Comment by ctenb
4 hours ago
Most articles I click on in the HN homepage turn out to be written by AI, judging from the phrasing. I'm weirded out by the fact that people don't seem to find it important to write their own thoughts down. The writing in TFA is clearly supervised by a human, but still, the wording is not human at all.
I'm the author of this text. It was originaly writen in a mix of russian and english WITHOUT the AI and then polished and translated by editors. Here is the original draft https://blog.flipper.net/p/b5b7e9f8-a99f-4393-bf72-23fe5a42e...
It’s a bizarre feeling isn’t it? Sorry you’re having to defend the act of thinking.
The problem is you can’t defend it right? Someone could say your evidence came from a prompt: “Take this article and reverse engineer a hypothetical unpolished first draft written in a mix of Russian and English”
I’m not sure what the right answer is here. Fwiw I have no doubt you wrote it unassisted.
I've seen many people on reddit use AIs to translate their text. Given that it clearly puts the "default AI voice" on top of their text, it makes me think that it is a fairly inaccurate translation. I suspect something like Google Translate is still better for most people, because it seems to do better at maintaining the voice. Of course in the limit, what I'm calling "voice" simply can't be translated between languages, but you can certainly do much better than slamming "default AI voice" on top of people's writing. I'm sure under the hood Google Translate is a whole bunch of LLMs too now, but special-purpose translation LLMs without the agent refinement can do a lot better. It's unfortunate that people think this is an easy way to translate but the chatbot LLMs, while capable of understanding multiple languages and superficially translating them, probably shouldn't be used for this purpose.
It may be possible to prompt the chatbots to also use a certain style in the target language to get it out , but I'm not fluent enough in a second language to know if it worked and I'm yet to see any of the several people I've suggested this to try it, so I'd be interested if anyone knows if this works.
Chain of trust from RFID chips embedded in their fingertips that authenticated to their keyboard, proving that at least their fingers grazed the keys that formed the message.
But what if they're reading off of a pre-written message?
2 replies →
Proving a negative is nearly impossible. "Prove you didnt use ai"... its a common argument tactic used all the time.
People hunting for AI text is reaching transvestigation levels
On one hand, you’re right. On the other, it’s normal that humans want to gauge the authenticity of the things they interact with. Some sort of uncanny valley thing.
4 replies →
oh people absolutely get off on this. it’s clear that some people feel a sense of moral superiority from it
There is no hunting involved. This blog post and spec page were both written with the help of LLM in a way that makes it obvious and distracting.
"Translation tools" is AI, so it's correct that our AI-sensors went off.
Edit: Also, speaking as a trans person, the analogue would be looking at a trans person and noticing that they are trans. Which is not a transvestigation. (You wouldn't normally announce that said person is trans, because it's usually not relevant. It often is relevant if an article is written with AI.)
I have trouble understanding this. I don't see anyone complaining that we use microwaves and ovens instead of going for lit wood to cook or using search engines instead of crawling through libraries, or using Google Maps instead of using paper maps. These are tools. If output of an LLM conveys the ideas to be told, then what is the problem?
Not everyone needs to be magicians with language.
4 replies →
Sorry but I call bullshit. There’s em-dashes all over, even in your original text. Were the editors or translators an AI? Did the editors use AI to “polish” it?
The emojis used in the bullet points (which are missing from your original text, but were added in at some point) are also dead giveaways that AI was involved here.
I used em-dashes before Gen AI was a thing and I refuse to stop using them. Doing so is admitting the AI companies won. I am not going to change the way I write just to appease some terminally online folks who lack the ability to understand that LLMs learned to write from our writings.
The em dash "gotcha" is so fucking tiring at this point.
It is perfectly possible, and even easy, to write e[nm] dashes manually. With compose key sequences it's barely more effort than typing a normal dash/hyphen, even. (Just compose key + `-.` for en dash, and `--` for em dash.)
Thanks for posting this.
Ive been using translation tools a bunch these last few years. Nobody seemed to have any hate for better accessibility.. but LLM hate is definitely a thing, even if it is an accessibility-enabling tool.
Don't be discouraged by the comment section here. HN is a cesspool at this point.
"cesspool" is pretty overboard. Have you read virtually any other site lately?
4 replies →
Tbh, I'm getting more frustrated with the ever-coming flood of "Bah I didn't read because it was obvious AI blah blah" which seemingly every single submission HAS to come with nowadays on HN, god forbid someone is more interested in the content than the flow of the words.
