doesn't follow. You don't know that they didn't read it! Only that they didn't write it. If the point of things like this is to be careful, then be careful.
The old social contract was that the writer had done more work than they were asking of the reader, but this is torn to shreds with LLM-generated text.
That is entirely invalid considering the existence of ghost writers and such.
No one reads a piece of text BECAUSE the writer had more work writing it than the writer will have reading it (this could end up not being true regardless of AI either way).
The “old social contract” is that readers trust the writers had something to say, and said it. The inner workings have always been irrelevant.
Trouble with genText is that it’s easy for your point to get lost in the narrative exactly because you didn’t write it—it’s like when you know something but can’t put it into words. Your own voice can get lost and so does the appeal to the readers. Nothing to do with effort, entirely to do with the content itself.
I don’t want to read your genAI text because I don’t trust that you managed to get your point across in prose, but I don’t mind reading an LLM’s output because I just wanted a piece of information.
The old social contract was between two humans. Effort was a consequence, not the target.
the code version of this is worse. an agent written PR looks exactly
like one the author read carefully. only test I've found is to ask
about one specific hunk, "why this and not X". you get a real answer
=in ten seconds or you don't.
If someone sends me LLM output that they read and thought was fine, then that's OK with me.
They might be wrong and it's not fine, and I have a few people I know who have a much higher tolerance than I do for LLM outputs, and I'm aware of that when they send them to me and I often reject it.
But it's a totally different category than someone who only skimmed what they sent me.
If someone sends me something they read and digested and think it's good, but I don't think it is, then I'm not upset with them. If someone sends me something without that first step, then I am.
(Of course they could tell me they didn't read it, and that's fine too, I can choose to read it or not.)
To be honest I don't particularly feel the need to give the benefit of the doubt to people who dump 10 paragraphs of AI on me. If they want to give a thorough explanation of it to prove they really understand it, that's fine and I can update on the new evidence, but I'm OK with my default baseline assumption being they didn't read it so I don't need to either.
people who are so unskilled in writing even a paragraph and need to have it generated for them are usually not skilled enough to notice the extremely rote, trite cliches that LLMs abuse over and over again, that make it glaringly obvious it's slop.
the the short quippy triplicate comparisons, the inability to contradict itself or reference itself, it's a stink you can pick up immediately and if I see it, it tells me you either can't tell, or didn't read it.
Well, they can serve it that way. Based on my experience, everyone knows who the person always serving slop on the team is except the person serving slop.
I don't understand this bandwagoning either. I'm a shit writer, there I dont write, but if I could prompt some AI w/ my thoughts and it generated something that passed for writing and wasn't shit, then you wouldn't be injesting my awesome insights b/c you think AI made them and dismiss it out of some kinda of hurrdurr spite? Makes no sense, and lets not forget the English-as-second language crowd, I guess they have nothing to say either...
here's my comment ran thru ChatGPT, I hope you dont read it!
"""
I don’t really understand the bandwagoning around this either. I’m a terrible writer, so I generally don’t write much. But if I could give an AI my thoughts and have it turn them into something clear, readable, and actually worth reading, why would that suddenly make the ideas themselves less valuable?
You’d be dismissing my thoughts—not because they’re bad, but because AI helped me express them. That just seems like reflexive “AI bad” nonsense.
And that’s before you even get to people who speak English as a second language. Are we really saying their ideas somehow count for less if they use AI to communicate them more clearly? That makes no sense to me.
"""
On a more recent note, today I'm really fed up with Opus 5.5.
Yes, it's the best model right now if you want cheaper than Astra, but they increased the verbosity again!
It just keeps dumping whole novels on me. Its writing might be an improvement over the ludicrous Opus 5 speak, but much of the grating way of speaking is still there.
It's been three days since Tuesday and I'm already tired of Opus 5.5.
Hard disagree, I’m finding Opus 5.5 so much more pleasant to work with in comparison to its predecessor. Much less of the invented jargon that I got from the last version, or Fable 5.1. Still a bit wordy, sure, but you can dial that down in the settings. I’m wondering what I’m doing differently from you, or is it just a Geschmacksache?
I wasn't aware of the setting - I changed it to "Concise" now, thanks.
Opus 5 was in another league. But often Opus 5.5 still has its moments.
