GenJam (short for Genetic Jammer) is an interactive algorithm that learns to improvise jazz. Computer scientist John “Al” Biles of the Rochester Institute of Technology, who also plays trumpet, designed the algorithm in 1993-94. GenJam listens to what he plays on trumpet and responds.

Biles often performs with GenJam under the name the Al Biles Virtual Quintet and is available for bookings. In his words,

GenJam listens to what I play on trumpet and responds interactively when we trade fours or eights. It also engages in collective improvisation, where we both solo simultaneously and GenJam performs a smart echo of my improvisation, delayed by anywhere from a beat to a measure. Finally, it listens to me as I solo and play the “head” of a tune and breeds my measures with its ideas, which steers its solo on a tune in the direction of what I’ve just played on that tune.

The algorithm “learns” what works through a feedback process where listeners identify results as good or not good. Biles is modest about the results, saying that “After sufficient training, GenJam’s playing can be characterized as competent with some nice moments.” The improvised melodies don’t always have the directionality that a jazz musician — or any composer, for that matter — imparts to a phrase. The tunes wander without a sure sense of where to land. But there’s no reason to think that such an algorithm can’t become more adept.

Further information, including a QuickTime recording of one GenJam session, can be found at http://www.ist.rit.edu/~jab/GenJam.html.