Music and Art Generation
DeepDream is a method we developed to visualize what deep neural net models were learning. The resulting images were visually fascinating, and suggested neural nets could be used for artistic purposes. Inspired by DeepDream, the Google Brain team launched the Magenta project, which directly addresses the question, “Can machines be creative?"
Magenta encompasses two goals. It’s first a research project to advance the state-of-the art in music, video, image and text generation. So much has been done with machine learning to understand content— for example speech recognition and translation; Magenta explores content generation and creativity. Second, Magenta is an attempt to build a community of artists, coders and machine learning researchers around machine learning tools for creation, via open source. Research areas in Brain surrounding creativity include generative models for media, learning in domains where we lack a clear measure of success (related to unsupervised learning), and reinforcement learning for real-time interaction with artists and musicians.
Visit magenta.tensorflow.org to track progress, including upcoming papers, blog postings and TensorFlow code releases.