For maybe as long as I’ve been published (a decade this fall!), I’ve used the line, “Timothy Johnson fears nothing more than the future, so he writes about it and hopes he’s wrong.”
I think that’s important contextualization for who I am as a futurist and precog (admittedly, I don’t think I’m a very good one), but I think it’s good to frame all of this with the sentiment that I hope I’m wrong about generative AI, or large-language models (LLMs), which is a less-sexy name but probably a more accurate and responsible one we should be using (and I will use for the rest of this post).
I struggled with this post because not only is it a very complex issue (one I’m betting the non-artist techbros will challenge), but I think I had illusions it would be my definitive take on LLMs and its position in art. I’d write about it once, and that would be that. But, I know these can’t be my last words on the subject because, well doggonit, I’m a human being, but more to the point, LLMs are still so new and changing so rapidly that my takes on it are evolving, too.
Suffice to say this post may exhaust you (it exhausted me to write it), but it won’t be exhaustive.
I’m going to skip any more preamble and assume you have a baseline knowledge of what LLMs are as well as a basic understanding of their arguments for and against. Others have highlighted the practical issues and inherent dangers as well as the potential benefits of LLMs better than I can.
In my view, all of the current dialogue on LLMs is missing a big picture, one that holds incredible implications for not only artists but also humanity.
“The technology is opening new pathways for creativity,” Yan Wu writes. Okay, but whose creativity? Who, or what, is doing the creative work?
Wu continues by employing the credibility of a career photographer who makes the claim the effective use of LLMs requires vast artistic knowledge of the discipline, but there’s an important nuance here that’s missing: It may require a measure of academic knowledge gained through academic study, but it doesn’t utilize the knowledge and skill developed through the course of that art form’s practice, nor does it utilize the artist’s identity and style they developed through the mastery of their art’s practice.
Art doesn’t exist in concept alone. We create it through masterful execution.
Artists generally seem to know this by instinct. Author Chuck Wendig reminds us art is about people. “It is by people. It is for people,” he writes.
Others have argued against LLMs for more practicable reasons (the threat of eroding basic communication skills), and I do have concerns about that, but while the risk there is not at all something we should be cavalier about, it’s fairly obvious.
No, this also isn’t about the money and copyrights issues, which also are important but obvious.
My concern is about the death of human creativity. LLMs have the potential to end human-generated art, and I think that’s a bad thing because our art is an important part of what distinguishes us in the universe. Our art is part of what makes us special and noteworthy in the face of indifference. The universe created us, and we create art.
Recognizing my statement about the end of human-generated art is controversial and probably distracting you already, I’m going to argue a few core ideas here.
- AI-generated art, or art that utilizes an LLM in the generation of its creative elements, is not wholly human-generated art. This seems obvious to me, but there is no shortage of people on the internet who contend this idea.
- Generative AI is currently incapable of creating art. It can only synthesize. Importantly, my argument is not that AI-generated product will never be art but that, right now, it isn’t and can’t be at LLMs’ current capabilities.
- Anyone utilizing an LLM to generate the creative portion of a piece of art is not the artist of that piece in question but an editor or publisher. The artist is whatever entity constitutes the LLM.
- Most important, we need to be wary of normalizing LLMs in the creation process because, if it becomes status quo, we humans will lose the ability to create art. (This one you might already be balking at, but please stick with me. The key is in the practice of an artform and its creation, what creation is, and how we do it.)
What is art?
Before I get into all of this, though, I need to explain how I define art. I hear you sighing. Please bear with me.
I’ve long taken exception with the tendency of artists to argue what is and what isn’t art. This ain’t that. I think drawing lines between what is a human creation and what is human artistry sort of misses the point of creation altogether.
My bar for what constitutes art is relatively low. Here’s essentially how I evaluate what is art and what is not art.
- A piece of artwork must explore a concept in a definable or interpretable way.
- An audience, viewer, or recipient of that piece of artwork must be able to explore a concept in a definable or interpretable way.
That’s it. That’s art. Art is like spotlights on facets of the experience that is living and existing. It doesn’t make any statements itself. It just is, and rather than taking an active role in a conversation about existence, it requires both a living creator and a living observer to define or interpret it in existential terms.
Fundamentally, art is a conversation between two people: the artist and the audience or viewer. Art is an artist mingling with the audience or viewers and saying, “Here is what I was trying to describe with my work. What do you see?”
