When algorithms create, who’s the artist?
MIT scholars Dr. Ziv Epstein and Professor Justin Khoo comment on AI art
Karie Shen with DALL·E–The Tech
An unexpected song played in one of my recent Ubers. Melancholic indie pop, a woman sang about slowly growing apart from a best friend, those small betrayals and growing silences. I was genuinely moved.
Curious, I pulled out my phone to Shazam the song. That's when I saw the artist: Xania Monet. It turns out the song came from an album that was entirely generated by artificial intelligence (AI).
The quality had surprised me, especially the specificity of the lyrics about the loss of friendship, a subject pop music tends to skip over in favor of romantic heartbreak. But now, when I replay the song in my mind, the rhymes begin to nag at me. Every couplet is linked together with the precision of an algorithm following patterns rather than a human reaching for the perfectly imperfect word. How funny is it that I only noticed in hindsight?
The moment captured something essential about where society stands with regard to AI-generated art. The questions it raises are no longer theoretical exercises for philosophy seminars. Instead, they’re urgent inquiries as AI-generated content floods our feeds, our galleries, our playlists, and our understanding of what creativity means.
The human element
For Dr. Ziv Epstein, a postdoctoral associate at MIT’s Schwarzman College of Computing, the answer begins with a simple premise: “An artist is a human who creates art.” It’s a stark line in the sand, one that immediately excludes machines from the creative circle.
“Art is a dynamic dialogue between humans, expressing intent, emotion,” Epstein explained. “It is participation in a scene, responding to the zeitgeist of the moment and its corresponding anxieties and imaginaries.” In his view, art isn’t just about producing beautiful objects. It’s about making meaning within a cultural context, a conversation that requires human stakes, human fears, and human dreams.
MIT Philosophy Professor Justin Khoo approaches the definition in a similar way. He sees artists as “people who create works which are for focusing the attention of others in particular ways.” When we engage with art properly, we enter what he calls “a mode of appreciation.” This framework leaves open intriguing possibilities while maintaining focus on human intentionality.
The creativity conundrum
Can an algorithm be creative?
Khoo points to philosopher Lindsay Brainard’s criteria: creativity requires making something new, through a process without a predetermined outcome, involving “the exercise of individual agency” and "deliberate critical reflection.” Whether AI systems meet these standards remains, in his words, “an open question.”
Epstein is less ambiguous. “Creativity involves the creation of something new, but is as much about the process as the product,” he stated. It’s about negotiating with a medium, wrestling with constraints, and making countless micro-decisions. Modern AI systems, he suggested, actually work against this. They may “homogenize creative production,” threatening both the diversity of creative output and the resilience of creative labor markets.
At their core, Epstein wrote, these systems are high-dimensional probability distributions. When you prompt an AI, you’re sampling from patterns already embedded in training data. These patterns reflect particular cultural assumptions, biases, and aesthetic preferences. Ask an AI to “be creative,” and you're invoking someone else’s codified notion of what creativity looks like.
The authorship puzzle
Perhaps the thorniest question is one of attribution. When AI generates an image from your prompt, who created it?
Khoo offers two competing frameworks. In one, the AI is “a very malleable all-purpose paintbrush,” like any other an artist wields. In the other, the AI is the actual artist, and the prompter is merely a patron, someone who commissions work but doesn’t create it. “I’m not sure which is the better way to conceptualize things,” Khoo admitted, and his uncertainty feels appropriate given how rapidly this technology is evolving.
Epstein sees another possibility that is “the dearth of the author,” a phrase from Santa Clara University scholar Max Kreminski. According to him, a prompter’s intent is “underspecified” with countless creative decisions delegated to an “AI-slop machine.” The result is a kind of ghosted authorship, where training data biases and trends replicate themselves without clear creative responsibility.
He warns against anthropomorphizing these systems, as they “undercut credit to human artists and increase credit to the technologists who built it.” It’s a power dynamic worth interrogating.
Liquid art in a solid world
Epstein introduces a compelling metaphor from artist and researcher Kate Compton who believes that we’re entering an era of “liquid art.” Just as mechanical reproduction challenged the uniqueness of paintings, algorithmic reproduction makes any single AI-generated image “but a drop in the ocean.” Compton coined the term “Bach Faucet” to describe situations in which generative systems produce endless supplies of content at or above the quality of culturally valued originals, rendering rarity and traditional value obsolete.
In this framework, “the model is the message,” as stated by Isabelle Levent and Lila Shroff. Any individual AI creation is less meaningful than the underlying patterns and processes that generated it. Treating a liquid artifact as a solid piece of art, Epstein suggests, misses the point entirely.
Yet he doesn’t advocate abandoning these tools. Instead, he proposes reconceptualizing them, not as answer machines, but as “hermeneutic technologies” and “serendipity machines.” Used thoughtfully, AI can inject unexpected randomness, creating “happy accidents” that spark lateral thinking. The key is cultivating your own voice first, then using AI’s glitches and surprises as raw material for reflection.
“AI is a cliché machine,” he said. “This is very dangerous but can also be generative when treated as prompts for critical reflection.”
The uncertain future
Khoo senses a “prevailing uneasiness” around AI-generated art, stemming partly from uncertainty about what a world “awash in AI-generated art might look like.” But he also sees historical precedent. Photography, phonographs, and remixes each sparked similar anxieties before people found creative uses and developed new evaluative standards.
Both scholars agree that art students should engage with these tools. Khoo encourages his philosophy graduate students to explore AI in their workflows, noting that the technology will only become more embedded in daily life. Epstein frames it more urgently: “Now, more than ever, is your time to find your personal voice and style.”
Perhaps the real challenge that AI poses is not whether machines can create art, but whether humans will continue to cultivate the distinctive voices, critical perspectives, and intentional decision-making that make art meaningful in the first place.