Do you remember where you were when you first realized that AI-generated music had passed the Turing Test with flying colors?
The past couple months in my life have been pretty full-on. Kamala and I had a very busy tour of the northeastern US, sectarians were trying to get a couple of the gigs canceled, and probably unrelated to any of that, my mother decided it was time to stop eating and die, which she ultimately did earlier this month, at home in Boston, with her daughter by her side (and her son walking back from a coffee run).
Aside from all that, some of the most intense experiences I've been having lately have been between me and my laptop.
Everything in the world seems to happen so fast these days, that taking a moment to think about the past, even the recent past, feels like a luxury that no one has time for. But if sometime in the future anyone is wondering what it was like for the first generation of professional musicians to come to terms with a world in which AI music-generation platforms could now pass the musical Turing Test with flying colors, then posts like this one will perhaps be your first draft of that history.
The experience has thrown me into a fairly constant state of philosophizing about the nature of being human, the nature of being a musician, and the implications for these things presented to us by our new reality. I was raised by two professional musicians from the world of classical music -- a concert pianist and a composer. Then since I was a teenager I have been steeped in the culture of the folk music scene, and steeped in various left and progressive social movements in various countries. My experience over the past few months of writing songs with what I call my AI band has caused me to think about all of that, with the sort of involuntary intensity that I associate with the way people talk about seeing their lives flash by in front of their eyes when they think they're about to die.
People often tell me I'm a prolific songwriter. Traditionally, if you record more than one album per year, you're edging in the direction of someone calling you prolific, so the bar is set very low. But by most standards, anyway, I do write a lot of songs. Since collaborating for years now with mostly remote musical partners like especially Chet Gardiner, which has also largely coincided with Israel's genocidal war on Gaza, I've been putting out an album of topical songs every two or three months.
Since I've been working with my AI band, over the past few months, I've written and "recorded" an average of about one song per day. And these aren't just songs, they are good songs. They are largely songs about current events, as good as any others I have written, except that I'm not the one singing the lyrics, and there is quite a bit more stylistic variation in every aspect of the songs, because with an AI band there are no limitations as far as available session musicians who play oud, shamisen, or whatever it is I want to hear.
There is no doubt in my mind that although I have written and recorded some pretty catchy songs with human musicians throughout my musical career, my AI band is at least as capable as me or any other human songwriter that I know of at coming up with memorably catchy melodies, refrains, instrumental riffs, ways of vocalizing that bring out the meanings and double-meanings in various lyrics, and everything else.
That AI-generated music is now generally indistinguishable from human music is a realization I have been involuntarily provoking with a fair number of people at this point. While I'm sure it's inevitable that this would soon have happened to them anyway, being the original bearer of the news comes with mixed emotions. I remember well how I felt when I first had this realization, several months ago in Australia, just before I started playing with the technology myself.
Last Sunday it was raining all day, and I haven't yet purchased a good canopy for such eventualities, so for our weekly neighborhood vigil for Gaza we didn't really have a setup that allowed for live performers to stay entirely dry. So I figured I'd at least set up my fancy speaker and play recorded music via my phone's bluetooth connection with it.
Among the eighty or so songs I've written with my AI band, ten of them have been about Palestine. I have those songs collected together in a playlist on Soundcloud, and I was playing that playlist. The production values with these recordings are so high, it's really great music for playing in such circumstances, as the sound effectively cuts through the rain and traffic. Plus a lot of what I've been doing has a hip-hop groove to it, so it's a good backdrop for keeping people moving, and feeling warmer and dryer than they may actually be.
One of the other regulars at the vigil, also a musician, asked me who the artist was that we were listening to. When I explained to him that I wrote the lyrics, but that no human was otherwise involved with producing the sounds we were hearing, he looked like he'd seen a ghost.
Wade went off to hold a sign for quite a while before I saw him again. "I feel like I was hit by a sledgehammer," he said.
