Maybe I am getting paranoid but the formatting and the language of that greentext seems off

>>108962930
git is a tech trend I hate. That thing is far too complex for what it does. Someone is eventually going to make a non complex version control system and git will die a deserved death. Team foundation server worked perfectly fine for me, I have no clue why we need git.

>>108962943
It's AI generated.
Hints:
> space between marker and text in greentext, like this
>overuse of snarky metaphors and similes like "every button now its own sovereign nation state" and "F16 preparing for takeoff"
>whole story feels somehow placeless and contextless
>the quick back and forth structure of wrong assertion and clever rebuttal, like "that's not X it's Y" but over a few lines, see "but bro" line, rebuttal, sarcastic wrap up

>>108964548
> space between marker and text in greentext, like this
I didn't pick this up at all, at least consciously.
The first thing that tipped me off was the weird poetry style breaks between sections of the greentext that I have almost never seen anyone ever do.
Then there were those stupid assertions as you mentioned
> > latency
> > version drift
> > remote = service discovery
but the main thing was
>reality:
no one on here has ever spoken like that, but LLMs can't help themselves with this "your assertion/reality" structure

That does make me wonder how often other LLM generated posts do manage to slip by me.

>>108964724
Yeah, that reality one is a good marker.
"You're not imagining things. It's a genuine indicator of LLM generated text."
It's still possible to identify LLM generated text but it has gotten harder. I think it's still pretty doable if you cross the initial "something isn't right" threshold and think to apply scrutiny, but that goes first as the output improves.
Something I have thought about is that as people use these more they start to adopt some of their mannerisms, so we might see people who talk like LLMs too, but are actually human and the thoughts originated entirely in their brains, which confuses things.
Similar to image generation, the details are quite consistent now, but there are often structural things that can give you clues if you focus your attention. Not one thing that instantly seals it though.
You can think about the models and the ways people speak and the posts we try to determine the origin of as different language patterns which all influence each other and it opens up some interesting ways of analysing things.

>>108964845
The models did learn to speak from redditors in the first place.
People here like to run it loose so patterned text stands out, but you just need a little fine-tuning and most of those LLMisms can get worked out and now you have a model that can generate believable posts.

It will only get worse: models will get better, people will get more paranoid, they will start second guessing what they read and what they write. Crazy world.
Somewhat related, in linguistics there are language isolates, languages that are spoken by one group and seemingly have no living relatives. One explanation, like for Basque and Etruscan can be that they were spoken by a previous group of people. But then you have places like America and Papua New Guinea where they all descent from one smaller group yet developed thousands of languages.
Why? Obviously there is linguistic drift. But another factor might be that some tribes created languages for internal communication that are distinct and only known to them for secrecy or identification purposes.
Same thing we see various sub-cultures do, some of it happens naturally, some of it is done on purpose.

I feel that instead of people adapting LLMisms, people might start to try finding ways around sounding like an LLM
The problem is that LLMs can be adapted with the afore mention fine-tuning, but that can be slow as it requires gathering data first so technically for a time people can stay ahead, but there will come a time where a paradigm shift happens when language models can adapt to new trends with very little data and very quickly, so methods like these will become obsolete at keeping ahead
Combine that with audio and video generation improvements and we enter an information crisis era like nothing before

It is said that language is a very low trust form of communication and why we don't see it more in animals, because it takes a lot of trust first
This information crisis might be what erodes this trust,
after 200 thousand years

>>108965256
>people might start to try finding ways around sounding like an LLM
using no-no words like nigger
anyway, good post.