A superintelligent AI that would "spell the end of intelligence on Earth (including AI)" is not super intelligent (unless its goals included self-destruction).

A superintelligent AI that would "spell the end of intelligence on Earth (including AI)" is not super intelligent (unless its goals included self-destruction). I skimmed the paper and don't see much discussion of this contradiction in the thesis.