I was informing it of a bug and it started its lengthy response with: "I confirmed less than I claimed" as if I had just accused it of being wrong or otherwise complained.
I found that one weird, and it reminded me that I am talking to Claude rather than an OpenAI model.
Yeah how many times have we seen this sort of thing on HN now? We get it, people hate AI writing. But screeds like this can be even more annoying as it's pure clickbait for the HN crowd as every time these sorts of posts rocket to #1 on the homepage just like this one is now. HNers say they don't like algorithmic feeds but they sure fall for the same sort of thing they decry others for falling for.
> Part of not allowing it to become normalized is complaining about it.
This seems like a fantasy to me, maybe you can offer an example where it played out like that for something that had a wide appeal or value to a lot of people. Ignoring the people complaining seems just as likely to be positive reinforcement.
I'm curious what value you think people are getting by being able to spew out large amounts of low to mid quality writing. I'm not saying we should complain about AI in general, just how some people use it.
> maybe you can offer an example where it played out like that for something that had a wide appeal or value to a lot of people
Sure, the French aristocracy had wide appeal and value for a lot of people, complaining played a part in keeping their behavior from becoming normalized even further.
how about the opposite: a massive resurgence of slur usage and inflammatory open bigotry in general conversations is going on right now, especially with kids, because people allowed it to be normalized without pushback. Some of the shit hear every day, online and irl, would've got you clapped from your job or expelled a decade or so ago. Even if you do find it offputting, without greater support if you stand up to it, many people don't, because in turn it'll cost them more than it costs the person saying it.
normalization works, and therefore it's important to not be a part of it and be brave enough to create an atmosphere where it's a bad idea socially. If you can't trust people to understand that plagiarism is bad, then I guess you've just got to make it toxic to be associated with.
IMO I'd rather have fewer enshittified unusable websites so filled with ads I can't actually read anything than some awkwardly written paragraphs that I can skim/move on. I'd rather the pitchforks have been pointed at ad revenue toxicity the past decade than the AI stuff being the focal. Just my opinion.
I came up with it before that but I'm not going to claim I'm the only one, or even that I didn't hear it somewhere and have it get into my weights and stochastically parrot it later.
I used to feel this way, but now, I want more people to send me AI slop because I'm tired of human slop. I am tired of parsing through two word git commits where the author didn't care enough to tell Claude to write the commit message for them. Sure, I would prefer to read your personal thoughts on the change, but if the choice is between no meaningful information and an agent's mostly coherent if a bit verbose description, I'll take the AI.
The same goes for AI generated code. I've seen some code that I'm pretty sure would have been better quality if my 90 year old grandma was steering Claude. The standard that I want to hold myself is derived from the open source world like SQLite. Sadly, some people just want to take the minimal path to ticket completion without regard for future maintenance or pride in their work. At least AI is token predicting off the code bases that I admire.
I can't even force myself to read Claude's multi-paragraph PR summaries, and instead just go read the code. They're barely coherent and incredibly verbose.
I'm struggling to get people to give me a good summary line. Even if I just ignore the commit message body, I still want a meaningful summary line so git blame will be accurate.
I joined this team not long ago, and I was trying to add a new feature to an API server. I saw a pretty obvious UnboundaLocalError that would occur given the right parameters. I hovered over the line and saw, "fix login". I then looked at the git log for that file and found three commits with the same text. No explanation of how login was broken, how it was fixed, and it introduced a bug. Some people care less and have less diligence than the AI agents.
I have a standard that I hold myself to which falls somewhere in the middle w.r.t. AI usage, but for the other people I work with, I want them to maximize their AI usage.
I understand the sensation of disgust when being offered to read a piece of AI slop. However, I'm actually curious about how our sensitivity towards slop will evolve. I imagine people will have varying degrees of tolerance.
Afterall, AI slop doesn't necessarily mean the prompter didn't read. It just means they're OK with the AI phrasing. It's also possible that they read the output and had the AI update with whatever edits they prompted.
Writing this comment makes me more convinced that if us users of AI want our written work to be taken seriously, then we should also share our prompts. I hope we quickly converge into a standard way of sharing the prompts that generated a given product.
I've been thinking the same. Imagine 5 years from now - nothing is slowing down. The pace that my SW teams can produce legitimately good code that my Sr devs are like "yeah looks good, comes with test coverage, tests look good, CI is green, ship it" - it's insane.
