I think it would be interesting to have google buy it. Pandora's strength seems to come from their massive dataset of song traits. Google could probably make good use of that data. They'd probably feed it into a neural net and see if they can teach a machine to detect those traits, if they haven't already.
I agree, Songza was a big win for them. I was really impressed at how quick they got their features moved into Google Music, not super typical for Google.
Google wouldn't do anything interesting like that if they bought Pandora. The first thing they would do would be to require a Google+ account to use it, which everyone would hate. Then they would keep changing the UI and the name over and over until no one knew it even existed or what it was called anymore, and then finally they would shut it down.
Google would probably integrate it with their existing play music offering. And I think they've backed off on pushing the Google+ logins due to the backlash.
Heh, the fully automated driving experience will be wonderful when the robots are driving us around and then music streamedninto the car is all robotically recommended to us.
now we just need fully automated robotic restaurants and we truly will be taken care of by technology.
You could also do was last.fm did and just base it on what songs other users were listening to. I found a lot of great music through last.fm and was never impressed by Pandora. I use neither now.
One of my dreams is to actually see how all of that works. I'm not in it for the music - I just really am a music nerd and would love to know how all that stuff is stored, pulled together, and queried.
For Pandora, they use a vector distance algorithm (with some other nudges to keep it from being repetitive, I'm sure). They break down a song on a scale of 1-5 across hundreds of "genes" and then pick songs based on how similar or "close" they are in the multidimensional space.
