It seems to me that those who can't see the fundamentally destructive nature of LLMs never fully grasped the miracle that is language in the first place. It is the delicate thread in which we weave meaning, and therefore human connection and purpose. Appropriating and poisoning it for...well whatever they think they're after, is a bonfire on a ship made of its own timbers.
33
Replies
11
Boosts
@mttaggart "Whereof one cannot speak, thereof one must be silent."
Ludwig Wittgenstein
@AAKL I hadn't, thank you
@mttaggart It's relatively new, from the 19th, and based on this, human language, as a vehicle for social interaction, will be deeply affected.
@AAKL @mttaggart fwiw we've almost already reverted to tersely annotated graphics (a.k.a, memes) as the primary form of comms anyway b/c the general population cannot fathom reading a paragraph of U.S. 12th-grade comprehension level text.
Language advancement was doomed well-before mass text generation by LLMs.
@mttaggart Had a co-worker ask me once -this was on a construction crew after relocating and restarting in a different career- why I bothered with getting an education. He was highschool educated, never spoke more than Pigin-English (in Hawaii). Around here thats plenty good.
As a rebuttal I tried explaining to him that we all each occupy more than the place we are in. We occupy a language, some people occupy several languages. That those languages change you on a deep level. That language shapes the way we (do or don't) understand the world, understand ourselves, other people. I learned calculus and statistics to change what I saw and how I saw it. Spanish gave me a bridge into histories, stories, and cultures I would not have been able to. But that's not the point here.
The point I'm circling around is that this former co-worker didn't grasp the subtle nuances I'd capitulated. To him language wasn't something he ocupied, it was simply a means to an end. A vehicle to attaining his needs. It needent hold a common understanding or bridge the gulf between him and everyone else. It was a tool like a hammer.
A hammer can frame walls, a hammer can open canned food. A hammer can defend you or it can attack people you hate. It was just a tool. So it is I believe with most regular LLM users.
@mttaggart Worth reminding people of how many Indians are paid tiny sums of money to mark violent and sexual images as part of the LLM training.
I agree with you about the miracle of language.
Where we differ is in what follows from that.
For me, AI has not poisoned language. It has helped remove a physical barrier (disability) between my thoughts and my ability to express them.
The meaning is still mine. The desire to communicate is mine. The human being at the other end is still a human being.
If language is the thread through which we create human connection, then in my particular life, AI support has allowed me to hold that thread more easily.
@Robinzz3 That's a valuable and important perspective, and I certainly can't speak to your experiences. However, would you agree that your careful, intentional use is uncommon? And if so, is the deluge of language generated without that care is deeply corrosive?
I’m not sure I know whether my careful use is uncommon. I suspect neither of us has enough information to know that yet.
But yes, I do share your concern about the sheer volume of language being generated without much human thought or care behind it.
Where I hesitate is with the word “corrosive.”
Language has always contained an enormous amount of rubbish, manipulation, advertising, propaganda and noise. AI can now produce that at an extraordinary scale, and that genuinely worries me.
But it can also help someone like me communicate more fully.
So perhaps the thing worth defending isn’t language from AI, but meaning from noise.
On that, I suspect we’re not very far apart.
@Robinzz3 @mttaggart My partner (a writer) and I were just discussing this post when this happened:
Her: I saw a post from someone about a Mick Jagger video and it was all AI. That’s when I discovered an entire channel dedicated to AI slop featuring famous people.
Me: There are many.
Her: Really?
Me: Yeah. Have you heard of AI psychosis?
Her: Yes.
Me: There are people so lost in it that everything they watch and experience is slop. AI music. AI news. AI girlfriend. None of it is real. What do you even call people like that?
Her: … … lost in the sauce?
“Lost in the sauce” may actually be the better phrase. 😄
What worries me isn’t that AI-made things exist. It’s what happens when someone ends up in an environment where almost everything they encounter is synthetic, personalised and endlessly reinforcing.
At that point the problem isn’t simply AI. It’s losing contact with anything that pushes back.
That feels much closer to what I meant by protecting meaning from noise.
What worries me isn’t that AI-made things exist. It’s what happens when someone ends up in an environment where almost everything they encounter is synthetic, personalised and endlessly reinforcing.
This is essentially my concern. However, I cannot see a way for the technology to exist in the context of our present society without it completely flooding the public zone.
