Part of my curriculum, just thought I'd leave it here as it's pretty schizophrenic.
The inclusion of Artificial Intelligence in the creation of the arts is not a new concept, but many artists fear it for its rapid development in recent years. However, in the field of electroacoustic music and audiovisual media, I do not view it as something beyond a tool, for it cannot create intent, unpredictability, nor can it demonstrate an acquired taste in art and culture.
While specialized “expert systems” artificial intelligence models, such as Deep-Blue, AlphaGo, and Stockfish boardgame bots, exist and perform exceptionally well at their focused tasks, trained on data and knowledge of the greats in their respective fields, I fear that taste cannot be taught or imparted. Rather, taste for art is something that cannot be replicated but is earned. The importance of the human element in art is that it is unique to each individual, where one might have a deeper interpretation of a work than the first impression it gives. For example, Dr Jiradej Setabundhu’s work, TS plays Balinese Underground, touches on the relationship between the composer and AI as a compositional device. In this work, Setabundhu explores AI music creation and the interactability of AI in performance through a pre-recorded media piece featuring audio and video of an AI “talking” to the performer, while quoting notable cultural figures like Taylor Swift and Steely Dan. Many of J. Setabundhu’s works often utilize a great performer miming pre-recorded media with nods to the cultural references of the music he grew up with.
I do not fear Midjourney nor Sora, for I know they are not capable of reproducing unpredictability or any perceived “mistakes” that are intentional. Artificial intelligence cannot produce an artist’s inner voice. To err is human. To have an intent is human. As an old Thai saying goes, “ผิดเป็นครู”, literally meaning that the person who makes a mistake later becomes the teacher, as humans learn from their mistakes. This perceived mistake, however, can become a device in a work that later imparts more context or depth to the piece once the audience has noticed it. Randomness, with intent, in the arts can never be replicated by AI either, as the agent is often trained to avoid mistakes and produce results based on the given prompt. On the other hand, stochastic processes in compositions (Xenakis, Tudor, Stockhausen) can be controlled musically by the artist working on the piece, producing a controlled yet unpredictable result. It is worth noting that a machine cannot replicate true randomness; rather, it attempts to mimic it, for transistor noise is more random than Gaussian noise.
Reinforcement learning, deeply inspired by Pavlov's conditioning, punishes the agent for errors and rewards it for producing the “best possible” result. However, we must question ourselves: do two artists have the same understanding of what the “best possible” quality of their works is? Does it refer to a polished work conforming to tradition? Or does it refer to a piece that defines boundaries and meanings of an aesthetic? Granted that the two artists in question have their own established style, it is certainly possible that the idea of “perfect” or even what is acceptable in their works is completely different.