Comradeship // Freechat

I can definitely agree to that, and I also think these are important questions to ask but a lot of the discussion I see revolves around artists specifically (and specifically illustrative artists), leaving little space for other stuff. So thanks for answering sincerely.

The function of the state is to reconcile class differences, but class differences are irreconcilable. How can you make a proletarian rationally agree that they should be paid less and the bourgeois more? All history of civilization is the history of the class struggle; in capitalism the state works for the bourgeoisie, and in socialism it works for the proletariat. So while I would also like for AI companies to struggle their own way on the market (they couldn't compete against China anyway), they wield enormous power over the state, especially the legacy companies, and get whatever they ask for. In the same way Musk gets tons of subsidies for SpaceX and Tesla - the point is to funnel money to the bourgeoisie.

Therefore objectively speaking I think China is the single major factor for destroying the oversized western AI industry. Project Stargate was announced (500 billion for AI over the next years), and literally a week later Deepseek came out and completely demolished that idea before it even took off. It was built for a fraction of the cost of GPT with a fraction of the hardware, and suddenly a 500 billion investment didn't seem like such a good idea anymore. In fact chinese models are largely open source and they are so cheap to run, they don't even charge anything. It's hard to compete against free.

What we'll see soon enough in the west however is monopolization of AI; fewer companies will remain and it will be mostly controlled by 1 or 2 company (probably microsoft and google, maybe meta). I can't really say what the consequences of this monopolization will be yet.

It's not like there's a lot of novel model creators currently either - the costs are too prohibitive and companies like openAI are hemorrhaging money and their 20$/month subscription tier is never going to fill in that massive bottomless hole (they'd need 35 million subscribers for it to break even), so they rely on both private and public funding. Even after releasing the semi open source gpt-oss which boasts 200 billion parameters, the most popular open source models remain z.ai, deepseek and qwen - all chinese models, and you can run a 20B model at most on consumer hardware (with a 1500$ GPU). So even that bet didn't take off, nobody is even using oss beyond the novelty it seems.