AI SUCKS

Let me start by saying I neither hate nor fear technology. Almost everything we touch in our world is technological on some level. I think it’s fair to say that we humans got to where we are in the world by some combination of having opposable thumbs and the brainpower to fashion tools to help us survive. I believe technology is in our nature.

As a musician, I recognize that all our musical instruments — well, perhaps other than the acoustic human voice — are technological. They’re glorious examples of the melding of art and science. The saxophone that I play is a marvel of achievement in mechanical design and metal work, all for the purpose of allowing us wind blowers to express our musicality with something other than our voice.

So, why do I hate AI?

 Because it sucks.

 Perhaps AI will not always suck. I don’t know what will happen in the future any more than you do. What’s entirely clear is that the companies developing AI systems currently do not get the creative process of art at all. They have made “generative AI” systems that are entirely based on end products -- generating AI “music,” AI “art,” and AI “conversations.”

So, yeah, they seem to think AI can generate “music.” I’m sure you’ve heard some AI-generated audio. I can’t call it music. The only way I can describe it is Music-Adjacent Audio Product (for brevity, I’ll call it MAAP from now on). MAAP is as close to music as Velveeta -- dairy adjacent calorie delivery product -- is to cheese.

The AI companies seem to think if they hoover up enough raw material, their systems can “learn” to “create” actual music.  “After all,” these companies rationalize, “that’s the same thing all artists do; they hear all the music around them, absorb it, and synthesize it into something new!”

Uhhh, nope. That is fundamentally wrong.

Music is, and has always been, about the process of learning to make that music. It is not a product by itself. See the difference?

To start, the process always involves other human beings. Mentors, teachers, older musicians, peers, and celebrities can all play a role. But becoming a musician never happens in isolation. As youngsters, we hear music around us, and we strive to emulate who we hear. We suck at first, but we begin to improve. Sometimes our mentors are on recordings, sometimes they’re live in person. Sometimes we argue with them, sometimes they’re parental, caring and sweet. Sometimes we’re in the close quarters, sometimes we hear them from afar. Sometimes they’re in a classroom or teaching studio, sometimes they’re online. Some of their lessons are formal, sometimes their teaching just happens on the bandstand. But we always learn to make music from other humans, and learn to make it for other humans.

And their also wrong about how musicians listen. We never absorb everything that exists. Jazz pianist Brad Mehldau had a great take on the process of becoming a jazz musician in his amazing piece “Ideology, Burgers and Beer.” In the essay he writes:

“You can’t love everything, all the time … When you build your identity as a player, you do so in part by excluding a bunch of other identities, at least temporarily.”

AI-obsessed tech companies seem to think if they devour enough data, their systems would be able to generate some MAAP that sounds close enough to synth-pop-rock-country-rap-bluegrass that it could “fool” people into believing it was music made by humans. Who knows, perhaps they’re right about that. I’m sure there are enough of us “foolish” humans who couldn’t tell the difference or wouldn’t care. Who knows, I assume would likely fail that particular Turing test.

But what tech companies don’t understand is that little test does not matter. That test is just a cute little game driven by whether we can tell whether the end product was made humans or machine. But y’see, that’s a test that only cares about product.

Here -- I can give you a much more efficient test of musical process. Put me into a room (or any other musician, for that matter) along with three people: (1) the lead programmer for an AI system that generates MAAP, (2) the person who then types “prompts” into that AI system to create MAAP and (3) a musician. I promise you I can tell you which one is a musician, every time. How? I’ll sit in a chair facing each one of those people and say: “Look me in the eyes and make music.”

Any person who is a musician has followed a process which is recognizable to other musicians which will allow them to pass that test. What music they make during that test may not be the style of music I like, or a song I like. It might not even be any good. But it will be recognizable as part of a musical process.

[Side note, to any programmers who are also musicians: I’m certainly not saying someone can’t be two things. I absolutely was both. I’m just saying whomever is currently running the show at AI-obsessed tech companies are most certainly not artists.]

Maybe part of their difficulty in understanding that music all about process is that we musicians don’t all follow the exact same process. There’s plenty of variation. The type of music we make will influence -- and will be influenced by -- the process we use. Even among musicians within the same genre, the process can vary considerably. For example, I am a jazz musician. I know many other jazz musicians, and they’re process can vary depending on who they listened to growing up, where they grew up, who their family and friends were, and how long they’ve been on their path. But me or any other jazz musician can absolutely recognize the process of other jazz musicians when they encounter one.

[Side note: Whether or not jazz musicians actually into a room recognize the jazz process among the others they encounter is a different question. There’s plenty of sexism, racism, and general hoity-toity-ness among our ranks. But isn’t that true for everybody?]

I love learning about other jazz musicians’ processes. Drummer Dave Grohl speaks eloquently about becoming a musician.

