You won't be able to choose at some point, and you probably already read plenty of AI generated stuff, it's a survivorship bias.
I have a SKILL that emulates how my supervisor prefers his answers and papers, he used to get triggered by generated stuff, but now it just writes in a style he likes which I got from a bunch of his papers and everything goes smoother. If you are bottlenecking someone they will just do the same to you.
I think the dev holdouts will have to give in on the PR reviews. They catch things and you are delivering less secure software to real people if the luddite desires trump real results for the sake of some false purity test.
This is how I've felt for a long time now. Sadly, I work at a company whose leadership is incessantly pushing "AI native" and I recently was denied a promotion in part because I don't share our VP's rosy view of AI. "Innovation teams" who generate all of their code with AI get tons of recognition while real engineers that are doing real work behind the scenes are ignored by leadership. I keep hoping this bubble will burst and sanity will return, but I fear that the hold AI has on executives' minds may be a permanent one.
Great. And you felt the need to scream this to the entire world? Look, if you don't like Generative AI, then don't use it. This isn't rocket science, and you're not making yourself look hip, cool, edgy, informed, suave, or anything else by yelling about it. Just go do you. Other people may do likewise. Others still might make different choices. And both cases are fine.
And never mind that the whole premise of this article is simply wrong. Look at the things GenAI systems are doing today and it's clear that they are more than just "parrots" stochastic or otherwise. Not even looking at the Navier-Stokes thing, but there are Youtube videos out there of PhD candidates in math who have tried out various models from a "can this help me with research level mathematics?" perspective multiple times, and quite often the answer is at least a qualified "yes". Does it really make sense to say that a stochastic approximation of "very average" content can be helpful in doing research level mathematics? I'm thinking "no".
My own anecdotal experience also suggests that whatever LLM's do, it amounts to more than being a mere parrot. ChatGPT was incredibly helpful to me over the weekend, for example, getting started on some neuromorphic computing stuff I wanted to do. It was amazingly helpful at debugging circuit problems, helping me setup some LXI/SCPI automation of my test equipment, doing literature reviews, yadda, yadda, yadda. Even better, unlike most human interlocutors I might want to interact with, it is fine with my weird "night owl" hours, happily debugging circuits at 2:00am, 3:00am, and on into the early morning hours. It never gets tired, never belittles me, tolerates my repeated questions, etc.
I get that some people are "anti-AI" for their own reasons, but I hope some of those people will at least stop and try to understand why some of us think modern GenAI systems are about the best thing since sliced bread. shrug
The stance is defensible but the article reads like it was written in 2023. Still using the term 'stochastic parrot', in 2026, when century-old mathematical problems are being proven by LLMs almost on a weekly basis doesn't make sense.
I'm very sorry to say it, but even reading LLM-slop is less tiresome, then reading yet another anti-AI slop article. It's a pretty low bar.
I could find 10 anti-AI posts on HN and they all will contain points from a very narrow pool of anti-AI takes. I mean, how many more times do you need to repeat them?
> stochastic algorithms trained to produce tons of "average," non-human content
As we can see, humans are also quite capable of producing "average" content. And by "average" I don't mean "average quality", I mean "average opinion of a certain group".
We are all stochastic parrots. Currently, millions of years of evolutionary pressure have produced a better algorithm than a few years of optimization pressure but we’ll see how long that remains true!
Evolution selects for fit organisms that have to survive and reproduce in the world, not language models waiting on prompts. The temptation to analogize technology to biology is just top hard to resist I guess. Problem with analogies and metaphors is they are never accurate, they are only similar in some respects.
I think it is worth asking whether anthropomorphization of LLMs does anything to contribute to a better understanding of them that isn't entirely post-hoc ("of course they were going to find CVEs, solve ARC-AGI...") or if it just ends up reinforcing simplified teleological views of neuroscience, evolution or psychology etc.
HN now has AI/LLM-enabled comment processing. Even if you don’t want to interact with the machine it will interact with you. You will eat ze bugs and you will be happy.
