Thoughts on an Artificial Intelligence Music

Brian Questa

July 6, 2020

Review of the piece Accumulator by the Autoacousmatic


Written for a course at the University of Music and Performing Arts Graz,

Applications of Artificial Intelligence in Interactive Music Systems with Artemi-Maria Gioti

Accumulator is a piece of music generated by Nick Collins´ Autocousmatic software, a program which uses machine listening in order to generate an acousmatic composition. The user enters sound-files which function as the source material for the program. The user also determines how many channels the resulting composition should be. Using this information, the Autocousmatic generates a piece of acousmatic music.

Acousmatic music is difficult to define. It is generally considered to be music made for presentation on loudspeakers, the source of which is not always apparent. This description, however, basically fits all of music today. Acousmatic might thus be more of a style, or a genre. Generally, however, in acousmatic music, the composer uses collected sounds, either recorded in the real world, such as through field recordings, or synthesized sounds, and often a mix of the two, to create a composition where traditional qualities such as harmony and melody are paid less attention to, for the sake of a focus on timbre, spectral variation, and the spatialization of the sound material.

The question of what acousmatic music is, is undoubtedly essential for the creation of a program which can create it. How it is defined would give us clues to the successes and failures of such a program. What is the success criteria of AI generated music? Is it how well it replicates the style? How cleverly it breaks the style? Our subjective qualities of judgement are brought to the forefront when thinking about music generated from artificial intelligence. The music points us to our own beliefs and prejudices, causing us to reflect on some of the most basic questions surrounding our definitions of music.

Accumulator can be found on the SoundCloud page dedicated to works by the Autocousmatic. Indeed, the piece, with its mix of field-recording-esque sounds and synthesizer bursts, including sudden changes in texture and filtering, reminded me of an early Luc Ferrari work (a well known acousmatic composer). Indeed, on a certain level, Accumulator sounds like all of the acousmatic pieces I have ever heard. I could easily imagine it in an acousmonium, or one of these halls, where the audience is surrounded by 60+ speakers. I can hear Accumulator swirling around the room, as if I were attending a concert at the International Conference on Computer Music.

I do have to wonder immediately, however, when considering my thoughts on the work, how am I affected by the fact that I know this was made by a computer? What are my biases? Can I get rid of them? I ask this because Accumulator sounds exactly as I expected it to sound like. Before listening, I considered the qualities that an AI generated acousmatic piece might have, and I indeed found them in Accumulator.

In other words, it sounds random. It sounds slightly within the given style, but also sounds slightly off. It also seems as if the sounds don't have any intention behind them. Indeed, listening feels like a waste of time. I feel as if I don´t need to hear anymore. I feel as if I got the entire gist of the piece within the first few seconds, and that there were going to be no more surprises. Indeed, a "surprise" would only confirm my opinions that nothing mattered.

But why? A reflection on why this music sounds this way might be interesting. Why does it sound random, jittery, even unreliable?

Perhaps even with abstract music, there is a hint, or a sign of human logic at work. I would define this logic as that thing which we as humans think we understand when judging, or experiencing, or listening to creative works. In other words, we speak a certain language, even if the language is non-verbal. In the case of design or music, for example, we can attempt to understand the way our object of contemplation, our piece, our work, is organized. We can imagine, if not directly ascertain, how the person who created the work has organized it, and what that person was trying to say.

If I think of the sonata form of the Classical Era, with its main theme, secondary theme, and with its development of these themes, the phrases within these themes, the “sentences,” the motives - all of these contribute to the feeling of a logical unfolding. It is not unlike a written essay, which itself would have a main statement, developing sentences, and a conclusion of some kind. Perhaps through our understanding, or through our subtle encounter with such logic, we come to find music pleasing, or understandable. At least we trust that the composer is trying to tell us something.

After all, do I really trust Accumulator?

If the analogy of language is continued, Accumulator feels like the words of the language are jumbled across the page “randomly.” It´s as if the words appear in an order which cannot transmit a message. Words were pulled out of a hat. What is left is the surface feeling of the language, the familiarity of its contours, the recognizable nouns and verbs. But, what is missing, is the feeling that someone is trying to communicate something, or at the very least, that someone is trying to make a reference to the very idea that something is communicable.

Yet, I am very aware of something. Many modernist compositions of the 20th Century were also accused of being “random," weren't they? They were also accused of meaning nothing, of being so inhuman, perhaps. Through repeated encounters with modernist works their logic become clear (at least to some). How can a computer be illogical anyway? If it has anything to do with the Turing System, it is indeed the outcome of some kind of pure logic. Maybe it is just speaking a language that we don´t. Maybe this language is more logical than our own.

Is the computer´s music akin to Schoenberg`s? Shocking at first, yet only requiring some initiation? Will we learn to understand it, or appreciate it, only is as time goes on? Perhaps. Perhaps not.

The key is hearing this music at this other level - the level of ideology and perception. The music can point, even negatively, to our own prejudices. Our tastes are the way through which we receive and process information. The music is “random,” “boring,” “tedious,” “stupid” - in so much as it is judged against our own ideological frameworks. If someone tells me their opinion on the AI music, I can see immediately and clearly their opinions on what music is. We out confronted immediately with our definitions, our most hidden definitions.

To me, this is the unique opportunity which AI generated music provides.

This site was designed with the
website builder. Create your website today.
Start Now