x
Bounding the impact of AGI — LessWrong