The Latent Space of Human Creativity
The Evolution of Creativity in an Era of AI Assistance
Humans are a species of growth. We are able to iterate, improve, and imagine new ways of engaging with the world around us. This ability is powered by many of our special talents: language, hand-eye-brain coordination, and maybe, most importantly, creativity. It’s the way we respond to stimuli in the world and adapt it to our needs, making possible all of the things we take for granted. While creativity is foundational to the progress of our civilization — from physics and engineering to medical science — many still think of it as a trait only inherent to “the arts.” This distinction is why many have had a visceral reaction to recent AI tools that can create text, images, and video. But what if these tools aren't the enemy of human creativity, but the greatest gift we could have ever asked for?
I’ve been obsessed with Midjourney over the last few years (43,000 images made and counting). Unlike most AI image makers, it has a model that rewards exploration and experimentation. Describing what you want to see is actually not the best way to use it. It’s more powerful as a creative canvas, where putting in different thoughts, half ideas, and quotes gives the system the freedom to find something that resonates with you, a collaboration between human intention, the quarterback, and machine interpretation, the receiver. Consider this example, a random prompt like “Where thoughts go when forgotten” provides the image below.
Midjourney also has a feature called “style reference,” which includes 4,294,967,295 unique “styles” or “aesthetics” that can be appended onto a prompt to give it a specific look and feel. Consider how "landscape” comes out with four different “srefs” below, each opening a portal into an alternate universe of artistic possibility.
Many will call this lack of precise creative articulation a cheat, a way to bypass the process of creativity. Instead, I find myself entranced by this seemingly infinite latent space, this territory that would take 680 years to fully map out (assuming 5 seconds per sref, going 24/7). I’ve built my own library of srefs to save, for use when I have an idea for my TikToks or just for fun. Many people call AI art “slop,” a dismissive label that reduces it down to garbage or something undesirable. But I think if one was willing to let their guard down, they’d find that the outputs of these programs can be beautiful no matter what your taste is. The “taste discourse” is a reaction to the fear that this AI slop will overpower us, with some claiming taste will become the only thing that can differentiate the truly creative. I like this recent piece by Aris C, who demystifies the allure of having good taste, showing how it’s more personal, a choice not to be judged by others.
And yet, the judgment of "slop" has reached a fever pitch this month, spurred by the recent Ghiblification trend through ChatGPT’s new image capabilities, where people can take any image and turn it into the style of Studio Ghibli. There’s been a lot of discussion about this on Substack, such as Erik Hoel’s piece Welcome to the semantic apocalypse. In it, he writes:
“While ChatGPT can’t pull off a perfect Miyazaki copy, it doesn’t really matter. The semantic apocalypse doesn’t require AI art to be exactly as good as the best human art. You just need to flood people with close-enough creations such that the originals feel less meaningful.”
Hoel captures the fear many people are currently expressing, the idea that this mass production of replication is diminishing something special about human creativity. Scott Alexander of Astral Codex Ten continues the conversation, viewing the trend of Ghiblification through the lens of technological progress:
“Maybe Progress repays us with interest for every medium it takes? Without mass-produced, mass-transmissible images, music, and bright colors, we couldn’t have Studio Ghibli. Dare we hope that, if anime becomes too cheap to appreciate, that very cheapness will open the door to new forms of art? But why should this always be true? If AI is better than all human artists, and you can run 100,000 inference copies at 10x serial speed in a data center, then why should anything be non-cheap ever again?”
