This really depends on why you are doing this.

The concept itself is trivial and not exactly new. It has been around for ages in a form of private areas on BBS'es (remember those?). It is implementing the concept in a way that compares to Facebook - that's where the challenge is.

Also, realistically speaking, the chances of making people switch away from Facebook are really low, even if Facebook shows an astounding lack of foresight and does not roll out a circles-like paradigm really soon. People are lazy and Facebook ties are sticky. Those who are not on Facebook are not there for privacy reasons, so they aren't likely to flock to your service either.

So it all depends what you want to get out of the project. In any case, if you can and decide to continue, it would be a great learning experience on all fronts - from technical to design to marketing - I can guarantee it :)

Facebook did try to implement something like this with their updated Groups functionality, but it still hasn't caught on. There's definitely a wide open gap for somebody to make a usable circles-like social network, but it would probably need to use the Facebook social graph as a starting point to gain any traction, and would probably want to aim for an exit strategy that ends with either Google or Facebook buying them out.

I tried working on a similar idea a couple years ago, but dropped it because I couldn't find enough of a draw to make even myself switch over from Facebook to my network, let alone other people. I wish the OP best of luck in executing this idea better than anybody else has managed to yet.

I would say something Google isn't launching because they can't execute well enough but are still very interested in is the perfect idea to keep going with.

Personally, I believe the secret to social circles is building mechanisms to allow users to better control their current social networks, as opposed to hoping for wholesale migration.

Is part of your plan to allow people to continue to use their existing services/graph, and trying to slowly move them more into your network?

I considered building an exporter to existing services, but I'm lured into to a new site by the idea of controlling the design.

I need to see the execution before answering that question. So, maybe?

Yes.