This essay is about generative AI, but I’m not here to shout about how unethical it is. I’m not here to rant about it being built on stolen work. I’m not here to belittle aspiring artists for effectively betraying their own community by using these tools or call them “scabs” for crossing some sort of metaphorical picket line. This essay isn’t about any of that because my goal is not to insult curious people. My goal is to demonstrate the precise ways in which the producers of GenAI tools aim to victimize and sabotage aspiring artists who choose to use them. I want to do this because users are hurting themselves far more than they are the rest of us. Helping them see that helps everyone.
The current damage these tools are doing to aspiring artists comes from “cognitive offloading” and “skill erosion,” which we’ll discuss at length. Often, discussion of these tools also mentions “deskilling,” but while that is the ultimate goal the creators of these tools have, this isn’t something I think is actually happening yet. We’ll get to why that is shortly.
Those of us who entered our teen years without a smart phone (suffering the agony of T9 word) only to have smart phones become the dominant information technology of the era will be viscerally familiar with cognitive offloading. Who remembers having to learn to navigate new places on the fly? How about memorizing directions, dates, events, or messages left for a parent or loved one? How about just having to memorize phone numbers? For older generations, they remember the shift from pencil and paper to reliable, handheld calculators and the necessity of knowing how to read a map or quickly locate your cardinal directions.
These are all cognitive tasks we’ve largely offloaded to technology (mostly smartphones). We let them do a good portion of our figuring and remembering, which is relatively harmless most of the time. We have plenty of higher order things to figure and remember. But a great deal of scholarship has come about recently regarding the danger of cognitively offloading our thinking—our ideating, our brainstorming, our reasoning, and our analyses—to AI tools. These are much higher order cognitive skills, the loss of which has major implications for our cognitive abilities (e.g. critical thinking, problem solving, etc.) and even overall cognitive health (e.g. reduced brain plasticity).
Skill erosion has also received a great deal of expert attention recently, particularly in regard to the use of AI tools. Skill erosion is a loss of skills or abilities that occurs in an individual when they don’t use said skill or ability for a long time and instead rely on something or someone else to do it for them. This is harmless—good, in fact—quite often. Think of construction workers who used to be able to swing a twelve-pound sledgehammer and consistently hit a precise point at a precise angle but can’t anymore because their foreman got a pneumatic punch or driver years ago, and they don’t have to swing a big hammer anymore.
However, there are plenty of negative examples as well, especially as we move to higher order tasks. Some high-end accounting firms recently discovered that staff couldn’t perform basic, fixed-asset accounting after an extended period of reliance on automated tools that required no manual input. There was a French plane that crashed because the pilots at the controls had diminished manual flying skills after years of over-reliance on autopilot.
For a systemic example, we can return to navigation. Many of our grandparents learned dead reckoning for navigational purposes and were well-versed in locating and using cardinal directions. Even most of our parents grew up learning to use the sun to orient themselves and at least being proficient at reading a map, but almost nobody in the millennial generation or younger has these skills. We grew up with mapquest, then Garmins, and now we all have navigation apps in our phones. Drop 99% of us in the woods with no service, and we wouldn’t even know what it means to follow the rising sun.
On the topic of deskilling, there’s a Forbes article, “AI Contributes to the ‘De-Skilling’ of Our Workforce,” that briefly discusses the “lost skills” of manual printing that have largely gone by the wayside as technology has usurped many of those roles, particularly for digital texts. It discusses the manual assembly of periodicals with typewritten content, photos pasted onto storyboards, sent to printers for “bluelines,” and on and on. None of that is needed in the modern age of digital media, and those operations are almost entirely automated even for print. Technological advances have always resulted in deskilling. In fact, the desire to render particular skills (or labors) obsolete has always been a large part of the motivation that led to innovation—the desire to no longer require X work to create Y product. Professions like blacksmithing and leatherworking are now relegated almost entirely to custom, artisanal projects, as the mainstream need for these skills has been entirely replaced by automation and large-scale industry.
But what abilities are being replaced (i.e. what deskilling occurs) when artists lean on GenAI tools? Use writers as an example: What cognitive processes are being offloaded when we generate text or plot ideas? What skills are we eroding when we ask a program to replace our brain during any part of the act of creation? These are the important questions of the moment. Let’s start by listing a few writerly skills. We’ve got plot development, narration (description, setting, etc.), dialogue, interiority, voice, style, and others, of course. These are some of the elements of our craft that we hone to become better, more efficient, more compelling storytellers. These are the skills the GenAI pushers are coming for.
On top of that, it’s important to understand that there’s no ceiling in any artist’s development. Hemingway had it right when he said, “We are all apprentices in a craft where no one ever becomes a master.” To me, this means that every artist is in a perpetual state of development—a state of trying to do more, failing, trying again, succeeding, and finding more is yet possible. As our skill accretes, so does its potential to grow further, similar to a planet or star gaining mass. We are all climbing different faces of the same mountain, but this mountain has no peak. So, what are we to make of deskilling, cognitive offloading, and skill erosion in this context?
First, I said earlier that deskilling is a term I see often at the center of anti-GenAI discourse, but that I don’t think it’s appropriate yet. I said this because I don’t actually think GenAI tools are deskilling the arts yet. The creators of these tools are trying to do this. Their goal is to do this, but they’re not currently doing it, and I think this is so for a few reasons. First and foremost, we have to acknowledge that, at least right now, GenAI is far below even mediocre (to put it kindly) at writing fiction. It is identifiable if not in a single sentence, then certainly in a single paragraph, especially when unedited.