No further explanation was necessary, but we talked lots more about it all anyway.
I guess when I'm talking about using an AI music-generation platform, I'm talking about an activity that only about 1% of the population has thus far engaged in. And I guess when I'm talking about writing and recording songs, I'm also talking about an activity only a tiny portion of the population is familiar with from personal experience.
Either way, it's worth discussing a little, with or without AI, why would someone play music, and specifically write and record songs, in the first place? The question may be more complicated than it seems, for a variety of reasons.
People play music in the first place because their parents signed them up for piano lessons, or because they were drawn towards playing the guitar because they noticed it was a good way to attract a potential date, or just because they enjoyed playing music. But by the time they might be writing or recording at a professional level, in whatever genre, what's it all about?
Of course in some genres it may be about producing the right kind of formulaic hit material so you might stand to make lots of money in the music industry. In other genres that's much less of a calculation.
What I've always run across, from people within my parents' classical/concert music circles as well as from veterans of the folk revival, and just from serious artists generally, is for real artists it's all about the art, and effectively communicating what that piece of art, that song, or that instrumental or whatever else, wants to communicate.
This orientation has always been in contradiction with the orientation heavily promoted by the music industry's PR machine for the past century and more, which involves making big stars out of specific artists, rather than the particular songs they're recording as such. It's the name of the artist, their likeness, the sound of their voices, their dance moves, their fashion sense, etc., that is the center of attention when it comes to the promotion within the industry.
Little do most fans of any given artist know, but most of those stars are singing songs pitched to their managers from songwriters that exist within the corporate ecosystem of the label. The instrumentation, the style of playing, what the band members look like, how they all dress, most all of that stuff is decided by someone other than the artist.
But all the while the PR machine focuses on the individual artists and how exceptionally brilliant they supposedly are, while they're promoting some version of a rags-to-riches story about an artist who writes all their own material, and bullshitting the whole way, a significant and serious subset of artists reject the whole thing, because we know it's always been all about the art, not the artist.
Perhaps more than anything else, being raised by a concert pianist and a composer helped for me to develop this perspective early on.
My mother could play anything on the piano, up to the most demanding, technical "avant garde" concert music. But she would always say it wasn't about impressing anyone with her technical abilities, it was about playing the music, communicating what the piece you were playing was trying to communicate. Effectively interpreting the intentions of the composer.
As a composer, and a brilliant pianist himself, my father would play approximations of symphonic pieces he was writing on a keyboard, to give himself and others an idea of what he was working on, which might only really be truly heard when it was performed by a symphony, or whatever type of ensemble he was writing for. Until then, it was music on paper, but not something recorded, that anyone could listen to. For it to exist in that form, people other than the composer would need to be very involved.
While the music industry was doing its thing and making stars out of various artists, for the industry's commercial purposes, so many real artists continued to pursue art for the sake of art, or for the sake of the power it had to communicate. Within the folk music revival in the early 1960's many people have talked about how what all the songwriters in the scene were hoping to do was to write a song that others would mistake for an old folk song. If you could write a song that people assumed was an old traditional, you had succeeded in writing a really good song. If no one might suspect you wrote it, this was a good thing.
As a composer, my dad worked with a lot of poets and those who dealt in the realm of words and lyrics. Words were never his thing, so much as music. There's a long tradition there of artists orienting more towards one or another aspect of the whole thing. Lots of songwriting partnerships, where one was more about the lyrics and the other more about the music.
As a songwriter myself, I have always found it easier to come by interesting lyrics than fresh music to go with them and make those lyrics really come alive. As I'm writing a lyric I can start to tell if it's one that could really become a good song, with the right musical treatment. I often write a good, almost complete lyric in twenty minutes, and then I'll spend hours, or days, trying to come up with a really good musical vehicle for those lyrics. That part of the process is just as important, and, for me, much more challenging.
