Humans as a PageRank Algorithm

When Filo and Yang started Yahoo! they had this vision of indexing the web...using humans. At one point I think they paid cadres of people to surf and categorize links. They couldn't keep up with the internet's rate of growth, and though they had listings for millions of sites, there were many millions more hidden and in the dark.

Then came Google with it's precise algorithms, which ranked and sorted the information in a mechanical fashion. This has been working great for years now, and Google's goal is to constantly modify those algorithms to expose even more information.

But something is happening. Everyone is calling it social software, or Web 2.0. Whatever. The point is, at sites like Flickr, del.icio.us, digg, reddit, and others I can't think of, users are categorizing images and links. If you need information on companies that can generate business logos for you, try looking for the tag logo on del.icio.us. Google can turn up some good results too, but the ones tagged by humans are usually very good.

It's especially useful for things like photos. Google's image search tries to be smart about how it picks up on what an image features, but it still overwhelmingly depends on the image's name and text adjacent to the image in the HTML layout. For instance, compare these pictures of Tucson: these are from Google and these are from Flickr. I suppose if you're looking for the Hyundai Tucson you like the Google results, but those are only appearing because people are likely to name an image of that vehicle by it's name. Tucson the place might appear in images named Spring vacation.jpg or My old house.jpg.

I have a deep respect for the search algorithms Google and others have, but you have to admit that users of these sites are doing amazing work, and they're doing it for free.