Creating Music Transforms Us

Have you made any music recently?

First…This….

By now you know that I’ve started to play around with AI tools and music. In this post, I’d like you to join me on a reflective short-term look back over that experience so far. It has been quite a ride. Here’s why….

Do you remember the last time you created something entirely new? I didn’t think I could, either, and I was right. I gave it a real effort. Unfortunately for me, I couldn’t come up with a single thing I’ve even created that was 100% entirely new. Everything I’ve composed, written, or (probably) improvised has been derived from something else. Sure: some of the connections to what the US Patent Office calls “the prior art” might be a bit tenuous, but they are there for everything I can remember ever making. Especially in music. This isn’t so remarkable, given that incredible genius-level thinkers such as René Girard have observed that all we human beings ever do is imitate one another.

In music, imitation is everything. You simply cannot make a living in music without imitating some other musician, covering somebody else’s song, or composing a slightly different version of somebody else’s song. Think of pop music — it’s popular because lots of people like it — and how all those people are kept salivating for the next hit song, and you’ll have some idea of what the composers of those hit songs must endure. It’s very formulaic, almost machine-like.

If that bursts your bubble, there’s more.

Now…This….

Ted Gioia has written extensively on popular culture in The Honest Broker, including the worrying trends in AI-generated music. Spotify, for example, is flooded with AI-generated tracks produced by factory bots that can riff on the popular music that drives Spotify’s bottom line at a cost that’s far less than actual human musicians charge. The machine-made music is almost identical to the organic music composed for pop artists by the very few talented hit-makers in the business. And, yes: they all compose for the top artists because they know how to write songs that sell. And AI does, too.

For the rest of us — the musical types like me who don’t make their living from music we perform, regardless of who wrote it — the world is a different place. That is, even with the all-human-made composition, recording, and performance tasks we might choose to employ, we’re never going to earn a pop-artist living anything like the few dozen of our colleagues who do. And that’s OK because we get creative in ways that aren’t based royalties from hit songs.

It’s not such a bad place to be!

Here’s Why

It might seem counterintuitive that NOT being a pop artist is good. For me, it is. Lots of reasons for that of course, and I want to focus on the ones related to AI. For me, as you know, this is difficult because I’m such a critic of and so concerned by the mis-use of AI that’s already happening in our world today. For me, using AI with music has been a really good thing.

The Player Piano

I’ve always been fascinated by player pianos. My very first paying musical gig was at an ice cream parlor on a player piano. One of my idols, George Gershwin, is said to have learned his piano skills by copying on the keyboard the notes a player piano “played.” Gershwin had great pianist skills, was a formidable composer, and knew how to market his creations into a world that was hungry for them. One of the additional ways Gershwin served up recorded music greatness employed a player piano.

In those days, composer/performers like Gershwin made piano rolls…scrolls of paper about twenty feet long with holes punched in them for all the notes needed to make a particular song happen on a specially-modified piano. Like others, Gershwin quickly discovered that a piano roll could be recorded more than once: the first “cut” might be the melody and accompaniment, played simply. Then, a second “cut” could be made over the first one, adding embellishment and complexity. There might even be a third cut of additional embellishment ‘way up in the treble registers of the keyboard. This “old time piano” sound defined the era.

Of course, in addition to multiple “cuts,” piano rolls could also be recorded by multiple pianists playing the same piano at the same time. Here’s one that will help you get the “visual” on what a multi-player piano roll looks and sounds like when it’s played back. To quote the Description from YouTube: “This is a piano roll from 1919, the year that Swanee became a big hit. The music is by George Gershwin and the words are by Irving Caesar. The two pianists are Mae Brown and Chet Gordon.”

The DiskClavier

I’m going to skip several decades of “reproducing piano” history here and jump straight to the modern-era DiskClavier, introduced in the 1990s by Yamaha. This device reproduces not only the notes, it also reproduces the nuance of the performance. Like the paper piano rolls, DiskClavier recordings can be augmented, and many of them have been. Yamaha is not the only manufacturer of this type of reproducing piano: Steinway has its own system, and there are aftermarket devices that can be installed in some pianos that do the same job.

