I missed your original comment but I think it was about writing blog posts in your native language? Please correct me if I’m wrong. I actually agree with you, makes sense that English is the most spoken language online but that really doesn’t mean you have to reach English-speaking audiences, it’s not a necessity or anything. I also think it’s very underrated and amazing when an author can translate their work themselves, no matter if the English translation has issues. Their own words are always better. Whenever I see grammar issues, spelling issues, whatever, I’m always like ‘score! not AI!’
That’s why I said spellchecker, because I agree, the grammar-changing totally alters not only the style but the meaning of a phrase/sentence. It was my most enraging point of conflict with Grammarly.
2 Likes
I personally have zero interest in reading anything that was composed with LLM assistance. People who are enthusiastic about LLMs tend to divorce the scaffolding of writing (e.g. outlines, topic sentences, general sentence structure, etc.) from the “ideas” themselves. They’ll happily give their LLM of choice a basic thesis, ask the LLM to create a well-structured outline based on that thesis, and then insist that anything they write on their own steam using that outline is reflective of their “own” work and their “own” ideas.
Anyone who has ever written an article for publication understands that outlines and ideas are often inextricable. Some of my best ideas have only ever occurred to me during the process of building a detailed outline. The way we construct an argument matters, and it’s a reflection of our unique ways of thinking.
One of the reasons why LLM-assisted writing is so boring to read is because LLMs typically adhere closely to predictable patterns and structures. I skim through dozens of dull LLM-assisted posts on Reddit every single day; these posts all sound like they were written by the same author, even though the “ideas” expressed are often very different.
LLM-assisted writing is little more than the gross average of everything in the LLM’s training set. It is beyond mundane.
6 Likes
yeah when I’m helping someone with their grammar, I put in every effort to keep the meaning the same, including asking the writer what their intended focus/emphasis/meaning was with a particular phrasing.
Computers can’t do that because they don’t actually understand anything.
4 Likes
In a vacuum, the act of using a generative service I do not think it’s a problem, my issue stems into all things related to honesty and capitalism, as well as a personal element. I also think it depends on if we’re talking creative writing or technical writing.
For example, I don’t think people using Generative Text to rubber duck off of or possibly get some input that one hadn’t considered, is a bad thing. You’re still using your own input and reasoning to decide what to do with the information you’re presented. I personally wouldn’t, I don’t care for it. (it’s too skewed by capitalism and existing power structures to be of any benefit to me to see perspectives outside my own, I’d rather see that of other people) Though I admit I find some novelty to the older models because they can give you some absolutely buck wild suggestions that make it kind of funny to consider. I like it better when it’s less ‘advice’ and more like spinning the roulette wheel of a word salad.
But I think using it to wholly re-write your work, removes your identity from it, which to me, that’s the whole point of making something yourself. To create something you can share with your unique perspective and way of speaking/presenting that only you have. It also removes your ability to practice how you want to convey your personal voice and is more likely to cause you to sound like everyone else. It is homogenizing.
I also can’t say I agree with the quote: Not many people know about camera angles and many people will interpret a picture in different ways. Likewise, people can interpret a sentence in many ways. We wouldn’t continue getting into arguments about what someone meant or did not if that were the case.
I feel similarly to those who make illustrations when its not their passion vs writing when it’s not their passion: There is value in the Flawed. Such as the communication I mentioned, being able to see from another’s perspective that sometimes is more raw and clear from someone not in the ‘practice’, because they aren’t motivated by the same rules as those often entrenched in the work, are. It gives you a whole new angle and there’s something so wonderful about that, to me. Using something generative makes you skip over that and removes the opportunity for both reader and writer, to learn how something new can be explored by the medium that can only be achieved by those that don’t know the ‘rules’.
On top of all this, I think the thing that gets left behind in the conversation around generative anything, is that we are so squeezed for time, we perceive the need to spend a great deal of time on something as a detriment, as was the issue in the initial blog. ‘Outlines take me hours therefore something else to off-board it to is better’ This is a capitalistic way of thinking, the narrow optimization of time, rather than seeing that time as a means to engage with our way of thinking and have a conversation about what is important.
