It is also far from obvious that values in generic large minds can easily be separated from other large mind parts.

This point alone would, if true, undermine the argument of Shulman's that Hanson discusses. It deserves more emphasis. And it seems correct.