Superintelligence

What would it mean for AI capabilities to substantially exceed human capabilities across consequential domains?

Orthogonality thesis asks a different question: what would such a system pursue?