An AI doesn’t have to have a purely logical structure (let alone a stupid one, e.g. structureless predicates for tables and chairs) in order to be able to logically prove important things about it. It seems to me that criticism of formally proving FAI by analogy to failed logical AI equivocates between these things.
An AI doesn’t have to have a purely logical structure (let alone a stupid one, e.g. structureless predicates for tables and chairs) in order to be able to logically prove important things about it. It seems to me that criticism of formally proving FAI by analogy to failed logical AI equivocates between these things.