The problem here was a lack of variety. The music was over-personalized to the tastes of someone who liked Beyonce's vacuity. You could argue -- and I would have to agree -- that the restaurant staff severely misjudged their customers and put on the wrong channel. My concern is that with the emphasis on playlists so heavily personalized, there is no right channel for any random group of people.
That brings me to the second problem I have with the MGP: I don't think the individual listener will fully (or even remotely) appreciate all the human involvement in the choices represented by such playlists. Such a huge human effort goes into each selection and each playlist arrangement -- and what comes out is that? A handful of tracks that your average kindergartner could pick out by playing a game of which songs sound alike? (That is, if your average kindergartner were allowed to listen to Beyonce or Rihanna. As a parent, I would discourage that.)
Even if the MGP were far more effective in combining variety with listener preference, that still wouldn't prove that all that human effort was necessary. A rival, like Apple, Spotify or Rdio or any number of other parties that are starting their own streaming services, could easily come along and, with a fully automated cheap imitation of MGP, eat Pandora's marketshare for breakfast.
Listeners won't appreciate the level of human involvement in the Pandora experience unless it is shown to them. They'll listen to whatever's out there and if they can get a good-enough experience cheaply and easily somewhere else, they'll go there. If a service is smart enough to couple that kind of automated functionality with some serious branding and artist sponsorship, MGP will cease to matter.Worse, faced with such competition, all those fine musical minds on Pandora's payroll could quickly become an unjustifiable expense.