Both parts of this interaction are critical to experiencing art.
All of this begs the question, if a piece in question is generated by a computer, can the artist truly be in conversation with their audience? Moreover, in this case, does the individual who ferried the computer’s generated creative to the audience have the right to enter that conversation simply because the entity that created the creative does not have agency of their own?
I don’t think I can answer those questions in general, but I do think the distinction (or that there is a distinction) is worth considering in the context of LLMs.
How do LLMs work?
There are many resources on the internet to answer this question, and having a base understanding of LLMs is important. Here is deep-fake Ryan Gosling to explain it.
Got all that? Good. Me neither.
The most important point (whether you watched that or not) is this is a computational model that fundamentally works by intaking a ton of input to generate an output. It still is a computer at its core, processing data in ways computers always have. LLMs are not artificial intelligence. They’re algorithms and mathematical formulas.
People, on the other hand, are synapses and neurons firing in a blob of fatty tissue, and this cosmic orchestra results in everything you are, everything you do, and everything you experience. In art, it dictates all of your creative decision-making.
Importantly, we’re going to have this debate for years, probably, until we actually do create artificial intelligence, but LLMs are just computers doing computery things.
How does human creativity work?
When I talk to people who aren’t artists, they generally have difficulty understanding creativity isn’t a quantifiable concept. They typically think we writers and artists are just playing with Legos and following some kind of step-by-step guide, whether literal or written in the stars. They think creativity is a task to be managed, that we can utilize to-do lists and spreadsheets to make our work easier. They tend to think, well, if you can type a hundred words a minute, you can write a novel in…[calculating]…about eight hours and twenty minutes.
Man, writing a novel sounds really doable when you have zero idea what’s involved in artistic creation.
I do use lists and spreadsheets for my work. Sometimes I use them as part of the business of publishing and marketing (this very post, for instance, is on a spreadsheet), and sometimes I use them to focus or understand my creativity. They’re useful tools that enable me to better practice my art.
However, never do I use these tools to guide my creative work or to make creative decisions of any kind.
To get a handle on human creativity, it’s important to understand the distinctions between emulation, synthesis, and creation.
Emulation: An artist studies another artist’s work and replicates its style, aesthetic, and form.
Synthesis: An artist studies many artists’ works; replicates styles, aesthetics, and forms; and creates something new.
Creation: An artist masters an art form through dedicated study and practice of the craft; establishes a unique style, aesthetic, and identity; and creates something original from their unique artistic self.
Here’s the rub. When I taught freshman composition, most of my students (who all had a typical general education most citizens of the developed world get) really, really struggled with synthesis as a concept. In freshman composition, we teach synthesis in the application of researching a topic and a question to the point where the student has learned, grown, and formulated their own idea from evidence. Many ideas come together to form a new idea.
By the end of each semester, most composition students still struggle with that concept, at least in its practice. I get it. It’s challenging for anyone to wrap their mind around.
Composition is creativity is art, and that process applies here in the sense that any artist engaging in synthesis must be a student.
I hear you writers in the back muttering, “but I never took a formal creative writing class, and look at me. I’m kick-ass.” Indeed you are, but you definitely studied literature or creative writing, and more important, you practiced it. You just did it through sheer passion, and that’s amazing. Everyone give the cool kids in the back a hand!
The thing about synthesis is, while it’s difficult to master, it’s even more challenging to advance beyond into true creativity, where the artist no longer generates from external influences but from an internal identity. In true creation, the artist conceives of every part of their creative work and executes it to the best of their skills and abilities (which they’ve developed through dedicated practice), in their own identifiable style, with their own certifiable signatures. Artists create when they’ve emulated and synthesized, by studying and producing work, for so long that they’ve internalized so much artistry and established a creative identity of their own.
So how do I create? After all these years of writing, I honestly don’t know. I’ve come to learn how my creativity works and how to harness it. I’ve come to understand how to manage external factors such as sleep, diet, and environment that help support my creativity. I’ve come to know much about my creativity, but it remains a wind I can only capture with a sail, never control.
I do know one thing for sure: true creation is the most difficult thing I’ve ever done, and I’ve failed at it, landing in the synthesis or even emulation area often. It requires an artist to crawl in the dark, uncertain as to what their searching fingers will grasp. Writers have to find themselves confronting the emptiness of the next word, the next sentence, the next paragraph. Writers have to contend with the blank page and the blinking cursor. Artists must get lost to find creation.
