The Experience of Creativity
Generative AI is not democratizing creativity.
Years ago, when I reviewed Jeff Tweedy | Starship Casual ‘s book How to Write One Song for the Chicago Tribune I thought it was only appropriate that, using his suggestions, I try to write one song.
I’m reasonably musical. I play the drums well enough to gig with some fellow geezers around town playing mostly Allman Brothers covers, and many many moons ago, back in Chicago was the original drummer in indie rock non-legends Quiet Kid, which played many shows at various cool clubs, including the one (Lounge Ax) co-owned by Jeff Tweedy’s wife, Sue Miller.
I have also been a mediocre guitar player for almost 40 years, handy around a campfire hootenanny that wants to collectively belt “American Pie,” but not really good for much else. My ear for pitch is above average, so within a limited range I can sing on key, but that same ear lets me know that I do not have what is known as a good vocal “quality,” so at best you want me in the background singing some ooh oohs and la la las.
But because Jeff Tweedy is very convincing about how it is rather straightforward to write one song, I did indeed write one song.
My song was not good, and no one has heard the song other than me. Fortunately, the penultimate chapter of How to Write One Song is titled, “What did you make? Is it any good?: Does it need to be good?” in which Jeff Tweedy reminds us that the meaning is in the attempt, not the result, and a key to development is being “OK with being bad at what you do.”
This may not sound like uplifting advice, but it is in truth excellent advice because of its accuracy, and while How to Write One Song does serve as a “how to” for songwriting, at its heart it’s more of a “why to,” with that why being to experience what it means to make an attempt at being creative.
One of the persistent narratives that has cropped up around generative AI is that it “democratizes creativity” by making things like text, image, sound, and video generation possible for those who do not otherwise possess the necessary skills to generate text, images, sound or video.
My instinctual response to this position from the get-go is that it is bullshit, and using these tools to substitute for human activity and experience is not democratizing creativity because being creative is an act of “being,” of “doing,” an “experience.” If that experience is removed, the act is no longer creative.
Obviously I’m talking about the difference between process (a creative act) and a product (a creative output). Generative AI allows those of us without the necessary skills or talent to create a simulated creative output, but this does not mean we have been “creative.”
A recent video from Rick Beato drove home this distinction in a way that finally crystalized my own thoughts into something more explicable and tangible.
In the video Beato walks us through a process he used to gin up an AI musical “artist” using a combination of ChatGPT, Claude, and Suno, a music-generating generative AI application. Meet Eli Mercer:
Eli Mercer is rather plausible looking as an indie/alternative artist: handsome, scruffy, a little punk (Nirvana t-shirt) and a little folky (jean jacket). I couldn’t help but think of this guy whose name I at first couldn’t remember, but is known for doing flips of the piano when singing his most popular song, so searching for that turned him up.
Beato goes to the Claude LLM with a song title (“West Texas Dreams”) and prompts it to write lyrics for an alternative song according to a prescribed structure (verse, pre-chorus, chorus, etc…) about “a college student dropping out to pursue their dreams of becoming a novelist in West Texas.”
Claude extrudes the lyrics, which Beato takes to Suno, prompting it with the desired style of the song (airy, alternative, acoustic) and in a matter of minutes we have, objectively speaking, a far better song that I produced using Jeff Tweedy’s method.
If you don’t want to watch the whole video, I’ve queued it to the start of the song below:
When I commented on the video at BlueSky, Keegan Lannon nailed what the song sounds like:
When you first hear the song you may think, “not bad,” because it is exactly that, not bad. It sounds like a song should sound. Eli Mercer sings in key, his vocal quality is fine without being distinctive. The chord changes (which Beato with his expert ear calls out) are sensible and predictable. The lyrics are also sensible:
(First Verse)
Lecture halls and fluorescent lights Textbooks stacked three stories high But I’m writing stories in the margins While the professor drones on by Got a letter in my pocket, says I’m failing three or four' But there’s a highway heading westward And it’s calling out for more
It’s a song about exactly what Beato asked for. I mean, exactly. Here’s the chorus.
I'm going to West Texas with a notebook and a dream Where the sky goes on forever and nothing's what it seems Trade my cap and gown for boots and dusty jeans I'm going to West Texas to write the stories that I've seen.
in More Than Words I have a long section exploring one of the most important skills for moving forward in a world with generative AI technology: taste.
