An artificially structured argument for expecting AGI ruin — LessWrong