A growing number of AI-generated songs are uploaded to Apple Music And Spotify, and music label Universal Music Group said it involved copyright infringement. Some of these songs are AI-generated covers of existing songs.
UMG has issued a large number of takedown requests and asked Apple and Spotify to prevent AI systems from removing melodies and lyrics from their services…
Background
AI-generated songs are created the same way as AI chat systems like ChatGPT and Bard. Chat systems learn by analyzing huge volumes of text, and music robots do the same with the melodies and lyrics of existing music.
Both have been subject to copyright issues. Chatbots can use whole sentences, sentences or more, from human authors. Similarly, musical robots can reproduce both the music and the lyrics of human artists.
UMG licenses around a third of the global music market, representing major artists like Taylor Swift, Elton John and The Weeknd.
Apple and Spotify urged to block AI-generated songs
THE FinancialTimes reports that UMG has sent an email to a number of music streaming services – including Apple Music and Spotify – asking them to block the scraping of lyrics and melodies.
The email, seen by the newspaper, says in part:
We are aware that some AI systems may have been trained on copyrighted content without obtaining the required consents from, or providing compensation to, the rights holders who own or produce the content. this. […]
We will not hesitate to take action to protect our rights and those of our artists.
A FT The source gave the example of how a music bot prompt might include something like, “Compose a song with lyrics that sound like Taylor Swift, but the vocals have to be in the style of Bruno Mars, but I want that theme be more Harry Modes.” The output may be original, but – like the sampling – it borrows from existing music.
One company that has created a songbot is Google, with its model known as MusicLM.
We introduce MusicLM, a model generating high-fidelity music from textual descriptions such as “a soothing violin melody backed by a distorted guitar riff”. MusicLM presents the conditional music generation process as a hierarchical sequence-to-sequence modeling task, and it generates 24 kHz music that remains consistent for several minutes.
Our experiments show that MusicLM outperforms previous systems both in audio quality and respect for textual description. Additionally, we demonstrate that MusicLM can be both text- and melody-conditioned in that it can transform whistled and hummed melodies to the style described in a text caption.
However, Google found that the model would directly embed existing music into its creations in about 1% of cases. The company says it won’t release the model publicly until it resolves this issue.
Apple had not responded to a request for comment at the time of writing.
Photo: Photograph possessed/Unsplash
FTC: We use revenue-generating automatic affiliate links. More.