Google wants artificial intelligence to make videos and music

Google announced at the Moogfest technology and music festival that it wants to investigate the extent to which artificial intelligence is capable of making music, videos and other art with a new project. The project has been given the name ‘Magenta’.

Google uses TensorFlow for its research, machine learning software that the company made open source in November 2015. Google Brain researcher Douglas Eck introduced the plan at the festival, but admitted that artificial intelligence systems are “a long way from creating a long storyline,” Quartz writes. By this he refers to the fact that there are currently no systems that can generate ideas of their own.

The Magenta project is intended to make the tools it develops available to the public. A first step is to release a simple program that allows users to import music in the form of midi files into TensorFlow. This should enable the system to build up its musical knowledge. Ultimately, it should be able to provide a listener with ‘musical shivers’ with entirely self-generated music, according to Eck.

Another possible application of such a system, according to Eck, could be a program that determines whether a listener is stressed based on a listener’s heart rate and then generates soothing music. Before getting involved with the Magenta project, Eck was working on the suggestion and search feature in Google Play.

Starting June 1, Magenta plans to publish more information on its GitHub page and track progress on a blog.