Can Machine Vision Replace Art Expertise?
This week, ART Walkway critic Panu Syrjämäki examines the limits of machine vision in the art world.
When we marvel at what machines can do—analyzing patterns, flagging anomalies, and even identifying forgeries—we often forget the flaws baked into their design. Machine vision isn’t neutral. It learns from us, and that means it inherits our biases, our blind spots, and even our mistakes. And in the art world, where meaning and emotion are as important as form and technique, the stakes couldn’t be higher.
These limitations—biases embedded in how machines learn—are at the core of what concerns experts like Panu Syrjämäki, Art Walkway critic and scholar in AI development, machine learning, keyword detection, and machine vision at Harvard SEAS. Guided by instructors like Laurence Moroney, Lead AI Advocate at Google, and Vijay Janapa Reddi, Associate Professor at Harvard, Syrjämäki has studied the intersection of technology and art in depth. His verdict? Machine vision is a powerful tool—but it can never fully replace the human touch.
The Flaws in Machine "Vision"
For a machine to "see," it must first be taught. This involves feeding it thousands—or millions—of images, each tagged with labels that help it recognize patterns. But this process, called training, is fraught with problems. If the dataset is incomplete or unbalanced, the machine inherits those blind spots. For example, a model trained primarily on Western art might fail to recognize techniques from indigenous or non-Western traditions. Worse, machines often miss details we take for granted.
“Imagine showing a machine only perfectly lit, whole paintings,” Syrjämäki explains. “What happens when it encounters a mirrored version or a cropped corner? It can’t connect the dots. That’s not the machine’s fault—it’s ours, for not teaching it to see beyond perfection.”
This isn’t hypothetical. Machines have flagged authentic works as forgeries because their brushstrokes didn’t match rigid patterns in the algorithm’s training. And without diverse datasets, machines are blind to entire cultural traditions. They fail to see the sacred symbols in Aboriginal Australian art or the ancestral stories woven into Native American beadwork. When machines don’t see these, they risk erasing them.
“Technology is incredible, but it’s also ruthless,” Syrjämäki warns. “It doesn’t forgive oversights. If you don’t scale an image, or teach it what happens when an arm or leg is cropped, the machine doesn’t fill in the blanks. It just stops recognizing.”
Why Machines Still Need Us
Despite these limitations, machines have proven to be invaluable in certain areas. AI has uncovered forgeries that even seasoned experts missed, traced trends across centuries, and provided insights no human could achieve alone. Tools powered by machine learning, such as those used to analyze brushstrokes or compare pigments, are remarkable aids to human expertise.
“AI can save us time and enhance our capabilities,” Syrjämäki notes. “But when it comes to art, it’s about more than just the science. It’s about the humanity. A machine might flag an anomaly, but it takes a trained expert to determine whether that anomaly is a clue to authenticity—or simply a quirk of the artist’s process.”
But there are places machines will always fail. Performance art, digital installations, and immersive experiences are about more than what’s seen—they’re about the audience, the interaction, the fleeting moment. Machines can’t capture the emotional resonance of an audience’s gasp or the electricity of a live event.
“Art isn’t just patterns and pigments,” Syrjämäki insists. “It’s emotion, intent, and connection. Machines will never fully grasp that.”
The Stakes Are High
Machine vision doesn’t just analyze art—it reflects us. And that’s the real danger. If we don’t teach machines to see the full spectrum of art, they’ll learn to ignore it. What happens when those omissions—whether indigenous traditions or experimental forms—become permanent?
“Art isn’t just about what machines can see—it’s about what we can feel,” Syrjämäki says. “If we forget that, we’ve already lost.”
The responsibility lies with us—not just experts but anyone who values art—to ensure that machines augment rather than diminish the richness of human creativity. Whether that means building more inclusive datasets, refining models to address flaws, or simply acknowledging what machines can’t do, the future of art is in our hands.
Preserving the Soul of Art
As AI continues to evolve, the challenge isn’t whether machines will shape the future of art—they already are. The question is whether we’ll let machines narrow that future into something smaller, colder, and less human. The soul of art lies in its ability to spark something within us, something machines will never feel.
“The heart of art will always beat in human hands,” Syrjämäki concludes. “The question is: Will we fight to keep it there?”
ART Walkway News