dcXhc
12 Mar 2003, 04:31 PM
Computer Hits
When a news release arrived announcing that a company in Barcelona, Spain, had developed an artificial intelligence application that could analyze a song and determine its potential to become a hit, it seemed to be a practical joke poking fun at the desperation and cluelessness of the music business.
But I regret to inform you that further research has determined that not only does this company, Polyphonic HMI, seem to exist, but there are already several major labels — Sony Music, RCA, and Universal UK — that are either using it or considering the option. Among its advocates is Muff Winwood, the former Spencer Davis Group bassist who is now an executive at Sony.
Tracie Reed, the vice president of the North American office of Polyphonic, said that the application, called Hit Song Science, was originally developed by Grupo AIA, a Spanish artificial-intelligence company. It analyzes the melody, beat, harmony, pitch, octave, timbre and other patterns in songs, and plots them on a chart that Ms. Reed said looked something like the Milky Way. Songs with similar elements tended to group together in clusters, she said.
One original idea for its use was to offer CD shoppers recommendations based on their favorite songs. But when executives at the company loaded every song in the Billboard Top 30 over the last five years into the program, they discovered that "lo and behold, they all had something in common," Ms. Reed said. What that something is, she added, is something that human beings cannot quite name but computer programs can recognize.
When the company discovered this, it decided to offer record labels the chance to compare unreleased songs to recent hit singles to determine their commercial viability. Ms. Reed said the company had already been paid to analyze unreleased songs by several major labels. She also said the company was offering a similar service to studios, so producers could get feedback on songs they were working on. "We think that Hit Song Science helps raise the bar for music," she said. "We empathize with the people who buy a CD and only like two songs. Let's get better music on more CD's."
A music executive who is working with the technology said the program also identified the "gold content" of a song. This is the part that is supposedly the most likable and can be sampled in television commercials, films and other songs. Several musicians and managers, when told about the technology, laughed at a future in which a computer program listened to their songs to determine if they were good enough to be released. Jordan Berliant, a music manager at Tenth Street Entertainment, said, "What creates a hit is that people have an emotional reaction to a song, in particular the lyrics. It's difficult to believe that a machine could gauge that."
Jaron Lanier, a computer scientist and musician who coined the term "virtual reality," said that the science side of the application seemed sloppy. "As for the music side of things," he continued, "I doubt pop music could get any worse, so using even a meaningless tool like this might result in some improvement."
Ms. Reed replied that the scientists at Grupo AIA were recognized as leaders in artificial intelligence.
When a news release arrived announcing that a company in Barcelona, Spain, had developed an artificial intelligence application that could analyze a song and determine its potential to become a hit, it seemed to be a practical joke poking fun at the desperation and cluelessness of the music business.
But I regret to inform you that further research has determined that not only does this company, Polyphonic HMI, seem to exist, but there are already several major labels — Sony Music, RCA, and Universal UK — that are either using it or considering the option. Among its advocates is Muff Winwood, the former Spencer Davis Group bassist who is now an executive at Sony.
Tracie Reed, the vice president of the North American office of Polyphonic, said that the application, called Hit Song Science, was originally developed by Grupo AIA, a Spanish artificial-intelligence company. It analyzes the melody, beat, harmony, pitch, octave, timbre and other patterns in songs, and plots them on a chart that Ms. Reed said looked something like the Milky Way. Songs with similar elements tended to group together in clusters, she said.
One original idea for its use was to offer CD shoppers recommendations based on their favorite songs. But when executives at the company loaded every song in the Billboard Top 30 over the last five years into the program, they discovered that "lo and behold, they all had something in common," Ms. Reed said. What that something is, she added, is something that human beings cannot quite name but computer programs can recognize.
When the company discovered this, it decided to offer record labels the chance to compare unreleased songs to recent hit singles to determine their commercial viability. Ms. Reed said the company had already been paid to analyze unreleased songs by several major labels. She also said the company was offering a similar service to studios, so producers could get feedback on songs they were working on. "We think that Hit Song Science helps raise the bar for music," she said. "We empathize with the people who buy a CD and only like two songs. Let's get better music on more CD's."
A music executive who is working with the technology said the program also identified the "gold content" of a song. This is the part that is supposedly the most likable and can be sampled in television commercials, films and other songs. Several musicians and managers, when told about the technology, laughed at a future in which a computer program listened to their songs to determine if they were good enough to be released. Jordan Berliant, a music manager at Tenth Street Entertainment, said, "What creates a hit is that people have an emotional reaction to a song, in particular the lyrics. It's difficult to believe that a machine could gauge that."
Jaron Lanier, a computer scientist and musician who coined the term "virtual reality," said that the science side of the application seemed sloppy. "As for the music side of things," he continued, "I doubt pop music could get any worse, so using even a meaningless tool like this might result in some improvement."
Ms. Reed replied that the scientists at Grupo AIA were recognized as leaders in artificial intelligence.