Notes on Existential Risk from Artificial Superintelligence