AI-1: Intelligence Emergent
"AI Series" -- experimenting musical kintsugi to symbolise something abstract and unfold a story
(TM writes:)
We spent more time to make and develop the music plan for this new series than we've ever done for any of our other works. The plan was written first by me and developed by TI, and now contains about 30 pages in Microsoft Word.
The concept is, according to the plan, "Symbolism, Kintsugi and Story-telling," where we express people and things (or, in other words, subjects and objects) related to artificial intelligence in symbolistic manners.
We also thought about developing the kintsugi and story-telling methods, which we endeavoured in the previous project called "Cube Series", titled as "Le Cube Dans Mon Rêve (The Cube In My Dream)," especially in its last track called "Le Rêve Restauré (Dream Restored)." It was so exciting for both of us that FMT should pursue such a concept for the next series, which was this.
According to AI, whom I asked to explain Kintsugi in short:
"Kintsugi" is a traditional Japanese art form where broken pottery is repaired using a lacquer mixed with powdered gold, silver, or platinum. This method not only restores the item but also enhances its aesthetic value by highlighting the cracks, celebrating the history and imperfections of the object as a form of beauty.
In essence, it is an experiment to use fragments that symbolise something to perform a musical version of kintsugi, thus unfolding a story, especially further developing "Le Rêve Restauré (Dream Restored)" into something that has completely different themes, ways and nuances.
Even though I've quoted texts AI generated not only above but also in our other past notes several times, I'd never asked it for anything related to production. It's been, hence, my first time to use AI, which is specifically ChatGPT 4 & 4o provided by OpenAI, for composition. I've just known that GPT stands for Generative Pre-trained Transformer. Like it tells, I had it created some specific phrases and brainstormed for symbols to adopt here, which I describe in the notes on each track in more details. I had "it" done that; "it" is the GPT called "Phrase Composer for FMT," which I specifically created and pre-trained with FMT's objectives, principles, styles and past compositions.
Yet, I'm not sure until when the series will last...
AI-1: Intelligence Emergent
TRACK DATA
Composition tool: MuseScore 4, Studio One 7 Pro
Recording tool (DAW): Studio One 7 Pro
Number of tracks: Original file 79, Additional String file 26 , Master file 21
Sound sources: Presence XT, Sample One (All built-in sound sources of Studio One), Musescore built-in sound sources
Composition and recording period: Apr 30 2024 - Jan 10 2025
Opus Number: 116
Main tonality: Nearly Atonal (& D minor in Section 4)
Main scale: Constituent Scales (& Dorian from D in Section 4)
Main time signature: 5-4 and 4-4
Main tempo: 109 - 111 beats per minute (& 108 in Section 4)
The Journey of the Track: From Concept to Completion
(TI writes:)
Twists and Turns in Development
This track has undergone numerous twists and turns to reach its current form. While the initial concept was clear, turning it into a cohesive piece of music that was also engaging to listen to presented many challenges. Without a concrete vision of the finished product, we relied on a process of repeated trial and error. Although I’m pleased with the version we ultimately released, I still can’t say for certain if it’s the "right" version.
Starting with Minimal Tonal Specifications
TM shared a similar uncertainty. When he sent us the score for this track, he had only a vague idea of the tonal direction. The sole specification was the inclusion of typing sounds that would run throughout the piece. As a result, much of the timbre evolved significantly from the original version TM created. The only elements that remained intact were the typing sounds, piano, strings, and some bass components.
Drum and Bass Configurations
For the drum arrangements, we utilized three different sets:
TR-909,
Musescore's Power1 (live drum style),
TR-808 and TR-606.
These were switched depending on the scene.
The bass was divided into four parts, incorporating a mix of cello, double bass, and four synth basses. In some sections, bass parts were repurposed as "high-frequency" elements, deviating from traditional bass lines. Additionally, parts originally written as guitar in the score were replaced with synth-sequence-like sounds for a modern touch.
Feeling the Insufficiency
After completing the track, I stepped away from it for over a month and listened to it repeatedly. While it was acoustically well-constructed, it felt like it was missing something. The concept seemed to outpace the music itself, resulting in a lack of the essential "groove" that makes music compelling.
Adding ‘Pop’ and a Touch of ‘Corny’
Revisiting the concept notes and score led us to conclude that the track needed a more accessible, pop-inspired feel, perhaps even a hint of "corniness." This approach was inspired by the 1990s technique of adding rap or vocals to serious, club-focused tracks, transforming them into radio and TV hits. Though often overlooked, this strategy felt necessary for this track.
Incorporating Rap
TM introduced a phrase from Pink Floyd's Another Brick in the Wall: Part II. To echo the lyric, “Education now copies people like it manufactures bricks,” we created a rap section featuring lines like:
Exact, the cold facts, and still they try to Xerox,
The leader of the new school, uncool.
The phrases were taken from an anthem by a well-known artist from a completely different era and genre than Pink Floyd. These lines were cut, processed through a vocoder, and scattered throughout the track. For fans of this genre, these references would be instantly recognizable. The theme of "creating copies" resonated with the current state of music distribution and its parallels in other fields.
Refining the Sound and Structure
The version with rap was left untouched for another month of listening and analysis. However, we found the sound quality didn’t fully harmonize with the instrumental track. To address this, we replaced certain synth parts with live string samples and re-orchestrated sections. The second half of the track was also restructured, extending its length slightly and introducing subtle dynamic shifts. This improved the balance between the concept and the music, while adding a layer of underlying playfulness.
