Artificial Creativity

How AI teaches us to distinguish between humans, art, and industry

I hinted at some thoughts about AI in my last Team Human monologue, and those generated as many emails as the main subject - which was learning to disengage from the pace of the internet. These ideas are related, of course: making a conscious, human choice about whether and how we use AI requires the same application of agency as disengaging from the pace of digital industrialism. 

But understanding the reasons and complications, and history, are pretty specific in the case of AI. 

I just had the pleasure and honor of speaking at a wonderful music industry program at Eastern Tennessee State University. They wanted a talk about how AI would impact the creative industries, and specifically music. And as I looked out at the bright young faces, I couldn’t help but begin with “First they came for the cab drivers, and I said nothing because I am not a cab driver…” 

Of course, only two older professors knew the reference I was talking about - a poem by Pastor Martin Neimoller, a one-time nazi supporter after his liberation from Dachau. I wasn’t trying to equate AIs with Nazis so much as to suggest that the way to come to an understanding of how AI may someday impact whatever we do is to look at the people it is impacting today. 

It’s no crime against humanity to replace most cab drivers with autonomous vehicles. That’s because the vast majority of Uber drivers are not doing it for the fun of it, but for the money. They’d rather be out playing softball or making music. If we really could substitute for our labor with machines - ones that don’t pollute or create even more work - it wouldn’t be a problem, as long as we all felt okay about letting people live and enjoy the bounty of a high-tech society without necessarily contributing to the industrial economy. Having more than enough of everything shouldn’t be a problem, and if technology really could make that a reality it’s not up to the tech bros to design the post-work economy.  

The creative industries are different, though. Or more to the point, creativity is different from the creative industries - as different as music is from the music business. AI’s are fine for industry. They’re a product of industrial age thinking, and reinforce its values. They do not create, they model. That’s what “large language models” means. Just like computers “model” a typewriter as a word processor, or a complex weather system in order to predict the future, AIs model writing or composing by looking at all the texts and songs and generating an approximate average. They model the most typical version of something possible. 

Now that’s fine for certain forms of entertainment. Maybe a Marvel movie or, better, a roller coaster. An AI can look at all the successful roller coasters out there, measure the effects on riders, and come up with increasingly idealized combinations of drops and loops and spirals. Aristotle considered “spectacle” the lowest priority in drama, kind of like cheating when compared with character or plot. Like Las Vegas fireworks instead of a Shakespeare soliloquy. But it has its place, particularly in an entertainment industry where formula is more important than personal expression or audience insight. It’s more like porn than a human relationship. 

This industrial approach to the arts is what brought us autotuning of vocalists, and an aesthetic that would condemn James Brown’s reaching up to a note as a form of noise, not signal. Technology removes the soul from human performance in order to make it more compatible with industry. Perfect. Not art, but idealized product. 

There’s still some value to AI in all this. But if AI is a threat, I think it’s more to the creative industries than to creativity itself. So, take the recording engineer. On a podcast, even like this one, interviews are recorded on two separate tracks - one for me, and one for the guest. We line those files up, but then an engineer - usually Luke, in our case - spends an hour or so “cross-fading” the tracks. That just means turning the volume of one person’s track completely off while the other person is speaking. It reduces any background noise. And so far, most automated programs have been bad at doing that. AI could easily do that task; it may already be doing it in some audio program, for all I know. 

Or take journalism. There used to be an entry-level sports journalism job for putting the overnight sports scores into language. The Mets trounced the Braves seven to nothing; the Orioles squeaked by the Cards in the 8th inning, 5-4…and so on. A bunch of numbers and stats are assigned appropriate verbs. The same goes for stock market results, or weather reports. Sports Illustrated, the Wall Street Journal and other publications have been using AI for a while now to do that sort of writing. 

The job I had back when, reading screenplays and writing “coverage” - basically a summary and evaluation - for a talent agency? That’s being done by AI now, and as far as the film industry is concerned I’m sure it’s doing a better job than an artsy graduate student like me did at the time. No pesky aesthetic values to get in the way of commercial evaluation. 