If you spend any time with AI “artist” Eli Mercer’s AI “song” we should agree with Rick Beato that while the song isn’t bad, it does indeed “suck.” It sucks because there is no spark of life. It is aural wallpaper.
Someone will read this and tell me that as long as people like it, who cares? Plenty of pop music is derivative and people eat it up. There is something called The Velvet Sundown, which is an AI-generated “group” getting millions of plays on Spotify.
I’m sorry, but they also suck.
What I think Beato’s video perfectly demonstrates is that there is a difference between derivative or even formulaic human products and what generative AI is doing with its distillation and literal copying of the original human outputs. It is truly lifeless.
Not coincidentally, the experience of creativity that Tweedy unfurls in How to Write One Song is nothing like what happens when you plug a prompt into an LLM. He shares a number of exercises that are designed to unleash the creative juices, and most of them involve turning off your rational brain. He demonstrates what he calls a “word ladder” where you write down ten verbs that are associated with something (he uses a physician) and then then ten nouns within your field of vision. (Tweedy’s objects are what you would expect for a musician in a studio: guitar, turntable, wall, window, microphone, carpet.)
I’m going to try this right now. I’m going to use a musician and then the objects in my office:
Musician verbs
Play Practice
Tune Jam
Strum Strike
Compose Sing
Harmonize ListenOffice objects
Books Photos Posters Computer Window Blinds Closet Couch Notebook Backpack
The next step is to connect the nouns with verbs that in Tweedy’s words “don’t normally go together.” From there, he’ll produce a little quasi-poem attempting to string some of the words together. I guess I gotta try that too:
tuning the backpack to the notebook jam on the couch harmonizing, strumming blinds while listening as the window book practices posters singing the computer and the photos strike play inside the closet
The exercise isn’t designed for writing finished lyrics, but more to suspend judgment, to riff verbally on the page, and perhaps to unearth a striking phrase that may be the genesis for something else. A major block to creativity is the urge to evaluate, rather than continue to explore. With this approach you are simultaneously writing a song and not writing a song.
I try to achieve something similar with writing. I collect materials, and then see how they might fit together. I think, I ponder, I poke around things I’m reading to see if anything sparks connections. I knew upon seeing it that I wanted to explore Beato’s video, but I also didn’t know exactly where I wanted to go. I sat with it for a couple of days and once at my desk, looked over to my bookshelf and there was How to Write One Song, and we’re off and running.
Having only written one song I can’t say if my word-ladder poem shows promise as a jumping off point for more composed lyrics, but there is some apparent imagery that sort of intrigues me, namely “the photos strike play inside the closet.” It doesn’t have any apparent meaning, but that doesn’t mean it’s meaningless.
What I wish I could properly convey is the type of absorption I experienced in doing this exercise, a brief period of deep concentration that’s untethered from sense making, which is quite different from most of the writing I do, which is a deep concentration pursuing sense making.
But what both have in common is that feeling of absorption, of going to a place that is not entirely rational, then emerging with something that works. Tweedy’s finished lyrics are particularly great at this.
“Company in my Back” (From A Ghost is Born)
I attack with love, pure bug beauty
I curl my lips and crawl up to you
I attack with love, pure bug beauty
I curl my lips and crawl up to you
… And your afternoon
And I've been puking
… I move so slow, a steady crushing hand
Holy shit there's a company in my back
I move so slow, a steady crushing hand
Holy shit there's a company in my back
There’s almost a childlike quality to the rhymes in that opening stanza, contrasted with the more aggressive verbs of “curl,” and “attack.” The pre-chorus is both cryptic and clear, we don’t know who are what is being referred to with “and your afternoon,” while “and I’ve been puking” couldn’t be more straightforward. The chorus, repeating the same couplet drives home the message.
The song is about a period in Tweedy’s life where he was struggling with migraines which led to a Vicodin dependency. As I read the song it’s about a man under pressure, the “company” perhaps being the band or family or all of the above.
Or it could be about something else. It’s a beautiful, surprising, evocative song, whatever it means, wherever it comes from. The way the melody and musical arrangement accompany the lyrics is sort of stunning, both simple and deeply layered.
It is also undeniably the byproduct of a creative process. What I desperately want people to know, particularly students, is the depth of interest and pleasure in doing this kind of work, work in which you go searching for something to express, and which need not be confined to the creative arts.