Yes, and that’s where I think your argument is strongest.
The technology may not inherently require the public space to be flooded, but our present incentives almost certainly encourage it.
We already industrialised advertising, clickbait, outrage and spam. Now we’ve dramatically reduced the cost of manufacturing all four.
So perhaps our disagreement is becoming rather small. My concern is less with the existence of the technology than with the economic and social machinery we’ve plugged it into.
@mttaggart I see it as pretty much the end of meaning-making which is done as a group exercise, and has a felt sense of connection, common purpose and fit to reality
Concrete example: “Claudish”.
That’s what some developers who use Claude Code and related tools call the weird text blobs it spits out when “explaining” what it does.
In my opinion, it’s even worse than the well-known ChatGPT verbal tics like “It’s not X; it’s Y” and overusing Rule of Three. Imagine taking the worst trends in both software developer jargon and corporate marketing language, inducing some model collapse so the vocabulary and grammar feel artificially restrictive and reductive, and finally wrapping it all up in a punchy, terse style that is neither clear nor pleasant to read.
Claudish has been around for a while (I’ve seen it creep into various documents at work over the last 6 months or so), but has finally reached a point where even hardcore Claude fans are noticing and getting irritated. https://slhck.info/software/2026/06/22/claudish.
@mttaggart not just language but writing and editing specifically. there is a comprehensive misunderstanding of what the process of writing *is*. “this is uncomfortable and I am not quickly getting to the result (e.g.: a word count); I must be doing this wrong”. no! the discomfort is the point. and not like in a “suffering is valorous” way but in a “resolving cognitive dissonance is painful” way. you’re ordering your own disordered thoughts! that is hard and it feels bad
@glyph @mttaggart It can feel bad and it’s hard and sometimes what you end up with doesn’t effectively communicate our thoughts to others whose language, experience and cultural context is just enough off from our own that our intent is not accurately (enough) reflected into their mind. It’s quite amazing what machine learning can do with human language, but it also is missing so much of what is going on because we often don’t know what is missing with our words until they fail.
@mttaggart @glyph Yeah. Our ability to communicate is so much tied into our ability to imagine a mind that works like ours creating the words. When I’m reading LLM generated stuff and I realize that wait this is nonsense because this wasn’t written (or even revised) by a person it’s so painful. Because I spent so much energy assuming it is.
@mttaggart @r343l even the imperfections can be extremely revealing! but with an LLM there is no revealing because there is nothing *to* reveal
@mttaggart “I must be moving these weights wrong, every time I go to the gym my arms hurt”
not a perfect metaphor since not all of what hurts here is just self-improvement, the ordering of the ideas also improves the ideas themselves, but, still, same vibes
@glyph Yeah for me "language" is broadly "the ordering of human thought into speech or writing" in all forms.
@mttaggart it is hard to talk about because the origin point for all their usage is still “language”. they write so many prompts. so many agents dot em dee. but it is … debased, somehow. the kind of thinking that goes into a prompt is distorted and bizarre. flaccid, careless, the perverse approximation of language that you get out of a chatbot is a strangely fitting reward for the perverse approximation of language you put in
@mttaggart @astraluma Part (the smallest part, TBH) of the damage that LLMs do to society is devalue the word. Words matter. And just because 2 things are semantically equivalent does not mean that they are equal - but LLMs act as they are.
@brass75 @astraluma I actually don't think it is the smallest part. By severing constructed language that appears human from human intent, and flooding our world with that intentless language, erodes the fabric of meaning that binds us
@mttaggart I'm going to slightly butcher this because it's late: IIRC Socrates (or Plato?) was against writing because it dulled the mind or something. It kinda looks like we could arrive at what they may well have wanted (less writing) with the full destruction of the value of writing because nobody will be able to trust written text.
@mttaggart Reading this is an absolute trip in the modern context
@tehfishman I used to use it as a counter to parents who were terrified of the internet. Joke's on me.
@mttaggart This is the closest I've heard someone say that if nerds would learn some humanities nuance then things would be better as far as LLM adoption is concerned. I don't know if that's the intent but it's where my brain took it. Possibly because I would agree with it.
@cR0w I actually don't think any amount of semiotics or epistemology would prevent or undo the craven nature of these people. But yes, perhaps the metaphysical danger would be better understood.