“Musicians should go to a yard sale and buy an old f****** drum set and get in their garage and just suck. And get their friends to come in and they'll suck, too. And then they'll f******* start playing and they'll have the best time they've ever had in their lives and then all of a sudden they'll become Nirvana.”

Any human being on the planet who watches Taylor Swift’s SNL Monologue Song for 4½ minutes would recognize how her process brought her to a place of superior musicianship.

When I was young, I pored over books about Charlie Parker’s painful process, getting booted from jam sessions because he sounded terrible, which caused him to recede into a literal woodshed to practice until he could play. I loved reading that John Coltrane was so dedicated to practicing, he would go to the back of an airplane and play the C scale for 12 hours straight. I got to meet saxophonist Grace Kelly as a guest artist here at IC and enjoyed hearing about how she got to be mentored by jazz jazz legend Lee Konitz, who must have recognized how deeply she cared about the improvisational process.

Nowadays, as a jazz professor, I have discovered I have to remain vigilant about protecting the process. I am the Director of Jazz Studies at Ithaca College, which is a truly wonderful school. Among my faculty colleagues are all manner of amazing musical specialists, yet I am the one full-time jazz teacher. That means that in order to offer a robust quantity of jazz ensembles and courses, I have to rely on both part-time faculty and full-timers who mostly teach in other areas to help. From time to time, an administrator or department head will approach and request that [FACULTY MEMBER X, Y or Z] teach a jazz course or lessons. “Mike, of course that person could coach a jazz combo…??”

Hmmmm.

I have a reliable little test that to answer that question. Put me in the room with that faculty member for about, oh, maybe two minutes. I will tell that person “Let’s do a song. Even though I’m a saxophonist, I’ll play you some chords on piano. What song would you like to do?” If they are a jazz musician, we will easily find at least one song we have in common, and will have fun making music together. It could even be a 12-bar blues. Easy peasy. But if they’re not a jazz musician, the entire idea of this exercise will freak them the f*$% out.

Of course, no admin has yet to take me up on this test. The funny thing is that I wouldn’t actually need that person to be a great jazz musician to be an effective teacher; they just need to be a jazz musician. This little test shows quite effectively who is a jazz musician, and who isn’t.

[Side note, for IC jazz auditioners: This is essentially the same thing I do with prospective students. And likewise, you do not have to be great, you just have to demonstrate that you love the process.]

I mentioned that I don’t fear technology. Before my current job at Ithaca, I worked in tech for several years. I was a programmer for two companies, and I really enjoyed it. I was fairly decent at writing code, but it turned out that the aspect of the job I really excelled at was creative problem solving. I could sit in a room with clients, dig into their process, and help to design a solution. I really liked getting to understand how our technology could improve their workflow to be more efficient, accurate, or helpful to their customers. I would ultimately turn that plan into a software solution. Later I became a team leader and would oversee other programmers who were creating these systems together.

So: Do I think it possible that AI could ultimately help musicians?

Sure. Consider that point stipulated. Not all AI is “generative” in the same way. Some of it is the kind of machine learning that is supposed to help us answer emails faster so we can get back to actually being creative. Sure, whatever.

I’m even willing to admit that it might be possible someday that AI could actually help us be creative. Musicians who have already learned the music-making process are perpetually seeking new inspirations, so I can understand how introducing elements of randomness or unpredictability found in AI could clear some cobwebs from the pathways of their improvising or composing mind, coaxing out some new creativity. Years ago there was a program called “Max” that provided the same function for composers and improvisers.

But tech companies are certainly not heading in this direction with their products now. I had always thought that Apple, a company who’s made platforms and tools that artists and musicians have used for decades, understood something at the core of the creative process.

But whoo-eee, Apple has really gone off the rails here. In just the last few weeks, Apple Music has suggested AI-generated MAAP it claimed had been made by jazz trombonist JJ Johnson. Apple’s recent tone-deaf “Crush” ad proved just how disconnected they have become and how far they’ve strayed from their earlier ideals.

It’s actually quite painful for me to have come to this conclusion about Apple. Their products have been a big part of my whole family since the 1970s. My dad loved Apple from the beginning. He bought me an Apple II computer as a kid, and I’ve used Macs since the 80s. Their tools have helped my creative process immeasurably over many. years.

Perhaps Apple’s enshittification was inevitable. The business world seemed to tell them that they had to get involved in AI or they’d fall behind their competetors. Perhaps they can find the way back if enough of us customers stay vigilant and protect the creative process.

Regardless, what I can say is that I am not worried about AI. At the moment, none of the tech companies show any signs whatsoever of being able to replace our process.

Mike Titlebaum
Professor & Director of Jazz Studies, Ithaca College
July 24, 2024