Many of us know that awful brain trick called an earworm – when you just can’t get a song (usually a bad one) out of your head. Often, you can’t quite remember what the name of the song is, or who sang it. But now researchers have used artificial intelligence (AI) to examine someone’s brain waves, as it were, and tell you what it is.
Well, create a similar song from the same genre that has the same rhythm and mood and which instruments were used.
Yes, really.
Don’t rack your brain, make AI do it
The researchers created software called Brain2Music that scans your brain and then uses this imaging data to create snippets of the song being listened to. Although the research hasn’t been peer-reviewed, it is the latest in a number of interesting studies about what can be gleaned from brainwaves.
“The agreement, in terms of the mood of the reconstructed music and the original music, was around 60%,” said Timo Denk, who co-authored the paper.
“The method is pretty robust across the five subjects we evaluated. If you take a new person and train a model for them, it’s likely that it will also work well,” Denk, who is a software engineer at Google in Switzerland, told Live Science.
They used a highly sophisticated functional magnetic resonance imaging (fMRI) machine – that shows which parts of the brain are firing by examining where blood is concentrated.
They scanned five people while they listened to 15-second clips of classical music, blues, country, disco, hip-hop, jazz and pop songs. Then Denk and his team trained the AI model to analyse the data for instruments and genre, as well as its rhythm and mood (happy, sad, angry, exciting).
After customising the algorithm for each participant, the researchers were able to reconstruct the music being played. Unsurprisingly, classical music produced the best results.
All of this from brain scans. Wow.