How is that something that defines humans? The machines break the rules as a matter of fact. It's also really a human deciding to have the machine post, they don't do anything of their own volition, so the person behind it is still the one breaking the rules, non sequitur?
I still don't see it as data point we can differentiate human-ness on.
Presumably that HN has an automated LLM-based filter to moderate rude or irrelevant comments, which is a little irrelevant to the article's wish not to see generated posts or comments, and not without its own problems besides: ML is necessary for large-scale moderation, but there are guidelines (like "don't post AI-generated text") which are IMO necessary but also just definitionally hard to enforce in an automated fashion
It's a reference to a very old meme. Where "eat the bugs" was a shorthand for the lower classes being fed bugs as a cheap source of protein. Has been re-mixed without attribution so many times that I don't know when it started, even Snowpiercer got in on it back in 2013. https://www.youtube.com/watch?v=J_jK9O4tEiA
I think here it's meant to imply that LLMs will be forced on the population at large as a form of social control.
> You cannot assume anything about generated code. Maybe it was simply hallucinated and, by chance, it compiled.
This is, ironically, similar to my mom's anti-mental-healthcare argument that the system producing a good outcome is not evidence that the process is correct.
It's everyone's choice what stances they take, and if being utterly opposed to any kind of AI interaction is their's, that's their choice, but just own that choice without resorting to absurdity to attempt to lend it a false sense of credibility.
LLMs don't produce working code through mere hallucinatory chance, they do it by mathematical approximation, which is no more or less valid than pasting together programs from stack overflow answers.
Idgaf if the person who coded a program had intention when writing a line of code I never see because it's a compiled binary blob, I care that the program works. And frankly, pretending otherwise is pretentious BS.
You won't be able to choose at some point, and you probably already read plenty of AI generated stuff, it's a survivorship bias.
I have a SKILL that emulates how my supervisor prefers his answers and papers, he used to get triggered by generated stuff, but now it just writes in a style he likes which I got from a bunch of his papers and everything goes smoother. If you are bottlenecking someone they will just do the same to you.
I think the dev holdouts will have to give in on the PR reviews. They catch things and you are delivering less secure software to real people if the luddite desires trump real results for the sake of some false purity test.
> Using AI is like a shameful drug addiction or a weird sex fetish
This is unfair to drug users and perverts
This is how I've felt for a long time now. Sadly, I work at a company whose leadership is incessantly pushing "AI native" and I recently was denied a promotion in part because I don't share our VP's rosy view of AI. "Innovation teams" who generate all of their code with AI get tons of recognition while real engineers that are doing real work behind the scenes are ignored by leadership. I keep hoping this bubble will burst and sanity will return, but I fear that the hold AI has on executives' minds may be a permanent one.
Great. And you felt the need to scream this to the entire world? Look, if you don't like Generative AI, then don't use it. This isn't rocket science, and you're not making yourself look hip, cool, edgy, informed, suave, or anything else by yelling about it. Just go do you. Other people may do likewise. Others still might make different choices. And both cases are fine.
And never mind that the whole premise of this article is simply wrong. Look at the things GenAI systems are doing today and it's clear that they are more than just "parrots" stochastic or otherwise. Not even looking at the Navier-Stokes thing, but there are Youtube videos out there of PhD candidates in math who have tried out various models from a "can this help me with research level mathematics?" perspective multiple times, and quite often the answer is at least a qualified "yes". Does it really make sense to say that a stochastic approximation of "very average" content can be helpful in doing research level mathematics? I'm thinking "no".
My own anecdotal experience also suggests that whatever LLM's do, it amounts to more than being a mere parrot. ChatGPT was incredibly helpful to me over the weekend, for example, getting started on some neuromorphic computing stuff I wanted to do. It was amazingly helpful at debugging circuit problems, helping me setup some LXI/SCPI automation of my test equipment, doing literature reviews, yadda, yadda, yadda. Even better, unlike most human interlocutors I might want to interact with, it is fine with my weird "night owl" hours, happily debugging circuits at 2:00am, 3:00am, and on into the early morning hours. It never gets tired, never belittles me, tolerates my repeated questions, etc.
