Google Makes Art with AI20/05/2016
The inspiration for Magenta comes from other Google Brain projects, including Google DeepDream. The Magenta GitHub repository states that: "Magenta encompasses two goals. It’s first a research project to advance the state-of-the art in music, video, image, and text generation. Second, it's an attempt to build a community of artists, coders, and machine learning researchers." In a recent blog post, Google describe the goal of Magenta as: “[To develop] algorithms that can learn how to generate art and music, potentially creating compelling and artistic content on their own.”
The 90 second piano composition was produced by Magenta after being primed with four notes - C, C, G, C. While the start of the piece is pretty basic, it does enter novel territory and seems like it's going somewhere interesting before it's cut off. According to Google, the 90 second sample is "an interesting amalgamation of sounds: simple, but full of complex musical ideas like repeated phrasing, form, and feeling."
Douglas Eck, project leader of Magenta, believes computers are more than capable of creating engaging and exciting music if given enough training. According to Eck, this project could even have therapeutic applications, offering the example of wearable devices tracking heart rate while making synced music to relieve stress and calm people down. While AI music is unlikely to surpass Bach or The Beatles in terms of emotional complexity, the feedback possibilities of this project are very exciting.
The Magenta project works in TensorFlow, a general purpose tool for distributed computation that is used heavily in AI implementation. TensorFlow offers the multidimensional arrays and neural networks necessary to generate original works of art and other intensive tasks. According to the team behind Magenta, “The design of models that learn to construct long narrative arcs is important not only for music and art generation, but also areas like language modeling, where it remains a challenge to carry meaning even across a long paragraph, much less whole stories,”
Google are not the only big name in tech to get involved in artificial art over recent months, with Microsoft's ‘The Next Rembrandt’ project learning how to produce a Rembrandt portrait using 3D scanning technology from 300 existing paintings. Formed using 148 million pixels from 160,000 fragments of Rembrandt’s works, this convincing portrait asks a lot of questions about the possibility of forgery and the nature of creativity. While human artists will never be in threat, creative AI could change the face of the art world in years to come.
Source Image: Tinxi / Shutterstock