A recent op-ed in the New York Times suggested that people “can’t quit” ChatGPT because it’s “fun,” but the fun it describes, e.g., “touching up the Mona Lisa” is a fun of novelty and distraction, rather than creativity and absorption. There’s nothing wrong with that kind of fun. It’s not actually possible to spend many hours a day in the kind of absorption genuine creativity requires. On an average day I can get in about four hours of good, concentrated writing. After that, I have to turn my attention to other things. I do Wordle almost every day.
But those four hours are the reason I keep at my work day after day, work I do with pleasure, under my own initiative, no overly concerned about what it adds up to in terms of money or a career. (At least while I’m writing. After writing there is some worry there.)
I saw a recent comment online about More Than Words that I can’t find anymore from someone who read it and they said something like, “this book isn’t really about AI,” which is exactly right. It’s about the importance of reading and writing to our own humanities.
Tweedy’s book isn’t really about how to write a song either, but about how to live in a space that makes room for being surprised by yourself.
Maybe in the end, this is the thing that outsourcing ourselves to generative AI most robs us of: surprise.
Links
This week at the Chicago Tribune I cover the community survey being conducted by the Chicago Public Library. If you’re a Chicagoan, you should definitely participate.
At Inside Higher Ed I note that thermostatic opinion has swung sentiments toward the positive for higher ed, which makes it necessary for them to seize the initiative with a compelling narrative.
Chicagoans should also note an impending appearance next Friday, July 25th by Hannah Pittard, author of the new novel If You Love It, Let It Kill You, at Exile in Bookville on South Michigan Avenue. I’ll be reviewing the book in my column next Sunday and should have a special Q&A with Pittard coming after that.
I’ve had it up to here with the debates about what’s happened to the commercial prospects of literary fiction, but Lincoln Michel is worth reading on the question, and if someone asks what I think, I just point to this.
Great piece by David Epstein about, well…just read it.
Brandon Taylor shared an excerpt from his introduction to the new Vintage edition of Jane Austen’s Persuasion.
By Greg Mania and via my friends McSweeney's “The EM Dash Responds to the AI Allegations.”
Recommendations
1. The Silent Cry by Kenzaburo Oe
2. Piranesi by Susanna Clarke
3. James by Percival Everett
4. The Master and Margarita by Mikhail Bulgakov
5. Rejection by Tony Tulathimutte
Christine O. - Chicago, IL
This list feels like permission to get a little weird: Trust Exercise by Susan Choi.
So, in an embarrassment of vacation riches, Mrs. Biblioracle and I are heading to Ireland next Friday for a trip with Brother and Sister-in-Law Biblioracle and I couldn’t be more excited, not just because I’ve never been, but because Ireland is one of the great literary tourist destinations of the world. (I’ve been rereading Dubliners in preparation.)
(If you have any tips for Dublin or Cork and the surrounding environs, please let me know in the comments.)
This means that there may be some interruption in the newsletter, something which has irked subscribers in the past, leading to cancellations, but in this case, should there not be a newsletter next week, I ask you to bear with me. That Hannah Pittard Q&A should be up even in my absence.
(But also, I may see something cool on the road that I can’t help but share.)
Thank you for your understanding, thank you for reading, and your overall support of this endeavor.
JW
The Biblioracle






Thanks for this! My husband and I were just talking about Tweedy and this book recently. I was also recently stuck on a chapter of a novel I’m revising and thought about dumping it into AI and then promptly dumped that idea and went for a run instead to let my mind marinate on the stuck-ness. The solution came to me. Maybe it was the Midwest humidity that helped unstick the sticky bit but allowing that creative process to work itself out was far more rewarding and original (I hope) than the AI alternative. Getting stuck sucks but AI can suck too for generating creative work that embodies you, as a human.
As for Ireland, we’ve been 3 times and would live there if we could. Get to Galway if you can. The Writers Museum in Dublin is closed now but the Little Museum looks promising. Walk about St. Stephen’s Green, of course. Temple Bar is touristy and you can find better places for trad music elsewhere. Get late night fish and chips from a chippie - know that some places leave the scales and bone in tact which can be surprising after numerous pints. The Guinness Storehouse tour is actually worth it for the view of the city at the top. If you’re driving to Cork, stop at Cashel (lovely town) or take a side trip to Kilkenny (also lovely) and if you have time, head to Kinsale when you’re in Cork. Slainte!
They are called New Grange. YThere are several pyramid sites. You can easily book a day or half day tour through Viator or other travel sites, if you are interested.