What artificial intelligence can do with music
Humanitarian or technician, creative person or master of routine processes - unlike humans, artificial intelligence successfully combines all these roles. A striking example of this is the music industry.
You can hear cool royalty free music anywhere: in a mall, in a cafe, or even from a nearby car window while standing in traffic. In order not to miss an unfamiliar song you like, all you have to do is turn on the recognition app. The name of the song and the name of the artist in them seconds give out the artificial intelligence. True, behind such a rapid result is thorough preparation: to quickly recognize a tune, the program must first memorize it. To do this, neural networks are introduced to a huge library of tracks, and then the algorithms convert the sound into a spectrogram and decompose it into time, frequency and intensity.
A spectrogram is a graph. The horizontal axis represents time, the vertical axis represents sound frequency, and the color represents its intensity at a fixed moment. A low signal is represented by a red bar at the bottom, and a high one - at the top. The result is a picture consisting of colored horizontal stripes. Analyzing such patterns helps to recognize music. The same neural network approaches are used in the spectrograms as in the image analysis.
Let's say a person hears a song on the radio and wants to know the title and artist. The recognition software builds a spectrogram of the sounding piece and sends it to its track library. It then compares the "picture" of the desired tune with the spectrograms of other songs and chooses the most accurate match. The artificial intelligence recognizes the melody even through serious disturbances such as road noise or repairs in the neighboring apartment.
By the way, a neural network is able not only to identify the artist and the name of the track stuck in your head but also to roughly determine its genre. For this purpose, artificial intelligence is taught to find patterns of different musical styles. Such specific characteristics are usually inaccessible to human vision and hearing. But machine learning makes it possible to calculate musical genres from spectrogram images.
It seems that finding the "right" track for your mood in billions of songs on your own is almost as unlikely as falling in love at first sight. But thanks to recommendation algorithms, perfect matches are not so rare. First the artificial intelligence searches for people with similar tastes, and then statistical formulas are connected: the number of likes, dislikes, listens and skips for a particular song.
Song recommendation works according to a simple scheme: if Vasya liked track X, and then Petya liked it too, then when Vasya likes track Y, it is worth recommending track Y to Petya too. When the algorithm needs to find the next song, a ready formula is applied to the set of potential songs. The most appropriate one pops up.
"Cold" content not seen in the playlists of the mass listener spreads more slowly. But thanks to neural networks, unknown artists and niche music still have a small chance to make their way into the stream of recommendations. If we simplify all the technical nuances, we can say that in such cases, artificial intelligence finds out how often a particular user listens to songs with similar spectrograms and periodically offers him to get acquainted with new tracks.
Neural networks also help generate music selections for fitness, walking, or sleeping. Content editors -select tracks for algorithms, and based on -their spectrograms, artificial intelligence expands thematic recommendations.
In the past, only composers could create melodies. Now it is possible without musicians. In 2020, the AI Song Contest for neural networks was held in the Netherlands for the first time. It was won by an Australian artificial intelligence collaboration with koalas, kingfishers, and Tasmanian devils. The song was dedicated to the wildfires raging on the continent. The sounds of the animals were recorded in short samples, one or two seconds long. The algorithm combined them with the hits of all the previous winners of the real Eurovision Song Contest, and then assembled the samples into their own tune.
This is not the only example of a successful creative alliance between programmers and neural networks.
Artificial Intelligence can create music in three ways. The first is by constructing from ready-made "bricks" of sound - samples. In this case, the algorithm simply arranges them in the right order on several soundtracks, and an electronic arranger puts together the finished track. The second way is to generate sheet music. It's like writing a manual for a musician to play a finished piece using it. And the third way is to record a "raw" audio signal. In this case, the neural network itself creates sound waves that resemble, for example, Mozart or the Beatles.
By the way, neural networks can write lyrics for songs, too. So far, such tracks sound rather strange, so sonographers should not worry about unemployment. In addition, the "computer mind" is devoid of feelings. It cannot get into the emotional context and convey the feelings that made the authors create.
Artificial Intelligence makes creativity available to all, and music helps it develop.