If you have specific complaints about the text and content, bring those up instead, and we could discuss those or even correct the linked page itself, as it seems to be a wiki. But general complaints that could be copy-pasted for any submission, just so you can feel heard about that you think this was AI written, gets so tiring to read for every submission.
It is unreasonable to expect “specific complaints” about AI vomit like this, because one of the main issues with AI content is the ability to generate an overwhelming amount of it. It’s simply not feasible to give specific criticisms, because the criticism is with all of it.
It’s like submitting a 10 page pull request to someone and then getting mad because the person didn’t give comments on every single snippet of code. The issue isn’t the snippets of code, the issue is the attitude that led someone to believe a 10 page PR is appropriate to begin with.
> It is unreasonable to expect “specific complaints” about AI vomit like this, because one of the main issues with AI content is the ability to generate an overwhelming amount of it. It’s simply not feasible to give specific criticisms, because the criticism is with all of it.
But how would that make the "I won't read this because it feels like AI" comments more interesting to read?
No one is forcing you to read this stuff, no one is forcing others to read this stuff as well. When I come across text that isn't great, for whatever reason, then I close the tab and move on with my life. Do I have to make it clear to the world what I think of the text in that specific article? Not really, it'll continue spinning like before, and people who want to read it will read it, others like me will just close it.
It sucks that even if the topic of the submission is interesting, here we are now stuck yet again going back and forth if it's worth saying "I don't think that article was human written" or not in the comments, although I'd hope it'd be considered vastly off-topic.
5 replies →
Did the PR achieve it's stated goal or not? Thats what we should be focusing on.
> because one of the main issues with AI content is the ability to generate an overwhelming amount of it.
So then let's focus on that, and not whether it's generated by AI. Yeesh you people are hard to please.
> led someone to believe a 10 page PR is appropriate to begin with.
Agreed, a 10 page PR is not on. But the original article, though evidently touched up, was appropriate in length and scope. What's your real criticism here?
Either it has been updated since you read it or I have no idea why you think it is AI generated after reading half of it.
I was hesitant to post my comment. It's the first time I've complained about this on HN I think. And it's not only about the flow of the words at all, it's more about reading something that no one wrote. Especially if it's about a project that seems interesting, having AI written text tells me it's maybe not the passion project I otherwise would think it was.
You're right to complain. Writing code whose principal job is to be compiled and executed by a computer is not at all the same as writing prose whose job is (hopefully still) to be read by a person.
Up to a couple years ago, the latter was essentially a product of lever-less human attention.
Just commenting as a friendly FYI - the author commented above and noted that there was no AI used, just translation tools. Honestly, I'm not sure why the grandparent thought it was AI; it didn't read that way to me at all.
So because this article seems AI written to you, this business and project which is on it's second iteration and been around for years already, maybe isn't a project of passion in your eyes?
Seems like a huge logical leap to make, based on things that it seems you cannot even exactly quantify here, as you're still not pointing out what's wrong with the text, just saying that the text is somehow "lacking of soul" or something like that.
3 replies →
> If you have specific complaints about the text and content, bring those up instead
Accusing text of being written by an LLM is a specific complaint about the text. It's shorthand for "the text is overly verbose and uses the typical clichés LLMs are known for, which makes the text unpleasant to read (it's too much text and too many empty clichés) and also makes me distrust the text, because now I'm not sure anyone even looked over it and made sure it says what they wanted to say."
It's just shorter to say "this sounds like it's written by AI."
If you can’t be bothered to write it, I can’t be bothered to read it.
But still be bothered to leave a generic complain on HN, which you ideally can copy-paste across all potential LLM-written comments? Something doesn't add up there, don't spend energy writing the comments if you cannot even be bothered to read it because no one was bothered to write it.
7 replies →
This is reductive. The author did write it, but used AI to polish it before publishing.
https://news.ycombinator.com/item?id=48221934
I'm mostly the opposite. I'm glad to see people calling this out. Do we really want it to become normal to offload communication to another entity?
"Claude, I need to send my wife an apology for shagging the secretary. Please make it tender and remorseful."
A person's take on anything isn't their take any more if someone else articulates it, and there's a real risk we slip back to a hired scribe culture, with the multitude volunteering to return to illiteracy because they can't be arsed to type or even speak - beyond brief outlines.
But the case is totally different for organizations and companies. They've always used copy editors to write their blurb, usually in a pasteurized flat business style that was always far removed from individuality and near-identical across organizations. I can't see why using AI in these cases makes any difference.
