Hacker News Reader

Comment by cj | Hacker News Reader

Examples?

Your media literacy training can start with skimming headings. When every heading sounds like Buzzfeed clickbait, emphasizing and contrasting numbers with a sort of faux-punchiness to them, or overusing "the" to make things sound more definitive than they actually are, or... on and on, it really couldn't be more obvious.

> Six emails, one button

> The hotline sends you to the office. The office sends you to the hotline.

> The bank could. Google couldn't.

> Three companies at one address

> A company wih 105 million in capital, five employees and zero website

> The address Google did not show its customers for two and a half years

> People who left, and people who came

> Germany: the court that ordered the data to be returned

> Ireland: where the other party to the agreement lives

> What disappears with the inbox

Or the "in brief" at the top, which is spamming numbers as though they add dramatic tension. Or, just like, every. fucking. sentence. You know, years of reading this shit can't be distilled into a comment. Have you not been reading articles for the last three years? LLMs have very distinct voices.

Jesus fuck, would you get off your high horse already? This is every investigative article heading since before Dario or Sam even finished high school.

Seeing LLMs everywhere is another kind of AI psychosis.

Can you show me one long-form investigative article from, let's say before 2020, that is nearly this dense with this BS? A single headline or a punchline here or there doesn't count; what makes it an LLM tell is that it's fucking constant, being abused and misused to add dramatic tension to things that aren't dramatic. The "humans do this too" defense is so tiring because you people never draw a distinction between a human doing something correctly as a rhetorical flourish and a bot spamming a learned trope over and over and over and over again in every single sentence without any regard for whether it makes sense to.

Or, maybe, people have skill issues?

A class of skill issues that never manifested a few years ago but which now manifest 30 times a day on the front page of HN. Riiiiight.

Hard to say what you're trying to get at here - are you saying people are shit writers and thus write in a shitty, LLM-sounding way, or that people are shitty writers and use LLMs to do the writing for them.

The latter is pretty inexcusable intellectually and disrespectful of people who will read the slop, and as the other guy said, it's incomprehensible (to everybody but you apparently) that shit writing would suddenly, in just the last few years and entirely by coincidence, mirror the shit writing shat out by LLMs.

> as the other guy said, it's incomprehensible (to everybody but you apparently)

also a skill issue

Additionally:

> The phone sent me to the address. The address sent me back to the phone. The loop closed in one day.

> This is not a mess. A mess looks different. This is a design (...)

> One account, four companies. Each does exactly what it was created to do, and each is fully within the law. (...) Nobody here breaks the law. And that is exactly why there is no one to go to. (Attempted conclusion)

> One empty chair remains in this puzzle. (???)

> The eloquent individual history comes from India. (eloquent???)

> What does this mean for the user? Simply that when the inbox was disabled on September 1 (...) (Talking in third person about himself?)

> Google can be sued in Poland. (No shit)

> Google itself publishes the rules for such an eventuality, which sounds almost like politeness. (WDYM it "sounds like politeness"??)

It's so bad how hard AIs try to sound smart.

You asked for examples multiple times, got a plethora of them, and then went silent. Profile is old but the entire 1st page of comments relate to LLM language use and in particular how to sound like a human being and not a large language model.

A bit of cheeky botting? AI agent training mayhaps?

Milestone achieved: I got accused of being a LLM for the first time!

Training checkpoint complete.

Segment 3 of 26.

Reply on news.ycombinator.com