Google Magenta Publishes The First Song Created By Its AI

Google has a new project called Magenta, and it is aiming to teach machines how to create art — including things like music, of which we've been graced with the first sample of its musical talents. It's a short piece and appears to contain only four notes (with some human-added drums and other instruments added in after a while), but it's not bad for being the work of algorithms.

Magenta hails from the Google Brain team, which is aiming to teach machines to create art and music. The project uses TensorFlow, and the researchers are publishing tools and models for the public on Github. According to the team's Douglas Eck, they will also be accepting code contributions 'soon,' and will eventually publish tutorials, demonstrations and technical papers related to their work.

The researchers have expressed two goals for Project Magenta, saying that first and foremost, it is being undertaken for the purpose of research in order to improve machine learning art and music intelligence. This is the same kind of AI foundation that is learning to understand natural human language, for example, and it could result in new types of art as seen through the eyes of machines.

As well, Magenta also seeks to "build a community of artists, coders and machine learning researchers," says Eck. That's why the team will be making tools available to artists, saying it that it doesn't know what musicians and artists will do with them, but that it is "excited to find out." Only a small group of researchers are onboard with Magenta for now, but contributors will be invited to participate once a "stable set of tools and models" has been created.

SOURCE: The Verge