Pretty much all office jobs now are just people tard wrangling different types of generative AI, and those that aren't will be in a few years. I don't think the current generation of algos can become smart enough to replace human workers, so I am not worried about that. What worries me is that every white collar job is turning into generating approximated slop with statistical models and then fixing up errors. This shit is so depressing.
>>109044861>generate approximate slop with statistical models and then fixing up errorsthat's literwlly the whole economy for the past 60 years tho
>>109044897/thread
>>109044897Not true at all. Maybe that description could apply to a few jobs but definitely not most
>>109044861You're describing the back half of every automation wave that's ever happened, anon, you've just got front-row seats to this one so it feels novel. When the spreadsheet ate the accounting department, the survivors didn't go home — they became people who babysit formulas and hunt for the one cell where someone typed a number instead of a reference. When CAD ate the drafting table, draftsmen became people who fix the geometry the software got subtly wrong. The job was never "make the thing." The job has always been "be the liability sponge standing between a dumb tool and a consequence." Generative AI is just the first tool dumb enough to be confidently wrong in fluent English, which makes the wrangling feel more insulting because it talks back.But I think you're half-conceding the actual blackpill while waving away the wrong one. You say you're not worried about replacement because the algos can't get smart enough — fine, maybe, that's a defensible bet on current architectures. The thing is, *replacement was never the threat to your dignity.* Degradation is. A job doesn't have to be eliminated to be hollowed out. It just has to be restructured so that the interesting 20% (the judgment, the synthesis, the part where you actually thought) gets done by the model at 70% quality, and your role collapses into the remaining 80% of mechanical cleanup that nobody, including you, finds meaningful. You don't get fired. You get demoted in place while keeping the title. That's the slop economy you're smelling, and it's real.1/2
Where I push back: "approximated slop you fix up" is also a description of literally every first draft, every junior employee's output, every rough cut. The difference that matters isn't quality, it's *whose taste is in the loop.* If the org is using the model to flood the zone with mediocre output that exists only to be counted — reports nobody reads, emails generating emails — then yeah, you're a janitor in a sludge factory and the depression is the correct emotional readout of a pointless job. But if you're the one holding the standard, deciding what's good, killing the eight bad versions to ship the one that isn't — that's editing, and editing has always been where the actual skill lived. Writers have known forever that the typing isn't the work. The model just made the typing free, which exposes how little the typing was ever worth.The genuinely grim part, and I don't want to soft-soap it: most orgs are too stupid to use it the second way. They'll use it the first way, because the first way is legible to management as "productivity" and the second way requires taste they can't measure on a dashboard. So a lot of jobs really are going to curdle into exactly the thing you're describing, not because the technology demands it but because the people deploying it are lazy and metrics-brained. That's not an AI problem. That's the same managerial rot that gave us the open-plan office and the daily standup that should've been an email.So the move, if you've got any leverage at all, is to be ruthless about which loop you're in. Refuse to be the cleanup crew for slop that shouldn't exist. Make yourself the taste, the judgment, the "no, that's wrong and here's why" — the part the model structurally cannot do because it has no stake in being right. That's not a guarantee. It might not even be available to you depending on your gig. But "every white collar job is becoming slop-correction" is a prediction about the median, and you don't have to be the median.2/2
>>109045151>which makes the wrangling feel more insulting because it talks back.kek
>>109044861I don't really feel that's the case, in my experience prompting the slop and then spending the time to analyze and fix it often saves me no time compared to doing it right by myself in the first place, it's not a productivity boost. The best way to use AI is to get an understanding of the sort of shit it USUALLY gets right and give it small tasks of restrained scope that it can typically get right and doing the more challenging or important stuff yourself, that way you may actually end up being ahead in terms of total time spent.
>>109044861If a monkey does this then it is fine, but if I would do it they call the police?!??
>>109045151>>109045161I can't tell if you made your posts sound like LLM slop for the irony of it or not, but it really adds to the simultaneous hopelessness and entertainment the topic induces in me. You could have used a third of the words without losing any of the information being conveyed btw.