When Yamaha loaned me a DiskClavier, I was able to mess around with new aspects of recording piano music. I quickly learned that the piano roll still exists in digital form, and that “seeing” the music in this way with digital editing tools meant that I could quickly and easily digitally change a recorded performance, then play it back with the modifications I made in “performance” by the DiskClavier.

Here’s how this looks in the edit window of today’s Apple GarageBand, taken from a project I’m working on right now:

See the piano keys on the left? Notice that “Transpose” slider? It is a great way to shift the key of a whole recording. The green bars correspond to specific notes, and their length determines how long that particular note will sound in the finished piece. Notice how the durations mostly begin together but are slightly different in many cases? That’s because my human hands released the notes at slightly different times. Some music editors would “fix” all the durations to be the same length, so that the playback would sound more precise. For this song, I didn’t.

From the DiskClavier’s recording (a MIDI file), a virtually unlimited range of modifications could be made, then played back on the actual piano itself, with nuance, changes in tempo, volume, and whatever else I felt like messing with in the MIDI editor of that day. Once the playback happened, I could overdub on it, just like the piano rollers did in their day.

Today’s DiskClavier has some nice built-in editing capabilities, and, in addition to the acoustic piano, it can play back sampled sounds of every other kind of instrument one can imagine. Yamaha has a whole stable of recording artists whose pianistic artistry is enhanced by other instruments that have been added to their recorded piano tracks, something like the effect of hearing a piano concerto played on one’s home piano with the entire orchestra “in the room” as well.

My several months’ experience with the DiskClavier were a gas. It opened me up to the modern version of what pop music could be. And yes: it was derivative of the technology from those old piano rolls from back in my ice cream parlor days. I still use GarageBand today — and it’s notation-intensive cousin Finale — connected to my MIDI keyboard, to “write” as well as record or play back music.

Hearing one’s own music performed live for the very first time by someone or something else is a real jolt to the system, both negative and positive.

Experiencing “Your” Music

Unlike the composers who write pop hits for a living, very few of us write music that gets “covered” by some well-known recording or performing artist. In my case, which has been largely about performing written-down music from the past, satisfaction has come from the opposite direction. Even for music that I’ve composed organically note for note, the experience of listening to playback falls short of other “significant” moments I’ve had with music.

Until now.

One aspect of music AI tools is what we used to call “rapid prototyping” in the design world. The IT world uses the same ideas today — quick turn-around of new software followed by alpha test followed by revision and customer release — to respond to an ever-changing landscape of requirements. You can probably guess that AI really helps make this process in design and coding more expedient than in the past; the same holds true for music.

I’ve written about this before here and here, and I’ve yet to share my response to hearing “my” own music for the first time.

Because of the rapid prototyping capabilities of today’s music AI tools, I can generate multiple versions of a particular song in less time than it takes to write about it. This is incredibly useful for setting lyrics to the most supportive genre, tempo, and instrumentation that fits best. When it comes together, it makes me cry.

Seriously. I weep when AI does something wonderful with words I’ve written. It’s part amazement because I know the effort of setting lyrics “by hand,” although the bigger part is simply emotional. Listening to the first playback of a new song can be overwhelming, especially if the lyrics I wrote hold some kind of special meaning for me. Sometimes it takes a few tries to prompt for the best genre, tempo, vocal style, or instrumentation, but when it’s right, it’s right.

I know this response. I have had it at what I call “significant musical moments” that I’ll remember for the rest of my life. Sometimes I weep at a symphony concert; sometimes at a piano recital. Sometimes it’s at the ballet; sometimes it happens at a musical. A musical moment like that always surprises me. I’ve wondered from time to time if the sax player in the jazz combo I’m listening to from a table just a few feet away from where he’s playing looks over at me and wonders why there are tears rolling down my face. I’m not a faucet, but I know what moved me like this, and music does it.

I have the same feeling listening to the “good” stuff AI makes for me.

