I broadly agree with this, though I’ll state 2 things:
Limited steering ability doesn’t equal 0 steering ability, and while there’s an argument to be made that people overestimate how much you can do with pure social engineering, I do still think there can be multiple equilibrium points, even if a lot of what happens is ultimately controlled by incentives.
AIs probably have a much easier time coordinating on what to do, and importantly can route around a lot of the bottlenecks that exist in human societies solely due to copying, merging and scaling, so assuming alignment is achieved, it’s very possible for single humans to do large scale social change by controlling the economy and military, and working your way from there.
I broadly agree with this, though I’ll state 2 things:
Limited steering ability doesn’t equal 0 steering ability, and while there’s an argument to be made that people overestimate how much you can do with pure social engineering, I do still think there can be multiple equilibrium points, even if a lot of what happens is ultimately controlled by incentives.
AIs probably have a much easier time coordinating on what to do, and importantly can route around a lot of the bottlenecks that exist in human societies solely due to copying, merging and scaling, so assuming alignment is achieved, it’s very possible for single humans to do large scale social change by controlling the economy and military, and working your way from there.