The robot -- named Shimon -- with four arms and eight sticks can play harmonies and chords on marimba. It also thinks much more like a human musician, focussing less on the next note, and more on the overall structure of the composition.
The researchers from Georgia Institute of Technology in the US fed the robot with nearly 5,000 complete songs -- from Beethoven to the Beatles to Lady Gaga to Miles Davis -- and more than two million motifs, riffs and licks of music.
Aside from giving the machine a seed, or the first four measures to use as a starting point, no humans were involved in either the composition or the performance of the music.
"Once Shimon learns the four measures we provide, it creates its own sequence of concepts and composes its own piece," Mason Bretan, doctoral student at the Georgia Institute of Technology, said in a statement.
"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," he added.
As long as the researchers feed it a different seed, the robot produces something different each time -- music that the researchers cannot predict.
In the first piece, Bretan fed Shimon a melody comprised of eighth notes. It received a sixteenth note melody the second time, which influenced it to generate faster note sequences.
This leap in Shimon's musical quality is because it is using deep learning, which is enabling it to create a more structured and coherent composition, the researchers said.
Shimon's debut as a solo composer was featured in a 30-minute video clip in the Consumer Electronic Show (CES) keynote and will have its first live performance at the Aspen Ideas Festival at the end of June, the researchers said.