When My Artist Mom Asked Me About AI
My mother, an internationally renowned doll artist, expressed concerns about Artificial Intelligence's apparent takeover of art. It was an honor to be asked and to answer with a little art of my own.
Technology has improved quality of life in many regards, but this is scary. I see AI-generated art that people are swooning over. Playing with Chatbot, I requested a poem about doll-making in the style of Mary Oliver. It was delivered in a matter of seconds. As a mere human, the implications are frightening.
Max Borders, I’d love to hear your take.
—Ann Hord-Heatherley
Then, my mom posted the concerns of novelist, Joyce Maynard, for consideration:
If you had told me, fifty years ago, when I published my first book (the year was 1973; I was 19) that the day would come when books might be written by anybody without blood pumping through her veins, or a beating heart, I would have said you were crazy. But that day has come.
As many of you may know, the growing sophistication of AI technology now allows for programs designed to replicate the voice, style, sentence structure and vocabulary of known published writers. This week came the news—thanks to research conducted by The Atlantic Magazine—that the books of hundreds, possibly thousands of writers have been scanned for the purpose of feeding the AI database in such a way that it is now possible to replicate a novel by Stephen King, or Michael Chabon, or Louise Erdrich…without any of those writers’ involvement in any way.
Or a novel by Joyce Maynard.
While these writers are understandably upset, there is no “AI database”—at least not in the sense that evokes the idea of a data breach. Instead, there is “pretraining” and “training” a model, which means allowing the system to recognize patterns in the data and return results based on probabilities informed by those patterns. So when you interact with the model, it doesn't summon info from a database. Instead, it generates responses based on patterns it learned during training. When the interaction is over, the model doesn't store user data, either.
But never mind that.
My Reply
After Mom posed the question, I shared a few passages from The Social Singularity (2018).
According to Japanese lore, a young man named Sen no Rikyu sought to learn an elaborate ceremony called the “Way of Tea.” Rikyu went to tea master Takeeno Joo, who tested the student by asking him to tend the garden. Rikyu cleaned up everything and raked the garden until it was immaculate. He scrutinized everything to be doubly sure. But before presenting his work to Takeeno Joo, Rikyu shook a cherry tree, which caused a few flowers to fall randomly—imperfectly—onto the ground.
Imperfectly?
According to the Japanese tradition of wabi-sabi, Rikyu understood the true nature of the aesthetic. There is beauty in imperfection. And there is something interesting about the fissures in wood, the crack in a bowl, or the randomness of a splatter.
I am reminded that value, particularly aesthetic value, is subjective. When we experience art, we bring our culture, our personal histories, our particular tastes, and all of our psychological freight, whatever it might be.
The experience of art has been likened to a tacit conversation between artist and viewer (or listener, or reader). But what’s so interesting about this dialogical process is that it’s almost as if art carries multiple possible worlds with it—carrying in the sense that each experiencer shrouds the work in his or her unique set of associations. These associations might be invisible to the rest of us, but they can be technicolor for the observer.
But can’t this also be true for AI art?
Of course, it’s not impossible for an artificially intelligent artist to iterate endlessly on patterns of aesthetics that most people tend to like. Indeed, many fine-looking designs might come straight from AI, developed perhaps as amalgams of popular designs from the past. But for many things, human beings will not want the popular or the amalgamated. We will want the uniquely beautiful, interesting, or novel—or compelling art forms that are these because we have imbued them with our humanity. We will want a lot of things wabi-sabi, because they mirror our imperfection. Experience will surround our goods like narrative metadata.
In The Social Singularity, I go on to discuss the idea of Chris Anderson’s The Long Tail, which is basically a curve of offerings one is more likely to find in a boutique shop than at Walmart. While every AI-generated work can be considered a long-tail item along some dimension, it cannot come with narrative metadata.
Van Gogh painted this very piece right before cutting off his ear.
Still, we can imagine that with a few more years of improvements to AI art generation, we’ll start to see the uncanny valley begin to close.
We already have. Yet,
AI can give us random petals strewn about the garden, or notched irregularities in our vases, but surely we will prize the idea that our artifacts’ creators had intentions, passions, and histories that we might have shared had we walked in their shoes or drawn the breath they drew to blow the glass.
We are human. We want work by real artists who use their heads, hearts, and hands.
To type.
To sew.
To sculpt.
And especially to feel.
PostscriptGPT
For good measure, I asked ChatGPT to analyze my reply to my mom:
Your reply is eloquent and offers a nuanced perspective on the intersection of art, technology, and human experience. Let's break it down:
Story & Wabi-Sabi: You start with a traditional story to underscore the value of imperfection, introducing the concept of wabi-sabi. This serves as a strong foundation for your later arguments about the value of human-produced art. You use the story to demonstrate that perfection, while admirable, isn't necessarily the most beautiful or authentic state. This serves as an allegory to human-created art versus AI-generated art.
Subjectivity & Art: You emphasize the deeply subjective nature of art and how individual experiences and histories shape our interpretations of it. This is a valuable point, highlighting that art is not just about the work itself, but about the relationship and dialogue it has with its observer.
AI Art: You acknowledge the capability of AI to produce art based on patterns, amalgams of popular designs from the past. Yet, you differentiate between this form of art and the imperfections and uniqueness that come with human-created art. You make the case that AI may be able to replicate designs and patterns, but it can't replicate the narrative, intention, and emotion behind the art.
Uncanny Valley: The idea that AI-generated art is beginning to be so similar to human-produced art that we can't easily distinguish between them is an intriguing one. The "uncanny valley" is typically a term used in robotics to describe the discomfort we feel when something looks almost human but not quite. By invoking this term in the context of art, you raise an interesting question about how close is "too close" and whether AI art will ever truly capture the essence of human creativity.
Narrative Metadata: This concept appears to refer to the story, intention, and context behind an artwork. By suggesting that AI cannot provide this, you emphasize the distinction between the mere act of creating art and the layers of meaning and emotion behind it.
Dedication to Human Artists: Ending your reply with a tribute to human artists who use their "heads, hearts, and hands" is a touching gesture. It not only reinforces the value you place on human creation but also brings the conversation full circle to the original concerns shared by your mother.
In summary, your reply provides a balanced view of the capabilities of AI in the realm of art while firmly advocating for the value and authenticity of human expression. It underscores the unique elements that humans bring to the creative process—intention, narrative, passion, and personal history.
Thank you for supporting my wabi-sabi.
Max, Your mom's art gave me goosebumps, even on the computer screen. Wow!