Should AI-Generated Images Be Copyrightable?
The talk from the HPC-AI summer seminar series, presented by Iain Cunningham, the leader of the intellectual property and data privacy team at NVIDIA, is fantastic. His focus? The impact of generative AI on intellectual property, a topic that’s as complex as it is fascinating.
Iain is an impressive public speaker, which brings to mind James Wagstaffe, my public speaking professor who also practices law. They each have their unique style, but both command the room with ease. Iain’s presentation was not only rich in content but delivered with such fluency it was genuinely inspiring.
This talk has stayed with me, lingering in my thoughts for days. It came in handy when I was preparing a presentation for my machine learning class. While others were planning brief overviews of cutting-edge techniques—given our ten-minute time limit so that they’re unavoidable shallow, which sucks—I aimed for something unique. So, I revisited Iain’s talk, thankfully available on YouTube, and a tip of the hat to Steve Jones, the remarkable professor who organized these sessions.
At the heart of Iain’s discussion was the question: Should AI-generated art be copyrightable? He began with a look at the history of intellectual property rights, particularly the debate over photograph copyright in the 19th century—a debate as contentious then as AI copyright is today. The evolution of technology inevitably brings about new legal and ethical considerations, and AI-generated content is the latest frontier.
One of the talk’s highlights was, after Iain’s reference to the Burrow-Giles Lithographic Company v. Sarony case, the storied copyright dispute over an Oscar Wilde portrait, he explained the Supreme Court’s 1884 decision underscored that creative motivation is key to copyright claims, which quoted:
“For example, if a photography club visited the National Mall and photographed the Washington Monument, each photographer would have a separate claim in his or her individual photographs… Each photographer made individual creative choices.”
Then, Iain used this as a springboard to discuss originality and creativity, juxtaposing Supreme Court reasoning with modern issues. His next move was a stroke of genius: he showed us a series of mediocre, internet-searched photographs of the Washington Monument, made us “appreciate” the originality and creativity of these casually taken pictures that enjoys copyrights, and posed a question that elevated the discussion:
“Because, by the way, can you tell which of these photographs is generated by AI?”
It was a moment that perfectly encapsulated the essence of great public speaking. But let’s pivot back to the core issue. AI’s capabilities have grown to the point where content can be generated with minimal human effort—just a text prompt can lead to stunningly vivid outputs. The machine’s ability to create ‘original’ work blurs the lines of creativity and human involvement in the process.
Jon Stokes captures this sentiment succinctly:
“People are now using AI tools to generate high-quality, original artwork with vanishingly little ego, time, or money invested. In fact, in most cases, the total investment in any of those three things is essentially zero.”
The human element influences, but does not dictate, AI. The generative process often begins with something as nondescript as Gaussian noise, incorporating a degree of randomness that makes each output unique—even with identical prompts.
The conversation about AI and copyright law is far from settled, and I anticipate even more vigorous debates as technology progresses. We must evolve with AI, adapting our laws to regulate and encourage creativity effectively. Yet, perhaps due to AI’s unprecedented boundary-pushing, we may also need to reconsider our very definitions of intellectual property.
It’s indeed a privilege to live in such a groundbreaking era. Cool, right?