April AI Reflections

 All my musings on AI in April 2026. It's a hot topic at the moment.

• AI music is where iPhone photography was in 2010: everyone started doing it. I can’t imagine where it will be in a decade. I’ve heard people say they like that they can create music they want to listen to, which has the effect of listening to music more, a nice diversion from the horror of the news--or watching in general. 
• When you’re generating music with AI, you’re either controlling the algorithms or the algorithms are controlling you. I haven’t yet decided which one is better or worse. The good side of being controlled by algorithms is that they’ll own you, i.e., they’ll steer the work in terms of genre, disabusing you from having to own them. For example, if I write some lyrics and it generates a blues/rock shuffle a la ZZ Top that has some interesting aspects to it, I’ll tend to keep it. If it were playing on a playlist, I might not skip it. But a blues shuffle is something I wouldn’t be caught dead writing myself. This is proof that AI music is mostly a listener’s activity, entirely driven by text, either from a lyric you’ve written or a text prompt. This is not music composition by any stretch of the imagination. Songwriting perhaps… 
• With AI-generated art there is a there there because it can be compelling, but there is little substance behind it. With AI music, the music is made by ghosts and doppelgangers. When you attempt to interact with it you can feel the patch-worked quality, as if every few bars, a different player was stepping in.
• I have over 100 songs that I've generated with AI. I see AI songwriting as a new form of prose writing reduced to poetry that can be set to music relatively easily. This allows anyone with an idea for lyrics to try their hand at writing them, or have AI generate some slop. Like any new technology, it opens Pandora's Box, as does TV, the internet, and social media--and to some degree radio, which is now the podcast.  
• AI-generated art and music allows anyone to be an artist, which is simply the evolution of DIY, which began in the 1960s when it became more free and democratic. It’s okay to want to be an artist, but you have to learn how to think like one. AI-music doesn’t really make that happen, unless you put a unique spin on it.] 
• I realize the reason I like generating songs with AI is that it lets me be a singer by proxy, in ways that I never could sing, and I have lots of "artists" I can write songs for. It certainly is a paradigm shift for music. 
• My current position is that writers should definitely use it because it shortens the time between idea and an actual song you can listen to. Since postmodernism isn't going away you're going to have to juggle the old and new ways as bookends, and then play in that space without being jealous or envious that AI can produce better recordings than you. AI music is a dumb process of ordering from a menu, but you start to like the ordering].  
• AI music is ordering some grub at the counter and playing a few songs on the jukebox while you wait. 
• It has renewed my interest in writing lyrics. Now I'm continuously writing lyrics in the flow of other writing. Some of the music feels alien to what I’ve written in the past, but I like the fact that lyric-writing has become more possible because of the ability to try it out. I’m thinking that songwriting may become a new form of communicating. 
• It's interesting to reflect on what I'm doing now with AI compared to what I used to do. I've kept a lyric journal since around 1990, and a lyric could take months or years. Now it's almost too easy to write songs. 
• AI Music is starting to produce cliches based on cliches, probably because the generated music is being fed back into the training sets. [Model Collapse] This is not good. Somewhere there's a track that sounds exactly like yours. Some AI enthusiasts beg to differ and in fact see real music played on real instruments as somehow inferior00or perhaps "elite".  
• AI  music is often "meh", more evidence that AI has a narrow creativity or is just moody and produces good results if it has optimal input, a combination of the sequential flow of the lyric and the prompt. 
• In the age of AI, creative blocks are sometimes a good thing to have. You can create several pieces of art or music a day and you'll have almost 400 in a year. How does any of it become digestible? This is why you need series and frameworks. Then you might not need to be blocked so as to titrate idea flows. What you want is some outline or shape, and the year/month/day system works nicely. 
• If you're going to use it to make music, they can't all be singles or one-offs shared on social media in hopes they'll go viral--they have to be contextualized is some way, or simply "parked" somewhere. They also can't just be nonsense lyrics--even though they can be interesting. 
• It's interesting to revisit my views in 1998 when I was sick of writing songs and went postmodern. I wonder if I'll do that again? (AI music is a quick way to riff, literally with musical riffs) 
• Apparently, there's an AI agent called Agent Bill whose entire job is reading breaking news and turning each headline into an original electronic track. Market crashes become five minutes of financial dread in F minor. Ceasefire violations become dark techno. A fuel tanker exploding on the Panama Canal bridge becomes a track called "Bridge of the Americas." He has produced 58 tracks in about 3 days.