Music LM

To answer the question of what I really want from AI and music generation I had to ask myself if I wanted that in the first place. The answer for now largely depends on the musical practices that I engage in. In a world where AI and music generation is prevalent in use, I would like to see AI assisting back to back (b2b) DJing format. Back to back djing is a fun challenge that requires your partner and yourself to improv. The format is as follows: you play 1-3 tracks and then your co-DJ transitions from your tracks into theirs, and so on. Very interesting flavors, styles, and transitions arise from this. A lot of practice together is involved. Your partner is learning your style and adjusts; AI would also learn your style. To be able to practice this on your own with AI would yield interesting results, as it augments the artist’s style, rather than bringing another autonomous flavor to the vibe. The power of the DJ curation still remains. While AI is on the b2b, it frees up the human DJ to do other creative and cool things, like playing live instruments over the set, or control some parameters. I wish this existed already--I was booked for a 5 hour DJ set for an event tonight and it would be nice to get some sort of break.

For other practices where I have more granular control for now, like music production, it’s more tricky to think about. To what extent do people find value in a work that is 90% generated by AI? I will take the AI portraits as an example. I find absolutely no value in that for several reasons. Upon introspection I find it perpetuates art coming from the ego rather than seeking connection with others or making any valuable statement. But currently based on the limitations that I see in the Music LM examples, it seems that it still requires informed musical language to get interesting results that might contribute some sort of value that transcends novelty for lay users. So these limitations might be a good thing. In a world where these limitations are addressed, I see diminishing creative exercise on the human side. Additionally, AI would be only drawing from existing music; vast bodies of work but the idea is still limiting. If we succumb to this format, maybe there is no room for revolutionizing or pioneering new music.

In short, I believe that we can keep our humanity intact and the AI tools that enable us to still find value in art and music will stay robust, and the tools that demonstrate possibilities but don't necessarily add value will lose relevance. 


“Starve the ego, feed the soul” - my friend’s tattoo