The English language has no words (yet) to capture that special combination of nausea and embarrassment one feels when they commit in good faith to reading a long piece of prose... only to slowly realize the whole thing is AI generated. But I'm feeling it more and more often these days.
I think it lands differently for different kinds of prose (e.g. informational, persuasive, etc.)
Persuasive prose, in particular, which is probably the majority of the things that get posted here, is less persuasive to the extent to which it includes obvious "AI tells". Even in the cases where AI is more articulate, (1) the emotional weight of the text feels manipulative when it is clear that the emotions were, at best, "vetted" by a human as opposed to having been produced by one, and (2) I think we have a built-in "bullshit-proximity" sensor, and have recently been trained to expect AI to be more willing to engage in bullshit than humans due to their obsequious disposition. (For the record, I think humans are also full of bullshit, so this one's more of a toss-up.)
Does that have something to do with OP's post, though? I mean, it's written like someone who's mostly formally familiar with English. Maybe your infamiliarity with formal English itself is factoring in, because the AIs are ALSO formal in diction?
I didn't get the same "AI wrote this" that you did, but I get a lot of people saying "AI wrote this" to stuff I definitely wrote by hand, all the way.
> Does that have something to do with OP's post, though?
Yes, the post reads to me as AI-generated, and then the author lightly revised the intro. They aren't evil for doing this. I just personally prefer reading things written by a human.
> I didn't get the same "AI wrote this",
The tells aren't always the same. For me LLMs have a certain indefatigable "bounciness" in their output that is impossible to miss. Everything has to be punchy. Everything shows the same level of (strangely elevated) excitement. At some point a human will do human things: put a little more detail here, vary their tone there, maybe lose the thread a bit and talk about something that's unrelated to the main flow. LLMs don't do any of those things–the writing is one speed from beginning to end. It's like riding a conveyor belt
> but I get a lot of people saying "AI wrote this" to stuff I definitely wrote by hand
I agree that there is a flip-side here where many people are eager to accuse AI usage for sometimes bizarre and facile reasons. And this is unfortunate, too. But it should illustrate the general frustration that many people are feeling about AI and its effects on social norms.
Genuinely curious, do you find HN to be AI-negative? (For whatever that mostly meaningless label is worth.) Would an "AI-positive / tech-positive" version of HN be just the same site, but with a smaller user base that holds similar opinions to you?
There's a wide spectrum of folks with different opinions on everything here, but I wouldn't say there's some lean towards LLM-hate. I'm not a fan of AI projects, but I can live with the (frankly) very high proportion of AI content co-habitating on this site.
For that reason, I don't think the root post here should be flagged really. That might get reversed, I think there's a flag-count threshold that triggers it automatically. But thinking HN is actively against AI projects stands in preeeetty stark contrast to the front page of HN! It's highly likely that people consider the concept of this metaframework to be bad, and they clicked the flag button instead of not interacting/leaving a curious comment, which is generally the way HN works.
Saying that, you should make one! Hacker mindset and all. It wouldn't be for me, but you might find a less confrontational atmosphere in your own space.
> I don't think the root post here should be flagged really.
If you're referring to my post ...then of course I agree with you!
I should clarify that I am in no way whatsoever opposed to AI usage. I use it all day, every day like most of us here and, if we are doing labels now, then I guess I would consider myself generally AI-positive.
What I don't like is when people use AI to produce low-effort content, and then package it in such a way that it masquerades as high-effort content. Of course what is low-effort or what is high-effort are highly subjective. But publishing an article that has your name at the top, it's written in the 1st person, and yet it's very obviously 100% copy/pasted AI output -- this falls on the wrong side of ethical for me.
I suppose OP could put "AI-generated" or some such disclaimer on the article. But I would rather people just revise the output, actually engage with what they're creating before they dump it on the rest of us.
I think the comments on AI related projects are filled with negativity and uninformed opinions and I too find the atmosphere around here pretty off-putting lately.
I won't say that HN is some bastion of objectivity, but I do find people tend to be a little more direct on here. Individual comments can be pretty harsh as a result, but I tend to put at least some weight in the "HN hivemind" simply because I know there's a lot of people in important fields participating here. You absolutely don't have to agree with that, but I've rethought some things before because of that "negativity" that in hindsight; were indeed not great ideas.
If it feels that AI-generated projects are getting negative feedback, it might be worth interrogating why that is. It might not be because they're made with AI, it could be because they're not a great idea/not well implemented.
> find the atmosphere around here pretty off-putting lately.
Hehe. I chuckle because I've been reading comments like this one for more than 10 years now. You might be right. But if HackerNews is dying then it sure is taking its time.
There's no objective measurement of “better”, which in itself refutes your argument, but I personally feels AI prose to be very annoying for mostly two reasons:
- it's the same everywhere. People could use AI to write in different styles by prompting and through few-shots learning, but they don't and as a result every AI generated prose looks like something I've already read. Its junk prose the same way way Mac Donalds is junk food.
- if someone doesn't make the effort of writing or even customizing the AI output, then what I'm reading is a low effort post and it's highly unlikely that I'm going to learn anything in the process. As a result as soon as I realize that I'm reading AI slop, I feel the immediate disappointment of having committed some time to reading something that wasn't worth my time.
> Historians will look back at all the silly dead humans in the past that icked and yucked about AI.
Like we look at the silly dead humans in the past who fought back against opium, cigarettes and leaded gasoline?
