Do-re-mi -fa-sol-la-AI… – the Republic

There’s jazz, pop, rock, blues, country, techno, salsa and bossa nova. And now there’s also AI-generated music, music created or composed with the help of AI algorithms or machine learning models.
Instead of relying on human musicians to compose or perform music, AI-generated music is created by inputting musical data into a computer program, which then uses algorithms to analyse, interpret and generate new music.
Obviously there are many genres, there are many forms that this music can take, including completely new forms, from simple melodies and chord progressions to more complex pieces of music that incorporate multiple instruments and styles. It is obvious that this is a relatively new field, where research and experimentation is still ongoing, however AI has already been used in a number of applications, from creating background music for video games and films to generating soundscapes for art installations and live performances.
The possibilities of use are many, first of all the composition: the AI ​​is used to create new compositions, both starting from scratch and analyzing existing pieces and generating new variations. One approach is to use generative models, such as neural networks, to create new melodies and chord progressions. Another is to use evolutionary algorithms to make phrases and musical structures evolve over time. Then there’s the performance: AI is used to simulate the performance of human musicians, with algorithms capable of generating expressive performances that sound as if they were being played by a human. This can include generating realistic sounds of instruments, such as guitar or piano, and even imitating the styles of individual players. In reality, we as music listeners already have a constant relationship with artificial intelligence: services like Spotify and Pandora use machine learning algorithms to analyze users’ listening history and generate personalized playlists and recommendations.
But AI is also useful in the field of music studies, it is used to analyze music and extract useful information, such as the tempo, key and structure of a song. This can be useful in a variety of applications, from automatically transcribing music to analyzing the emotional content of a piece of music.

Of course, there are territories and musical fields, such as jazz or rock, which seem so distant, in conceptual and expressive terms, from this type of research, as to seem “protected” from the invasion of AI. Instead artificial intelligence has also been used to create and analyze rock music. For example, a team of researchers from Sony Computer Science Laboratories in Paris used it to create an original rock song, entitled “Daddy’s Car”, back in 2017, with a program called Flow Machine, inspired by the style of the Beatles. The system was trained on a database of over 13,000 songs, including those by the Beatles, and was able to generate a song, which objectively had many of the characteristics of the original Beatles style, which was widely successful, with over 2,000 9 million views on YouTube alone. Another example is the use of machine learning algorithms to analyze the structures and patterns of existing rock music. This can help identify common chord progressions, melodic patterns, and other musical characteristics typical of the genre. Information that can be used to generate new rock music that sounds authentic and true to the genre. AI was also used to simulate the performances of rock musicians. For example, researchers at the Georgia Institute of Technology analyzed the playing styles of rock guitarists with machine learning algorithms and then used this information to generate realistic-sounding guitar performances. This technology is used to create entire bands or even just virtual “guest” musicians for live performances or recordings.
Spotify closely follows these evolutions and has long ago opened the Creator Technology Research Lab, to help artists in their creative journey and study the possibilities of applying AI in music. Today artificial intelligence creates, produces and plays entire albums. On Youtube, a simple search with “artificial intelligence music” is enough to obtain a long list of machine-generated songs, as well as on Spotify, Amazon Music and Apple Music there are already many pages dedicated to virtual artists. Like that of the Dadabots, an AI project created by two Berkeley students.

Today on newsstands with ‘La Repubblica’

The new era of artificial intelligence

March 23, 2023

But we must not think only of easy special effects or imitations. Generative music has been with us for a very long time. Even if it is not designed by an AI, what animates Brian Eno’s installations is art created automatically by sound machines, following the instructions of its creator. In short, the possibilities are many, and we are only at the beginning of an adventure that could reserve beautiful surprises.
* And this article was created using, of course, an artificial intelligence, that of ChatGPT.