Kıvanç Tatar, a newly minted Doctor of Philosophy from SFU’s School of Interactive Arts and Technology, has developed an innovative interactive AI music agent named MASOM for live performances during his Ph.D. research.
Trained with musical styles from famed electroacoustic composers of the past, MASOM has been jamming alongside human performers in an audio-visual performance project named REVIVE in multiple Canadian cities since 2018. On Nov. 30, 2019, REVIVE will put on a show in Vancouver at Performance Works on Granville Island featuring MASOM playing with Tatar and Phillippe Pasquier, and real-time visuals generated by Remy Siu.
Embedding interactivity in music AI
“With REVIVE, we don’t claim that we revive the composers on stage, but we revive their aesthetics by combining them with new tools of AI,” Tatar explains. “I would like to have my aesthetic roots in the past while gazing towards the future. The composers that we focused on in the project created fixed-media pieces; there was no interactivity. I think it is exciting to come up with an autonomous interactive music system incorporating their aesthetics.”
MASOM, he says, is distinctive in its interactivity while most of the current music AI programs are either generative or merely reactive. The creation of MASOM was inspired by the notion that ‘music is nothing by organized sounds,’ an idea proposed by electronic music pioneer Edgard Varèse in the 1940s.
Kıvanç Tatar has developed an interactive AI music agent named MASOM.
| Photo courtesy of Berkay Secil
“In my research, I was more interested in a generalized model of music so I can generate any music,” Tatar explains. “If music is nothing but organized sounds, then how do I organize sounds? I came up with an answer in two parts: first I organize sounds by timbre, by differentiating one sound from another, then I organize my sounds by time, by putting one sound after another in time. This is in any style of music.”
Brought on by his research of a generalized model of music, Tatar says at the moment he is also interested in modeling a general space of sounds using AI and exploring the possibility of creating new sounds using new technologies.
“What would be a synthesized sound of a piano and a plastic bottle? Maybe we couldn’t create that sound before but we are able now. It will be really interesting to use such a tool and explore these new sounds to make compositions, or interactive artworks, or to play them live on stage,” he says.
Currently a postdoc fellow at SFU’s School of Interactive Arts and Technology and a visiting researcher at the Institute of Computer Music and Sound Technologies at Zurich University of Arts, Tatar has also been working on a collaborative project between the two universities called Zeta, an immersive interactive space that incorporates audio, visuals and touch screens using deep learning techniques in AI.
With an experimental approach and a highly inquisitive mind, Tatar says he is also applying AI tools to make art pieces by training the computer to imitate styles from paintings of well-known artists. Whether it is art or music, Tatar says he is interested in discovering new aesthetics by using new AI tools.
An interdisciplinary man
A bit of a Renaissance man, Tatar studied engineering in university in Turkey while taking 15 elective music courses – ranging from instruments to music theory to concert hall design – at the same time. As a musician, he plays trumpet and electronics and has a keen interest in experimental music with a few compositions of his own.
“The aim of my career is to integrate science, technology, engineering, contemporary arts and design to research interdisciplinary topics to create transdisciplinary knowledge,” he says. “Scientific and artistic research are in a mutual relationship in my practice. I create new tools and technologies for artistic practices through scientific research. Real-world artistic applications are then the test scenarios of these technologies. These test cases in artistic practices often raise new research questions or ideas to generate know-how within the scientific research.”
On the creative potential of AI, Tatar stresses that despite the fact that AI can accomplish many creative tasks now, society should not approach these AI models as identities on their own and either fear the power of AI or fully absolve the creators’ responsibilities of potential negative impacts of their tools. For instance, he is strongly against the creation of autonomous weapons. According to him, AI tools are not independent of their makers in any step of the creative process from the conception of the idea, to the engineering of the tool, to training the program with selected data, even down to choosing the model output and aesthetics.
To learn more about Tatar’s work, please visit www.kivanctatar.com.
