1.Understanding Music through Modern Technology
c) Autonomous Agents – An Accompanist in VR
Classical music used to be played by musicians, created by composers, and listened to in music halls. In folk music, one, or several musicians interpreted popular songs. Today, music, both pop and classical is often made with the assistance of computer tools, if not computer generated altogether.
In these situations a risk exists that the "added value" generally introduced by the human touch, differentiating the music interpreted by humans from the exact reproduction by a machine, is lost. This "added value" can be acoustic, perceptual, or cognitive, direct or introduced as a complex interaction between individuals in a creative process that is frequently not fully understood. An example of such "added value" could be, for instance, the adjustment of timing and dynamics by musicians, the intrinsic adjustment of the timbre to suit the orchestral sound and the intonation to fulfill the best consonance possible while remaining true to the well-tempered scale when necessary.
In order to achieve a reliable interpretation system in every layer of the music production system traditionally performed by humans, we propose the idea of using the concept of autonomous agents known from artificial intelligence in music. The main idea is to make the computer (program) behave as an autonomous agent – a virtual person or thinking being that can react on input as a conversation partner. These agents can be used in many roles in the music production chain, such as notation helpers, music instruments, virtual musicians or even automatic composers.
Ideally, a seamless co-existence of humans and machines in a single "orchestra" should be achieved - the humans could play and create music together with machines. In order to help overcome the interface issues between a machine and a human, the agents need to be made sensitive and adapted to the "out-of-band" processes going on during a performance - the non-verbal communication between the musicians and the perception of the various cues. The agents have to be also able to communicate their own information back to the humans in an understandable but not disturbing form, ideally mimicking the behavior of a human musician. One of the goals of this project would be to give a "human face" to the machine, to maintain the communication channels the musicians are accustomed to.
Such autonomous agents could be used both as a creative tool, enabling the exploration of possible uncharted musical territories and also as a teaching/learning tool, allowing a musician to rehearse and train, while receiving the real time feedback from the machine on his performance. These agents are to be created to enhance the learning tools of music, to understand the importance of the non-verbal skills of musicians and to elaborate the new creative expressions of music.
(Kristoffer Jensen, Lars Graugaard)