Back to posts #743 and #745, I happened to re-listen to George Hotz last night who says that the danger of AI isn't its misalignment with humanity accidentally
paperclipping the world, it's tool-wielding humans who are misaligned.
He posits that the only way the world gets paperclipped is allowing a single AI safety board to consolidate authority over AI rules. The way to avoid paperclipping is to democratize AI to distribute its strengths across the population (a 2nd-A sort of argument).
https://youtu.be/dNrTrx42DGQ?t=3384
His TinyBox is engineered to perform to the normal constraints consumers have. It has to run from a standard wall outlet, and not raise suspicion from anyone monitoring power consumption. It's meant to run without assembling anything or major configuration.
SA today mentioned that AI generated entertainment would not be illegal, regardless of how disgusting the content is.
https://youtu.be/Aq7BfBTFVQ0?t=744