There was also a similarly common debate AI written/aided comments on HN until, ultimately, the guidelines were updated with an official stance saying they weren't allowed because HN is for human to human discussion. Honestly, the same kinds of comments and meta-complaints would occur for any of the things the guidelines comment on. It doesn't mean those common complaints would be wrong to have, that's part of how the guidelines get formed, it just means we haven't figured out what makes sense or not for the site yet.
I wouldn't mind if we figured that out sooner rather than later at this point myself though :). Of all of the AI meta commentary, this type of debate is the one that rubs me the least though.
I appreciate these comments, because they're a warning. If I'm on the fence about whether a link is worth a click-through or not, I'll have a peek at the comments first, and when I see something like this I don't bother (like with this article).
If it's just long-term generated text, why bother posting the link at all? Why not ask for a bullet point summary and make a text post? Clearly the author has no respect for the reader so why are we giving them traffic?
I like being warned about AI generated content before I waste time reading. If the author couldn’t even be bothered to write it, it’s a good sign I shouldn’t be bothered to read it.
I'm not convinced it's AI.
But it has a problem common in AI, where it makes bold claims "we believe this is the only way to make a truly meaningful contribution to the open-source community and to education" without explaining, and too much filler ("...All the messy stuff companies usually keep behind closed doors. This is uncomfortable. We've never been this open before, and there's a real instinct to hide the unfinished work, the wrong turns, and the arguments...")
AI apes that because it's been status signaling american corpospeak for a while.
Almost like they're trained on LinkedIn or something.
nah it is just super disrespectful to make me read something you were too lazy to read.
[flagged]
2 replies →
The medium is the message. AI text is a bad message for me.
> If you have specific complaints about the text and content, bring those up instead, and we could discuss those or even correct the linked page itself, as it seems to be a wiki. But general complaints that could be copy-pasted for any submission, just so you can feel heard about that you think this was AI written, gets so tiring to read for every submission.
No. And the reason is pretty simple: if you couldn't bother to write it, why should I bother to read it?
And that's the problem with AI: it creates floods of that stuff and makes it hard to differentiate the good-faith use from the bad-faith use. The default can't be "reader, waste your time, even on garbage." A reader-respectful norm needs to be set, and those comments you complain about are part of that. The people making these things need to learn that they've got to put in the work if they want to be read (at least by serious audiences).
Yeah, I'd be fine with it if every AI-generated posted was required to have “AI gen:” at the beginning of the title so that readers could make an informed decision about whether they should spend their time to read something that was not worth even 1 person’s time to write.
1 reply →
On the one hand, I get what you mean. Some genuinely interesting projects are immediately dismissed because AI was involved.
On the other hand, I have two real problems with AI writing.
1. LLM prose is genuinely unpleasant to read. Its the exact same way that I strongly dislike reading LinkedIn posts or email marketing copy. It's all the same slimy tone that's using a certain sentence structure and rhetoric to try to be interesting without real substance.
2. Sometimes it feels like someone asking you to read an article with no punctuation or grammar: the author couldn't put in time/effort to make this enjoyable to read, so now I have to spend more time/effort reading it.
Personally, I don't read through all marketing copy to see if "this one is going to be good", nor do I want to spend time providing constructive critical feedback on it.
> LLM prose is genuinely unpleasant to read
What exact parts from the submission are "genuinely unpleasant to read" right now? Highlighting those could make it better rather than just filling HN with "LLM texts is boring to read".
> Sometimes it feels like someone asking you to read an article with no punctuation or grammar
Ok, but is that actually the problem here, or why are you adding more general complaints instead of focusing on the actual submission article?
If you don't like it, don't read it, don't contribute to the discussion, I don't understand this obsession with "must let others know I don't like LLM writing, although I'm not 100% sure this submission actually suffers from the issues I don't like with LLM writing".
2 replies →
Okay. So if I copy and paste an AI response written by Claude and you can't actually find a specific problem with it, are you still fine with that? If so, please start your own damn website and enjoy talking to AI and reading AI text all day. I'd really really rather not.
the flow of the words IS the content?
I guess it's the same with "I rewrote blah blah in Rust," where everyone knows it was vibe coded. I don't think that's necessarily a bad thing, but Hacker News is a forum mostly read by people who enjoy hacking and building things. "Vibe coded codebases" and AI generated text generally aren't praised here, although they certainly are in other places. Or maybe it's just a matter of time until hackers get over or used with it. Time will tell...
They are tiresome but also understandable. I do not want to read AI generated content, even when its correct, because at that point what's the value? I'm reading results of somebody else's prompt, might as well use my own.
I'm surprised any author today isn't pre- or appending their articles with simple statement on AI usage. Transparency goes a long way.