While none of use really loved those drudgery jobs at recording studios, newspapers or agencies, they were important to our development. It was how we learned the recording software - got those controls in our bones. Or where we learned filing deadlines, and how to vary sentences and rhythms to make them more interesting. Reading and evaluating ten screenplays a week taught me more about screenwriting than any course I took in film school. When we relegate all this engineering, reading, and writing to machines instead of human apprentices, how does the next generation arrive at a level of mastery? Great surgeons tell the resident, “close for me” after the operation. That’s not just because they’re lazy, but because the resident needs opportunities to practice sutures. 

There are highly industrialized, commercial, or maybe “repeatable” and easily simulated roles that professional artists play, too. 

Sure, a record producer can use an AI to perform the role of a session player on an album. They can instruct the AI to come up with a solo “like Slash” would play. And the AI will dutifully listen to all of Slash’s solos and construct the most average Slash solo ever. More typically Slash than Slash himself. That may be fine for elevator music or some TV commercial. But what of the session musicians who eventually become the next Toto or Billy Preston? Where is the culture in which they are going to stew? 

Plus, sometimes session musicians are crazy talented, themselves. Ever hear of Wrecking Crew? They were the band for the Byrds, Bob Dylan, the Monkees, and the Beach Boys. Yes - the Tannerin (something like a theremin) in Good Vibations? That was the Wrecking Crew’s invention. Or Muscle Shoals, the session band behind Aretha Franklin and the Rolling Stones? The ideas generated by these people don’t come from AI’s averaging-out past hits.  

As far as industry is concerned, getting rid of people and replacing them with AI is the business opportunity. That’s happening on all levels of the AI phenomenon. AI is the last gasp of a dying digital hype cycle. The ultimate meta. Just as the stock market goes “meta” on itself with derivatives and derivatives of derivatives, digital businesses level up into abstraction as a substitute for providing any real value. Web 1 (the dotcom boom) becomes Web 2 (the social media boom) becomes Web 3 (some combination of crypto and virtual reality). When all of those fail, we get AI which is supposed to turn the entirety of the internet into its content as it churns out some new meta-version of digital industrialism that can create exponential growth.  

But every modern business model is depending on the same old industrial age premise of making money by getting rid of people, or at least people with skills. That’s what the assembly line was for: get rid of skilled craftspeople, and create a factory system where unskilled, low-paid workers can be trained in minutes and replaced just as fast if they get pesky or try to organize. Business plans for digital businesses only succeed in getting venture capital if they can prove they don’t need human workers in order to scale infinitely. AI offers that possibility. 

I’m not so sure it will work as easily for a cultural product like music as it did for consumer goods. There’s something anonymous about AI music and art that is ultimately unsatisfying to anyone paying attention. Music is a medium — something that conveys meaning from one person to another. If there’s no one on the other side, what is it? 

I remember when rave started (what’s now called EDM or electronic dance music) we all loved how there was no one on stage. For those of us used to rock and roll shows with long-haired dudes playing masturbatory guitar solos on stage, the anonymity of early techno was refreshing. The gatherings were so impromptu that sometimes we couldn’t even see the DJ from the dance area. We were more interested in each other, and felt that whoever was spinning the records was feeding off of our energy, anyway. Even then, for better and for worse, the DJ booth ended up becoming more prominent, the DJ’s names became “headliners” for events, and superstar DJs like Calvin Harris won residencies at Las Vegas for over a million dollars a night. On the one hand, this is simply capitalism co-opting another commons. But it’s also our natural, healthy tendency to seek out the human being on the other side of any cultural creation. We engage with art for a sense of connection to the artist. Someone has to be home. 

The way we relate to that someone, however, will still be determined to a large extent by the dominant media of our age. As McLuhan colleague John Culkin once said, we create a tool and from then on, it creates us. The tools we use in the arts change our aesthetics and our experience. Early industrial age media technologies like the printing press changed the way we understood literature. The highly personalized handwritten manuscript became the easily reproducible printed book. The flow of handwriting gave way to the mechanized units of typesetting. Every M would be the same, no matter its context.

Not coincidentally, that same wave of innovation also brought us the keyboard instruments. Again, each note - say an Ab - could only be tuned the same way, no matter which key the instrument was supposed to play. As any string musician knows, an Ab is not the same note when it is functioning as a G# in the key of E. Keyboard instruments required the invention of what is known as “equal temperament” - a compromise of fine tuning for mechanical keyboard instruments to approximate the notes of the music score. In some ways, that can be understood as the precursor to MIDI and digital music, which depend on these discrete, exact metrics in order to execute a song. 

