Safety

[King-Darius]

How can local LLM's be both safe and accessible in a way that will prevent future legislation just banning them causing lock-out for the open source community?

[mscottgithub]

I am curious about this question as well. The idea that of legislation banning any speech annoys the heck out of me. Anything that can be done to prevent that scenario from unfolding is good. But I suspect no matter what we do tyrants will be tyrants.

[byrongibson]

In the US there's already legal precedent protecting software as speech, originating with the Crypto Wars (https://reason.com/video/2020/10/21/cryptowars-gilmore-zimmermann-cryptography/). It will be difficult for the govt to pass a bill outlawing open source AI, since it likely won't pass Constitutional muster, and especially when Marc Andreesen, A16Z and the Accelerationist community start counter-lobbying against the AI doomers.

Regarding safety, I don't think local AI's are anywhere near the level of AGI yet. Only the most powerful datacenter AI has a chance at becoming AGI, and that can certainly be regulated.