How artificial intelligence creates music

The use of software and IT tools in the musical field (the so-called computer music) is certainly nothing new: think for example of the increasingly pervasive use of autotune. However, with the rapid development ofartificial intelligenceeven the music field (in addition to the medical field, that of electronic devices, chatbots, etc.) has a new tool to deal with.

At present, in fact i AI generators – as Boomy, Soundraw, Aiva, Flow Machine – are revolutionizing the world of music production because they are capable of generate melodies, lyrics or entire songs following the instructions given by a user in terms of style, genre And personal tastes. An example of such an application is song Daddy’s Char of the 2017created using the software Flow Machine and launched by the label Sony. The song matters to this day beyond 3 millions of views on Youtube.

How AI “creates” music and lyrics

Similar to generating text or code using ChatGPT, you can train a artificial intelligence generator in the music sector. In essence, these are algorithms trained on enormous amounts of data relating to the music industry: every single song is then analyzed by the generator to try to recognize its peculiarities in terms of genre, rhythm, tone And style.

Once the model has been trained then, any user can interface with the generator by specifying the characteristics that the song must possess artificially composed: it will in fact be possible to request songs that follow the style of a specific artist or that recall musical trends of a certain historical period.

At the same time as learning about melodies, artificial intelligence tools analyze the lyrics of the songs that are included within the dataset from which they learn. Thanks to the analysis of the texts these tools are able to find patterns typical of an artist or a musical genre: words most frequently used, frequency and structure of refrains and the use of words more or less modern are found by the software in order to identify their characterizing elements.

After the learning phase, therefore, it will be possible to ask the generator to produce a song on the style of a certain one artist or create compositions that mix various styles and genres: this is how the so-called is created today music synthetic.


Limits and criticisms of the use of AI in the music field

Even in the music sector, the debate on the use of these instruments is more heated than ever in this period. As in all artistic sectors, in fact, the creativity and the genius of the individual artist or author have historically represented the fundamental parameter on which to evaluate the quality of a work. The pervasiveness of intelligence artificial in this context it represents potential discontinuity compared to this paradigm. The issue is controversial precisely because, at the moment, AI generators do not show real creativity like that of humans, “limiting themselves” to reorganizing patterns learned through the training process.

Many artists and authors they took sides against the use of such tools considering them potentially harmful for the economy of the sector based heavily on the compensation deriving from the writing and interpretation activities of authors and artists. The platforms Of streaming musical – like Spotify – have instead decided to create laboratories to develop tools of artificial intelligence to assist the creative process of artists. On Spotify’s official website you can read:

New technologies have always given rise to new ways of making music. (…)
Our new Creator Technology Research Lab will focus on creating tools to help artists in their creative process, with the goal of helping artists of all sizes realize their creative vision.