The thing about robot musicians is they’ll never trash a hotel room, show up to a gig drunk or break up a band after meeting an eccentric performance artist.
At least not yet.
The music-computer lab of George Tzanetakis at the University of Victoria is ground zero for the rise of robot musicians. But don’t think of it like anthropomorphic C-3PO strumming a mandolin – it’s more like a series of mallets linked to solenoids on a drum, backed by sophisticated software. Tzanetakis effectively wants to teach music to a computer, and for the computer to pick up on musical cues while jamming with humans.
“When you play sound, a musician hears what is happening. We are trying to add the ability to understand music to an artificial agent that performs,” says Tzanetakis, an assistant professor in the department of computer science, and in the department electrical and computer engineering – and the school of music.
“The idea is to make the system musically intelligent, to have robotic musicianship.”
It remains early days for robot jazz bands. Mechanical musicians are largely limited to percussion instruments due to increasing engineering complexity as each mallet and its actuator is added to the mix, and to the software that crunches the music in real time to adjust the tempo and volume.
“We want the musicality of the system to learn how to slow down, speed up, understand if it is too loud. We are still working on making sure it doesn’t get drunk,” Tzanetakis jokes.
Of course computers, even advanced learning systems, don’t understand music or sound or much of anything else – they just crunch data in sophsticated ways. Gabrielle Odowichuk, who worked in UVic computer-music lab as a masters student, said completely computer-driven music seems unlikely. Computers still can’t connect on an emotional level.
“You need a human component or its boring. All robots need a human component,” says Odowichuk who worked on gestural control of sound. “Often you can program in more randomness to make it more human. But when it comes to conveying something and connecting with an audience, computers lack expressiveness.”
Tzanetakis sees robots and computer musicality as the natural evolution in how technology influences how music is made. To him, it’s no different than the person who first strung wires and called it a piano, or plugged in a wah-wah pedal.
“Using computers and robots is the same process,” he says. “Musicians accept synthesizers, drum machines and DJ sets. If anything a robot is more innocent than a drum machine. It’s just another technical dimension.”
How this work will eventually influence the broader music world is hard to predict. Tzanetakis, a 37-year-old native of Greece who earned his PhD in computer science from Princeton University in 2002, tends to be about five years ahead of current popular technology.
Tzanetakis’ early work on acoustic signal processing laid the groundwork for popular apps such as Shazam, which can recognize songs by holding a smartphone up to a speaker. He’s helped develop melody-matching systems that identify songs by users humming or singing. His music sensing algorithms are at the heart Smule’s AutoRap app, which creates a rap song out of any set of sounds, and the Ocarina app, which transforms an iPhone into a flute.
Of course, when he began working on what’s called audio fingerprinting – systems that analyze, dissect and find patterns in acoustic signals – smartphone technology didn’t exist.
“If I do my job right, stuff I’m doing now will be commercial in about five years,” he says. “What is commercial now is stuff myself and others were doing 10 years ago. Ten years ago there was no iPod, no iTunes. The world was a different place.”
Tzanetakis expects the next big creative leaps to involve computer systems that can analyze and extract individual instruments and voices from complex music scores. The human ear can pick our individual conversations in a noisy party – computers cannot.
“It’s really a hard problem, far from being solved, but is actively researched,” he said.
Despite being at the forefront of intertwining computers and music, Tzanetakis, a jazz musician and pianist before he became a computer scientist and engineer, admits he doesn’t always practice what he preaches.
“I’m a sax player. I have typically avoided playing with technology.”