Ah, sorry, that's not going to happen in your lifetime. Facial recognition takes HUGE databases and Monstrous processing power. If Apple object to Girls in swimsuits imagine how ape**** they will go over any application allowing anyone to fetch their details.
Not going to happen.
I was thinking about some facial recognition scenarios some time ago. You wouldn't need a database of the faces themselves, only of the features extracted from their face (biometric data). The biometric data is comparitivly small compared to the size of a facial image. When I was finishing my degree the last project I was working on was a system that would learn how to classify photographs. The dataset I gave to train the system was composed of several thousands of photographs. Compared to the size of the photographs the pattern recognition data was small (only a few megs). For a facial recognition system the faces for training a recognizer could be gathered from social networking tools. Facebook comes to mind. I
While a sequention search through a large collection of facial data could be timely depending on the amount of data that you've collected. The complexity of the search can be lowered by several orders of magnitude by taking other factors into account. If the recognizer makes moderate success in distinguishing between males and females the average success rate would be improved. If the program takes geography into account then the success rate could be improved some more. I won't go into a full list of how to improve the outcome of a facial search here but it can be done.
For processing power if one didn't want to gather their own network farm it can be rented. Microsoft just released their pricing for their cloud database and processing services (Azure)and there are other providers available too such as Amazon (I've still got a trial account I've yet to use).
Some one could put a prototype for such a system together quickly. Neurotechnology has a Windows Mobile Facial Recignition SDK. You can try it for free for 30 days. Omeron also licenses their facial recognition technology. The SDK can actually do a fair amount of recognition on a moderate powered Windows Mobile device on it's own. I've tried it ouw by feeding it photographs of my family members and then the software on my phone was able to successfully recognize them. That's just using the software in a stand alone environment with what ever power the Windows Mobile device I was using at the time could give. Now let's say you made a facebook crawler that was giving faces to the same algorithm running in a distributed computing environment. A phone could send a photograph of some ones face to the service and a few seconds later the service could respond with the identity of the person.