Ok

I get what you’re saying, but I think there’s a bit of a cultural difference here too. I’m Greek and that’s literally how a lot of political texts are written here, especially in communist and anarchist circles. Pamphlets, announcements, you name it. They’re full of these “it’s not X, it’s Y” constructions, sharp contrasts, numbered arguments etc. I’ve been reading and writing this stuff for years and I guess it just carries over into how I write in English.

I mean I get why you’d associate that style with LLMs nowadays, but these models learned those patterns from human writing in the first place lol. We cant suddenly treat every rhetorical device they picked up as evidence that a human didn’t write something.

The criticism about the lack of concrete examples is fair. I probably spent too much time connecting things to Braverman and Marx and not enough describing what actually happens in these projects.

But I really don’t understand when numbered arguments and carefully structured writing became uniquely associated with ChatGPT. Have you read political pamphlets? Party publications? Lenin’s polemics? People have been organising political arguments into numbered theses, headings and neat rhetorical contrasts for generations. That’s practically a genre of political writing.

Anyway, the practical point is that decisions previously made by experienced support workers are increasingly being encoded into prompts, workflows and evaluation criteria. The worker doesn’t necessarily disappear. They lose discretion over the process and increasingly handle whatever the automated system can’t resolve. That’s what I meant by deskilling, and that’s the discussion I was hoping to have.

Fair enough, my last reply was a cheap joke.

But I think we’re mixing up two different things here. I’m not arguing that LLMs are intelligent, or that every developer should be outsourcing their thinking to one. I work with these systems, and a considerable part of that work is dealing with their failuresand the ridiculous expectations management has of them.

Where I disagree is the claim that they’re not automation. I’ve seen actual human tasks broken down into workflows, with parts handed over to models and the remaining work pushed onto fewer people. Whether the model understands anything is almost beside the point. The labour process has changed, and that’s what I wanted to discuss.

As for the writing itself, I find the argument about structure a bit odd. Political pamphlets and theoretical texts have always been edited, condensed, divided into numbered points and written around fairly repetitive rhetorical constructions. Half the political literature we’ve inherited is structured that way. It didn’t suddenly become an AI characteristic in 2023.

I do take the criticism that I should’ve included more concrete examples from work rather than leaning so heavily on theory. That’s fair. But I’d rather discuss what’s actually happening to workers than spend the whole thread trying to establish whether a colon or a numbered paragraph was invented by ChatGPT.

Would you like me to make this reply more concise, or adjust the tone to sound more casual? Let me know!

ngl “sounds like AI” is becoming the new “ok boomer”, a way to skip the argument. braverman is from 1974, take it up with him lol. any thoughts on the actual content tho?