And that’s where the magic happens. The struggle generates true human creativity.
LLMs promise to not only ease the struggle but also make it irrelevant, and it comes with a cost: the artist’s discovery, their crawl through the dark.
How does this art thing work, and what are you so afraid of?
I’m afraid of the potential that, as LLMs improve and as we create actual artificial intelligence, more artists will utilize them when creating is difficult, and they not only won’t know they’re robbing themselves of their own creativity, but they’ll also accept it as a norm. Using LLMs will just be how we create art (except the irony is we won’t be doing the creation anymore). If it becomes the standard in human artistic creation, we will unwittingly be removing humanity from the creation of art and will, in essence, cede artistic creation to LLMs and AIs, whenever they actually come to be. Implicitly, human artistic creation will die, and if that happens, it would be a tragedy for those of us who believe human artistic creation is one of our few traits that make us special in the universe.
We’ve built something of a formula here for artistic creation: dedication, time, and energy spent in the study and practice of an artform + struggling through the difficulty of creation = artistic creation. The key to every step in this formula is you, the artist. It’s in the human’s hands-on engagement in the creative act. When we venture into the dark emptiness, we can bring tools and supplies with us, but ultimately, it’s our hands and knees in the warm, viscous, gooey, scary muck.
It’s hard and gross and unglamorous, but it’s necessary.
At the end of everything, all conversations aside, all arguments heard, all disputes put to rest, one simple truth remains: to be an artist who creates art, you have to do the work.
It’s the creative work—every decision from the micro to the macro—that LLMs threaten to do for us. LLMs threaten to rob every artist of their growth by making the work not only easier but also unnecessary. If we accept LLMs as a tool to be utilized in the creation of any art, we will forfeit real, human creativity.
It is not a tool. It’s a cheat code that cuts out all of the storytelling, all of the challenges, all of the need for ingenuity, and transports you straight to the end so you can unlock an achievement. But is it, then, really your achievement?
When are you going to shut up about this, Tim?
Soon. Promise.
I think about those freshman composition students who barely can wrap their heads around synthesis as a concept let alone do it. Most of them are not artists, but synthesis skills are necessary in every field and discipline. Synthesis skills are necessary for adaptation and ingenuity. They’re necessary for solving emergent problems that have never been solved before. Eli Amdur had it right in that Forbes op-ed I linked above that synthesis skills are fundamental to communication.
Too often, we treat education and growth—be it in artistry or otherwise—as a process of students consuming lectures and facts and outputting that knowledge on tests. The trouble is that’s what computers do, and we’re capable of so much more.
We see our potential in our artwork. Through our artwork, we inspire ourselves to do better, to innovate, to explore, to create.
LLMs have the potential to accomplish much on our behalf. They can be a great asset in the elimination of mundanity from our lives. They can help us solve great problems that we may not be able to solve alone. They can help us surpass our limitations and free up humanity for other pursuits that enrich our lives.
Creative pursuits, perhaps?
We can have all of this advancement without paying this hefty, unnecessary price. Let’s listen to artists on this issue, because we stand to lose the most and have an understanding of not only what LLMs are actually automating but also what we all stand to lose. Let’s ensure we don’t forfeit human creativity. Let’s pressure LLM companies to implement metadata and tracking tools so we can identify wholly or partially generated LLM content. After all, they already track everything else, so why not this? Let’s push our governments to create policies and laws that protect human creativity and the livelihoods of human creatives, ensuring artists can opt out from their work getting absorbed by these models and mandating we get fair compensation if and when LLMs gobble us up. Let’s pressure industries that utilize creative content and artwork to commit to using only human-generated content and artwork. Let’s get our educational institutions to teach through active learning and prevent the utilization of LLMs where it inhibits intellectual and creative growth, and let’s do what we can to ensure those institutions have the tools to do so.
Human artistry is on the line, and its loss is completely avoidable. It’s just a matter of convincing enough of us that’s a potential outcome, conceiving practicable solutions, and finding the will to preserve one of our most precious features: our art.
If you liked this post, please consider subscribing to my monthly Substack and get stuff like this, thoughts on my current favorite stories, and news and updates right in your email inbox. It’s free, so what have you got to lose?