What do I think about network states?
Metadata
Highlights
- One of the fascinating and under-discussed ideas from the rationalist and friendly-AI community is functional decision theory. This is a complicated concept, but the powerful core idea is that AIs could coordinate better than humans, solving prisoner’s dilemmas where humans often fail, by making verifiable public commitments about their source code. An AI could rewrite itself to have a module that prevents it from cheating other AIs that have a similar module. Such AIs would all cooperate with each other in prisoner’s dilemmas.