> Should the freedom to harm others also be guaranteed as freedom?

everyone should have the ability to harm others, with the understanding that if they actually use it for evil there will be consequences. as in "i can guarantee freedom of speech, but not freedom after speech."

its the same principle as guns in america. you can own a gun and shoot it as much as you want but if you (try to) kill someone you go to prison unless you prove it was self defense.

the difference is with a physical weapon you can do things like registration or red flag laws to reduce the risk, but a llm is made of information that can be copied without anyone knowing. that means any laws to regulate ai at the level of models are unenforceable without totalitarian control over all computer hardware. that would kill free speech and real criminals would still find a way.

i think the right answer is to give up control over model capabilities and regulate uses. leave individual people and communities out of it. create an absolute safe harbor for open source and self hosting.

as you said, we need to focus on the ways ai concentrates power in the hands of those who already have a lot of it - big business and the state. that is where the biggest, and maybe the only, real harm comes from.