Hur skiljer sig graden av AI-bias mellan olika musikgenrer? : En studie med Suno AI

Sammanfattning

This study investigates how AI bias among listeners varies between three music genres: pop, rap, and drum & bass. The study is based on a survey in which participants listened to 10 songs from each genre, which were either AI-generated using Suno AI or created by humans. All participants rated, commented, and guessed the production method. The results show a significant degree of AI bias in the rap genre, where songs assumed to be human-made received higher ratings, while songs assumed to be AI-generated received lower ratings. Only 60% correctly guessed the production method in the rap genre. In the pop genre, AI bias was less pronounced, but participants in the study still tended to give higher ratings to songs they thought were created by humans. DnB showed the least AI bias, with an accuracy rate of 69.3% in identifying the production method. In DnB, several AI-generated songs received high ratings even if the participants assumed them to be AI-generated. The study suggests that AI bias may be more prominent in certain genres and that listeners' assumptions about the origin of the music influence their judgments.