon!
No, most people don't know about this or care about it. People have other things to do.
AI writing will be impossible to distinguish from human writing in a year or two. The weird crappy way it writes now will be adversarially driven out of it in the next LLM releases, and probably gone in one or two more.
No more lists of 3 things, no more emdash, no more vacant live laugh love level vapid niceties.
I don't think or at least hope so. The corporate desire to sanitise LLMs for "safety" (for PR/liability) and making it agreeable (for engagement) will hopefully leave AI with a recognisable non-human style.
>No more lists of 3 things, no more emdash, no more vacant live laugh love level vapid niceties.
So only humans will have these things then?
Just realtors; realtors describing properties they are listing.
AIs will still have a "default" style that most people will leave because people are lazy, and that default style will be spottable.
I wish I knew how to (ask an AI to) characterize this style. Is it always three examples? Nice formatting and bullet points? Evenly issued, lengthy and didactic explanations?
Following the guidance of another article that was just posted here, I’m going to just share a ChatGPT conversation…
https://chatgpt.com/share/687cd655-1cd8-8010-926d-645cefe928...
When I write, I now go out of my way not to use lists or em dashes. I learned before LLMs, that using lists in writing was lazy. Even when I do use an LLM for writing, I tell it to explain everything in paragraph form. I start off with my own outline and then “discuss” what I want to emphasize.
https://brevity.wordpress.com/2014/04/21/10-reasons/
The second thing about LLM generates text especially when doing technical writing, is that anytime it explains something, even with a lot of context of what I’m working on, it always adds lines about “benefits”.
In my use case something like “this provides a secure, adaptable environment…”. I have had to remove wording I had like that on my resume before LLMs and definitely don’t put it in my own writing now.
Ironically enough, out of the four examples ChatGPT generated about “what makes a good leader” in the link above, the example of “AI generated”, is the one I would lean toward in my own writing.
Also, putting AI assisted writing back through a new session with ChatGPT and asking it does it sound AI generated multiple times, and taking its suggestions will make it sound less like it’s AI generated.
When I do write something “thought leader”ish which unfortunately is part of my job now as a staff architect, I give a lot of real world examples that couldn’t easily be fabricated by AI and lean on those examples as I make my points.
[dead]
AI writing is good, only people who cry about AI are people afraid it will take their jobs, or people who think AI slop is the default output as opposed to the product of a human with shitty taste using AI.
All these tears and tantrums over AI are so tiresome.
I've definitely stopped using hyphens - so, yes!