I get that some people are "anti-AI" for their own reasons, but I hope some of those people will at least stop and try to understand why some of us think modern GenAI systems are about the best thing since sliced bread. shrug
The stance is defensible but the article reads like it was written in 2023. Still using the term 'stochastic parrot', in 2026, when century-old mathematical problems are being proven by LLMs almost on a weekly basis doesn't make sense.
I'm very sorry to say it, but even reading LLM-slop is less tiresome, then reading yet another anti-AI slop article. It's a pretty low bar.
I could find 10 anti-AI posts on HN and they all will contain points from a very narrow pool of anti-AI takes. I mean, how many more times do you need to repeat them?
> stochastic algorithms trained to produce tons of "average," non-human content
As we can see, humans are also quite capable of producing "average" content. And by "average" I don't mean "average quality", I mean "average opinion of a certain group".
We are all stochastic parrots. Currently, millions of years of evolutionary pressure have produced a better algorithm than a few years of optimization pressure but we’ll see how long that remains true!
You are not just a stochastic parrot. Your speech is not just "next likely word", but a product of your internal model of the world.
Evolution selects for fit organisms that have to survive and reproduce in the world, not language models waiting on prompts. The temptation to analogize technology to biology is just top hard to resist I guess. Problem with analogies and metaphors is they are never accurate, they are only similar in some respects.
Yes, thank-you for informing me I was using an analogy.
I think it is worth asking whether anthropomorphization of LLMs does anything to contribute to a better understanding of them that isn't entirely post-hoc ("of course they were going to find CVEs, solve ARC-AGI...") or if it just ends up reinforcing simplified teleological views of neuroscience, evolution or psychology etc.
HN now has AI/LLM-enabled comment processing. Even if you don’t want to interact with the machine it will interact with you. You will eat ze bugs and you will be happy.
> Don't post generated text or AI-edited text. HN is for conversation between humans.
This is HN's newest guideline / rule, seems the opposite to me. Does it still happen, yes, is it as bad as 6 months ago, no.
The most human thing is posting something that breaks the rules.
How is that something that defines humans? The machines break the rules as a matter of fact. It's also really a human deciding to have the machine post, they don't do anything of their own volition, so the person behind it is still the one breaking the rules, non sequitur?
I still don't see it as data point we can differentiate human-ness on.
What do you mean by this?
Presumably that HN has an automated LLM-based filter to moderate rude or irrelevant comments, which is a little irrelevant to the article's wish not to see generated posts or comments, and not without its own problems besides: ML is necessary for large-scale moderation, but there are guidelines (like "don't post AI-generated text") which are IMO necessary but also just definitionally hard to enforce in an automated fashion
It's a reference to a very old meme. Where "eat the bugs" was a shorthand for the lower classes being fed bugs as a cheap source of protein. Has been re-mixed without attribution so many times that I don't know when it started, even Snowpiercer got in on it back in 2013. https://www.youtube.com/watch?v=J_jK9O4tEiA
I think here it's meant to imply that LLMs will be forced on the population at large as a form of social control.
> You cannot assume anything about generated code. Maybe it was simply hallucinated and, by chance, it compiled.
This is, ironically, similar to my mom's anti-mental-healthcare argument that the system producing a good outcome is not evidence that the process is correct.
It's everyone's choice what stances they take, and if being utterly opposed to any kind of AI interaction is their's, that's their choice, but just own that choice without resorting to absurdity to attempt to lend it a false sense of credibility.
LLMs don't produce working code through mere hallucinatory chance, they do it by mathematical approximation, which is no more or less valid than pasting together programs from stack overflow answers.
Idgaf if the person who coded a program had intention when writing a line of code I never see because it's a compiled binary blob, I care that the program works. And frankly, pretending otherwise is pretentious BS.
I'm praying that you're not a developer
Outright refusal of the tech based on outdated takes is such a boring, luddite opinion
[dead]