There’s a story up on slashdot about computer scientists at Carnigie Mellon using Flickr to train software to determine the geographic location of a photograph just by looking at the elements within the picture. They’re using holiday photos people have posted on Flickr to do this, so at this very moment some chunk of LISP code could be checking out your teenage daughter on the beach.
The program attempts to mimic the way that humans can deduce the location of an image by searching for visual clues, such as similarities to pictures or locations they have seen previously.
On seeing a coastal scene, for example, humans can instantly work out that the image is not taken at a land-locked location and rule out the vast majority of options.
Using these techniques, the program was able to guess the right location within 200km, 16% of the time. This may seem like a low success rate, but it is nearly 20 times what random guessing would achieve, and extremely accurate considering the huge complexity of the problem.
Getting computers to mimic human intelligence has turned out to be WAY harder than anyone thought it would, but this kind of statistics based weak AI is finally starting to happen. I can’t imagine what its going to do to society and the economy when AI becomes a commercial product.
(Anybody like to guess where the photo was taken?)
*UPDATE. Poneke and George Darroch were pretty damn close, guessing Agra or somewhere in North India. Its the Jama Masjid mosque in Old Delhi. The red Agra sandstone gives it away.
Other guesses were the Blue Mosque or Versailles. It’s too small to be the first and nowhere near ugly enough to be the second.