For written work, i think we're so new that over a couple years two things will happen
1 - people get kinda used to it (frog in pot?)
2 - the AI will keep getting better and reducing these cringey sounding tells
What will be tough is finding that line of "does this paragraph add value to X" but that applies whether a person wrote it or an AI.
I hope that labs are actively working on the voice and writing styles of their models.
I don't want to read slop in large part because their prose is overly verbose and complicated.
If the models were writing simple sentences in mainstream English, mirroring the vocabulary used by my colleagues, I'd mind less than the current status.
I think they definitely are. One of the Opus 5.5 selling points in the release announcement was something like "less jargoney / rambling communication"
Might be a long road, but at least we'll have no shortage of good instagram memes in the meantime
Sometimes at work I get slop emails from some VP or other, and the problem isn't the tone or use use of AI per se, it's that 2 sentences of meaningful content is diluted into 3 paragraphs for no discernible reason.
I can say first hand after having years of my life blown up by an (eventually) AI slop powered vexatious litigant and the thousands of pages of slop filings and slop evidence forgeries I had to read... I feel a physical urge to give anyone who slops me a black eye.
Probably not a typical response-- but perhaps it's useful to keep in mind that every day there are many people being victimized by attackers using AI (and shitty corporations.. wait I repeat myself). So in addition to the ordinary signals of indifference, ignorance, and carelessness that hitting someone with obvious LLM slop implies you'll also be accidentally taking on the voice of some people's abusers, and pick up the spillover effects of their impressions.
This can happen randomly, of course, -- picking up a chose phrase that someone else used-- but slop isn't just a random match. Right now an AI written piece (especially from claude) will identify itself over and over and over again in a way that wouldn't happen with another writer that happened to share some cultural background.
There has been an ancient form of slop that the highest form of authority has read for hundreds of years, and its a presidential brief... The President doesn't reads all the stuff that goes on, he doesn't know, he is handled the a slopified version of reality. Now the presidential briefing is at everyone's hand.
Reminds me of my own manifesto, which I've written since getting sick of dealing with the hype and gatekeeping machine that surrounds AI based development right now.
Reminds me a little bit of that guy who made his company called "Polsia" about like one-shotting full companies or something. I think he had actual investors, would have loved to be in their internal staff meeting when the intern walked in and said "you know that spells AI slop right"
What I mostly find issue is with people pushing incorrect, buggy, and low skill human work when AI would do the work better. It's sad to see oldschool pros clutching to their pearls; These days I often find myself trying to push them to do something that AI _can't_ do as well, instead of forcing a comparison.
Care about people but always through caring about the product and making people capable of earning their dignity.
>If it’s not worth writing, it’s not worth reading.
that's why I keep all logging thoroughly disabled across all systems.
Saves a lot of room.
I don't get these new logic statements everyone is coming up with against AI systems.
If the excuse is "I meant literature!" well then I argue that where that line is drawn in the sand is exactly the question these kind of black and white expressions strolling entirely past.
(and that's actually the important question; where does it become neglectful for a human to throw an AI into the drivers seat wrt human read media?)
This is an outdated take. It used to be the case, but now detectors have low false positive rates. In other words good precision but not necessarily good recall for detecting AI. So some AI text evades detection, but when it flags something as AI, then it's usually right.
I believe that one was discussed with the author on HN, and that they used AI to 'copyedit'. Unfortunately that's a trap because if you make a general copyediting request most LLMs will absolutely slopify your work-- and make edits that are unambiguously bad as well as just making it look like lazy llm output.
The better use is to ask it to point out grammatical and spelling errors, or phrasing which is unclear.. then go fix yourself (or direct highly targeted fixes that don't let it slopify). Never hand the clanker the pen for your writing unless you're looking to subject everyone to your Honest Take That Recontextualizes Everything.
IME, it's best to ask an AI to criticize your work and suggest changes without providing actual rewrites, and make the changes you agree with in your own voice.
> If it’s not worth writing, it’s not worth reading.
I mean, you didn't write that, you're just repeating a meme for the 1 millionth time. People have always been saying things they didn't write, and AI is just one new way.
I just wonder how many people smugly cry 'slop' and then immediately go ask chatgpt about something they want to know about. I suspect it's a surprisinly large amount.
