Kelland Thomas, associate director of the UA's School of Information (Photo: Emily Litvack/UANews)
Kelland Thomas, associate director of the UA's School of Information (Photo: Emily Litvack/UANews)

UA iSchool Project to Test Artificial Intelligence

Are computers capable of collaboration with humans? Professor Kelland Thomas plans to find out with the MUSICA project, funded by a $2.3 million research grant.
Aug. 19, 2015

Kelland Thomas, professor and associate director of University of Arizona's brand-new School of Information, recently began a research project to build a robotic jazz musician that tests the bounds of artificial intelligence.

His MUSICA project — which stands for MUSic Improvising Collaborative Agent — received a $2.3 million research grant from the Defense Advanced Research Projects Agency, or DARPA, a branch of the Department of Defense, to be utilized over the course of five years.

Jazz music and DARPA may seem an unlikely pairing, but the endgame has little to do with music. It's about collaboration. The project will address the question of whether information systems, such as computers, are capable of collaborating with humans. If Thomas' MUSICA program can effectively build a computer system that improvises in real time with a human musician, the answer may be yes.

Paul Cohen, program manager at DARPA and professor and founding director of the iSchool, commented, "I'm very concerned that we treat computers as servants, and because we treat computers as servants, we're not exploiting everything they could do for us."

Because "being able to view machines as colleagues" is a central goal of DARPA's Communicating With Computers program, "It was sort of a no-brainer that (this) kind of work would be selected," Cohen said.

According to Cohen, computers not only process large amounts of data in ways that a human cannot, they also do not have biases.

He points to economic policymaking as an example. The number of variables to consider, the amount of literature to pore over and the polarizing nature of some political systems make it a job for computers, he said.

MUSICA, which ultimately aims to create a robotic collaborator, taps into that potential. It uses jazz music as a vehicle for testing a computer’s ability to be a colleague rather than a servant. Cohen and Thomas agree that addressing some of the greatest challenges of our time — curing cancer, investigating climate change and ensuring global food security — could require this technology.

Information systems helping us get our jobs done better is not novel. 

For example, Uber, which owns not even a single taxicab, has been wildly successful by brokering information.

"Information systems match people who have a need with people who have a capacity," Thomas said.

If computers themselves had agency to make informed decisions using open-access data, it could be a slam-dunk for humans, too.

As for taking the jobs of human musicians, Thomas promises that his robot will do no such thing. In fact, he has enlisted the help of some musician friends to come into a studio for a jam session. He will record video and transcribe and analyze their playing to use as a model for the robot, which will look fairly unimpressive to the untrained eye — like a big, clunky processor.

Asked if the robot will have musical preferences of its own, Thomas responded, "It'll be interesting to see if it develops something like taste. I anticipate that these kinds of things may emerge and surprise us, but that remains to be seen."

He said the MUSICA project represents the potential of the UA's School of Information.

MUSICA "combines artificial intelligence, music and engineering," Thomas said, "and the big idea is that we need to move to an educational model that is intrinsically interdisciplinary. The iSchool is poised to contribute to great research that gets at the heart of important problems of the 21st century."