Is that wrong?

You’ll find a new song nearby: what say you?

But Bill….

I know that some of you will object to this based on the assumption that I didn’t make all of it from scratch. Do you give that same critique to jazz musicians who “improvise” solos over a specific set of chords with a specific melody, even though those solos might be repeats of solos they played to that music many times before? There are certain rules everyone must follow — more or less — to produce a satisfying result and, yes, sometimes brilliance happens, but is that brilliance truly original?

Did Mozart or Beethoven just go to print with the very first version of a sonata or symphony? You can bet they really worked on their compositions! Sure, they were the best at doing what they did in their days, but Mozart still relied heavily on Bach’s ideas, as Beethoven did with Mozart’s, in order to develop a unique sound and style that defined their musical era for all time.

This process holds true for every band you’d like to name, even the jam bands who show up with no reference music and make multi-hour concerts happen: it’s all derivative.

And we have had some incredible music over the ages, haven’t we? Are we going to diss Gershwin for releasing a piano roll that no single performer could play? Or Yamaha for embellishing a recorded piano performance with symphonic instruments?

Is the average music consumer really all that interested in who wrote the music they love, or how the process of creating that music came to pass, provided the resulting music “does it” for them? I’d have to look up the names of the top pop music composers to actually know them, although the fact that there are fewer composers of hits than there are pop musicians performing those hits ought to be a bone of contention for the savvy music consumer.

As we’re learning with the “large language model” AI systems, it becomes easy to spot their output. I suspect that this will soon begin to dawn on the savvy music consumer, and at that moment they will achieve a more discriminating, nuanced appreciation for the music they love.

Coda

Which, in musical terms, means we’re getting close to the end….

The way I’ve experienced it, if no less than George Gershwin inspired so many pianists with his art and the available tools in his day, why not use today’s tools to strengthen the art with which composers deploy today’s music?

If Girard was correct (and I have no reason to doubt him), and all creative effort, whether by human or machine, is derivative, what is so wrong with using AI to facilitate the creative process, especially if the result creates an emotional response for listeners?

Sure, I could do this process the old way: compose and write the parts, play or sing them myself (using different patches for different instruments and maybe AutoTune for the vocals) or hire musicians to record them, mix the whole thing down to production-quality, and release it formally for actual royalties (which for me means creating an album on CD Baby which then goes into worldwide distribution). It’s expensive and time-consuming to do this.

Or, I could spend a few hours making great lyrics, messing with how AI sets them to music, and post the result to YouTube for you to hear with all the proper disclaimers that Gershwin never made in his day and that most singers who still use AutoTune don’t make these days.

I’ve chosen to do the second thing. And, at least as of this post, I feel strongly that monetizing AI compositions is still a very open and ethical question. The strength in the AI approach to music is that it gets the ball rolling quickly, allows one to change horses in mid-stream without a penalty, and can have very emotionally-satisfying results. It’s weakness is the same as the one composers have experienced for centuries: what can I write that is truly new, while still giving listeners enough of what they need to have something familiar to hang on to?

What if a machine can create something truly transformative for you? You already trust apps that do that; why not music?

---

Over the course of more than 40 years of paying attention to how music works on us, Bill Protzmann has rediscovered the fundamental nature and purpose of music and accumulated a vast awareness of anthropology and sociology, as well as the effects of music, the arts, and information technology on human beings. Bill has experimented with what he has learned through performing concerts, giving lectures, facilitating workshops, and teaching classes. He first published on the powerful extensibility of music into the business realm in 2006 (here and abstract here). Ten years later, in 2016, he consolidated his work into the Musimorphic Quest. In this guided, gamified, experiential environment, participants discover and remember their innate connection to this ancient transformative technology. The National Council for Behavioral Healthcare recognized Bill in 2014 with an Inspiring Hope award for Artistic Expression, the industry equivalent of winning an Oscar.

In addition to individuals, Musimorphic programs support personal and professional development and wellness for businesses, NPOs and at-risk populations.