Of course, in the matter of certain formalities where maybe a certain kind of rigidity and structure is presented, I can’t fault people too much for using generative tools to edit their work. Especially if we’re talking a capitalistic work environment that doesn’t benefit from ‘outside the box’ thinking. But I do think too much use over time will start to change your way of thinking and that can water down your personal, passive voice over time and thus leading to the parts I mentioned before about being homogenized. But at the same time, many of us can’t afford to spend that time. And I am sympathetic to that.
I appreciate at least, the blogger in question does make effort to go over and personally re-evaluate and write the things provided to them by Claude. But, there may come a time where they decide ‘productive’ is more important than personal voice, and stop any of their personal processes and just let the LLM have it with the assumption they ‘trained’ it enough, and in the end, I feel that’s the real ‘danger’.
This all lends to why I have no desire to use the tools, myself.
4 Likes
Yes! The homogenizing is what I dislike so much about it too. Even if we somehow fixed all the environmental issues and ethics around training data, we’d still have an issue with people’s unique voices becoming homogenized into whatever got the highest score when the model was trained. It’s a combination of designing by committee and creating new normativities.
And in isolation, that high-scoring result might look pretty good. But when everything looks like that, it becomes worse, and increasingly so over time. I can recognize a site that was vibe coded the same way we can all recognize emdashes and lists with emojis and “its not just x - its y” and so on as LLM generated. It’s the lack of diversity that makes the text and images and apps and everything else made with AI so much worse, both stylistically and in content.
3 Likes
You’re absolutely right! (Haha, sorry. I couldn’t resist.)
I’m reminded of this passage from 1984:
Julia was twenty-six years old. She […] worked, as he had guessed, on the novel-writing machines in the Fiction Department. She enjoyed her work, which consisted chiefly in running and servicing a powerful but tricky electric motor. She was ‘not clever’, but was fond of using her hands and felt at home with machinery. She could describe the whole process of composing a novel, from the general directive issued by the Planning Committee down to the final touching-up by the Rewrite Squad. But she was not interested in the finished product. She ‘didn’t much care for reading,’ she said. Books were just a commodity that had to be produced, like jam or bootlaces.
[…] She had even (an infallibIe mark of good reputation) been picked out to work in Pornosec, the sub-section of the Fiction Department which turned out cheap pornography for distribution among the proles. It was nicknamed Muck House by the people who worked in it, she remarked. There she had remained for a year, helping to produce booklets in sealed packets with titles like Spanking Stories or One Night in a Girls’ School, to be bought furtively by proletarian youths who were under the impression that they were buying something illegal.
‘What are these books like?’ said Winston curiously.
‘Oh, ghastly rubbish. They’re boring, really. They only have six plots, but they swap them round a bit. Of course I was only on the kaleidoscopes. I was never in the Rewrite Squad. I’m not literary, dear – not even enough for that.’
So many websites and social media posts are now little more than boring, “ghastly rubbish” produced by engagement-harvesting machines. The subject matter differs from site to site and post to post, but they all ultimately sound the same.
3 Likes
I agree with the consensus that AI-generated writing homogenises the human experience into a corporate-approved homunculus. The beauty of reading is seeing the individual behind the writing, like how our fingers have their own prints, or how dog noses have their own prints. I want to see the spark of that person’s soul in their writing, I want to see what makes them feel alive.
If someone has to use AI, it should be for the sole purpose of improving the text from a cognitive standpoint. Note that I do not mean to flatten the poor thing so violently it makes the Hemmingway website beg for mercy. Such tasks require due diligence from the user.
This is just my method, but at most I use it to help smooth my cadence and ask me questions so I can expand less-developed sections. For the cadence bit, I nest half a million clauses per spoken sentence and this does not translate well to text. I like talking but I like being concise while I do it. For the expansion, I am somewhat of a mute and this can translate to text, so having prompts to expand my arguments is encouraging. It is not that I’m unconfident in my articulation, but rather I (without getting into the gory details) am prone to dissociating, even if I’m not overly stressed. However, I do not want the robot to sand down my words to an inoffensive beige powder. My words must remain mine. If you see an em-dash in any of these posts, assume I’ve been killed, someone is wearing my skin and call the bobbies immediately.
