Post #3865047 · Elektrine

2026-07-16 18:12 UTC

@ferret The fact that you, literally in a server called workersofthe.world, CANNOT CONCEIVE of issues with AI-generated code other than codebase quality is absolutely baffling. Maybe before wrapping yourself in such aesthetics you should actually give a shit about the workers? The data centers that generate images and music are the same ones that generate code. They destroy communities and burn the planet for profit. They have totally reversed all the minuscule gains we had done against climate change. The tech companies that provide the LLMs are willing collaborators of mass surveilance and genocide. Anthropic, the company behind the most popular programming LLM, literally murders children. Microsoft, OpenAI and Palantir are willing contributors to the gaza genocide. By using their shit regardless you are validating them. LLMs are trained mostly by African underpaid workers sitting through extremely horrendous working conditions to catalog the massive amounts of data needed. LLMs are a fascist political project to deskill and devalue labour as extremely evident by all the tech companies having mass layoffs and the rhetoric of every single company pushing for them. And more importantly... Coding LLMs are also made with stolen human-made code??? This is like an extremely well known fact??? Recommended reading: https://tante.cc/2026/04/21/ai-as-a-fascist-artifact/ https://ali-alkhatib.com/blog/defining-ai https://henry.codes/writing/it-doesnt-matter-if-it-works/

Replies (1)

@ferret@hub.workersofthe.world 2026-07-16 18:29

@@witchgirls@woof.tech Only one of those is relevant given what I have already said, which is that, yes, LLM code is based on training on human-generated code. I think most programmers would agree that coding is not a creative endeavour on the level of art and music; it is a tool with a solution for a problem in mind. While you can definitely call a character someone's creation, you can't so much call a block of Python someone's creation in the same way, as it's very likely that, given the same problem, another programmer would create an almost identical code block to solve it. "It uses the same data centers—" Yes, it can. It doesn't need to, and depending on what model you're using, it doesn't. Not everyone is running a subscription to Claude Code. "The tech companies—" Then don't use them. I don't. "LLMs are trained mostly by—" For images and so forth, what we call "generative AI", yes. Underpaid African workers are not going through generated code and testing it for quality. "LLMs are a fascist political project—" Again, yeah, so don't support the five or whatever giant fascist companies. The same way I can choose to run Linux rather than Windows, and I can remove Google from all my devices, and I can leave Facebook and X — and still use an operating system, a phone, and social media, while divesting myself from that moral and ideological trash pile, it is possible to have LLMs that are not part of all that crap. It isn't all Gemini, Claude, Meta, GPT, Copilot, and Grok. You can use Deepseek. You can run your own model. And as I was saying before, if you're just using it to generate code, it doesn't even take much to do so. And, if you actually listened to Torvalds' interviews, this is exactly what he says: That AI is fine, but people need to get the hell away from relying on the big companies that sell it as a service.

Open ##4921148