OK, put down your pitchforks until you hear what Magenta has created as its first song… a 90 second piano medley.
No really, all robot-fearing musical luddites among us need not break into a cold sweat. Magenta isn’t here to replace musicians (just like Skynet, right?)… ahem. OK. Well, the intention isn’t to replace musicians, rather to give us creative tools to further dumb down improve the music making process.
So, before we delve into what Magenta is and wonderful it is being open-source, etc, etc, let’s take a listen to the first song it has created:
Now, we’ve got that out of the way, on to the tech!
Magenta, we’re told, has two primary goals. As a research project to “advance the state of the art in machine intelligence for music and art generation.” Magenta is all about developing algorithms that can “learn how to generate art and music, potentially creating compelling and artistic content on their own.”
Secondly, Magenta is about building a “community of artists, coders and machine learning researchers. Magenta is built on “open-source infrastructure around TensorFlow for making art and music.” The TensorFlow team tells us they are interested with providing audio and video support, with tools for working with MIDI. “For example, we want to make it super simple to play music along with a Magenta performance model.”
You can learn more about Magenta in this blog post from Google. http://magenta.tensorflow.org/welcome-to-magenta
So, much like existing tools we have available in most all DAW software and 3rd-party audio plugins, Magenta is designed to accompany us in the music-making process rather than replace the musician all together. I'm also now pondering the impact Google Magenta might have (both positive and negative) on the plethora of music software developers. Of course, Magenta isn't the first instance of machine learning for musical composition, but it might be the most prominent to date.
It’ll be interesting to see how Google’s Magenta project progresses. Being open source means we’ll get a whole lot of interest from coders and musicians. And you can follow along with by checking out the Github link below.
Read more: Enlightening article on AI for Composition