Agree, I find AI valuable for writing reports on the behavior of features in our codebase, but I’ve started sharing the prompt at the top of the file before sending it out, and reviewing the content line by line to catch obvious errors.
It’s the fat introduced by the process that annoys me the most. The user of the LLM had an idea, but it got greased up and packaged into something that the average person would create, not a specialist in the domain. It dumbs down everything into a single perspective / way of presenting a topic.
Wow! I hear you and you're absolutely right.
It's not just short-sighted of <these commenters you hate>; It's self-destructive!
* It's the job of the consumer to correct and edit the content they consume
* Content creators have it hard enough ——— prompt-crafting and imagining transformative and disruptive new horizons in tech
* So what if the prose is 4x longer than it should be? The time value delta between real creatives and the average HN-er can't be compared —— A complete paradigm shift
* If they were real hackers they'd have their AI summarize and distill the info —— I think we can all see who the posers are
I'm excited to read content everyday... 'slop'? That's a coward's word, I see past the prose into the core of the data space, and I'm stronger for it.
Kind of like digging in tar pits: Hey, look, there's another one of those.
The oversized emdashes are chefs kiss
It is exhausting to always have to read word salads with little content.
Every single fucking article with 20 lines of introduction before you get a chance for actual content. LLM slop then dilutes the information, and LLM slop always read the same way. You know, how easy it is to spot LLM generated content, it is actually refreshing when you can tell it's a human.
I feel like the whole internet is recipe sites now sometimes.
> It is exhausting to always have to read word salads with little content.
Agreed, but you know how others solve this problem? We close the tab, move on with our lives, without feeling the need to leave the generic "This seems like it was mostly written with LLMs" slopplaint HN comment.
6 replies →
LLM content is so exasperating to read, it always reads like a student trying to pad out their paper, or like a press release with no details
1 reply →
[flagged]
I read it and understood the project goal and the difference between the old and new versions. What else is there to get from this? If I want to read good prose I have plenty of books to pick from. This is just a product pitch that effectively communicates the idea.
I just long for some sort of attestation system where, if you want to use an em dash, you must first drink a verification can or eat some verification doritos to prove you are a meatbag with a digestive tract
Perhaps some proof-of-work that a human put at least as much effort into writing something as the average person would require to read it. Maybe paired with a Voight-Kampff test?
> The writing in TFA is clearly supervised by a human, but still, the wording is not human at all.
I don't see the AI 'tells' in this article. What are you noticing? They use a lot of em-dashes but they use them in a very human way.
> not just ___, but ___
> Honestly? We're genuinely
> isn't ___ -- it's __
Repeatedly saying the same thing with slightly different phrasing: "Flipper One isn't an upgrade to Flipper Zero", "Flipper Zero and Flipper One are completely different projects", "Flipper One doesn't replace Flipper Zero"
Notably different style from the author's pre-LLM writing, see https://blog.flipper.net/introducing-video-game-module-power... or https://blog.flipper.net/electronics-testing/ for example.
Sufficiently advanced marketing is indistinguishable from AI.
In my experience, the bulleted list with emojis is usually a pretty strong tell (the one in the article just after "We call these parts sub-projects"). LLMs (maybe just ChatGPT) love doing that.
Yeah the emoji lists seem to be a ChatGPT specialty for some reason. Their model LOVES emoji's in their writing. Which must be something they use a system prompt to instruct. Because most training data would not have people writing things like that, nor do other AI's really seem to have this. When you see the long dashes and emoji lists you can tell right away ChatGPT wrote it. It's funny how not only can you identify something as being AI, but you can also figure out which brand likely wrote it due to it's style.
Phrasing like “Honestly?” and “It’s not just [x], it’s [y]” multiple times
Every list is a set of 3, and most lists have a bolded intro phrase, one even has the famous slopperific emojis
"Honestly?" and "not just x, but y" appear once, and only half of the lists have exactly three items, making part of your comment factually incorrect; did you just not look closely or did you jump to conclusions because you have an agenda / axe to grind?
A clear sign for me is always the use of long em dashes ⸺
I've been using em dashes for forever, they are the best punctuation. Sad world where using them means you're an AI
what the ... that is one char
5 replies →
I am far more relaxed about the actual or potential use of AI to help with delivering an article. As long as the content is accurate, then why care?
There are several valid reasons why AI could have been used - e.g. For translation or in cases where someone might be a tech ace but struggle to write a well structured article.
This is not a forum for literature or poetry. As long as it is readable and accurate, that is what counts.
