http://www.digitaljournal.com/tech-and-science/technology/robot-uses-deep-learning-to-write-its-own-music/article/495253

Robot uses deep learning to write its own music

Posted Jun 17, 2017 by Tim Sandle
A robot, designed by the Georgia Institute of Technology, has used deep learning, artificial intelligence and big data to write and play its own music.
Shimon  the robotic marimba player  can listen to  understand  collaborate with  and surprise his hu...
Shimon, the robotic marimba player, can listen to, understand, collaborate with, and surprise his human counterparts.
Image courtesy of Georgia Institute of Technology
The robot, called Shimon, looks a little odd. It has four arms and eight sticks. Appearances aside the machine can write and play its own compositions. The pieces are generated using artificial intelligence and deep learning. To achieve the impressive feat, scientists gave the robot some 5,000 complete songs. These covered a broad spectrum, everything from Beethoven to the Beatles to Lady Gaga to Miles Davis. In addition the robot was presented with in excess of 2 million motifs, riffs and licks of music.
The robot has produced two compositions using a marimba. They are short but they meet everyday classifications of 'music'. What is remarkable is the advancement in technology that led to the robot being able to do this in the first place.
Composition one:
Outside of a human sending the complete music and the mix of riffs, no human was involved with the composition or the performance of the music. The researchers, long term, believe "that real-time collaboration between human and robotic players can capitalize on the combination of their unique strengths to produce new and compelling music."
Composition two:
Shimmon is able to recognize classical chord progression and influences of artists.
The research is led by Mason Bretan and it has take seven years to achieve the two compositions. However, most of this was in development and the pace of composing is expected to accelerate. In a research note, Bretan says: "Once Shimon learns the four measures we provide, it creates its own sequence of concepts and composes its own piece."
He adds: "Shimon's compositions represent how music sounds and looks when a robot uses deep neural networks to learn everything it knows about music from millions of human-made segments."