HIST 390 Blog

I really enjoyed reading about how Pandora is doing its “Music Genome Project”. It reminded me a lot of Claude Shannon and how he talks about turning things to information.  In this scenario, Pandora is hiring people to map music based on qualitative and quantitative variables.  This make it in theory easier to give listeners what they are looking for in a particular station on Pandora radio.  I think that in theory this sounds great in order to improve accuracy.  For instance, reviewers were asked to rate the song quantitatively in the length of the song, as well as qualitatively in terms of how smooth the vocals are.  However, one problem that some people had in response to this, and one person said particular, was that it caused too much homogeneity in their listening, taking away the element of ebbs and flows in the song’s differing tempos and other qualitative factors.  This person went on to likened it to how a mix tape usually gives you a variety of sound while also giving you enjoyable music.  I really agree with this statement and am unsure of how you can program to find music that gives you both likeness and difference that people seek in their music.  It can be complicated because people want different things to be alike and different as well.  Sometimes my reasoning for liking one song over another can be for an obscure reason such as certain lyrics at one point or that I don’t like how they build up the chorus.  However, I may really love an extremely similar song to that one because of changes that are very small.  I am not sure how you could differentiate or predict that for me using computer algorithms.