Like everyone else, I have been thoroughly fascinated by the improvements in artificial intelligence (AI) we have witnessed over the past few months. Last year, I used GPT-3 to write one of the articles in this column, even if only to see if the content I generated using nothing more than instructions to an algorithm could get past the watchful gaze of editors.
When ChatGPT was released late last year, this sort of technology suddenly became widely accessible through a conversational interface that anyone could use and ask for answers to just about any question—from writing songs in the style of an artist of their choice to creating functional software regardless of whether or not they have had any prior experience with coding.
This is great when refactoring code written by non-native Python developers who used a different naming convention.
Notice how ChatGPT not only gives you the updated code; it also explains the reason for the changes.
4 of 16
268
But as interesting as it has been to experience how generative AI can be used for creating text-based output, what has really blown my mind is the way in which it has transformed the visual arts. I’ve been watching with amazement—and the teensiest bit of envy—as ordinary people with no formal artistic training create truly stunning images using generative AI tools like DALL-E and Stable Diffusion. On the Midjourney Discord, every channel is an uninterrupted stream of inspirational creativity—prompted by human imagination and brought to life by AI.
Stylistic Renderings
To me, one of the more interesting features of generative AI is its ability to create images in the style of famous artists. I’ve found this to be a particularly useful hack that novices like me can use to create images with a given mood despite not being an AI-prompt ninja.
But then I got to thinking—for the AI to be able to create all this output in the styles of so many different artists, it must have been trained on hundreds of samples of their work. Who gave the permission for that? Surely, these artists must be outraged at how just about anyone can now create images in the unique style that they took an entire lifetime to make their very own.
Last month, three of those artists filed a lawsuit against OpenAI, Stability AI and Midjourney, alleging violation of the three ‘Cs’ of copyright—consent, compensation and credit. They argued that not only had the image training data used by these AI engines been scraped off the internet without the consent of the original creators, none of these artists had either been offered any compensation or been given any credit for the final output. Getty Images initiated a different legal proceeding, claiming that these AI companies “chose to ignore viable licensing options and long‑standing legal protections in pursuit of their stand‑alone commercial interests”.
These lawsuits raise interesting legal questions about how an old law should be used to regulate a new technology.
Copyright Protection
The reason copyright law was created in the first place was to offer artists the assurance that should anyone attempt to “copy” their images, the law would have their back. Surely, the spirit of this protection can be extended to images created in the unique style of an artist, particularly when the output is so realistic as to be able to pass-off as something the artist actually created.
As a matter of fact, generative AI does not copy any of the artworks it has been trained on. Every output it generates is an original work that has no one-to-one correspondence to any part or whole of any image of a given artist. And yet, the AI manages to somehow infuse into its artwork the impression that it was created in that ‘style’.
I am unable to conclude that this ‘impression’ is enough to constitute the offence of “copying” under copyright law. After all, nothing prevents a human artist from creating an artwork inspired by the style of another (so long as this ‘inspiration’ does not suggest a brush-stroke for brush-stroke reproduction of the original). This is how most young artists hone their skills—by looking to senior artists around them for inspiration, copying the techniques from one and stylistic representations of another. This is why most modern copyright laws include fair-use provisions that expressly allow artists who have sufficiently transformed an existing work of art to be immune from the claim that the resulting transformed work is guilty of infringement.
This is the balance that intellectual property law has to strike—it must find a way to protect artists from copies that erode the value of their original work, but at the same time encourage them to continue to be inspired by others, so that we all continue to benefit from the creative cross-pollination that is art’s defining characteristic.
AI Copyright Holder?
The US Copyright Office has already declared that AI generated art is not entitled to intellectual property protection as it lacks the “nexus between the human mind and creative expression”, which is necessary to invoke copyright protection.
I believe that this official viewpoint has to change in the not-too-distant future. While it is true these tools make it possible for just about anyone to create images that would have otherwise been impossible without dollops of talent and a lifetime’s worth of hard work, I have learned through personal experience just how much effort and skill it takes to fashion prompts in exactly the right way to get the desired result.
These may not be the skills that have defined artistic talent so far, but as far as I can tell, they are going to become the skills that will define artists in the years to come.
