French music service Deezer has revealed that on average, 10,000 tracks generated using AI are submitted to their platform every day, making up around 10% of all music uploads. Deezer discovered these numbers using a tool developed to recognize the music created through generative models, such as Suno and Udio. The service is now working to detect “deep-fake vocals” with the aid of the same detection tool that made their alarming discovery.
Deezer clarified that their detection tool works most effectively when ‘trained on data sets’ from a specific generative AI model but is less effective at detecting AI-generated tracks when scanning unfamiliar data or models. Deezer’s CEO, Alexis Lanternier, expressed pride in the development of a technology that increases transparency for creators and fans while safeguarding the rights and revenues of songwriters and artists.
Lanternier mentioned that Deezer is banking on an upcoming AI-blocking capability where these AI-synth Pop songs will be the least recommended to potential listeners. This move was prompted by concerns from a study they ran in December predicting a possible 25% loss to musicians’ revenues due to AI-generated music by 2028. Deezer has pledged to detect AI music that mimics artists and popular tracks by 2023.
Generative AI has the potential to impact music positively, and its use must be guided by responsibility and care. However, this potential must safeguard artists’ royalty payments, a situation where Deezer has called for transparency. Deezer also urged the need for responsibility in harnessing AI’s power to positively impact music creation and consumption.
Deezer is challenging other online streaming platforms to institute similar tools to identify music made using AI and Bitcoin Music. So far, Spotify, Apple Music, and Amazon Music are yet to come forward about AI and Music
Read the full article from Mixmag here: Read More