Final Adjustments and Completion
In December, TM requested a significant reduction in the rap sections. Initially, rap and vocals were present throughout the song, but we pared them down, leaving only the key phrases. With this adjustment, the track took on its final form. Even so, the question remains: is this the "right" version? Only time will tell.
Composition: First collaboration with AI
(TM writes:)
This work is the first one in what I call "AI Symbolism Series," where we express artificial intelligence in symbolistic manners based on the kintsugi and story-telling way. Also, it's the first one we've worked on with help of AI. (Even though I know that there are AI-composing applications, I have no interests in them at all as the products are too "standardised.")
In our plan this piece was supposed to have five sections and eventually it does like these:
Section 1: "Dialogues between humans"
The first introductory section is about online dialogues between two humans by symbolistically playing the sampled sounds of my typewriting on a PC keyboard and taking a role of percussion, a bit like Leroy Anderson's "The Typewriter" or Jarosław Kapuściński's "Oli's Dream," though very different.
The dialogue phrases in which two parts react each other are designed on a method I call "constituent scale." It consists of the component notes of a certain tension chord, especially more-than-five-notes chords, but plays them independently as a sequential phrase rather than a simultaneous chord playing altogether. (In addition to that, there is another reason that the tension chords or "tension scale" along them don't necessarily generate tension in my view.)
The first dialogue phrase starts with, just to give you an example, a constituent scale of six notes like G, B, D, F#, A and A again, and then, similarly, a constituent scale of C9. Later there appear other ones of eleventh chords, minor nineth chords, and suspended fourth (add9) chords.
Those shift like G9, C9, A9, C9, G9, C9 and A9 in every 1.5 beats -- or every six 16ths, whereas the opponent responded with C9, B-flat 9, E9, G9, C9, B-flat 9, and E9. That's how it goes. This is perhaps the most quirky and attractive one of many phrases I wrote experimentally for this series. To me it seems somewhat like talking in words. Each one of the constituent scales symbolises one clause or word the person is meaning, and a whole phrase is designed as a cluster of meanings, that is, an example of what I considered to be a form of Music Symbolism.
Section 2: "Dialogues with AI"
Then, it's developed into pairs of one constituent-scales phrase and a two-bars set of a chord phrase. This is another symbolistic way of what I may call symbolisation by contrast, one development from our previous work "Le Rêve Restauré (Dream Restored)."
The other contrast here is made between human dialogues in Section 1 and ones with AI -- more precisely, contrast between the phrases where chords are broken down into constituents and simultaneously sounding chords. Although Symbolism in music is quite challenging, it's enabled by making contrast with its previous section, which is quite equivalent to action to make context.
Section 3: "Evolution with Baysian Estimation"
Although you might feel it's suddenly peaceful and soft, basically this section starts with the constituent-scale phrase that appeared first in Section 1. The difference is that the bass and piano follow the scale religiously and play along the chords. Since the chords are nineth in major tonality, the atmosphere becomes peaceful more or less. But the section in overall is not necessarily tonal because the chord progression doesn't follow a single tonality.
Then, the motif that generative AI composed takes place. I asked it to "estimate the next series of notes according to the phrase so far" and it did that. Actually, the phrases it composed don't have the same structure as the ones I made, ie, constituent-scale phrases. Rather, it tends to create something very freely.
At the end of this section it composed a triad phrase just before the break; that was not my request but something it did on its own. How was that? The quality of AI's composition is not bad but just... ordinary in my view, but sometimes exhibits unpredictable or accidental brilliance.
Such fast development was NOT my request either, which I thought symbolises the evolution of AI.
It sounds like the time of 6-8, but it's not. That's because one pattern within the phrase has six notes and the bass and piano follow it; that's just it. One pattern takes 1.5 beats and one phrase consists of six patterns -- so, it's like nine beats.
Section 4: "Employment & Capitalism"
This section is another Symbolistic one. I covered Pink Floyd's "Another Brick in the Wall: Part II" and arranged it. Applying the song was AI's suggestion. I asked it for "examples of concerns or anxiety about something newly emerging" and one of the answers was it.
Hey, AI, can you brief it please?
"Pink Floyd is a progressive rock band formed in 1965 in the UK. They are known as pioneers of psychedelic and progressive rock, characterised by their unique sound, complex song structures, and deep lyrics.
"Another Brick in the Wall Part II" is one of their most famous songs, released on their 1979 album *The Wall*. The song critiques the rigidity and dehumanising aspects of the school education system, with the famous line, "We don't need no education."
Section 5: Human cognition
The theme here is that using AI is just to reflect the user him/herself.
I see AI as a mirror-like existence. Its responses change depending on the perspective and attitude of the person interacting with it. In other words, AI functions as a reflection of human thought and emotion, and the quality and depth of what can be gained from it vary greatly based on how one engages with it.
For example, approaching it with curiosity and a spirit of inquiry can make AI a mirror that brings out new perspectives and deep conversations. On the other hand, if it is used only superficially, it will return only shallow responses. Rather than having intrinsic value on its own, AI serves as a tool that reflects human attitudes and awareness, fully revealing its potential within the context of a relationship.
This section is created based on such a perspective.
Visual
(TM writes:)
It's also a work created in collaboration with AI; I asked ChatGPT 4o for an abstract image intentionally without telling it about the details. I also asked it to adopt Hasui Kawase's touch and texture.
Comments