Aligning agents from outside, through institutions, can draw on centuries of tested practice in governing people (separated powers, audit, markets, rules for commons), while aligning them from inside rests on interpretability and training methods only years old; and agents are more auditable, designable and compartmentable than people, so that practice should transfer to them at least as well as it worked for us.

qualifiesUntitled source

It goes without saying that we haven't fully solved these problems with humans, so I would not advocate for a naive Panglossian view.

groundsUntitled source

they transform alignment from a psychological problem of internal model disposition into a constitutional problem of institutional design

groundsUntitled source

Computer scientists are rediscovering a lot of political science and economics from first principles, but luckily they get to skip the bloody revolutions and can apply well-honed game theory principles to systems that are more auditable, verifiable, designable, and compartmentalizable than humans are.