2023-06-19 22:00:00
Last week, Meta, the company led by Mark Zuckerberg, released MusicGen, a generative AI tool that creates music from text and melodies. Tech Church assured that artificial intelligence she was trained with 20,000 hours of music including 10,000 high-quality licensed music tracks and 390,000 instrumental tracks from ShutterStock and Pond5, major archival media libraries.
But how does it work? The dynamics are so simple that it impresses. Upon entering the platform, the user describe the style of music you want to generate. For example, “a happy song from country with acoustic guitars. This order can be accompanied by a melody in mp3 format that you want to use to guide the creation of your song. Finally, click on “generate” and so, as if by magic, the AI creates—or at least tries to— a 12 second demo with the instructions indicated.
A 12 second demo can be generated with the Meta tool Credit: Screenshot MusicGen.
According to a note by Mattias Bastian for The Decoder, MusicGen is unique in its ability to handle both text and music prompts. “Compared to other music models like Riffusion, Mousai, MusicLM and Noise2Music, MusicGen performs better on objective and subjective metrics that check the agreement of the music with the lyrics and the plausibility of the compositionBastian explained.
However, all generative AI models and systems bring difficulties. Meta researchers have published a paper paper in which they describe their research. In it, they include a discussion section in which they delve into the ethical aspects of MusicGen.
“We made sure that all the data we trained the AI with was covered by legal agreements with rights holders, in particular through an agreement with ShutterStock”, explain the researchers. In addition, they clarified that they are aware of the possible “lack of diversity” in the data set they used since most of them are Western-style music.
Another challenge highlighted by the researchers is that generative AI models can present unfair competition to artists. “We can ensure that all people have the same access to these models”they stated. “We hope they can be useful for amateur and professional musicians”
Furthermore, according to a note from Music Business Worldwidethe rise of AI platforms that allow users to create large volumes of tracks with the push of a button has also revealed the potential of generative AI for fraud of the streaming.
Although it is evident that generative AI brings many benefits, it also implies risks in many ways: ethical, professional, moral, technological. That is why Every day the word “regulation” grows more together with artificial intelligence. If not, who will be the next pop star? An AI bot?
This content was originally published on RED/ACCIÓN and is republished as part of the ‘Human Journalism’ program, an alliance for quality journalism between RÍO NEGRO and RED/ACCIÓN
1687212959
#Meta #launches #MusicGen #tool #creates #unreleased #music #lyrics #melodies
