It’s the hit that won’t die. Science explains why.
Did any of us expect the Song of the Summer to come from a completely unknown 20-year-old rapper named Lil Nas X, prominently feature Billy Ray Cyrus, and be based around a sample from Nine Inch Nails? Or that the music video would feature full-on cowboy getups? Or that the decidedly hip-hop track would top not only the country music charts, but also the Billboard Hot 100, where it set the record for longest-running No. 1 single—an astounding 19 weeks and counting—of all time?
No. No, we did not. Alas, “Old Town Road” went there, and the world will never be the same.
But still … how the hell did that happen?
To find out the answer, researchers at the University of Southern California created an AI-based tool that assesses how we process and perceive music, which may shed some light on Lil Nas X’s unprecedented cultural takeover.
First off: Is this song even really country? Remember: Billboard initially removed the song from its Hot Country chart back in April because it did “not embrace enough elements of today’s country music to chart in its current version,” according to a company statement.
But according to Timothy Greer, a Ph.D. student in computer science at USC who built the AI tool, the answer is yes. Kind of.
The system determined that “Old Town Road” is country, due to its lyrics (horses, tractors, etc.); rock, thanks to the guitar chords and presence of NIN’s Trent Reznor.; and pop, when you mash it all together. Of course this thing caught fire—it’s a genre-bending song that tickles the brain.
“The lyrics are steeped in the country genre, but the chords and the instrumentation don’t sound like country at all,” Greer said. “The algorithm highlights the complexity of music, both in terms of how the music is constructed and how it is perceived, in other words, how people process it.”
To analyze the song, Greer created three models to help predict a song’s genre:
- using only chords
- only lyrics
- chords and lyrics
Greer used 190,165 different music segments from over 5,300 pop songs as training data to teach the system how to identify genre. While using only lyrics or only chords can tell different stories about a song from a more comprehensive analysis of both, Greer said that each method renders some useful results.
Most genre prediction tools rely on a song’s full audio file, which requires retrieving and processing a high-quality recording. Greer’s tool, however, can determine the genre of a song through just chords and lyrics. Since both are found online easily, that makes identification a much simpler process.
“We always say there is no hard-set rule for human experiences of music,” said Shrikanth Narayanan, Greer’s supervisor and coauthor of the paper. “AI and machine learning can provide a lens from which to look at this very human experience.”
This article was written by Courtney Linder and was published by Popular Mechanics on 19/08/2019