The word choices we make when talking about generative AI are interesting. We talk of generating music, not of making, playing or creating music. So I asked an AI what it thinks about another AI. (I guess this is the world we live in now.) I posed ChatGPT the question: “Is generating music the same as making music?“ It told me:
(Side note: It really fascinates me how very wise an AI sounds. It is so eloquent that I would believe anything it tells me. Very dangerous tool for propaganda purposes. Back to the topic of music and AI:)
The characteristics of human music to emphasis here are “emotions, intention, original“. For now, MusicML is not able to create music that stirs emotions. When listening to the examples the music sounded correct but it left me emotionally untouched. It is hard for an AI to create emotional music since it does not understand what emotions are or how music should progress to make us emotional. Another limitation apparent from the examples of MusicML are the vocals. The words that are sung are gibberish and thus the songs do not tell a story and are hard to relate to or sing to. I believe that this will change as the technology evolves. But even then I would imagine it to be a little like ChatGPT where the words make sense and it says all the right things but it lacks authenticity or originality. So what is originality and how do humans achieve it? I think it is a combination of intellect, cultural background, social interactions, and lived experiences. However, originality might be illusionary as all the opinions we have about topics are based on knowledge and opinions that we have heard or learned before. All our thoughts are shaped by information and interactions we were exposed to. As an example: I would most probably be thinking very differently about AI and music-generating AI if I hadn’t taken Music and AI and hadn’t listened to Ge’s and other student’s thoughts. Therefore, I think that in all this debate about originality or bias in AI, we must also consider that humans are equally biased. What sets us apart might be our intellect, our capacity to think critically and our exposure to much more diverse information than just a set of specific data relevant for a specific task. We are using all our experiences from all situations of life to make music which creates an authentic expression through music. An AI is restricted in its amount of experiences while we are not.
So what do I want from AI music generation? When looking at the melody to genre examples of MusicML, I think that AI is very good at helping us generate the same melody or variations on a melody using different instruments and genres. It is great at generating an endless loop of music that will work well as background music in a Café or on a Study lofi playlist. What I am more apprehensive about is AI live performances. Imagine going to a venue with some friends to listen to an AI generating music on stage. That seems a little odd. In a concert I am looking for human connection, for emotions and for authentic expression. I guess I don’t want a future in which is the sole creator of music. I want AI to be a tool - great at providing a survey of sounds and a starting point for creativity, but I don’t want it to mark the end of human music making. We humans make music for other, we make it to express ourselves, we make it because it brings joy. I believe that in the future we will share this act of music making with AI, but I am hoping that we can create with it instead of it taking over.
Final reflection: What I wonder with AI: We humans are lazy. We ask ChatGPT for opinions since forming our own opinion is exhausting. So once we start relying on an AI to think for us, to create for us, will human creative abilities diminish?