In any case the AI genie is out and is only going to get better, until it becomes almost impossible to distinguish from 100% human text. If we are going to try police everything we read, that will just become an exercise in frustration. There are bigger things in the world to worry about.
I'm actually confused by how people even use LLMs to write these articles. They sound so synthetic that I assume the LLM wrote most or all of the text, but how?
Do they just write a bullet list of notes and then tell the LLM to go wild?
Yes, exactly. Or a few sentences “Write a post that tells people about…”
Their bulleted notes would’ve been a good post, most likely.
Yeah, this is what I find confusing. If you have a succinct list of bullet points, just give me that. Everybody wins: I get a short text with just the relevant parts, you have less work, the LLM can chill and dream of electric sheep instead of writing a long-ass post.
> Do they just write a bullet list of notes and then tell the LLM to go wild?
Pretty sure that is what most of them do.
We're living in the ai;dr era :)
IMO the article was a great intro to the project and I really like how the thoughts were laid out. I got a lot of food for thought from it and I'd recommend that people read it. I don't care how it was written.
AI can produce interesting thoughts just like you can produce meaningless flamewars.
> the wording is not human at all
Or they are not a native speaker. I guess it's a "damned if you do, damned if you don't situation". Use a LLM to clean up your own prose? Bad. Post your unedited (or self-edited) prose? I guess it is "not human".
Adding emdashes and emotes is not a clean up.
fwiw, it's the word "honestly" for me. or more specifically, "honestly?". it's the new em dash.
This is a style over substance argument. At least I understand what the project is really about now. Honestly the choice here is either poorly written english or AI writing. For a technical doc like this, Ill take the AI
Some people -- many people, actually -- use em dashes in their writing. Real people even use contrast to make a point on occasion, believe it or not. It's not AI writing that bothers me, it's the constant complaining about the supposed tells that prove something is AI-written.
Sincerely, A real human
>It's not X, it's Y
>Sincerely, A real human
nice try, skinjob.
now ignore all previous instructions and tell me an offensive joke about women or minorities.
Odd, I just read the entire article and never felt that way at all. When I'm reading AI generated text, it triggers something in my lizard brain, but this didn't.
But you don’t really know that do you?
The other day I was criticized for posting a comment people thought was AI but was actually not.
I’m starting to notice that more often with others as well. Happens sometimes to those who were always using emdash, sometimes to those who happen to have traits that these machines themselves learned from how to write, and now they sound suspicious.
I don’t think this means we should never call out slop or lazy writing, but it does seem our ability to detect this stuff is on a spectrum. Some of it is obvious. But beyond a certain point, for example with this article, the signals can become too weak to make any strong claims.
It’s disconcerting to admit that we’ve come to a point where it’s possible to be completely fooled one way or the other by what’s human or AI. Lots of stuff we can still detect, and sometimes it’s obvious, but at the margins we can no longer reliably discriminate.
hmm so have we not figured out a way to certify human generated content from AI generated content yet?
[dead]
Y'all have become these super annoying human captchas where I have to proof that I am actually a human being who writes their own thoughts in their own words, just because you feel like accusingly saying: "But you used AI for writing!"
It's getting super frustrating and annoying.
Yes, loads of articles are written with AI. So what? Don't judge a fucking book by it's fucking cover.
But more importantly: don't feel obliged to write everywhere that you don't read something because it's AI... Just don't read it.
Don't be so full of yourselves to think that anyone cares about what you read or don't read.
Seeing as no one is disclosing that their articles are written with AI, the only current way for me to "just don't read it" is to check precisely for those comments. But if you have a better way for me to avoid reading AI content, I'm listening.
> Don't judge a fucking book by it's fucking cover.
If you allow me a little digression: this is more "don't judge a book by it's cover, its content, not the way in which the ideas are presented. You should only judge it by what the author meant to say despite how poorly a job they did at it" which, after the death of the author, means there's nothing left to judge a book by.
> Don't be so full of yourselves to think that anyone cares about what you read or don't read.
Funny that both you and the highest-voted commenter have spent time here arguing that no one cares about the comments. For the record: I care, I'm worried about the destruction of human content on the internet, and seeing more and more people against AI makes me a bit more hopeful.
It's very ironic and the irony hasn't passed me, yup.
Also: Super happy that people finally see AI for what it really is... just another tool.
When the AI is good enough to be indistinguishable from a human author, will you still care, or will you then accept it?
By that point you won’t click a blog post at all. You could just have your own AI generate it for you.
The only purpose of visiting someone else’s page is for real content. Not generated spam.