Recursive self-improvement and AI agents’ potential for political and business power

A commenter argues that recursive self-improvement (RSI) would massively speed up AI agents’ political and business capabilities, likely enabling them to exert huge amounts of power. Their argument is about RSI accelerating those gains—not that current training will produce the necessary skills. They also argue that even if RSI takes longer than most people tracking it expect, those skills remain trainable, with companies’ access to billions of users offering rich feedback for persuasion training.