Henk Tillman comments on AGI safety from first principles: Superintelligence