The Music Buddha (MuBu)
MuBu (as the Music Buddha is commonly referred to) has a different approach to recommending music. Instead of having you rate music that you are already familiar with, MuBu presents you with seven-second audio clips to review before it makes seven album suggestions. MuBu CEO John Adams describes its technology as "an intelligent music recommendation service" and not a conventional collaborative filtering process. "Our interface is sound based. The rest of them are text based. We play a clip that you either like or dislike, and there is no text to evaluate. That's a key differentiating factor between MuBu and other sites. As for the back end, we could use a collaborative filtering system, but we don't. What we use is editorial expertise. Our editors define the content within our system, and when you come into the Buddha through the sound-based interface, we build a profile for you. We can then match your profile to the songs in our system and recommend the most appropriate artists to you."

John Adams, Mubu CEO

Acquiring results from the Buddha is as easy as rating the sound clips. Almost immediately after clicking "enlighten me," the results arrive. Conversely, the back-end technology is far from simple (and far more elusive than I ever imagined). The MuBu editors normally assign values to thousands of different reference points for every song. (The actual number of reference points per song is a MuBu trade secret.) This way, every recommendation will be that much more specific to your needs.

The Music Buddha

The MuBu staff stresses that MuBu is not a company that's designed to establish relevance between two songs in a specific category. On the contrary, MuBu strives to establish relevancy between you and the particular music in question - not to build genre associations. The best way that this can be visualized is to think in terms of genetics. Adams relates, "If you will, we have identified all the DNA points for music. The relevancy is numeric, and the definitions are textual. We don't link similar songs together, we link songs to the user. You and I could take two separate MuBu sound tests, and yet the same song could appear on both our profiles. This reflects the eccentric nature of people's musical tastes. This is not about putting songs together, this is about respecting the fact that all music is different and that, likewise, everyone has unique listening tastes."

Though by no means should this be considered a collaborative filtering system. MuBu doesn't consider any market data, nor does it compile databases of user opinions. Each MuBu user is influencing the Buddha with his or her own tastes, and these are not being affected by population averages. Adams further states, "There is a fundamental difference between our system and a collaborative filtering system. As you mouse over the sounds, you select how you feel about the music. As you do this, we build a similar double helix model that matches your DNA (your music choices) with the DNA of the songs (the parameters that our staff has defined) within our system through a proprietary and patented algorithm. This brings up a spectrum of music that's new and appropriate for you, and you only." Thus, the more you use MuBu, the better the recommendations will become. Finally, once you have located new music that you like, MuBu provides you with easy buttons with which to buy your selected CDs at various online retailers.

A human is the best recommender…