Critical Response #1: What do you (really) want from AI music generation?

by Andrew Lee

Music 356: Music and AI


Being I'm a music composer myself, I must say that a AI music generator makes me feel quite frustrated, but I'll try my best to make this reflection more open-minded. I'll start off imagining myself in a world where musicLM is already in prevalent use. In terms of the music industry, what would this look like?

Assuming the general public believes that "the creator of an artwork" and "the process of creation" is what gives music value, then it'd make sense that the field of music creation is still dominated by humans. I'd imagine that certain lower-budget projects may employ musicLM for the sake of cost efficiency, but since human-created music is more "valuable," most big projects will probably still use human-created music. In this scenario, I think the use of musicLM is acceptable, since human musicians are still generally needed. I do think that the music industry might get a bit more competitive, since there may be a reduction in music commissions.

However, if the general public views music solely as a source of entertainment, society might use musicLM for most projects, since it's cost-efficient, and its quality might be just as good, or even better, than human composers. Would the music industry no longer need music artists then? Or perhaps the human music industry will focus more on developing new styles of music, since AI can only produce music of exisiting styles? In such case, would music be a rapidly changing world where there's constantly new unheard-of sounds? How would the public react to this then?

Next, I want to consider the social implications of this world. We discussed in class that music also encompass things like culture and history. With an automated music generator, how would this change the way people think about music? Would music still be such a complexly human thing, or would it begin to feel more detached and superficial? Can we listen to AI-generated music and still think "Oh, the artist must be going through something rough when they made this" or "This song reminds me of my hometown"? Or perhaps it'll be the opposite, where the AI allows us to realize more about the social, cultural aspect of music through its objective mimicry? Now, more specifically on music creation, how will people accept this form of data-based music creation process? Recently, there's been an "No AI Art" protest going on in the art world because artists feel that their art has been used without consent. Personally, I side with these artists. I'd imagine something similar to occur in a MusicLM-dominant world. Will there really be a way for musicians to live in peace with MusicLM?

So, what do I really want from AI music generation? Before I took this class, my answer would probably be "I don't want AI music generation at all," because I couldn't imagine a way to co-exist with this technology without it completely replacing music composers. And because music creation is such a valuable and enlightening activity for me, the thought that the world might no longer need this activity was simply overbearing. However, what fascinated me the most when reading the course description was "how AI might augment, not replace, human creativity." And after discussions about oracle versus tool, I now want AI music generation to be a tool for me to make music in new ways. The assignments we've been doing so far, I appreciate that I'm still the one making the creative decisions, and that's what makes it okay for me. I think there's potential in this new domain of human-computer music creation, and I want to see more of what humans can create with this.