Change My View

LLMs and generative AI will, and already has, replace "a number" of jobs.

LLMs have revealed something that shocked lots of job sectors: Quite often, having absolutely shit quality is good enough. 15 years ago, you could get two types of translation: Excellent, and None. So if you didn't want "none", you had to pay for Excellent, and as a result there was a very large sector providing Excellent translations of all sorts of materials. LLMs have revealed that many, maaaaany people don't want Excellent translations, they just want not-None. Suddenly, the market for Excellent translation plumets, and the market of not-None skyrockets. This has happened for quite a few different sectors. Especially the ones that work with text, like content writing, translations, etc. Then it happened for images. People don't generally want amazing artworks, they want not-a-blank-space.

Yes, LLMs generate slop, and will always generate slop, but we've recently been confronted by the reality that there's an absolutely immense market for slop.

But there's a different matter: LLMs are expensive, and nobody wants to pay for them. You're not paying what it costs to run the LLM you're using. If you're a private user, you're not paying what it costs to answer questions, generate cake recipes or make an easter-bunny picture. If you're a huge company paying millions for tokens, you're not paying what it costs to run the LLM. Tokens are hugely subsidized, but even they weren't, LLM companies couldn't afford to run. They're going to collapse 0.01 seconds after the venture capital stops coming.

Will the major collapse of large LLM companies end LLMs? Nah, of course not. Smaller, local LLMs will work just fine for many purposes. But almost every large scale LLM implementation is currently running on an industry that might as well be shoveling cash into a steamboiler.

So yes, "a number" of jobs will be replaced. It won't be a huge number, but contest writers, translators, basic artists and people in similar industries will be replaced, at least temporarily.

What will happen is that as the post-covid job bloat is reducing, the economy is slowing down and hitting its next crisis thanks to the US adventuring in the middle-east, there will be LOTS of people getting fired. And many of those companies will not say "We're firing people because the company isn't making money and we need to reduce our output" they will all be proclaiming "We're firing people because of AI, please don't think we're doing poorly!"

This is a hard one to debate because "a number of jobs" isn't really a statement you can counter. Yeah it'll probably replace somebody, the question is how significant will the impact be and will those people be able to get other jobs/will any jobs be created by it. Currently it seems to be creating far more jobs than it's costing, mostly just building places to put servers.

What you're probably looking for: Generative models are just an algorithm, arguably even a fairly simple one. We're already seeing the limits of how far we can push the current paradigm of model structures so without a big change in how they work things may stagnate. The problem with the current state of the art is that the average person can, if not always specifically call out generative model output, definitely note the mediocre quality. The most promising use case is writing certain types of code because it's repetitive and predictable in structure, but even there they reliably barf on moderately complex tasks where you can't easily let them run forever closed loop (as in, they can't self-test the program because it generates output that isn't text). Most other use cases fare far worse so if we're not seeing mass software layoffs I'm not convinced we'll see much damage elsewhere.

The only other case where one could argue it's truly all-powerful is scams because the scams deliberately target the bottom of the bell curve already and the models just make it easier to automate. So, does putting scammer foot soldiers out of a job count?

It does the thing it's for.

People have to get used to the idea of computers doing things they previously couldn't. Neural networks are a whole new kind of software, driven by examples instead of comprehension. They're already powerful enough for gigabyte-sized models to write code, animate cartoons, and photoshop images. Results vary wildly, but they're improving all the time, and it doesn't have to be genuinely intelligent to drag whatever you provide closer to whatever you describe.

Local models will be expected software. An OS lacking an offline do-as-a-I-say chatbot will be like shipping without a word processor and a music program.

Which is why - in line with rule three - it's gonna become a professional tool. Spreadsheets drastically changed accounting, to the point small firms could suddenly be one guy and an Apple II, doing the week's work by lunch on Monday. What happened next was not fewer accountants. What happened next was not less accounting. To whatever extent an LLM's ability to code is like a junior coder, you can hire a fully sentient human being to manage a bunch of junior coders. Or maybe it'll be a human instructing a single model, but getting an hour of fiddly changes within seconds, then spending ten minutes ensuring it did what you friggin' asked. For video game art, we can simply return-with-a-v to team sizes that fit in an elevator, and release games once a year for sensible quantities of money. Will a workflow with qwen-image-3 be exactly what an artist intended? No. But neither is a square meter of concrete that required three weeks of notes and revisions, inside a team with a hundred fucking people, for a live-service game that became a tax write-off after two months.

The people most worried about this technology are animators. Animation suuucks. Animation takes for-fucking-ever. Which is why everyone who fantasized about nuking Adobe HQ, after the Animate erasure fakeout, is liable to come around and quietly start using a program that simply fills in between whatever they draw by hand. Did it get things wrong? Alright, draw more stuff in-between. Once people can manually animate a guy, and then draw a medal on his jacket and have it Just Work for every other frame of the shot, you'll hear a lot less hue and cry about the stupid models that make a static picture dance real sexy.