Dates are of interest to us mainly because they're one of the few ways we're allowed to differentiate one name heading from another. If we were able to differentiate entities and their relationships to resources and to each other without finding happenstance bits of information about them, then our interest in birth dates might fade. I'm more interested in information that would help me determine whether author X wrote resource Y, and in being able to identify resource Y with author X uniquely, than I am in knowing when author X was born.
What's really needed is the implementation of a distinguishing piece of data that's always available (e.g., the LCCN) in a structured identifier so that we no longer need to build undifferentiated personal name authorities for persons we would happily distinguish if only we had a birth date, fuller form, etc. Once the uniqueness of a person's authority record is switched to a machine-processable identifier rather than the current name heading, that identifier can be used more successfully to locate information about the person via linked data stores--e.g., affiliation, other authored titles, etc.--thereby making the decisions about who likely wrote what simpler. Authority records could move from being sketchy, stand alone records that often don't tell us enough for the decisions we need to make to being the focus of a richer range of information drawn from disparate sources through use of a unique, structured identifier.
We need to keep our eyes on the ball, and I don't think birth dates are the ball.