How an innate music instinct can manifest in the human brain without special training. (Image courtesy of KAIST)
DAEJEON, Jan. 17 (Korea Bizwire) — A research team at the Korea Advanced Institute of Science and Technology (KAIST), led by professor Jeong Hawoong of the Department of Physics, has elucidated how an innate music instinct can manifest in the human brain without special training.
This breakthrough, announced on January 16, utilized an artificial neural network model to understand the underlying principles of this phenomenon.
Despite the diversity of ethnic and cultural backgrounds, music has become a universal language with universal elements. To explore this, scholars have sought to understand the universality and distinctiveness of music across various cultures.
A 2019 study published in the renowned scientific journal ‘Science’ found that all ethnographically distinct cultures create music, often with similar beats and melodies.
Neuroscientists have discovered specific regions in the auditory cortex of our brains dedicated to processing music information.
Professor Jeong’s team used an artificial neural network to demonstrate that the cognitive function for music can spontaneously form through the learning of natural sound information, without any direct training in music.
The team trained the artificial neural network using a large database of sounds provided by Google, teaching it to recognize a variety of sound data.
Interestingly, they discovered neurons (the basic units of the nervous system) within the network model that selectively responded to music. These neurons showed little response to various other sounds like human speech, animal noises, environmental sounds, or machine noises.
However, they formed spontaneously and demonstrated a high response to various types of music, including instrumental and vocal pieces.
These artificial neurons showed response properties similar to real neurons in the brain’s music information processing area. This trait was not limited to a specific genre of music but was commonly observed across 25 different genres including classical, pop, rock, jazz, and electronic music.
Suppressing the activity of music-selective neurons in the network could significantly reduce the accuracy of recognizing other natural sounds, suggesting that the ability to process music information may aid in processing other natural sounds.
This implies that ‘musicality’ could be an instinct formed through evolutionary adaptation for processing sounds in nature.
“These findings suggest that evolutionary pressures for processing natural sound information could have contributed to forming a common basis for music information processing across various cultures,” noted Jeong.
He expects this artificial simulation of human-like musicality to be used as a foundational model in developing music-generating AI, music therapy, and music cognition research.
Kevin Lee (kevinlee@koreabizwire.com)
