A friend from work introduced me to Music-Map, an interesting little application that creates a visual mapping of musical artists it thinks you may like, based on your input of one or more artist's names. For example, I typed in DJ Kevin Yost and learned that 9Lazy9, Slowly, and Soulstance are in close proximity, indicating that I may like their music as well. The closer two artists are, the greater the probability people will like them both.
The site is driven by Gnod, a self-adapting system that collects a variety of user inputs and draws associations between them (e.g., you tell it 3 bands that you like and it uses an adaptive learning engine to make inferences between your inputs and those of others). Music-Map lives at Gnoosic.com - a site devoted to music; it has two sister sites - Gnooks.com and Gnoovies.com - for books and movies, respectively.
While the user interface of the map is cool, the information Gnoosic is delivering isn't much different from what collaborative filtering-based sites have provided for years (most notably, Amazon.com's "Customers who bought this title also bought..."). It harkens back to my first job in the Internet space at a company called Charles River Analytics that in 1997 commercialized a personalization engine called OpenSesame (later acquired by Allaire) that identified similar relationships (also between music, books and movies).
One of the biggest challenges at Gnoosic is that it's tough to link to audio clips from any of these bands to confirm their relevance; Music-Map offers up a random assortment of Amazon links and Google ads that may or may not be representative samples of music. And some of these bands are obscure (since they're entered into the database by users), making it even harder to locate more info about them online (hence the lack of a hyperlink in the Slowly reference above).