The biggest story of the year — the story we should all be paying attention to — is the increasing power of artificial intelligence.
Computer code can write itself, chatbots can generate academic papers, and, with a few keystrokes, a website can produce an image worthy to be framed on any wall. Everywhere we turn, AI is outputting text and images that mimic (and often surpass) humans’ abilities.
There’s so much to be concerned about in these developments, especially in the realms of plagiarism and labor replacement, with artists and writers particularly worried about their job prospects drowning in the infinite sea of AI-generated graphics and essays.
However, after taking stock of AI’s current limitations, I don’t think that artists and other creatives are in danger of extinction anytime soon. To my mind — and to many theorists, critics and media lovers — the most compelling artistic expressions contain some sort of original idea that reflects a lived human experience, and contemporary AI models lack exactly this capacity.
In short: Even if AI bots can create passable text blocks and captivating graphic designs, they cannot fabricate the sort of genuine art that speaks to our humanity.
At their most basic, current AI models are engineered to reorder information they’ve seen before. While different bots accomplish this task in slightly different ways, each system is designed to observe huge amounts of data, then find predictable ways that the data is ordered.
In the case of visual art, a bot studies which colors and shapes tend to occur near one another, while text bots identify how words and topics are organized. These programs then output a convincing painting, poem or essay that re-forms the shapes, words, colors and topics that they’ve observed within their data sets, always using some version of the organization found in their original data.
Surely, many of our daily tasks involve simple reorderings of past information, from book reports to obituaries to police sketches to legal briefs. In each, an author does some research and constructs a product based on already-existing information. We should be prepared for AI to dominate these types of recombinational tasks in the coming years.
But none of this is exactly art.
Having an original idea, expression or epiphany — having an experience that no one taught you to have — is a deeply human act and also impossible for AI — at least in the engineering behind today’s most-used bots.
While an AI-generated poem, story or painting might reorder existing information in new, believable and compelling ways, it is using preexisting building blocks drawn from its data set. This means that profoundly new, paradigm-shifting artistic choices — think: the visual juxtapositions of Frida Kahlo, William Shakespeare’s origination of new vocabulary or the jarring imagery of Toni Morrison — are nearly impossible for an AI to invent.
Instead, for AI to create anything like these artistic expressions, it must observe human creativity first.
One further simple but crucial fact separates AI bots from genuine originality: They do not experience the world around them.
To be sure, much of human knowledge is learned from reading, digesting and recombining already-existing ideas — the book reports and face sketches of our stored knowledge.
But the deepest part of human understanding comes from experiencing life around us, with our imperfect bodies scraping up against the outside world as we clumsily learn to live together. No one teaches us the enjoyment of sunlight against our skin or the despair of losing a parent or the fulfillment of growing old with your husband. These feelings arise from simply living in this world, and when we process these feelings through painting or poetry, we are germinating art that refracts components of the human experience.
When a computer writes an essay about love or a poem about loss or paints a warm, glowing sunrise, it is not creating. Instead, it is regurgitating — reporting — aspects of its data set that involve records of past love, loss and light.
There’s a huge distance between merely collecting data about an experience versus actually having that experience, and there’s a chasm between poetry that echoes and reports data and that which expresses a lived exposure to the world around us.
When a song, novel or painting truly moves me, it entangles my own experience into the story being told by that work of art. With only secondhand knowledge of the human experience, AI’s output will never actually express the existential sensation of living in the world.
Again, there are many reasons to be excited and concerned about the ascendancy of AI. But I am not worried about AI making art.
While there may be a day that AI can generate truly new ideas and can encounter the world with the viscerality and directness of a human, I see a great distance between contemporary technology and that eventuality.
At present, AI can do many things — it can make beautiful visuals, compelling essays and interesting poetry. But it cannot participate in the fleshy, awkward, complex and unique human condition.
And because it cannot experience, it cannot create something truly new and meaningful. That solidly remains the domain of us humans.
Chris White teaches music theory at University of Massachusetts at Amherst and Harvard University. His research uses big data techniques to study how we hear and write music, which is the subject of his book “The Music in the Data.” He lives in Massachusetts with his husband and son.