AI and the Question of Authorship
When a machine generates an image from a text prompt, who made it? The question is not new — photography asked it first — but the stakes feel different now.
Photography's arrival in 1839 prompted a crisis of authorship that took decades to resolve. Was the photographer an artist or a technician? Was the camera a tool or an author? The questions seem quaint now — we have no difficulty accepting Cartier-Bresson or Cindy Sherman as artists — but they were genuinely contested at the time.
Generative AI raises the same questions with greater urgency, because the gap between human intention and machine execution is wider. When I type a prompt and receive an image, what exactly did I make? I made a description. The model — trained on billions of images made by human artists — made the picture.
The copyright question is currently being litigated in several jurisdictions, with no settled answer. But the artistic question is perhaps more interesting: does it matter? If the resulting image is beautiful, moving, or meaningful, does the mechanism of its production change its value?
I think it does matter, but not in the way the most anxious critics suggest. The issue is not that AI-generated images are inherently lesser — some are extraordinary. The issue is transparency: viewers deserve to know what they are looking at, and artists deserve credit for the work that trained the models.
The more interesting question, to me, is what AI cannot do. It cannot have a body that moves through the world and accumulates experience. It cannot be surprised by its own work in the way a painter is surprised when a passage comes together unexpectedly. It cannot fail in the productive way that failure teaches. Those gaps are where human art-making still lives.