As I’ve written elsewhere (in books like Life Inc and Throwing Rocks at the Google Bus) this printing press moment is also when we got hourly wages and a clock on the highest tower of the town. Human beings were to be understood as units of labor, and as interchangeable as typeset letters or keyboard notes. And that has come to define the quality of every  industry, music included. Think of the recording industry, with the artists behind glass, responding to a “producer” operating the board in a control room. The competitive advantage of a recording, in the most pop of pop music, ended up having less to do with the artist than the producer. 

AI may just bring this industrialization to its logical extreme. If artists and producers are both replaced by AI’s, how does anyone distinguish themselves? If all the book publishers, music makers, art producers, and film studios are using the same AI technologies to create their products, how does anyone gain a competitive advantage? They can’t. The only way to win is to become the only player. The commodification of its product requires a business to become a monopoly. 

No wonder all the tech bros’ business tactics are aimed at total domination and complete monopoly. That’s why they are asking for “regulation” that effectively locks in the current AI companies. They already realize that the endpoint of AI industrialization is equivalency. Just like the printing press and equal temperament turned every letter and note the same, AI will commoditize every industry in which it is deployed. The only way to make money is to control the whole market. 

I keep trying to think of AI in terms of augmentation; AI as some sort of helper. The ultimate Paperclip Man (you have to have been around for early Windows to remember that little pop-up system help character). An AI developed in that spirit would help me by doing all the grunt work of my writing, so I could do the creative stuff like crafting the scenes and moments. But as I’ve tried to use it, I find myself doing all the grunt work, feeding my characters and their histories, and source materials into the AI, so that it can do the fun work of thinking up the scenes. 

And that got me thinking about the way AI doesn’t really replace labor at all. It simply moves us humans further down the hierarchy of skills. It’s like the dumbwaiter, which didn’t really save Thomas Jefferson’s enslaved servants any labor. It just spared Jefferson and his guests the indignity of actually interacting with the enslaved human servant. They were still there, just more hidden. Likewise, an AI may take the job of a driver, a writer, or a lawyer, but it requires the work of legions of people to mine for the rare earth metals, get the cobalt out of lakes, tag all the data that it’s fed, and so on. So it doesn’t simply take our jobs, it down-skills our labor. 

AI doesn’t simply take our jobs, it down-skills our labor. 

There was one micro-moment, however, when the AI showed me something valuable. I asked it to write some scenes of my graphic novel that I had already written, just to see if it had any good ideas for me to consider. And one scene that it wrote ended up almost exactly the same as the one I had written. It was a strange feeling to read it, where I went from feeling good about having finally gotten the AI to write like me, to feeling disgusted by what I had written, myself. If the AI looked all over the place and “averaged out” the most appropriate scene, it means that I was writing in a perfectly average way, myself. I had arrived at the same cliché as the machine! 

That’s when I felt like I finally recognized the value of this technology in creative work: not to come up with the ideal commercial product, but to show us perfect examples of what to avoid. They can act almost like cheating detectors. They can identify when we have stopped being truly creative, and have fallen into regurgitating tropes. If a human does what the AI did, it means the human needs to take a break. 

It also suggests a unique role for AI in arts and entertainment. Instead of using AI to commodify all the arts and entertainment that went before it, what about celebrating AI as its own form of entertainment? When we use AI to make a movie or a book, no one gets to see the AI at work, interact with it, or experience those weird uncanny valley moments that make AI so interesting to its developers. We see the finished product, which is about as interesting as looking at wallpaper. 

I don’t like looking at the pictures people have made with Midjourney or Dall-e, but I know they had a really profound experience iterating those pictures with the AIs. Looking at someone’s Midjourney picture is like listening to someone talk about their acid trip. You had to be there. Likewise, the value of AI is the opportunity to play with the AI itself. 

AI can still be a creative partner rather than an extension of the same industry that’s been squashing creativity all along. We can use it to help us distinguish between our own artistic innovation and our unconscious reversion to the mean - our own tension between creative and industrial urges. And we can still unleash all sorts of new creative, playful experiences for our audiences - not by using AI to simulate the artistic media of the past, but by crafting accessible, transparent, and direct experiences of artificial intelligence itself.