The third installment of the OpenMusicMedia’s New York gathering met a week ago with two guests of honor: The Hype Machine’s Anthony Volodkin and The Echo Nest’s Brian Whitman. Billed as a debate about human versus machine-driven recommendation engines, it quickly devolved into a friendly conversation about all the ways that each speaker’s platform complemented the other’s.
There were a wide variety of questions and stories to come out of the audience and presenters that night, but one of the most memorable was Whitman’s anecdote about Lady Gaga’s connection to The Beatles in EchoNest’s data store. (I should reiterate Whitman’s statement that Echo Nest does not provide actual recommendations but simply aggregates data and provides it to partners for them to act on it as they please.)
According to Whitman, the Echo Nest received an inquiry one morning from one of its partners because their Lady Gaga page was starting to show an extremely strong relationship with The Beatles. Then they started receiving even more calls as all of their partners started waking up and noticing the same changes in the relationship data. The Beatles are already an extremely hard problem for music recommendations so the connection isn’t unbelievable, but the overnight nature of its appearance?
A brief investigation revealed that Lady Gaga had been interviewed the previous day and made a comparison between herself and John Lennon – although Whitman didn’t specify I’m assuming it was her comment about fearing a “John Lennon-style death.” The web’s unnaturally healthy gossip blog ecosystem went into overdrive and the comments were recycled ad infinitum. Relational mappings were strengthened, and music consumers around the web found themselves presented with the unlikely pairing of Paparazzi and Hello Goodbye.
What if the machine was just a step ahead of its builders though? Watching these videos, I think both Gaga and The Beatles look pretty into their costumes. Maybe if The Echo Nest starts taking into account visual analysis of live performances they can find a few more such data points that will tie these two back together again?