There’s a difference between using and sharing. I don’t want to see your notebook. I want to see the finished document. To some of us, the difference is intuitively obvious, so we’re offended by what we’re getting at work these days.
I'm very envious of your workplace if the stuff before AI was somehow well written. If I had AI 10 years ago I would've been mostly using it to interpret most of the corporate speak e-mails and poorly written technical briefs.
Those are different though. One is just reading your own slop, the other is trying to make others read your slop. I don't love either one, but it's clear that one is a lot worse.
If you couldn't bother to read it, why should I?
doesn't follow. You don't know that they didn't read it! Only that they didn't write it. If the point of things like this is to be careful, then be careful.
The old social contract was that the writer had done more work than they were asking of the reader, but this is torn to shreds with LLM-generated text.
No one reads a piece of text BECAUSE the writer had more work writing it than the writer will have reading it (this could end up not being true regardless of AI either way).
The “old social contract” is that readers trust the writers had something to say, and said it. The inner workings have always been irrelevant.
Trouble with genText is that it’s easy for your point to get lost in the narrative exactly because you didn’t write it—it’s like when you know something but can’t put it into words. Your own voice can get lost and so does the appeal to the readers. Nothing to do with effort, entirely to do with the content itself.
I don’t want to read your genAI text because I don’t trust that you managed to get your point across in prose, but I don’t mind reading an LLM’s output because I just wanted a piece of information.
The old social contract was between two humans. Effort was a consequence, not the target.
They might be wrong and it's not fine, and I have a few people I know who have a much higher tolerance than I do for LLM outputs, and I'm aware of that when they send them to me and I often reject it.
But it's a totally different category than someone who only skimmed what they sent me.
If someone sends me something they read and digested and think it's good, but I don't think it is, then I'm not upset with them. If someone sends me something without that first step, then I am.
(Of course they could tell me they didn't read it, and that's fine too, I can choose to read it or not.)
Also the distinction between whether they read/wrote it is nil. Either is disrespectful.
> You don't know that they didn't read it!
If someone sends me AI slop to read then I am going to assume and behave as if they didn’t read it until proven otherwise.
It is the responsibility of the sender to proactively say “I read this and stand by it as at least mostly correct and reasonable”.
They are also free to say “AI slop have not yet proof read”. Then I have an informed choice.
the the short quippy triplicate comparisons, the inability to contradict itself or reference itself, it's a stink you can pick up immediately and if I see it, it tells me you either can't tell, or didn't read it.
here's my comment ran thru ChatGPT, I hope you dont read it!
""" I don’t really understand the bandwagoning around this either. I’m a terrible writer, so I generally don’t write much. But if I could give an AI my thoughts and have it turn them into something clear, readable, and actually worth reading, why would that suddenly make the ideas themselves less valuable?
You’d be dismissing my thoughts—not because they’re bad, but because AI helped me express them. That just seems like reflexive “AI bad” nonsense.
And that’s before you even get to people who speak English as a second language. Are we really saying their ideas somehow count for less if they use AI to communicate them more clearly? That makes no sense to me. """
Did you even read it through carefully once? If you couldn't bother to craft it, why should I read?
If I'm reading reference material/documentation or an update, I care a bit less so long as it isn't tedious.
> Not anti-AI. Pro-giving-a-damn.
AI isn't the problem. The problem is that people are sending text without proof reading.
Yes, it's the best model right now if you want cheaper than Astra, but they increased the verbosity again!
It just keeps dumping whole novels on me. Its writing might be an improvement over the ludicrous Opus 5 speak, but much of the grating way of speaking is still there.
It's been three days since Tuesday and I'm already tired of Opus 5.5.
Opus 5 was in another league. But often Opus 5.5 still has its moments.
I was informing it of a bug and it started its lengthy response with: "I confirmed less than I claimed" as if I had just accused it of being wrong or otherwise complained.
I found that one weird, and it reminded me that I am talking to Claude rather than an OpenAI model.
This seems like a fantasy to me, maybe you can offer an example where it played out like that for something that had a wide appeal or value to a lot of people. Ignoring the people complaining seems just as likely to be positive reinforcement.
I'm curious what value you think people are getting by being able to spew out large amounts of low to mid quality writing. I'm not saying we should complain about AI in general, just how some people use it.
