Those boxes and boxes (and gigs and gigs) of old photographs make for fun moments of nostalgia, but they can also lead to plenty of confusion, too.

To wit: That interesting-looking house you shot on the family summer vacation 10 years earlier... but is that John Quincy Adams' childhood home or John Tyler's?*

Who knows. But when memory fails, why not turn to the web?

That's the promise of an upcoming technology in development at Google, whose developer says can identify 50,000 unique landmarks worldwide based on a snapshot alone and correctly identify them without the need for humans to get involved.

Using a data set of 42 million images stored on web-based photo sharing sites, Google's Jay Yagnik first spent time algorithmically separating famous landmarks from your average backyard, in part using GPS tags to help identify landmark-rich locations (such as Paris or Washington, D.C.).

The engine then took the collected photos of identified landmarks and analyzed the images under as many angles and lighting conditions as possible to put together a composite picture of each place, so to speak.

The engine currently has an accuracy of 80 percent, and Yagnik says his team is trying to improve on that in advance of an actual product release for consumers to use.

I love this idea, and I'm hopeful Google will keep working on it. In fact, I hope they expand it beyond just looking at old buildings. Could the engine someday be adapted to figure out where a landscape photo (sans man-made structure) was shot, based on the size of the mountains in the distance, or the type and density of foliage in the foreground? And what about people? Say I find a photo of a celebrity online but can't quite place their name. Surely Google could look at those cheekbones and figure out who they are, right?

*It's Adams' birthplace.