Generative AI and Art as Product
Recently, the introduction of widely accessible AI interfaces has dramatically changed how people and businesses conduct their work. Suddenly, work that took hours to do now needed only moments, requiring only the basic scaffolding to be created. Developer code, Excel formulas, or even complex equations could similarly be generated by description alone- when I needed to calculate an arc length to visualize my journey Rowing to the Azores, I was able to have a derivative calculated from an equation I didn’t even have to share with the model.
For all of the new opportunities AI has provided users, however, none have been nearly as controversial as Generative AI. The ability for models to create new content based off of a simple prompt is a fundamental shift in our world that has made many weary of what the technology can do. Many of its detractors point to it as a danger to existing creatives, whose livelihoods are imperiled by these generators- after all, why pay for an in-house designer when a machine can do the work near-instantly? This controversy recently re-ignited, as OpenAI’s latest model allowed users to create art closely resembling the style of Studio Ghibli. It was particularly controversial from an artistic perspective, as one of Studio Ghibli’s founders has strongly rebuked generative works as “an insult to life itself.”
However, despite participating in numerous discussions surrounding the quality or value of AI art, I’ve found myself frustrated at how the battlelines are drawn. Those favoring AI champion its ability to let anyone generate content with minimal effort, while artists decry it as a dangerous replacement for “real art” and lacking merit. Personally, I see the introduction of Generative AI works as a pivotal moment in Art not seen since the 19th century. But, unlike the slow growth and adoption of older technology, these new capabilities have become so widespread, so quickly, that it’s caused chaos as opposed to transition. If photography was a sea-change moment, Generative AI is a tsunami, and artists are at the beachhead. Let’s dive in.
This AI-generated image is an example of a “Studio Ghibli” ChatGPT prompt. The image, made from an engagement photo of my fiancé and I, took just moments to make.
From Modern to Modernity
To understand the situation we find ourselves in today, we first need to look to the past, starting 500 years ago. After the Renaissance reshaped the definition of “great art”, future generations of European artists were trained to recreate and replicate the style of the Old Masters, with a focus on idealism and realism. Movements emerged over the following centuries that toyed with lighting or detail, but abstraction was never given particular value or considered masterful. While there are multiple reasons that drove this, a significant factor was that paintings were one of the only ways of replicating the real world. For instance, if a person’s likeness could only be captured with paint, then highly accurate portraiture would be in high demand, as these paintings would help secure a person’s legacy.
But, with the introduction of the daguerreotype in the 1830s, there was suddenly a way to capture the real world in mere moments, and without all the years of painstaking art apprenticeship. Photographs captured a perfect likeness in a fraction of the time, and at a fraction of the labor cost. As photography grew into its own industry, it challenged artists to think of how their work could represent more than the world as it is. From this challenge emerged Impressionism, and eventually the whole of Modern Art. Today, while portraiture still exists, it is often only employed by those wanting specific painterly qualities.
So why is this relevant to Generative AI? The answer lies in the relation between artist and client. While some portraits were certainly commissioned for their quality, at the end of the day, many portraits were Art as Product, rather than Art as Expression. This distinction — that the value of some art is mostly in its function to the user, not its message from the artist — is one that is rarely discussed, but critical for understanding the moment we’re in.
Up until the mid-19th century, portraits like the painting to the right were the only way for a person’s likeness to be replicated, short of resorting to sculpture. This sort of portraiture is an example of Art as Product- once photography became more widespread, demand for portraiture declined because most portraits were commissioned to replicate a likeness, not for the portrait artist’s particular skills or artistic expression.
Art as Expression, Art as Product
One of the hardest questions to answer, when it comes to art, is the most basic of questions: “What is art”? If a banana taped to a wall can be worth $6 million, it’s hard to say anything created by human hands cannot be, at some level, seen as art. I would agree with that, and personally see art as any pursuit by a person to express how they see themselves and the world around them. This is Art as Expression. For many detractors of Generative AI works, they recognize this inherent quality of Art as Expression, describing the fever-dreams of LLMs as soulless, unnatural, and certainly not art. However, this definition only really matters from an academic, not consumer, view. And the reality is that, while artists may see their sales as purchases of their artistic talents, many consumers only really cared about the product, and not the Expression at all. If all someone wants is a banana, they’ll buy it for $1 at the corner store, instead of millions for the banana + Art as Expression price.
This is the concept of Art as Product. If consumers have a desire for something, and an artist is the only source for that thing, then consumers will support those artists out of necessity. But, when a cheaper or more practical alternative comes about, they’ll shift to it as their consumer intent was on the Product made by the artist, not the Expression given to that product. When portraits were the only way to capture a likeness, portrait artists were in demand, but the demand was much more often for the likeness, not the artist. With photography, consumers shifted their demand to the more efficient method of acquiring the Product of a likeness, and portraiture demand dwindled. And photography is not the only example of this. Pre-industrial revolution had innumerable Potters and Smiths, crafting items that every household needed. Few cared how these Products were made, though, and with mass production these industries drastically shrank to meet the much smaller demand for pottery or smith Expression.
Let’s look at a personal example of Art as Product vs. Art as Expression. I’m an avid, long-time Dungeons & Dragons player, and enjoy having visual representations of the characters that I play that are unique to me. I’ve never particularly cared much what those images looked like, but did want them to have a few recognizable features that made them distinctly mine. In short, I was largely searching for Art as Product, and in the past, I have contacted artists to produce this sort of image because I could not acquire the likeness through more efficient means. I wasn’t looking for an artist to create something expressive; if I could will an image into existence, I would be thrilled.
And with the advent of Generative AI images, I can now do just that. For a recent character of mine, I was able to put in a few specifics and get an AI-generated image that felt unique to me, and satisfied the low bar I needed to be satisfied. Were this a more significant piece I would likely still contact a digital artist, but that is because I would need their Expression and artistic talent, not a straightforward transaction of money for an art product.
This does, unfortunately, mean that some artists will lose revenue streams, primarily those whose work was largely Art as Product. When a consumer is presented with the choice of a (nearly) free AI image, or a $50+ commission, those looking for a product will go with Generative AI 90%+ of the time. Art as Product artists are the modern day portrait artists, and will have to adapt to the shifts of available imagery that AI represents. For these artists, I see two distinctly viable paths to continue their work:
Focus on Art as Expression. The one thing that AI cannot do is create truly unique, expressive work. Instead of selling productized artworks in competition with AI, artists will need to shift toward fewer, but more significant, commissions. The sort of work where AI is simply not desired, or acceptable, to the client.
Work in tandem with AI to lower costs. This path may be hard to stomach, but the combination of speed and cost savings in AI imagery makes it impossible for Art as Product artists to reasonably compete with AI head-to-head. Instead, there may be paths where an artist utilizes AI to create a jumping off point, and finish the product themselves to add specific details and qualities to the final work. As custom-trainable LLMs become more available, artists may even be able to feed AI their portfolio to ensure the art most accurately replicates the artist’s style.
In the above images, I have representations of two of my Dungeons & Dragons characters. On the left was a commissioned piece, and on the right an AI-generated image. Personally, I value this sort of representation as a product, and as such turned to AI as it was the more economical option. Were I to want a more detailed or specific piece of artwork, I’d still seek out an artist, as my goal was to acquire an expressive piece of art, not just a product.
