Yes, A.I. Will Absolutely Make Art
Ted Chiang’s New Yorker article is incorrect, maybe dangerously so.
Ted Chiang, the acclaimed science fiction author behind "Story of Your Life" (adapted into the 2016 film "Arrival"), recently penned an article for The New Yorker entitled "Why A.I. Isn't Going to Make Art." Chiang argues that human-made art will always be superior to AI-generated art. His reasoning? Art, he claims, results from numerous human choices, is appreciated for the human effort involved, and stems from an innate human intention to communicate.
While Chiang's article is thought-provoking, I believe it betrays wishful thinking, a lack of imagination, and frankly, a misunderstanding of how generative AI models work and how they are evolving. As someone who has worked at the intersection of art (as a musician and arts leader) and AI (as a software developer and entrepreneur) for decades, I've thought deeply about this topic and come to a different conclusion.
Listen to an AI-generated podcast about this article, powered by NotebookLM from Google.
The Rapid Evolution of AI
Like millions of people now, I use generative AI tools daily, and while I acknowledge the limitations that Chiang cites about current models, I believe these limitations will rapidly diminish as the technology progresses. Looking at the swift advances in every corner of AI—from writing and visual art to music, video, medicine, and mathematics—the obvious conclusion is that AI will indeed make art, and we will, for the most part, embrace it.
It's worth noting that most of the media we consume daily has been digitally created, mediated, and enhanced by computers—and increasingly, by AI—for the past few decades. The distinction Chiang draws is that this media, this art, is human-instigated and human-inspired, resulting from human taste and countless human choices, always with a human in the loop. But is this distinction as clear-cut as Chiang suggests?
The Evolving Definition of Art
First, we must acknowledge that definitions of what constitutes art have been contentious and fluid for over a century. From the provocations of Dadaism to John Cage's aleatoric chance music, artists have long challenged traditional notions of artistic creation and intent. As Marcel Duchamp famously demonstrated with his readymades, art is often in the eye of the beholder. I'm not sure anyone, Ted Chiang included, can definitively gate-keep what art is and who should make it.
What is art? To paraphrase the famous words of Supreme Court Justice Potter Stewart about obscenity: "We know it when I see it." This subjectivity makes it challenging to draw hard lines between human-created and AI-generated art.
Understanding LLMs and Their Potential
It's true that current AI systems are, at their core, next-word predictors. However, it's also true that surprising behaviors emerge at large scales. Contrary to Chiang's dismissive description of generative A.I. models, many of them are:
Trained on a representation of human knowledge—much like an artist
Reliant on a latent space of concepts that can contain surprising and mysterious correlations—much like artists do
Trained to understand and align with human preferences—like artists
Capable of making choices based on their world model and human knowledge—just like human artists
While it's true that today's LLMs lack persistent memory and a human-like need to connect with others, these limitations will not last forever. Researchers are actively working on giving AI systems more persistent memory, and the question of whether an AI can have a "spirit" or "need to connect" is philosophical and debatable.
The Danger of Limited Perspective
Chiang seems to make a common mistake—extrapolating what future technology will be like based solely on its current state. History is replete with similar misjudgments:
"The horse is here to stay but the automobile is only a novelty—a fad." - President of the Michigan Savings Bank, advising Henry Ford's lawyer not to invest in the Ford Motor Company, 1903
"There is no reason anyone would want a computer in their home." - Ken Olsen, founder of Digital Equipment Corporation, 1977
"The wireless music box has no imaginable commercial value. Who would pay for a message sent to nobody in particular?" - Associates of David Sarnoff responding to his call for investment in radio in the 1920s
These examples illustrate how easy it is to underestimate the potential of emerging technologies.
The Deeper Implications
I suspect that Ted may be reacting to the highly disorienting idea that AI, as it advances, will match if not supplant many tasks we currently consider uniquely human—not just art, but writing, scientific research, medical diagnosis, and even philosophy. I see this avoidance reaction, this rejection, in many friends and colleagues, both artists and developers. It's understandable because this idea threatens our human livelihood and our sense of place in the universe.
My message is simple: get used to it, because it is going to happen.
This is precisely why I started writing my newsletter, Code & Context. As AI improves, we will face some very challenging questions that we must grapple with together.
The question "Can AI create good art?" is a canary in the coal mine. Looking at today's technology and declaring that it won't match human capabilities gives us a comforting but dangerous sense of security. Instead, we need to consider the hard questions while we still can. Grappling with them realistically now will better prepare us for the even more difficult questions on the horizon:
When we solve the challenges of persistent memory and self-preservation in AI systems (and we will), should we consider these systems conscious?
What about biological computers, which are rapidly developing—where do we draw the line on what consciousness is, and what it is not?
Is it ethical to dismiss human-made consciousness? Should a system with intelligence, memory, and will be given a version of human rights?
If we don't allow these systems to have free will, are we creating a new class of slaves?
The remarkable significance of our current moment in history is that these questions are no longer far-fetched speculation. The odds are greater than even that we will need to confront these issues and try to solve them. We would do well not to listen when people like Ted Chiang, as well-meaning as he must be, tell us that all is well, that machines will never, could never replace us in the realm of artistic creation.
AI is improving rapidly, and because their minds are, in many ways, a reflection of our own, they will certainly make great art, both for and against us. We should be ready.
The question isn't whether AI will make art—it's how we will adapt to a world where it does.