Google's "Project Magenta" reunites a group of researchers from the high-tech giant's "Brain Team." Their aim is to teach artificial intelligence how to create compelling human art. The team has now released the first piece of music composed without any human intervention and been written via machine learning.
Project Magenta is using Google's open-source TensorFlow AI engine. By publishing it online, Google is opening the alpha version of the code for anyone to use.
Research in this area is growing fast and is global. However, at present it lucks standard tools. By sharing their machine learning toolbox, Google hopes to create an open standard for transforming machine learning into products and exchanging research ideas.
The AI-composed song was uploaded on a website online. The piece of music released by the Magenta team is a 90-second short clip of piano. The researchers have added a drum beat to give context to the computer's harmonic rhythm.
According to Digital Trends, the song is an interesting amalgamation of sounds. It is full of complex musical ideas such as repeated feeling, form and phrasing.
On the Magenta Project's website, researcher Douglas Eck said that machine learning has already been used for translation and speech recognition but has been explored much less "when it comes to creating compelling artistic content." With new machine-learned tools for artistic creation and music, new opportunities are opening for the future collaboration between AI and human artists.
As Billboard reports, the song created by Google's Magenta Project is just a four-note prompt. The researchers note that the most difficult part of Magenta is not composing a song but creating a compelling song that makes people want to listen to it.
The Magenta team is exploring this new realm of possibilities, hoping that artists and musicians will use the AI tools for new creations. The team will continue to present the AI's engine artistic creations to the public.