Virtual pleasures

Closer home

Experiments in using artificial intelligence to make new K-pop music are not limited to the genre's home country, nor to the music industry alone. At the Hong Kong University of Science and Technology's Department of Computer Science and Engineering, Andrew Horner's students have made a number of such attempts. The results include a mash-up music video featuring major hits by two of the hottest K-pop acts - Fire, a song recorded by the boy band BTS for its compilation album in 2016; and the girl band Blackpink's song Playing with Fire, released the same year. Students used AI tools to separate the sounds produced by each instrument in both pieces before recombining them.

"Like a good K-pop artist, AI is good at separating musical components as well as recombining them in fresh and interesting ways," Horner says.

His student, PhD candidate Chris Law Man-hei, sounds hopeful about the prospects of AI-generated K-pop. He says he believes the rapid advancement of AI technology as well as the variety of choices available to K-pop fans will inspire greater openness toward appreciating AI-generated music.

Xu Xinyang, also pursuing a PhD in computer music, says the application of AI can reduce the cost and speed of making K-pop videos available in multiple languages. She hastens to add that being the provider of the main idea and structure of the piece, the role of the human music composer is crucial when it comes to music created with AI input.

However, music generated via machine learning, more often than not, sounds highly similar to the training data. "It's like a hobbyist trying to sound like Ed Sheeran," Xu says. "People hate such copycat behavior."

Horner says that the key to making intelligent use of AI in creating new K-pop tunes "is to keep enough of the iconic hooks so that the audience knows that it is K-pop, while varying the music in ways that are emotionally and musically satisfying."