AI and Composition
February 19, 2019, 18hs.
Tokyo University of the Arts, Music Environment Creation Department, Senju Campus, 7 Hall.
1-25-1 Senju, Adachi-ku, Tokyo 120-0034
*Reservation is not required
In the present, Artificial Intelligence (a.k.a. AI) is commonly used for automated driving of cars, automatic translators, etc.
Also, the news that AI has won the World Championship of chess has already been discussed. In addition, it is also used for music, as Google’s composer AI “Magenta” is one of the best-known examples.
In historical composition, total serialism, even stochastic music, etc., existed from the twelve-tone technique. Moreover, fractal music, genetic algorithms, fuzzy theory also became very popular in the 80’s. Algorithmic composition, which automatically generates all music data, tends to shift to composition assistance programming which eventually produces libraries according to the style of composition and flexible programs according to the tendency of composition.
These not only evolved as tools, but also, somehow affected the development of music or how to compose it. In one way, does AI's “deep learning”, which has been frequently spoken in recent years, seems to have some influence on the progress of composition?
Perhaps, are we just analyzing the music that existed in the past, and exaggerating what was just a reorganization of it, in relation to the fashion?
This year, the research of Goto’s Laboratory at Tokyo University of the Arts, was conducted based on the theme "AI and Music Composition”.
It started from a slightly oblique point of view of how to handle music with such AI. From the position of a composer, it’s been specially focused on how AI should be used for the development of music. Moreover, development and theoretical research concerning AI's technology alone were not the only focus. Also, it’s been limited to composing by AI, doing a musical score and having a person to perform it. At the same time, computers are not used on stage and other elements, such as the internet and sound design, are avoided at all. Only to be devote to the question: "What kind of new idea can be found using AI purely for music?" Surprisingly, time has been spent trying different tools, but questions about what effects does "deep learning” had on the progress of the composition, continue. Also, are these algorithms that we are discussing as part of AI, can be used to truly exceed the conventional “algorithmic composition”? Such a radical and frank question does not always have a negative meaning, but a constructive sense for linking to a new idea. However, the concert does not intend to provide a definitive answer,
but a substantial opportunity to lead to new ideas.
Therefore, the purpose of technical AI and to search for music of “deep learning” which is being done now, it seems will become something quite different. The objective here is to try to develop art. using in some way AI.
Artificial Intelligence and Music Composition (first half)
Presentation about Artificial Intelligence and Music composition (Suguru Goto)
Suguru Goto Duali II (World Premier) Recorder and String Quartet (１vn Kamei)
Haolun Gu Serenograph II (World Premier) Recorder and String Quartet (１vn Kamei)
Manjie Enchanted (World Premier) Recorder and String Quartet (１vn Kamei)
=== Break ===
Concert for recorder and string quartet (second half)
Tomiko Kojiba Fragments 2 (2018) recorder and string Quartet (1 vn Yohanan)
Salvatore Sciarrino Lettera degli antipodi portata dal vento (2000/2008)
Recorder Solo Toshio Hosokawa Fragment II (1989/2012 recorder version)
Recorder and String Quartet (1 vn Yohanan)
Duali II / Suguru Goto
Composing with AI (or, strictly speaking, having someone compose instead of you) it contrasts to the act of thinking intentionally with the human brain. So, I thought about what AI can do, by giving the priority to the things that humans cannot do. Usually, composing is performed by one person, and it is unlikely to be performed by more than one.
What I mean with this is not in the situations where the composition is divided into parts, like composing each section by multiple composers, or one person composes it and another one arranges it. What I mean is having to composers with completely different tastes and work techniques sharing a single piece of music on a sensory level and working at the same time. To put it another way, it is naturally impossible to have heads and hearts of human beings connected by cables, like in the internet.
So, we have no choice but to make communication by language. Although it looks like a science fiction movie, but if two composers with different tastes and techniques can really connect their heads with cables and do one composition with their own abilities, it probably will be worthy.
In this work, the attempt was made to make a composer who assumed different personalities, that is, an AI that assumed two different brains, on one computer, and to create one composition in collaboration within them. For example, one is a person that will try to work on complex compositions and the other person is someone who doesn’t like classical music, or a person who composes innovative music and the other person is not interested in music, such contrast is a completely different way of assuming. Specifically, with one constraint programming of AI creating complex music data, the other one with strong individuality AI, the not liked portion of pattern recognition will be filtered. It is a method of evaluating, learning, and remixing another person with another personality.
Serenograph II / Haolun Gu
This is the second music work from the “Serenograph” series and has been written inspired on the theme "Fragments and particles”. A faraway place from me, a place sterile and full of radiation. I consider that the inner voice, born from extremely quietness becomes music. Using a special method that we can call it as “sound effect”, more similar to a noise than a musical note, as it has no pitch, the idea of music, from the center of itself, raises. For this reason, some ideas and motifs were created and artificial intelligence based on OMax, was used to help with the concept of the work to be developed.
Manjie / Enchanted
This time I produced this work regarding on my research theme “Artificial intelligence and composition.” Using software such as Open Music, Max / Msp, Omax, and Audio Sculpt, I tried various ways in the process of composing music, such as process the structure of sound in the initial stage of the composition, the automatic generation system related to the structure of the work, and the automatic generation system of music in the development of the work. I intent to materialize this research theme and the concept of this work by combining Mongolian ethnic music recording materials and music materials recorded in Japan with AI technology and creating different sounds in the work.