"Unsettling" might be a overstating it a bit. The first slideshow[1] explains (on slide 38, Limitations) that face recognition techniques still have a long way to go. More clearly stated on slide 39:
"Face recognition of everyone/everywhere/all the time is not yet feasible.
However: Current technological trends suggest that most current limitations will keep fading over time"
However: Current technological trends suggest that most current limitations will keep fading over time
I'd say that is pretty unsettling, but then I worry about privacy. Given how facial recognition has advanced in the last 5 years, it's plausible that you could have an Android app to do their SSN trick with, say, 50% accuracy in the next couple of years - certainly 5.
No that really isn't plausible. Read their original SSN prediction paper [1] and you will see that the highest they got their SSN prediction (with < 1000 attempts, which is quite a few) was 8.5% for SSN's assigned after 1988. For 1979-1988 that number was 0.8%. Only in the smallest states with issuance after 1996 do they get their best numbers of +60% matches.
That's only the SSN portion. When you add in the facial recognition, the requirement of a facebook account with pictures and a correlation to place of birth, you're going to see those numbers plummet.
At best this is a clever hack that highlights the risks inherent with publicizing your life on places like facebook.
You're still only talking about current technical limitations though. Cameras are getting better, the recognition is getting better, and the trend is for people to have more and more information publicly available online. Maybe it won't happen in 5 years, but I can't see how it won't happen relatively soon.
I think you're missing the fundamental limitation of this: the SSN identification.
Even if I give you the ability to go from picture to real name (which to many is still fairly jarring) given only "open" databases at 100% accuracy you still can't come near your 50% mark. The leap from real name and place of birth to SSN is the limiting factor, and the reason for this is the randomness inherent in the assignment of the last 4 digits of one's social security number. Given the ability to guess > 10% of that range (they have access to all SSNs who are deceased as well), the researchers could still get only 8% nationwide accuracy on recent SSNs.
You would need a fundamental change in that algorithm to get close to the numbers you want. This isn't a technological limitation, it's a algorithmic limitation, and you can't really put time frames on these kinds of breakthroughs.
Remember that this is done using clear full-frontal pictures, FRT drop significantly in accuracy with even the smallest change in angle. Not to mention sunglasses, (facial) hair obscuring parts of the face and different lighting conditions.
Current techniques that try to deal with such variation need several (clear) pictures from different angles to build a model of a face to be able to recognize a face.
I'd go with decades rather than years before the limitations mentioned are resolved.
Try installing Picasa and spend a little time tagging your family photos. It's a little startling how good it is. It consistently tags me regardless of whether I have a beard or my glasses in any given picture. Likewise, it quickly IDed my parents in 40 year old family photos that they scanned and sent, based on recent images. Mrs. Browl and I enjoy photography as a hobby, I'm pretty sure Google could ID either of us with 90%+ accuracy from a photo taken on the street. With Facebook you have both the computers and the crowdsourcing at work, so their tagging may be even better.
I don't have iPhoto, but I'm a little scared by how good Picasa is. I have about a prosumer level of knowledge - not enough to develop image-processing SW, but enough to enjoy reading SIGgraph papers.
They fund a lot of things, but yes, I don't think I'd like to be on the receiving end of "You bear a distinct resemblance to someone we'd like to explode."
Current technological trends suggest that most current limitations will keep fading over time"
That may be so, but history and theory both tell us that the remaining issues (after eliminating "most current limitations") are going to be hard-ass bugs that prevent the technology from actually working, and that isn't even accounting for the Red Queen Syndrome.
"Face recognition of everyone/everywhere/all the time is not yet feasible. However: Current technological trends suggest that most current limitations will keep fading over time"
[1] http://www.heinz.cmu.edu/~acquisti/face-recognition-study-FA...