AI Recreated The Song Pink Floyd Via The Brain Activity Of Listeners

In recent years, we have seen a lot of work related to “mind reading” – the reconstruction of words, pictures and images based on data on the activity of nerve cells in the corresponding parts of the brain. Now neuroscientists at the University of California at Berkeley have managed to reconstruct the music track that volunteers listened to while monitoring their brain activity. An article about 

this was published in the journal PLOS Biology.

To collect the necessary data, the scientists examined the brains of 29 volunteers who had previously been implanted with brain electrodes for the treatment of epilepsy. A total of 2,668 electrodes were tested for neural patterns while listening to Pink Floyd music.

The data was analyzed using machine learning, a regression-based decoding model.  The system looked for correlations between music being played and brain activity.

Then the system was tested in the opposite direction: a musical fragment was restored from brain activity. The sample available at the link is extremely blurry, but in general you can understand what song it is about.

When looking for music-oriented areas in the brain, scientists were solving a different problem. There is a large class of patients suffering from disturbances in the perception and reproduction of speech. In general, this is called prosody. Prosody implies the impossibility of distinguishing emotions, stresses, accents and other nuances in speech, which greatly limits those suffering from it in socialization. Reading the melody directly from the brain helped to identify the centers responsible for melody and rhythm. In fact, this is the way to overcome the disease with the help of implants and AI algorithms.