The August issue of The Scientist ran an article on using online tools to find a job in which I was quoted. The article summarizes five tools for prospective job seekers: SciLink; NotchUp; VisualCV; Jibber Jobber; and Meeting Wave.
It's no coincidence that these are all social networking tools. The problem is that only one, SciLink, even makes an attempt to appeal to scientists specifically.
What will make a successful scientific social networking tool? My guess is that the ones that succeed will, as their primary benefit, make it significantly easier to find collaborators, grants, and jobs. A service established for any other purpose is bound to fail, no matter how much I (and others) wish this weren't true.
Strangely enough, making it easier to find collaborators, grants, and jobs is exactly the function that scientific publication serves today, when viewed from the perspective of those submitting papers.
Although the traditional scientific publication business model is in deep trouble, the future is very bright for publishers willing to make a clean break with the past.
Sound impossible? Here's one approach: bring original scientific content together with social networking features. Using recycled content from journals just wouldn't cut it; old-guard publishers and outsiders alike have been there and are doing that (without creating much that can't be found in a different form elsewhere). Original content - perhaps content "unpublishable" by today's standards because, for example, it's too narrowly focused or describes unsuccessful experiments - would be key.
Whatever the outcome, this kind of system would radically redistribute rewards and outcomes in scientific publication. Which is one reason why you won't find an established publisher willing to touch it with a ten foot spatula.
We're still a long way from that future (on a technological timescale). I'm not aware of a single service that has truly embraced the idea.
If you aspire to build a successful social networking service for scientists, regardless of the angle you take, focus on the boring basics: finding collaborators, jobs, and grants for your users.