Agreed. The biggest challenge for Facebook or any other social network is reducing the signal to noise ratio -- finding the 12 people for whom the post about Billy's birthday is tremendously meaningful (Billy's Gramma for instance) and keeping it away from everyone else.
Google's right that the people most qualified to help in this effort are users. It is posters themselves who know the most about the content being shared and which people will be interested in it. Moreover, posters have every incentive to try to get their content to the right people... they're sharing it for a reason, to be seen, and spam accrues to their reputation. Hayek in purest form.
Google's mistake I believe though is asking users to predefine categories. The point of connecting tiny latent groups (the set of people who may care about Billy's birthday, the set of who may want to come out to the movies with me on some particular Thursday night, etc) is flexibility. To have to rigidly structure my communication channels ex ante defeats the purpose. How many posts will really fit neatly into the set of circles you establish? How often will you have to go through modifying your circles and how cumbersome will this process be (even if you can add old circles to new ones, are you going to have to be manually combing through friend sets to get them right)?
Twitter's boon, an external design gift from its early user base, is hash tags. Hash tags allow users to identify what a post is about, and then go even further by allowing recipients to participate in the process as well, identifying themselves as target audience through keyword search. Certainly this process can be meaningfully augmented with preset acquisition or sharing channels (such as followers or circles), with algorithmic suggestion (critical for discovery and helping us find the things we don't yet *know* we'd be interested in), and with demographic or behavioral levers ('share this only with my friends living in NY', 'share this only with people who click on more than x% of video links').
But an approach based on rigid preset structures I believe is clunky and inadequate. Moreover, for the company that has championed not only tag cloud organization but the very essence of paring content with those who might be interested in it any given time on an ad hoc basis (ie search), surprisingly ungoogly.
