Optimization and the Intelligence Explosion — LessWrong