I hate AI write ups!
It's always the same sentence structure which has "not this or that, but such such." The AI doesn't even make a proper sentence because it makes very terse ones, which makes it sound too dramatic than it needs to be, because it leads to a lot of pauses. That said, AI makes terse sentences because it doesn't know how to use commas to connect related ideas.
I'm not inherently opposed to AI, but seriously, I dread what will literacy be like for decades to come because even art and writing is being outsourced to AI. This will blunt the the most basic writing skills of humanity! I'm starting to sympathise more and more with John from "Brave New World" as time goes on. The "savages" have more soul than those who live with technology.
In George Orwell's 1984 weren't all of the songs and books and movies written by a "machine"? That's how they spread the doublespeak!
LLMs are only going by the statistical trends of their training sources, with a few parameters to help push it in different directions. So when you push through lots of variances in writing style through a black box, Because that's not how it forms output. Now, you can daisychain together a generation and filtration system that will give a better result, but too much of that will lose context and so the core meaning in the piece is in danger.
The better product will always come from the human using the AI as one of several tools to help them form something, rather than simply asking for the AI to spit out something polished. Which is how many are using it, and are either missing that the output is crap, or it's good enough for their purposes, i.e. low content or spam.
I do also wonder at the eventual effects of all this, as languages do evolve and change via usage, and we are using things differently even from a few years ago thanks to AI. It's not so much as preserving the "past", but just protecting better communication, even if it does (and will) change some.
You hate poor-quality write-ups that nobody put any meaningful effort into. The issue isn't the LLM but the user who thinks they can just offload the entire creative process onto the AI. That's how you get generic outputs. Plenty of people use it in a way that genuinely improves their writing. It's just that writing like that doesn't register as AI and it skews people's perspective. It's the good old Toupee fallacy all over again.