RE: https://mastodon.social/@404mediaco/117592857554097967

I mean. You can also just not use LLMs.

Insane to me how many people think open-weight models like Qwen just come out of thin air:

• As if they aren't trained on smuggled NVIDIA GPUs and deeply reliant on NVIDIA remaining alive
• As if they aren't reliant on distilling American models
• As if they aren't built via Alibaba Cloud scraping the shit out of your websites
• As if they aren't reliant on the same exploitative labour relationships in poor countries, for annotation and training as the Americans use
• As if they aren't released by Chinese tech giants which do the same, and worse things, than the American ones
• As if they don't use massive data centres for training, mostly outside China or in poor regions of China, so people can pretend "oh China has no data centres". People are literally protesting these right now in Indonesia

Absolute L take from 404 Media

#NoAI