The Work of Art in the Age of Generative AI
A pragmatic view into the reality of AI-generated art and how it will disrupt creative industries — and what artists &creators need to…
We are on the threshold of the most important change in productivity since harnessing fire allowed our hominid ancestors to eat better, spend a lot less time chewing, and gather enough nutrients to evolve bigger brains.
Now, artificial intelligence is augmenting the power of our brains, and it seems that no field of human endeavor will be unaffected.
Transportation and warehousing jobs will largely be replaced by robots and autonomous vehicles. Fast food preparation, janitorial and many manufacturing jobs will follow. In an economy where, in December 2022, there are still nearly two job openings for every unemployed person — it seems we need robots to meet consumer demand and do the work that people won’t do.
All forms of labor will ultimately be a function of electricity, and electricity is only getting more abundant. Renewables are already cheaper than fossil fuels, and getting cheaper; and in a decade or two we might even harness nuclear fusion, the power at the heart of stars, to generate virtually unlimited energy.
Labor will not be limited to physical labor, but intellectual labor as well. While most people were thinking about the blue-collar work that would be replaced by AI, they overlooked an entire sector: our creative industries.
If your work is the manipulation of symbols, text and images — then AI is coming for you sooner than many of these physical jobs will be replaced. That’s because these jobs don’t require any special hardware — it runs on the computers we already have.
Nowhere is this change producing greater anxiety than the world of art.
My daughter is a budding artist. And I’ve worked alongside many artists — who I consider to be friends, amazing colleagues and esteemed professionals. So everything I write here is written with love in my heart. Some of what I will say will be alarming, some of it may even appear insensitive.
My goal here is to lay out the forces that are at work, so that we can grapple with reality as it is rather than as we wish it might be.
What is Art?
Art, the kind that taps into unique human perspectives on the human condition, plays with new forms of media, reflects the lived experience of emotion and physical being — all of this art will continue. I’ll leave it to academics to define what art really is; for me, it’s enough to know it when I see it:
I know that Flower and Journey are art.
I know that Las Meninas is art.
I know that Spirited Away is art.
I know that Don Draper’s pitch for the Carousel is art.
You might have a different definition of art than me, which is fine — and you may consider your own creations to be art, which is wonderful. I’m not interested in being a gatekeeper.
Everyone who wishes to create art, ought to have the opportunity to do.
Art versus Commercial Graphics
But most “art” is not art. A huge quantity of professional work is not art of the type I described above. It is commercial graphics production done as part of a business transaction. In that sense, it doesn’t belong to a special class that is different than the work various robots are about to replace. This isn’t to say that people don’t enjoy their work, or value it highly (having invested enormous amounts of time into honing their craft) — or to imply that “true art” is intrinsically more important than commercial graphics.
Commercial graphics is a form of work within our system of capitalism.
And since we’re on the subject of capitalism, let’s talk about about intellectual property. Part of the backlash against AI-generated artwork is the objection that it exploits intellectual property belonging to artists. The argument is that since models are trained on copyrighted works, that artists are being ripped off.
This is a complicated issue, but let me summarize some of the problems with it:
The AI models are pattern-recognizers, not copyists. The models observe patterns (brushstrokes, curvature, edges, organizations) across many works, and create a statistical model that render these similarities. In that sense, it is vaguely similar to what any human artist does when learning art: they observe works, learn the methods behind them, and internalize the practice.
There have been cases where the diffusion models have regenerated some works — perhaps an overfitting problem (one that appears worse in small datasets). But even in those cases, which appear uncommon — this is a bug and not an intended feature of the technology. I suspect improvements will address these issues, or entirely different approaches will pull ahead that yield less problematic results (GANs do not have this issue).
But let’s just assume that there are some copyright problems that will get worked out over time, and people who prefer to opt-out of having models trained on their work will have the ability to do so:
The technologies will not be stopped even if training is curtailed. Whether it is legal to train these models from copyrighted artwork is one that will no-doubt be tested by litigators; and if those attempts fail in the courts, then perhaps it will result in new laws and regulations.
