The current state of AI-music generation services : Can non-music professionals and music students correctly identify what music is human made respective AI-generated?

Sammanfattning

The primary purpose of this project is to gain a deeper insight into the current state of artificial intelligence, or AI, within the realm of music, what music generation services are currently available, what they cost, how much input is needed from a human for the service to generate a piece of music and what can be done with said musical piece post-generation. The secondary purpose is to determine whether it is possible for non-music professionals and music students to correctly identify which of two musical pieces is AI-generated and which is human made. The method used to research the questions surrounding the primary purpose of this study was literature study, whereby various scientific articles, AI-music generation service’s home pages and textbooks among other things, are used to gain a deeper understanding of the current state of AI-music generation services and AI in general. The method used to research the question regarding the secondary purpose of this project was the gathering of data from an online survey filled out by both non-music professionals and music students. The research revealed that there is currently a wide range of AI-music generation services available, each with different pricing tiers. The research has also shown that with the higher paid tiers, there are more benefits than the lower paid tiers, or the free tiers. The seven researched services have similar ways of generating music and they all require the user to choose criteria such as genre, mood, tempo, and key to generate music. The survey results revealed that the majority of respondents were able to identify which song was AI-generated, and that the majority of respondents preferred the human made song over the AI-generated song