It works as long as majority of the CS population dreams of a Google job. There are so many candidates that having a high false negative rate really doesn't matter. However, as Google slowly becomes less popular, it's current interviewing procedure might not work anymore.
I think that having high-quality open source project is a reasonable indicator of a person's technical abilities. For companies with a smaller stream of candidates, it would be silly to ignore such sources of information.
By the way, I am always surprised how obsessed people are trying to get into Google. After some persistent bugging, I went through a phone interview. I was invited for a follow-up interview, but decided that they could not offer the kind of position I was interested in. (No, I don't want to work 18 months in site reliability engineering when my interest lies with natural language processing.)
I'll start doing stuff on Github when I genuinely feel like it (probably soon, I have a few pending personal projects that would make sense there), but recently it feels too much like you should do stuff on Github because it's going to impact your future. Similar to how you should get a college degree, should nurture your LinkedIn profile (not sure anyone cares about that anymore), and so on. Bleh.
(Non snarky response) I dunno. I joined Google in 2008, and as far as I know Google was always like that. So, unless you claim developers have preferred Facebook etc. to Google for more than 5 years now, it's probably not related to the interview process.
http://www.glassdoor.com/press/wp-content/files_mf/133228554...
Also base rate is higher at Facebook, and interview difficulty is rated lower at Facebook.
Somewhat disappointing, as I'm a huge fan of Google.
Google interviewers are interested in how well you can code (or design a solution) in a whiteboard for 45 minutes. I'm not claiming that it's the best way, and who knows, it might be the absolutely horrible way to interview candidates, but that's the way it is here.