MixedContent

Huge props to Lee LeFever for having a big ‘ol BBQ this evening. Some neat ideas floating around. I was chatting with Tom Conrad from Savage Beast about their cool service. They hire 40 professional musicians to listen to songs and then grade each song according to a list of 400 attributes. The thing I like about hiring an army of experts (as opposed to the current en vogue method of collaborative community filtering) is that the
‘rich-get-richer’ effect that often accompanies these services is side stepped. The thing is that Savage Beasts’ whole process ignores popularity, focusing solely on the musical fundamentals. Ideally, I think this type of approach to ‘manually’ creating a taxonomy of music should be combined with the ‘what’s popular’ view. The goal of this algorithm would be to try and mitigate the effects of positive feedback loops, while still reflecting some sense of what’s popular.
But really, I just want to take a look at the Savage Beast database. The thought of a group of experts marking up the recorded history of popular music is just pretty cool. Recommendations are good starting point, but this data could also be used for automated composition or for ‘hit-screening’ an artist’s new single (Will this single be a hit?) or one million other ideas I haven’t thought of. Savage Beast, I want an open API to your data!
Also, I have to admit that in another universe, my lazy-sleep-to-noon musician alter ego would love the dream job of simply listening music all day.