I have sometimes successfully risen to the occasion on that front, and often not. I have often settled for music that was too predictable, and didn't serve the lyrics the way music can do. I have often written lyrics that never found music. Of course, other people have this whole thing in reverse, but that's how it's always been for me.
During the pandemic, when I had a lot of time on my hands and extra income, I did a whole album together with Jane Reynolds/Virtual Bird (It's Been A Year), where I sent them lyrics and they turned them into catchy songs.
At many other points, touring with people who sang harmonies with me, and the years to date of remote collaboration with Chet, knowing that songs I write will often be heard with harmonies, or they'll be recorded with the kind of cool things Chet is likely to do to stuff I send him to work with, my tendencies as a songwriter go towards thinking about what things might sound like with vocal harmonies, or thinking about leaving more sonic space for Chet to work with as he adds layers of instrumentation to a given song.
Working with my AI band does bear various things in common with the experience of recording with other people, and even more in common with the project with Virtual Bird.
But more than anything it bears more in common with the notion of being a lyricist working with session musicians of my choice, that may or may not differ completely from song to song. I am otherwise completely removed from the recording in the sense that I am not the recording artist here, and neither are any other humans with their various talents and limitations.
Being a songwriter with a lot of experience as far as writing a song that is going to work, that serves its purpose well, one of the things that I've always been so impressed by when working with the many musicians I've worked with to produce a whole bunch of albums over the past few years is how often it happens that I send off a song lyric to Virtual Bird or a guitar-and-vocal recording to Chet, and the first thing they do with it is perfect. Both experienced recording artists and performers themselves, they know what works and what doesn't, and any changes I might make to their first draft is usually going to be cosmetic.
That's never how it goes with my AI band. The first draft is never the keeper. But after refining prompts and editing lyrics and spacing and such, within ten or twenty or thirty iterations, invariably, we land upon an interpretation of the lyric that just nails it.
And all of those thirty iterations happen in the space of ten or twenty minutes. And it only takes that long because it takes me a little listening to determine what needs to be adjusted next.
But it's not the outrageous speed in going from lyric to brilliantly-recorded song that is the most intoxicating thing about working with the AI band. It's the potential to be anyone, playing anything. As with being a composer, I can write for whoever I want to, whatever instruments. One song may be acoustic hip-hop with oud, the next newgrass chanson skiffle. And after singing as a male for my entire life, I find writing for a female voice is something I haven't grown the least bit bored with yet.
I'm quite certain that it is also because I can now write for a female voice who can deliver lyrics in much more popular musical idioms than the ones I generally play myself that some of the songs have already gotten picked up and shared more widely than I suspect they would have otherwise, most notably "No Contract No Coffee" among striking Starbucks workers.
Whether or to what extent that happens with other songs is of course unknown, but whether they are good songs, well-written and well-recorded, with catchy refrains, etc., in my humble opinion, there's no doubt.
Because some AI-generated music has gotten very popular, and this has made big news stories, some people assume if we're playing with AI music generating platforms, we're headed towards wealth and fame. This has certainly not been the case for most musicians using AI music-generation platforms, since most pro-level musicians already do that, and the future is obviously going to be more full of AI slop than anyone can imagine.
Looking not at Ai Tsuno on the dominant streaming platform, Spotify, but at David Rovics, if you click on different albums in the discography and see which tracks have received any plays over the past month (which is what those numbers represent), you'll see that the vast majority of my songs are in the single or double digits. Looking at all of these tiny numbers, if you start wondering where those overall monthly listener numbers are coming from (currently around 18,000), you'll see that it is just a handful of songs that each get a couple thousand streams. Just those few songs that made it into the algorithms and the playlists and get circulated on the platform. The rest, such as most of what I've recorded over the past decade, is largely invisible.
Ai Tsuno will probably be similarly invisible, like most good music is and has always been, since there was a music industry. But if a given song, recorded by humans or AI, may serve the cause for which it was written, then it's probably good that it exists.
