In my work I often implement facial recognition technologies in government applications like border control and law enforcement.
I was aware that the technology was moving into civilian applications but it hadn't struck me how useful this could be until today when I discovered a new feature in Picasa Web Albums, Google's photo sharing application. When you upload new photos it automatically detects any faces and then tries to match them to people you've previously identified.
With reasonable image quality the face finding accuracy is excellent. The only images that didn't work were old low resolution ones from my first ever digital camera and those where people are wearing ski goggles or the lighting is really bad.
The facial recognition algorithm does not appear to be cutting edge, but in this context that doesn't matter because the number of faces the image is being compared to is quite small - I'm guessing that most users would have less than 25 people in their list - and the system only needs to identify the two or three most likely candidates to be very useful, which is exactly what it succeeds in doing. It's not quite in the same category as comparing a face to 40 million others to make sure someone isn't trying to register twice!
The neat thing is that you can then easily filter your photos by person, and because Google assumes that people in your photos are likely to be in your Gmail address book you can also quite easily send your family and friends links to the photos that they are in. The one thing missing is the ability to find photos containing all of the selected people e.g show me all the photos with me and my wife together.