Hacker News Reader

Comment by tannerc | Hacker News Reader

What exactly indicates the post was AI generated?

i've been using AI for as long as GPT has been out, so if you can't see through the rambling, overly complex to make you sound smarter kind of text, as well as the written patterns that are always used ad nauseam like "this thing isn't JUST this, it's THIS" -- i dunno how else to prove it to you. IYKYK.

I have also used GPTs since 2020. I am also a writer. Much of the writing equated with “generated by AI” is so precisely because it’s broadly trained on real writing.

So the claim of “AI slop” without proof is little more than heresy. It would be helpful to have any evidence.

It’s not about just the writing in one example, it’s about writing patterns—which are common—being equated with AI simply because they’re common.

if you're a writer, and you're using GPT for so long and you can't see it as obvious, i dunno what to tell you at this point. i guess LLMs are trained particularly on this guy's writing.

So you have zero evidence to convey something you claim is “obvious”? Got it.

here's your evidence, tanner https://xcancel.com/PlumbNick/status/2016590894485385347#m

I'm not going to say i told you so, but you should utilize these tools more before you start arguing, especially when it's so goddamn obvious

If you read his original draft you can see how much of it was still carried over, as well as how his original writing conveys much of your same arguments that an AI wrote the final text.

I don’t think your point is as strong as you believe it is.

Lastly, I work directly with AI models and utilize all popular generators every single day, so I don’t know why you think you’re the expert here.

my point was that its AI slop. whether his original point is intact doesn't matter to me. the fact that you're now defensively doubling down and steering the conversation into a direction which serves you better is just cringe. i bet you're a pleasure to work with. c ya later nerd.

If you can’t explain it, then you don’t actually understand it.

i just don't feel like enumerating all of the common patterns ai slop produces. again, if you don't see this as obvious, i can tell you're clearly not using this stuff often enough (which might be a good thing)

The thing is, these stylistic patterns existed before AI, and weren’t completely atypical. Maybe you’re using LLMs so much that you’re over-associating them with AI now. Or maybe the author is using LLMs so much that he’s unconsciously adopted some of the patterns in his own writing.

i think the much higher probability isn't that this guy wrote literally like an LLM before LLMs came out, but rather that he just used an LLM to write all of it. You can see even more of these examples directly on his campaign site.

And despite the downvotes, you were correct.

https://xcancel.com/PlumbNick/status/2016590894485385347#m

thank you. even without this tweet, i was willing to die on this hill. certainly feels nice to place these obnoxious HN know-it-alls into their place.

> certainly feels nice to place these obnoxious HN know-it-alls into their place.

You don't have to take the time to explain your reasoning if you don't want to, but "obnoxious know it all" is not a stone you should throw while at the same time refusing to explain yourself and saying anyone who can't see what you see is necessarily missing the obvious.

it's too difficult honestly. there are a lot of the classic easy traps -- "it's not just X, it's y" which are a dead giveaway, especially when they're used like 3-4 times in one essay. But the harder to spot ones, IMO, are ones where the overall tone is unnecessarily complex. E.g:

"When replacement is cheaper than retention, the decision gets framed as strategy instead of consequence."

This sentence is tight and on paper reads well, but it's robotic. It's kind of like taking a dead simple if/else statement that's pleasurable to read into a one line ternary statement. Technically a one line sentence, but now I have to re-read it like 5 times to understand it. The flow is dead.

Another example:

'AI becomes the excuse, not the cause. It’s the clean narrative that hides what’s actually happening: experienced workers being swapped out through global labor substitution while leadership talks about “efficiency” and “the future of work.”'

Starts off with a short & trite sentence (LLMs loves this if you don't steer it away). The other thing LLMs _love_ to do unprompted is: "It's the X: _insert_next_loaded_statement_here"

It's hard to get my point across, and I hope you kinda see it? I'm not a linguist, but these patterns are literally in every piece of LLM writing I've ever seen.

Again, you don't have to explain yourself, just don't be rude about it. It's hypocritical to call someone obnoxious and a know it all while you are engaging in schoolyard behavior and refusing to allow them to challenge your reasoning.

Saying nothing is an option. Other people who agree with you will be happy to explain their reasoning. Or maybe they won't and the conversation quietly fades away. Both are preferable.

give me a break. have you read the other comments? asking for proof in the most smug attitude possible. it's the definition of obnoxious HN commenters. and that's not even counting the one guy that wrote "you sound and write like a bot", got downvoted and deleted the post. i don't need to take any high roads here--it's the internet. As far as being "rude" it's a solid 2/10.

I'm not saying "take the high road" as much as "don't wrestle with a pig." It certainly isn't appropriate to call you a bot. But they probably insulted you to provoke you, right? Why give them any additional ammunition?

That's just my two cents, ultimately it's your business.

Well he literally confirms it was from ChatGPT in a later reply, so there's that.

And his original draft is conspicuously missing the telltale "it's not X -- it's Y" and overall breathless dramatic flair that people like the poster you're replying to (correctly) picked up on.

https://xcancel.com/PlumbNick/status/2016590894485385347#m

I've been writing in a contrasting style like that since probably 5th or 6th grade.

... I wonder how much the writings of a lot of autistic / borderline folks impacted the LLM writing style.

> i dunno how else to prove it to you

A prompt to generate similar output would be a good start.

how's this for your prompt, pal https://xcancel.com/PlumbNick/status/2016590894485385347#m

hopefully that's enough of a good start and a good end for this conversation

$foo isn't just $bar, it's actually $baz

Reply on news.ycombinator.com