It is important to understand that the core disagreement is not whether Fedora should support AI development
Sad. Even the kernel seems to be going all in now: https://www.neowin.net/news/linus-torvalds-declares-massive-ai-fueled-code-surges-as-the-new-normal-for-linux/ I do hope there’ll be a discussion one day, so far no response yet: https://lore.kernel.org/lkml/e12330b9-c29e-45ca-9375-9e3d13426d85@horse64.org/T/
This wasn’t about integrating commercial AI, or even AI generated code, but supporting local, open weights AI.
Fedora has approved its own AI Developer Desktop initiative with a unanimous council vote. Three Atomic Desktop images are planned, two of them CUDA-enabled, and none of them would be phoning home to cloud services. source
Personally I saw it as a positive, when only
criminalscorpos havegunsAI etc. Local only alternatives should be encouraged.It is all corpo controlled though. The open AI is still using models created by corporations.
This is the correct way to do AI, use it when possible locally, without gate keeping its secrets and without violating user’s privacy. If AI is here to stay, I would much rather have alternatives like this than couple tech oligarchs dominating the whole field.
Its not AI that is the problem. Its the current investment and business model around the AI tech industry that is the problem.
And it manifests in way more places than just AI.
What’s wrong with the investment and business model? For me the big issue is the training data and licencing. As well as the issue that it genuinely destroys the value of labour at a time when living costs are high.
What I have observed in busienss is that smart people do smart things with AI. Just Like a well skilled person can do great things with a spreadsheet.
Stupid and lazy people do stupid and lazy things with AI, while looking smart. Just like that guy who build crap spreadsheets, but knows how to make them look pretty.
Most of the labour destroying practices come from market capitalism and shareholder culture. AI is just the excuse.
The IP theft is part of the business model in my mind, but I take your point. Its similar to the early days of the internet and the dot com boom and bust.
No this does not seem related to market capitalism or “shareholder culture”, yes some of the firings excused as AI disruption can be attributed to that but we must understand that every economic system wants to get the most output with minimal input. The disruption from AI isnt something that I think has happened yet. If the tech just affected 1 industry I wouldnt call it labour destroying. This is not just LLMs but vision models and general robotics all at breaking through at once.
Yeah, and the brain rot and that AI code is dumb and ruins projects.
So much more seems wrong than just the business model.
It’s that the asshole ddos archive site?
You mean the business model of investing big money into an industry that haven’t turned a profit because it increase the stock price of everyone involved until it reaches a critical point because it wasn’t designed to be sustainable in any form?
Exactly. Much like crypto, the productivity and capability gains are there, but the toxic shareholder value culture hijacked it for its own purposes.
Yes people invest in developing technology I dont know what rock youve been living under. They dont invest because investment pumps the stock that would be illegal and the market is so big theres very few who could move it. Instead they invest because the technology getting better raises the stock. I dont think anyone would claim AI hasnt gotten a ton better since the early days of chatgpt 2
As for sustainable, Inference is profitable but training costs a lot. Agentic models are driving up inference usage by 10x so its printing money they just need to own enough compute to provide the inference while training their next model. The market is growing like crazy, anthropic 30x their revenue in the last year going from 1b to 30b revenue and they expect to be profitable next year. Their latest model cost 5b to train.
The business model is fine. They’re over the hill and the tech has proven to be useful… unfortunately.
