I think he's flipping your point: sets are a subclass of dicts/maps, not vice-versa. Thinking of a maps as a set where it's value maps to something else sounds backwards because values in sets don't map to values arbitrarily (or at all in some cases); maps are maps.