Shimon is a four-armed, bug-faced, marimba-playing robot built by the Georgia Institute of Technology, that composes its own tunes. The robot has learned to make its own music based on a dataset of 5,000 tracks, including songs by Miles Davis, Lady Gaga, Beethoven and The Beatles. Ph.D. student Mason Bretan has spent over seven years in building the robot. Only the first four measures are fed to the robot as a seed or a starting point, after which Shimon composes the music and plays it.
Bretan says, "Once Shimon learns the four measures we provide, it creates its own sequence of concepts and composes its own piece. 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." Shimon is an experiment to explore the possibility of artificial intelligences creating music that humans will find appealing.
Shimon has moved beyond thinking about just the next note in a composition, and has an overall picture of the song, similar to how a human would approach a composition. The compositions by the robot sound like a fusion of jazz and classical music. Every time a seed is fed to the robot, it generates entirely new music that cannot be predicted by humans. The researchers cannot pick out the particular tracks that lead to the compositions, but can notice influences, such as Mozart.