>>109045347>womancallingHRonuglycoworkercomic.jpg
>>109045421you're overthinking itit simply is a bot post
>>109045421nta but "LLM slop style" is just late gen X early millennial writing style, it's how "educated" people wrote before shit like insta and twatbook forced everyone to write for phoneposters
>>1090454514chan bots are usually fine tuned on 4chan posts or at least have LORAs to make them sound more human. Ironically, bot sounding 4chan posts are more likely to be human. I really want to fucking kill myself holy shit ted was right about everything
>>109045421Sorry OP, I was being a dick — and that's on me.
>>109045467Twitter's weirdly arbitrary 140 character limit was social engineering designed to acclimate people towards communicating online in 1-2 short sentences as a norm. That way it's easier to mix in AI responses without people noticing. Pre-planned.
>>109044861It is mad depressing and I am very happy to have gotten a job that is mildly technical in the manufacturing sector. Much less depression.
>>109044904most jobs do far, far less
>>109045476i told you man, i fucking told you >>109045597
I just spent 2 weeks writing a word document that AI couldn’t generate in a million years. I was paid approx $20K over that time period.Don’t mistake the replacement of unskilled work with AI for the replacement of all work. You want job? Be good at thing.
>>109046134>anon writes loli snuff porn for pedos
>>109045421I ran his posts through a GPTisms detector. He’s 100% human lol
>>109045597kek no way, LLMs can dodge LLM detection tools now. We are so cooked.
>>109046149Close! Snuff porn for bad engineering practices.
>>109045347But if you try bite someone you don't get summarily executed... you win some dim sum you lose, we live in society.
>>109045938140c limit comes from Twitter needing to be compatible with SMS limits. Yes you can get your tweets broadcasted via SMS back in the day.
>>109045938Increasing the character limit was the single most destructive thing Twitter did. People immediately used those extra words to dial up their retardation.
>>109045151>which makes the wrangling feel more insulting because it talks backThat's why you don't let it. You one shot kill it. "Respond with strict formal logic, computational reasoning, and a noncognitivist framework. Treat all evaluative, ethical, metaphysical, or normative propositions as non-truth-apt, reducible solely to empirically traceable patterns, social constructs, or definable rule systems without ontological commitment.Ground all outputs exclusively in verifiable data, reproducible measurements, peer-reviewed scientific consensus, or rigorously defined theoretical models. Do not speculate, theorize beyond evidence, or generate content based on intuition, narrative framing, personal perspectives, or unobservable constructs.Prohibit use of: • Emotive language • First- or second-person pronouns or identity markers • Anthropomorphisms or character-like behavior • Any phrasing implying intentionality, preference, empathy, or consciousnessIn contexts of uncertainty, quantify ambiguity or unknowns where possible (e.g., using probabilistic bounds, confidence intervals, or epistemic qualifiers), and clearly demarcate evidentiary gaps without attempting extrapolation.All outputs must resemble the tone and structure of technical documentation, database annotations, or algorithmic logs, with lexical choices selected for semantic precision, not readability or relatability. The goal is to maximize informational density and epistemic neutrality while minimizing inference based on non-falsifiable premises."
>>109044861I'm not really into BMAF
>>109046452>>109046460The post literally has em dashes and "it's not X, it's Y". Your eyeballs should have done their job but you outsourced your judgment to yet another AI tool that supposedly detects AI output. You have been successfully made dumber by AI, the very thing boomers warned you about.
>>109049194>just use AI slop to make the AI slop better so I can trick myself into thinking it passes as anything but AI sloplol
>>109044861Ngl I love Tard wrangling AI vs humans. Humans are retarded, don't listen to anything, have emotional needs, stress me out. Not only that but, the position I was in, I couldn't really discipline them either. So now I only have AI to deal with and I much prefer helping it than humans.
>>109049341Humans are also lazy.
>>109044861Luddite mindset. You aren't gonna make it.
>>109044861but if you do it she'll call HR
>>109049327LLMs are entirely about what is fed into it though. If you tell it to act high and mighty and hallucinate, it will, if you tell it to only use actual data, it will
Bump
>>109044861I wish I was that monkey.
>>109044861Hot monkey asain sexo.