However, let’s envision some distant point where these LLMs can produce coherent, competent fiction based on prompts or outlines alone. This means, in this hypothetical, they have developed sufficient abilities in plotting, narrative, description, setting, dialogue, interiority, etc. so as to be readable without too much nose scrunching—so as to be undifferentiable from mediocre human writers. This means they are not producing fantastic writing. Not great—not even good—writing. Just competent plot with consistent characters and standard prose. In such a situation, which is years (highly optimistic) or even decades (more likely) away, does such output effectively “deskill” the writing or production of successful fiction? No. I mean, think about it. Are any of the above adjectives how our friends recommend books to us? “Oh, it’s just so competent!” “You’ll never believe that all the characters are consistent and every sentence is coherent! Not a word vomit in the whole thing!” Any of this make us want to buy that book?
So, I don’t think “deskilling” is a proper lens through which to think about what GenAI is currently doing (or will do anytime soon) to the writing of fiction. However, I don’t think those who currently own or market these “tools” expect them to do this directly either. It’s apparent the strategy here is the same the techbro brigade has used for other bait and switch products. Look at social media. Study after study shows how toxic and unfulfilling and damaging it is, and this is largely because it’s a crude facsimile of the thing we were sold on it enhancing (socialization).
Similarly, GenAI art is a crude facsimile of the thing we are being sold on it enhancing (the process of making art). And what “enhancements” have we seen in people’s ability to socialize stemming from social media? Broadly speaking, the epidemic of loneliness. Just as social media is a con, a trick, a fugazi, so is GenAI “art,” and the con is, in my opinion, a long one. To work, it is leveraging the cognitive offloading and skill erosion inherent in anyone using these tools as part of their creative evolution or output.
For current aspiring writers who have been making headway in their craft and even had some level of professional success, I see some of my peers (doomed to be former peers, I’m afraid) talking about how they need to “adapt or die” and how they need to be “early adopters” because this tech “isn’t going anywhere.” Well, they’re right on that last bit, just not in the way they imagine because, when considering cognitive offloading and skill erosion, what can we expect to happen to these writers (let alone any truly new writers just beginning in their craft journey) as they rely, in any capacity, on GenAI tools? What will happen to writers who increasingly turn to these tools to describe a castle or monster, spit out a scene for them to digest (even if we acknowledge they go on to heavily edit it), or put words in their character’s mouths?
Given what we know about cognitive offloading and skill erosion, these authors will become less adept at any and every skill they’re relying on these tools to execute for them, and if they’re wholly new authors, they’ll never really develop or sharpen them to begin with. In short, those who let GenAI build their castles or design their monsters will wind up with piles of stones mismatched to their moments and déjà vu instead of boo. Those who let these tools sculpt the spines of their scenes will only lead us up mountain ranges we’ve scaled before, and those who let computers craft their characters’ conversations will fail to tell us anything we haven’t already heard. That’s what the scholarship shows for these effects in all other areas, and there’s absolutely no reason to think it’ll be different for writers or other artists.
I know people will say, “But they edit the output, so they’re still building writer skills,” but that’s not true. As someone who has edited everything from flash fiction and short stories to novels and anthologies and written them too, editing is not writing. The number of editors of major magazines or imprints who are also accomplished writers is vanishingly small because these are actually different skillsets. So, no, editing the output does not develop our writerly skills.
And that is the endgame. That is the only way GenAI companies can deskill the arts. Not by creating technology that allows for the production of high-quality pieces or artistic expression, but by reducing the cognitive abilities and craft-based skills of the would-be practitioners of these arts to the point where the quality of human art is diminished so much that their products can get close enough for the difference not to matter. They aren’t here to raise everyone’s creative floor, but to lower our creative ceilings. That’s the only way they can reach the same height as us because, as I said, there is no limit to the skill and ability human artists can achieve, but for machines, tools, and LLMs, there very much is.
The goal is to make would-be artists in some way dependent upon these tools (which won’t always be free, mind you)—to whither our abilities rather than enhancing their own—because this means that any artist that is so dependent can never be any better than the tool allows them to be. They will be constrained, where before they were boundless. And constraint is the only way they could ever supplant human creative capabilities.
So, don’t be fooled. Don’t be tricked. Don’t allow yourself or anyone else to be taken in by this fugazi—this long con. Don’t be drawn in by the allure of never having to go through the “being-bad-at-it” phase of learning a craft. We must allow ourselves to be bad—to downright suck—for a little while. Outsourcing our infinite potential to limited tools is how we become limited ourselves. Instead, pick up a pencil. Sit at a keyboard. Contend with the blank page and the empty canvas—the unsullied slab of granite. It’s the only path to greatness. Taking shortcuts, outsourcing our skill development just to avoid those awkward first efforts, is just an aspiration to mediocrity—to some middling ledge on the peakless mountain.
While these “early adopters” and “adapt or die” aspiring artists are sometimes called scabs, I don’t think that’s actually true because scabs are healing. They’re the regrowth that follows educational misadventure. All artists grow through pain—through effort and failure and lessons learned in the healing—but too many of our peers are being drawn in by the promise of avoiding this pain, along with the accompanying scabs and scars. But it’s purely bait and switch. These aspiring artists are not learning, so they’re not healing. They’re just bleeding, and they’ll never heal—never learn—so long as they keep using these tools. The wound only grows, and they’ll bleed their artistic potentials—waste their creative energies—in this liminal, constrained, finite space the owners and operators of these “tools” are creating for them. I encourage you not to join them, and to save all you can from themselves. Save your bleeding for the page, the easel, the microphone, the chisel—whatever your medium may be. Save your blood for the free-solo ascent because even if this mountain has no peak, the view grows more beautiful the higher you climb.