I enjoy Alexander’s essay, and believe he reaches a hopeful conclusion, but want to interrogate that last sentence above. “Cheapness” may not be a bad thing. The idea that technology breeds new forms of creativity is one with a never-ending adjacent possible (a concept from Stuart Kauffman, which describes how the space of potential ideas is built on the foundational blocks of society’s current cultural and technological capabilities). Throughout history, each technological disruption threatened artistic traditions before ultimately expanding them. The current pessimism assumes we've reached the end of creativity, the death of the adjacent possible, when in fact, we may be passing a threshold into the greatest era of creativity since the renaissance. While the discourse fixates on Ghiblification, the creative explorers are on their own path, not replicating existing styles but playing with tools like the new Midjourney version seven. This v7 update interests me far more than the ChatGPT one. Steph Ango frames what's truly revolutionary about AI tools in a short essay from 2022, where he calls tools like Midjourney “a camera for ideas”, a tool for synthography:
“A revolutionary new kind of camera was recently invented. Instead of turning light into pictures, it turns ideas into pictures.”
“Synthography can capture moments that did not happen and moments that could never happen.”
These tools don't simply reproduce, like Ghiblification would make us believe, they help us visualize the previously unimaginable. And this transformation extends beyond images, the same implications exist for the written word. Amelia Wattenberger writes:
“I invite you to try it. Instead of using chatbots to steer toward specific answers, try using them as partners in that natural cycle of taking in, transforming, and sending out that drives creative thought. Not to get somewhere specific, but to participate in the endless exchange of ideas that moves thinking forward.”
Venkatesh Rao, who has been experimenting with AI-assisted writing under the banner, “sloptraptions,” says “I’m an AI+human centaur now” in a recent piece. I like calling myself a cyborg, but it’s the same idea. These perspectives suggest a fundamental truth: these systems aren't replacing human creativity but augmenting it, creating hybrid possibilities that neither human nor machine could achieve alone.
To understand why this distinction matters, let's consider another creative domain entirely. Humor. Years ago, I worked with sentiment analysis software that could quantify not just whether content generated positive or negative emotions, but how funny it was. By analyzing these humor scores against other content variables, we could predict what would make audiences laugh. I bring this up because humor is seen as a creative act that cannot be measured down to numbers; it’s purely subjective, right? And yet, these measurable patterns in something as ineffable as humor point to a deeper truth about creativity itself. Although creativity is an emergent phenomenon, emerging from the emergent phenomenon of consciousness itself, it is just another way of processing information. From DNA to bits in computers, information is the fundamental level of reality that we can currently understand. While placing creativity in some mystical realm — as popularized by figures like Rick Rubin in recent years — can be a powerful metaphor, it may obscure the fact that creativity is also a pattern-driven process, especially for those more mechanically or analytically inclined. It’s another mode of information processing: emergent, yes, but grounded in the same substrate that underlies DNA, humor, and code.
At its core, this debate hinges on a fundamental question: where does creativity truly reside? Creativity can happen in one’s mind, but it only becomes something truly magical once it is brought into the world, which requires the use of a tool of some form, whether that’s a pencil, brick, or computer program. The pencil doesn't diminish the artist; why should the algorithm? Storytelling as a creative art is constantly evolving: oral → handwriting → printing press → word processor → LLMs. This chain reminds us that creativity has always been a ladder to greater adjacent possibilities. Each new tool didn't diminish human creativity, it transformed and expanded it. I like this tweet from Nick St. Pierre:
And this from Lawrence Yeo in More to That:
“In the end, AI is there to assist your creative potential, not to replace it. Learn how to tell great stories and lean into the skills that emphasize your unique identity, as that’s where the best parts of your creative spirit will shine.”
Look, I get it, artificial intelligence is not something to just accept without constructive thought and debate. But the arguments dismissing it entirely are antithetical to the soul of the human spirit. We are an inherently paradoxical species, "a god who shits" as Ernest Becker says in his book The Denial of Death. Our divine nature is mechanical in nature; we build tools and use them to make better tools and art to appreciate within the context of these tools. Perhaps our most godlike quality isn't creation from nothing, but rather our ability to transform existing materials — whether pigments, sounds, or now, latent spaces, into expressions of meaning. In embracing AI as a creative partner rather than an adversary, we might discover not the end of human creativity, but its true potential.