But I expect these cases and rules to fail to prevent the industrialized use of generative AI. There are a number of reasons: the ample corpus of artwork by creators who are no longer alive and out of copyright; there’s plenty of commercial graphics owned by companies (not artists) who will be happy to license it; and the companies building generative technologies could also hire artists to produce content where gaps remain in the training, and train from those instead.
These technologies will not be stopped, and they will not be canceled — no more than you can stop the efficiencies gained by artists doing paint-overs on photo reference, or from applying digital tools in Photoshop.
For most of recent history, creating commercial graphics has required a deep commitment to craft: many years spent honing skills with illustration, learning the tools, practicing, and expression across a wide range of media.
Nothing will stop you from continuing to learn these crafts, just as one can still build furniture entirely from hand tools.
It is a uniquely human experience to make something with your own hands and your own mind, and everyone ought to have the opportunity to do so — and if some of the rosier predictions of our future economy of abundance are realized, perhaps everyone will!
For some artisans, applying craft skills in industry will continue to be valuable: because humans will continue to explore unique visions of art and creativity; or perhaps because the aura associated with human-crafted artifacts will become more valuable as machine-generated versions become more abundant.
In other cases, the “value” will be the experience of learning and indulging in a craft — not the commercial transactions it can produce.
And that’s what’s at the heart of the anxiety pulsing through the world of art right now. For the vast majority of working artists, it isn’t really about copyrights and intellectual property (those are mostly owned by companies or are moving out of copyright). It is about having a craft that one truly loves, have made a massive investment in, and want to continue doing. And you want to be paid for it. It’s an ikigai.
Composition > Production
We are moving to an economy which will favor composition over production.
If you want to pursue art (without the scare-quotes) as I’ve used this term above: you can, and I doubt that any generative AI will prevent you. That’s a form of production that seems like it will be safe for quite some time, even if it is limited to a very small number of producers who can make a living at it (as has always been the case with art).
If you’re one of these producers, you may still wish to master the new tools at your disposal. You may iterate, riff, or explore new directions faster than ever before. You’ll paint with ideas instead of pixels.
If you do, then it is the skills of composition that will be important. Here’s a deck I compiled to capture some of what you need to know:
Technology is Accelerating
In the industry I’ve spent most of my career — computer games — much artwork started to move offshore years ago. Artists who didn’t master new technologies, learn to work inside 3D engines, or more technical aspects of digital art couldn’t compete with the costs of a production house in a far-away land.
Today, we’re seeing the natural progression of this movement: if this work can be off-shored, and it can be automated even more capital-efficiently — then it will. Those who master the new skills will survive. That means learning the aspects that are more technical, more complex, and the types of composition I wrote about above.
For those of you who feel these technologies are threatening or worrisome: unless you’re convinced that generative technologies will be regulated out of existence (something I find highly unlikely), then you owe it to yourself to master the compositional skills that will allow you to stay relevant in the marketplace of skills.
For others reading this, you’ll find it very exciting. Maybe you have an idea to incorporate generative AI into your production process so you can disrupt much larger competitors who will be too slow to change their methods. Or even better: you’ll build whole new products where generative AI is at the core of the experience itself, bringing us whole new ways of living in the world.
Or you’ll ride on the counter-trend: you’ll be the artisan selling handmade human products in a world of automation. There’s a market for it in ceramics, in oil paintings, in furniture, in jewelry, in food — so why not in digital products like games and online experiences?
Whichever path you’re on, I know the future seems intimidating because these technologies are moving faster. The rate of change is not linear; it is compounding. It is happening: not only for artists, but for every kind of creator. The opportunity to scale-up our creativity exponentially is before us.
The reference to “aura” being associated with handmade art — and the title of this essay — owe a debt to The Work of Art in the Age of Mechanical Reproduction by Walter Benjamin.
An example of applying compositional skills to generative technology appeared in my post on LinkedIn about creating a massively multiplayer game from a text prompt.
Some of my foundational thinking is contained in Composability is the Most Powerful Creative Force in the Universe.
I wrote about the exponentially-accelerating technology of AI in Artificial Intelligence and the Metaverse.