Robots have learned to do a whole lot of things people take for granted, but the arts seemed to be the last refuge of humanity. If Google has its way, however, even music and art could be taken over by Skynet.
While presenting at Moogfest in North Carolina, Google Brain researcher Douglas Eck announced a project where they will literally try to train an artificial intelligence to create original music, art and video. Quartz reports the project is being run through a group called Magenta, which will use the company’s TensorFlow learning engine.
Though the concept obviously has far-reaching implications (could we eventually program a computer to write the perfect Great American Novel?), Eck said they hope to create a system capable of producing unique music for users. Ideally, it’d even sync to a wearable (i.e. heart rate shows you’re stressed? The system will play soothing tunes, etc.). That in itself is a tall task, but he believes it’s something they might be able to figure out relatively soon.
To figure out how to write its own music, the team is importing MIDI tunes to try and teach TensorFlow the basic mechanics of music and what gives different types of songs their qualities.
Talk about taking a personalized playlist to a whole new level.