They created an algorithm that is 97% predictive of a song’s success.

Have a Formula Great song it would be great. Logically, composers always tried to obey, and record companies knew which songs to trust. But it’s not that easy. musical tastes they’re very subjective and it’s hard to tell which songs will succeed and which won’t. However, there are some tricks. For example, if the song is for a contest such as the Eurovision Song Contest, one can compile statistics on what was most liked in previous years. So at least you can know where to focus the song. But there is another, even more concrete way to find the perfect song: analyze brain reaction who is listening.

After all, when we like a song, the areas associated with it are activated in our brain. mood and energy levels. They will vary depending on the purpose of the composition. There are songs that aim to make us happier, while others aim to separate our souls. And wow if they get it.

Considering all this, a group of scientists from Claremont UniversityHe developed an algorithm capable of finding the perfect song by analyzing the brain activity of several people. They have achieved excellent results with 97% accuracy. But how exactly did they do it?

Artificial intelligence to find the perfect song

For this study, the authors expected the participation 33 peoplewhich were subjected to two tests. The first consisted of an interview about his musical tastes. As for the second, their neurophysiological activity was measured during listening. 24 different songs.

With this data, they conducted several statistical models, aimed at finding a combination of stimuli of a perfect song. Then, using an artificial intelligence algorithm, they compared these statistical models, looking for the best one. One of them was able to predict the songs that would later become successful, with accuracy 69%. Then they trained the algorithm on the information obtained during the experiments. That is, artificial intelligence was used to associate song parameters with positive brain activity. So, they went up to 97%.

In short, with the neurological data of 33 volunteers, they were able to find out if they would like the song. millions of people.

More applications in the future

This algorithm is trained with neurological data can help record companies and songwriters hear the perfect song. If what they are looking for is a commercial topic, of course.

However, the authors of the study believe that in the future things could go even further. And the point is, if devices that measure brain activity could be made as small and portable as wearable more could be done individual recommendations much more accurate. Currently there are platforms such as Spotify who recommend songs based on our tastes.

These tastes come solely from what we’ve listened to before. But, if it were possible, they would not only analyze what we heard, but also How did we feel when we heard this?. In this way, the recommendations will be much more successful. In fact, there is no such thing as a perfect song. What is masterful for one person may be insipid for another. There is a taste in variety, but the brain does not lie. This would certainly be very interesting app. We do not rule out that this will become something everyday in the future, which would leave Black mirror at the height of costume history.