The reality of superintelligence in the current development of AI is increasing gradually,as confirmed by the three approaches to superintelligence-generalized intelligence,embodied intelligence,and interactive intelligence-and there is a risk of superintelligence itself away from the human control.Therefore,alignment as a coping solution is of particular importance.There are two common misunder-standings of the concept of alignment we need to be urgently corrected in the alignment understanding.The value-based misunderstanding leads the discussion into the"value embedded trap";the technological misunderstanding is essentially technological,which not only obscures the real problem but also brings even greater harm.When superintelligence comes,we should not only consider the future of human beings,but also the fate of machines.Contractual ethics is a good choice.The"contract"between humans and superintelligence is not just a relationship of boundaries and constraints,but also a synergistic relation-ship of mutual molding and understanding in coexistence.In this relationship,human beings not only see themselves through the mirror of superintelligence,but also understand and shape themselves better.
NETL
NSTL
万方数据
