Building or Blocking? AI and the Beloved Community.

AI generated image of AI as a puppeteer

Today, I’m writing about AI. It’s a hot enough topic to approach with some hesitation. But it has implications for the work of Summit Stages and it’s been on my mind a lot, for reasons that will soon become clear. 

The explosion of world-altering possibilities that hit us with ChatGPT and is nowhere near slowing down, can be unsettling. It raises a lot of questions, in a lot of areas. There are environmental concerns about AI, unknowable implications for the job market, and even metaphysical questions about the line between artificial and self-aware intelligence, and how we know when it’s been crossed. Just over a month ago, my daughter sent me this Instagram reel, with ChatGPT detailing how, if it wanted to, it would take over the world (warning: there’s some strong language from podcast hosts Yaz and Feras in reaction, but that’s only after ChatGPT stops talking. They are speechless during its spiel).

Is this a real danger? Maybe, and maybe not. Perhaps it depends on whether we figure out appropriate and balanced use? 

For this article, I’m going to zero in on one area: the use of AI in the creation of music. The big question that’s been running through my mind and heart for about the last month is whether creating music with AI is cheating. And writing this newsletter is part of how I’m hoping to get to an answer. Here’s my story, and the questions I’m still grappling with. I’d love your thoughts if you’re willing to weigh in. 

A couple of months ago, I purchased a year-long subscription to Easy Peasy AI. It promised me that it would learn my style of writing, and I was hoping that I could use it to dramatically reduce the time I spend writing my newsletters and focus more time on other projects to move forward the mission of Summit Stages. 

I was disappointed. It did not learn to speak in my voice. And even if it had, I would not trust it with writing these newsletters. Here’s why: because AI doesn’t offer original thinking. It does a decent job of gathering up and summarizing the ideas of others and presenting them in an organized manner (only decent, because it doesn’t know how to evaluate people’s claims so it can be wrong). But it’s not insightful. It doesn’t know how to explore a work of art’s message in the light of a set of values. And it doesn’t care, so its presentation of information is not particularly engaging. The best it can do for me is gather up research, and even some hard-to-find articles that I can quote from. But I have to tell it to give me links to every article it cites. And then I have to check them for myself, because it sometimes attributes a quote to the wrong person. 

It does a lot better filling in the gaps for me in areas that are not my strength. For instance, it’s getting markedly better at creating art that can illustrate a piece of my writing. (FYI - I have not noticed that Easy Peasy does a much better job of this than does Canva. There are some areas where the interface with Easy Peasy is easier to navigate than is Canva’s, but image quality with Canva is either improving dramatically, or I’m finally learning how to write the right prompts). 

AI created illustration of an AI visual artist

As I’ve been using AI for illustration, I have not found myself stressing about the real live artists who need to make a living, and who I am not paying for my art. The truth is, I don’t have a budget for art. So if I need an illustration that isn’t a photo from a musical, it’s AI or it’s Unsplash. I usually go to Unsplash first, but they don’t have a massive collection of royalty-free art that meets my needs. So then I resort to AI. And this has not given me qualms of conscience, at least, not until recently. 

Here’s what happened in the last month. 

First, I broke out the Launchkey 49 that I purchased maybe a decade ago, in hopes that it would make composing easier. Only I’d never learned how to use it. It was time to fix that. Unfortunately,  I discovered there is a steep learning curve, not to mention the problem of latency. That’s the term for a disorienting pause that happens between when I press the key on my midi-keyboard and when I hear the sound from my computer. It’s supposed to be mere milliseconds. But I’m getting somewhere around a ½ second of delay, no matter what fixes I try. For me, that makes it really hard to hear my rhythms, which is discouraging. 

Then, composer/librettist Gunnar Andersen and I started working on our Still Learning podcast (launched three weeks ago). Gunnar suggested we needed a theme song. I thought: good thing he’s been doing music production forever, because I don’t have room on my plate to write a theme song. But it wasn’t like he had room on his plate either. You don’t just pull out something you’ve written in the past for this. You want a song with an overarching message that’s specific to your branding. 

Fortunately, Gunnar was just completing a course in music and AI. He’d learned he can unblock his creativity by giving AI a prompt, seeing what’s wrong with what it offers up, and then having the perspective with which he can write something wholly original of his own.  So, the day before we filmed our first potential episode (“potential,” because it wound up being a learning opportunity which will never see the light of day) Gunnar sent me an AI-written song on the theme “I’m Still Learning”. I thought the music was decent. The lyrics were entirely uninspiring. So uninspiring that they got me thinking about what they ought to be. My experience was very much like Gunnar’s. All of a sudden, I had a message burning inside me that I wanted to share, largely as a reaction to the non-message AI had given me. The AI generated tune gave me a framework for the lyrics, which came together quickly. I sent my original lyrics to Gunnar the next morning. And maybe 15 minutes later, he sent back two new versions of an AI-generated song, complete with really lovely vocals. It was astonishing! 

Suno.ai home page

Thus began my foray into writing lyrics that Suno.ai turns into songs. I’m still floored by what the program can do. I’ve learned that I can input my own melody and lyrics into Suno, and it will create an arrangement and vocals, in minutes. Sometimes, it gets creative with my melodies in ways I like. Other times, it takes liberties that are irritating. I’m still playing with learning the controls that might reduce those liberties. 

What’s most exciting is that I am suddenly able to write songs and have them ready to share with others, in a format that is pleasing, in less than an hour. I want to report about what I’ve been able to do with this power. 

