But if about half the people in the world are female and there are more male than female developers, then isn't it safe to assume that the likelihood of a random male being a developer is going to be higher than that of a random female?
When picking a random person from a phone book, yes. But that's not what this whole discussion is about. In virtually any real situation you don't have a completely random female. In most relevant situations you don't even have even close to equal amounts of men and woman in your sample.