> maybe you can offer an example where it played out like that for something that had a wide appeal or value to a lot of people
Sure, the French aristocracy had wide appeal and value for a lot of people, complaining played a part in keeping their behavior from becoming normalized even further.
normalization works, and therefore it's important to not be a part of it and be brave enough to create an atmosphere where it's a bad idea socially. If you can't trust people to understand that plagiarism is bad, then I guess you've just got to make it toxic to be associated with.
I think maybe just AI;DR...?
https://news.ycombinator.com/item?id=49849134
https://marketoonist.com/2023/03/ai-written-ai-read.html
The same goes for AI generated code. I've seen some code that I'm pretty sure would have been better quality if my 90 year old grandma was steering Claude. The standard that I want to hold myself is derived from the open source world like SQLite. Sadly, some people just want to take the minimal path to ticket completion without regard for future maintenance or pride in their work. At least AI is token predicting off the code bases that I admire.
I joined this team not long ago, and I was trying to add a new feature to an API server. I saw a pretty obvious UnboundaLocalError that would occur given the right parameters. I hovered over the line and saw, "fix login". I then looked at the git log for that file and found three commits with the same text. No explanation of how login was broken, how it was fixed, and it introduced a bug. Some people care less and have less diligence than the AI agents.
I have a standard that I hold myself to which falls somewhere in the middle w.r.t. AI usage, but for the other people I work with, I want them to maximize their AI usage.
The irony...
AI;DR (AI; Didn't Read)
https://news.ycombinator.com/item?id=49336573
Afterall, AI slop doesn't necessarily mean the prompter didn't read. It just means they're OK with the AI phrasing. It's also possible that they read the output and had the AI update with whatever edits they prompted.
Writing this comment makes me more convinced that if us users of AI want our written work to be taken seriously, then we should also share our prompts. I hope we quickly converge into a standard way of sharing the prompts that generated a given product.
For written work, i think we're so new that over a couple years two things will happen 1 - people get kinda used to it (frog in pot?) 2 - the AI will keep getting better and reducing these cringey sounding tells
What will be tough is finding that line of "does this paragraph add value to X" but that applies whether a person wrote it or an AI.
Surely that's something that can be trained for.
Might be a long road, but at least we'll have no shortage of good instagram memes in the meantime
Probably not a typical response-- but perhaps it's useful to keep in mind that every day there are many people being victimized by attackers using AI (and shitty corporations.. wait I repeat myself). So in addition to the ordinary signals of indifference, ignorance, and carelessness that hitting someone with obvious LLM slop implies you'll also be accidentally taking on the voice of some people's abusers, and pick up the spillover effects of their impressions.
This can happen randomly, of course, -- picking up a chose phrase that someone else used-- but slop isn't just a random match. Right now an AI written piece (especially from claude) will identify itself over and over and over again in a way that wouldn't happen with another writer that happened to share some cultural background.
Reminds me of my own manifesto, which I've written since getting sick of dealing with the hype and gatekeeping machine that surrounds AI based development right now.
https://www.just-prompt-motherfucker.com/
It should be line https://motherfuckingwebsite.com/
Do they really not understand that their website uses the most definitive form of design slop?
You really think you coined this,"Ben"?
Also, building this in the bog sloppy way? c'mon.
edit: didn't realize i verbatim repeated what everyone else said
Care about people but always through caring about the product and making people capable of earning their dignity.
That said, wouldn’t it be funny if this site, too, were AI generated?
that's why I keep all logging thoroughly disabled across all systems.
Saves a lot of room.
I don't get these new logic statements everyone is coming up with against AI systems.
If the excuse is "I meant literature!" well then I argue that where that line is drawn in the sand is exactly the question these kind of black and white expressions strolling entirely past.
(and that's actually the important question; where does it become neglectful for a human to throw an AI into the drivers seat wrt human read media?)
The better use is to ask it to point out grammatical and spelling errors, or phrasing which is unclear.. then go fix yourself (or direct highly targeted fixes that don't let it slopify). Never hand the clanker the pen for your writing unless you're looking to subject everyone to your Honest Take That Recontextualizes Everything.
Would you extend that approach to the conversations you have with AI? If not, why not?
I mean, you didn't write that, you're just repeating a meme for the 1 millionth time. People have always been saying things they didn't write, and AI is just one new way.