2 Likes
I admit I’m perplexed that you express such a negative perception of using chatbots to alter text right before saying you use chatbots to alter your text.
What kind of output are you getting out of that exercise? Have you saved any examples?
1 Like
I completely disagree with using AI as a writing aid.
You become good at things by doing them. This is true of grammar, writing in a second language, or any other number of things. Relying on AI to do that work means having a crutch that stops you from fully honing your skills. A text riddled with mistakes is still better than something “fixed” by AI, because at least it’s an opportunity to learn.
LLMs also have the unfortunate tendency of flattening text, resulting in a generic style. This is especially terrible in terms of literary texts. Something that many people miss, despite generations of teachers trying to point this out, is that the merits of literature are not confined to a text’s narrative. There are matters of style, of using words and punctuation and syntax to achieve a specific effect.
If people simply need some quick checks, there are already tools for this that aren’t LLMs. A bunch of word processors have had spellcheck tools for decades now!
Regarding language barriers, I’ve experienced them, but there too it can be helpful to put your skills to the test and develop them. I recommend trying to write things yourself first and foremost, making use of the many free dictionaries online and the helpful articles and discussions that inevitably come up if you Google a tricky phrase. Even if you run things to a translator, it’s good to look at it with a critical eye. Many things don’t translate 1:1 so the results can be very poor. This will only change if we make AIs that can read people’s minds to grasp exactly what they mean…
2 Likes
How is my process perplexing to you? I stated my reasoning as to why. I use a robot to help me smooth my cadence and make my text more cognitively accessible, not to totally replace my speech. Those are completely different uses and I do not see it as any different from using a tool like Hemmingway.
I do not have any examples on hand, as per my post, it is a guiding path that inspires me to write more, not lull me into indolence and churn out slop. There are no intended examples, it is a prototyper that I ultimately have the final say on.
I mean when you ask it for questions, what questions do you get back?
1 Like
I was also actually wondering about this part. Doesn’t this add/replace words or entire sentences? If not, can you explain what it does to your writing exactly? You mentioned expansion quite a bit and to me it sounds a lot like generating words/sentences that are not yours but related to your own writing but please correct me if I’m wrong.
I use a spell check. Does that count? 
Honestly, I think it’s hard to know where to draw the line. Ages ago, long before LLMs and AI became a thing, I used Grammarly in my writing process. Ultimately, I felt it stripped out most of what made my writing sound like ‘me’. I was reading Kerouac at the time, and remember wondering what The Dharma Bums would’ve been like if Kerouac had used Grammarly. So I decided to drop it.
After that experience, using AI — which I do use extensively in other ways, I might add — as part of my writing never really made sense. In a sense, it’s probably a form of vanity. I’m way too picky about how I want to come across in written communication to be comfortable outsourcing writing to a random token generator.
2 Likes
(@Lars-Christian, I’m on mobile and somehow missed the reply button to your post. 43 injured 72 dead.)
Hah, imagine if any literary classic was written solely in Grammarly!
I don’t consider you vain, I like your conviction. It’s your vision and you know what’s best for it, likewise with the projects where you do use robots, which is not my business at all. Given this then, and if I am not assuming too much, we seem to feel that things don’t have to be entirely handmade to be human.
2 Likes
Do you want it to…?
I’m fairly indifferent to the answer, to be honest 
It was more of a rhetorical question, given the fine lines between spell checkers and something like Grammarly, which I’ve always seen positioned as a slightly more advanced spell checker. And, in my experience, there’s very little practical difference between using something like Grammary and AI in your writing. It sanitises and strips writing of its character.
Then no, spellcheck doesn’t count. And on the subject of Grammarly, have you seen what they’ve been making headlines for?