Post #3101114

@zeenix@toot.cat I keep stressing this point, and I wish more people concerned about LLMs took it into consideration - instead of going full anti-LLM mode because of (mostly valid) concerns (loss of privacy, being on the mercy of big corps etc.), I think it would be more productive to focus how we can decentralize the entire technology and reduce the cost of hosting models by users. That is already sort of happening. I'm super happy with my Radeon AI Pro 9700 and hosting Qwen3.6 there. There are also boards like StarPro64 that draw little energy and can run 7B models. This is the better way than full dismissal and doomerism.