Flickr has just released two new, amazing features: "interestingness" and "clustering."
Interestingness creates a daily map (with the usual Flickr clean and easy user interface) of the most interesting photos posted to Flickr, as determined by an algorithm that considers how many users have added the photo to their favorites, how many have commented on it, and what the explicit and implicit relations are between the viewers and the poster (so that getting your wedding photo favorited by 300 attendees doesn't embarrassingly vault it to the front of Flickr's page for the day).
Clustering is a major improvement to Flickr's "tagging" feature, whose main failing to date has been the ambiguity in tags — some people tag photos of bushes with "bush," others tag pictures of the President with "bush." Clustering to the rescue — it automatically finds the congruences in tags and groups them according to the discovered relationships, so "cute" gets broken in to "cute kittens," "cute puppies," "cute babies" and "cute smiles."
This is brilliant stuff — I can tell already that I'm going to lose a lot of productive hours to playing with it.
(Thanks, Stewart!)
