Why This Team Is Teaching Robots to Play Jazz

As the band gets into their groove, the guitarist gets ready to rip a solo he’s practiced for endless hours. The crowd knows he’ll pull it off perfectly. That’s because he’s a robot; it’s always perfect.

This is the future that researchers, funded by the US Defense Advanced Research Projects Agency (DARPA), are hoping to produce, according to Tech Insider. A team, lead by Kelland Thomas from the University of Arizona, is trying to teach artificial intelligence software how to jam to jazz music, in the hopes of one day designing robots that can play real instruments and make sweet music.

Thomas’ team will build up a database of music from jazz legends like Miles Davis and Louis Armstrong. Then it’ll force the software through machine learning to go over the music until it knows it inside and out. The software will use that knowledge to theoretically improvise on its own over jazz music.

“A jazz musician improvises, given certain structures and certain constraints and certain basic guidelines that musicians are all working with,” Thomas told Tech Insider. “Our system is going to be an improvisational system. So yeah, it will be able to jam.”

While DARPA has funded a lot of research in robotics over the years—including the recent DARPA Robotics Challenge—learning jazz wouldn’t seem to have many useful applications in the world of defense. But the core idea—knowing how to improvise solutions to structured problems—could be applied to a wide range of issues. It could help robots from the Robotics Challenge better react in emergency situations, or perhaps even on the battlefield.

As Tech Insider points out, the DARPA project isn’t the first attempt to create a robot band. In 2013, German scientists created the band Compressorhead to play Motörhead and Led Zeppelin covers. While Jimmy Page might not have much to worry about yet, we could soon have a future where robots generate and play our chart-topping hits. Though that might not be so different after all.