I am a disability support worker, which means I help a developmentally-disabled adult with all the activities of daily living. She always has a lot going on in her mind, and she often has difficulty expressing herself in a way that leads to her feeling heard. Six years ago, she and I wrote lyrics together about the frustrations she regularly experiences, with a chorus about her choosing to cope with them by singing, because she loves music. I put it to a simple melody, and her team and I have used it as a simple support strategy ever since. We’ve sung it acapela. It’s not a particularly inspiring piece of music. But it has proven very helpful. 

About a month ago, I was getting serious about strengthening my music-writing skills. I brought an iPad to work and we sat together while I began adding chords to our melody. But it was slow work because I’m on the learning curve. And we didn’t get nearly done.  

This past weekend, we just plugged the lyrics into Suno and we now have a bunch of versions of her song that we can play, depending on what kind of music she feels like singing to. We’ve got the Europop version (she’s a big Abba fan), a nostalgic-sounding country treatment, a “blue-eyed soul” serenade that sounds a lot like Elvis, a tune from a 50s diner, and a vulnerable-sounding soft rock. We even adapted it into heavy metal and Gregorian chant. She sings along to all of them. And they are hers. She helped write these words. They are her feelings, in her language, and they sound like something you’d hear on the radio. 

So of course, we’ve been writing more. She talks about what’s worrying her. I put rhythm and rhyme to her words and experiences, plug it into Suno, and presto. All of a sudden, her feelings have validity. They are out there where she can process them. 

I’ve written and published two more songs for our Still Learning podcast, in order to convey messages more clearly than I could with just words. The last one, I uploaded my own melody and chose a version that came pretty close. I loved it. And that felt incredible. 

co-writing a song

Also, last week, a friend was blindsided by some intense and distressing emotions that she was terrified to process. We talked. The more I listened, the more I felt that her feelings – and the purpose of them – needed to be put into a song. I came home, started pulling weeds in my yard, and by the time I came inside, the first verse and chorus were already fleshed out. We’re going to finish this song together, and I believe it will be a profoundly healing experience. 

So all of this has really warmed me to AI as a tool with which we can do a lot of good that was previously unavailable to us. It’s also prompted some serious soul-searching about how I can use it in a way that builds, instead of blocks, the Beloved Community. To be more precise, I have a huge collection of poetry that I’m just sitting on, and that I could now turn into songs. I’m in the process of revising Neverland: Wendy’s Story, which includes writing new songs, and getting new arrangements of the existing ones, with a more pop feel. And I’m asking myself how much I want to use AI in both processes. 

I see risks to both overuse and underuse. The problem with overuse is that replacing human collaborators with AI circumvents the magical harmonizing of hearts that happens when people musick together. The problem with underuse is that I don’t have the budget to pay for vocalists or music production for projects that don’t have a realistic prospect of bringing me income. So, if I don’t use AI, I won’t do those projects. I miss out on the learning opportunities, because the best way to get good at something is to practice it a lot. Every time I write lyrics that Suno puts to music, I learn more about what lyrics are effective. I also learn about musical genres and principles of orchestration as I ask ChatGPT to give me a prompt for music in the style of such-and-such powerful song, and also as I listen to the various versions it produces to learn what works and what doesn’t. I’m listening to music A LOT. Which is another key to becoming a better musician as well as lyricist. 

One piece of good news is that AI has serious limitations that help to reduce the risk of overuse. Just like with writing prose, with music, AI does a decent job of gathering up ideas that are already out there and offering them up in an organized format. It can create some pleasing music that follows established formulas. And that’s helpful for someone like me, who’s learning those formulas. But it lacks passion, brilliance, and the capacity for original thought or insightful analysis. Even when you give it some very specific prompts, it’s prone to getting some things horribly wrong And it has a very difficult time with anything complex.

Audimee home page

For example, I’ve spent a couple days trying to get from Suno a demo version of the new opening number I’m working on for Neverland: Wendy’s Story. But despite giving it very specific prompts for what the feel of the different verses should be, it still struggled. And it simply ignored all my prompts about who was singing what. It really likes to use male voices for lyrics that obviously belong to females. And vice versa. I’ve just been told that, with a little work, Audimee can fix that for me. Wow! But even so, originality, intentionality, and a loving attention to subtle details are the domain of human artists. 

So here’s where I’ve landed today: I don’t have the information or expertise to address the environmental and metaphysical issues, but I do feel like I can respond to the question of whether collaborating with AI for songwriting is cheating.  

AI can facilitate, not replace, collaboration with people

1. AI can be a super helpful tool for doing good in a way that builds the Beloved Community through music and art. I want to multiply my impact for good by taking wise advantage of the available tools. It just needs to stay a tool, which it will do, so long as my focus is on building the Beloved Community. 

2. I need to be actively looking for ways to collaborate with insightful humans who have passion for the work we are doing. I will help nurture a culture where we look for ways to create synergy by working together, within the constraints of time and budget and with an openness to using the tools that can make the impossible possible. 

3. I want to use AI as a tool for learning, not a lazy way out. I will keep developing my musical skills independent of AI. This will help to prevent the “Dependence” scenario that the Yaz and Feras reel at the top of this article warns about. 

4. I will continue to pay deep attention to the things that AI can’t replicate: values, passion, and people.

What are your thoughts? Do you see AI music generators as a threat to music producers and vocalists? What ideas do you have for structuring our use of it in order to strengthen our efforts to build the Beloved Community? 

---

Thanks for your support for Summit Stages! If you liked this story, please consider sending it on to a friend.

If you are not already subscribed to my weekly newsletter, I'd be delighted if you'd do so here.

And a deeply, heartfelt thanks to those who've contributed to my tip jar Your support is greatly appreciated!