The AI Art Revolution: What is the Future of Artists?
With new tools comes an uncertain but promising future for artists.
When I first heard of AI art software, I didn’t have many expectations. Painting detailed artwork (digital or not) is a significant time investment that takes a labour of love and patience — I thought surely an AI wouldn’t come close to emulating the skill and creativity of a human artist. 5 minutes later after trying a free trial of an AI art software called Midjourney, I was left with the following result:
It was breathtaking. Not only was I blown away by the detail and quality of the image, but just how little time, money and expertise it had taken to create it. This was the equivalent of at least a full day’s work for a seasoned digital artist, and only then I realised this could be a lot more than a fun tool to play with — this would be something big. Here I’ll discuss what AI art is, how it’s generated, some pros and cons of AI art, and its implications for the future of art.
What is AI art?
In its current form, AI art is using a trained AI algorithm to generate artwork from a text prompt. You describe what you’d like the AI to make, it chugs along for a minute, then delivers you an image.
Nearly all the images used in this article are generated by an AI called Midjourney, but there are a host of other AI art programs, such as DALL.E2, DALL.E mini, Stable Diffusion and NightCafé. These are just a few of the many out there — I recommend trying some for free to get a feel for what they’re capable of first-hand.
How is AI art created?
Let’s see how the magic is done. Put simply, they scour the web for hundreds of millions of images that match the description given to it, using captions and alt text to help it ‘understand’ what it is you’re describing. It uses this to create a library of hundreds of characteristics — such as how much red-ness, round-ness and shiny-ness, things have. This builds its ‘understanding’ of what things look like, so when you ask for a group of llamas at a birthday party, it roughly understands what you mean. Finally, it takes a bunch of random pixels, and then generates a completely new image by refining the pixels over and over using its ‘understanding’ of your text prompt, until it delivers your artwork:
Vox does a better job of explaining it here in just a few minutes:
The significance of explaining this is that AI art doesn’t copy any one person’s artwork — it uses them to build up a general ‘understanding’ (or knowledge), and then generates pixels until it aligns with all of the characteristics it thinks the picture should have. This process is partially random, so you’ll never get the same image twice — every artwork is unique.
How easy and expensive is it?
If you tried one of the AI programmes mentioned above, you’d see it’s fairly easy to make some very decent art, full of detail with rich landscapes, architecture, and creativity. Once you look into it some more, there’s a lot of depth to using prompts optimally to generate even better images — this is known as ‘prompt crafting’.
Learning all the quirks of an AI and how to best communicate to it will give you noticeably better results than a novice user, so there is certainly some skill involved in the process — but it’s undeniable that even a novice can deliver remarkable results very quickly.
As for pricing, I’ve currently got Midjourney’s standard $30/month plan — but many are free or offer free trials, so casual users will have no problem finding excellent software for little or no cost at all.
Will artists become obsolete?
The biggest concern that AI art raises is how it will affect artists — especially digital artists and graphic designers — now that there is a free and simple tool that can generate excellent artwork. The gut reaction is to stipulate that artists will soon no longer be needed as AI will be able to do it quicker and cheaper without compromising on quality. However, I think what’s more likely to happen is that the job of the artist will change slightly, but there will be more jobs for artists, and more artists overall — let me explain why.
First let’s talk about how AI art lowers the barrier to entry, and why this is such a good thing. The cost and accessibility that AI art delivers is unparalleled in the creative space, which means it’s available for almost anyone, all the time. Did you love creating art as a child, but ran out of time or money to pursue your passion as an adult? Perhaps you can’t afford expensive drawing tablets and powerful computers to run drawing programs, or lost your fingers in an accident, or weren’t even born with hands. Well, now you’re able to create art regardless of your circumstance! It’s a lot of fun to see how different prompts and parameters change the final result, and you can quickly pour several hours into perfecting a few of your favourite images by generating countless variations and tweaking your prompt.
“But that’s not really art” cries an indignant voice — “that’s just asking a computer to do it for you”. It’s easy to dismiss AI art as ‘too easy’, ‘cheating’, or ‘not proper art’, but that begs the question — what is art really? If it’s aesthetically pleasing, makes you feel happy, or adds value and meaning to another’s life, why shouldn’t it be art? How much of the creative process do you have to be directly involved in, for it to be art?
Where does the boundary lie between that which is art, and that which isn’t?
With the development of AI models that can generate content which very
convincingly imitates human textual, visual or musical creations, many of our traditional, as well as contemporary, theoretical and practical understandings of art might become challenged. [Cetinic, E and She, J 2021]
There’s a lot of commotion about AI at the moment, but many digital humanities experts like Eva Cetinic and James She acknowledge that it could become the norm to have AI play an integral part in the creation of painting, music and literature as their capabilities rapidly improve.
Many think it’s unfair that the user didn’t make the subject themselves— but what of photographers, who often also don’t make the scene themselves? Are those who capture beautiful landscapes, adjusting focus, brightness, saturation, composition, and a host of other metrics to deliver stunning results not artists because they didn’t make the countryside themselves? In one podcast about AI art, someone spoke about their grandma who enjoyed taking leaf prints and framing them in glass — do we disqualify this as art because they didn’t create the subject as well? Surely not! In the same vein, just as a photographer uses their camera to produce art, a musician their instrument (especially those using digital tools), and an animator their animation software — digital artists are equally justified to use AI software like Midjourney.
In the past, portrait artists who had dedicated their lives to skillfully painting faces were in an uproar when the camera was invented, and similarly, many will want to protect the craft that they’ve worked so diligently on — and understandably so. There will be a portion of people who will be negatively impacted by AI art, but significantly more will benefit from having fewer barriers to expressing themselves creatively. When digital cameras came about, photographers with expensive dark rooms and equipment scoffed “well, anyone thinks they can be a photographer now”, and there is a tendency to do the same with AI art. And it’s true. Anyone can be an artist now, and that’s a good thing.
How will the role of artists change?
More creative expression is good and all, but our understanding of what an artist is will shift — artist and designer Rob Sheridan believes that instead of being focused on technical skills with a brush or pen, they will likely be closer to a curator of art. There will likely still be some drawing skills required, but the future of art will involve a significant process of selecting, adjusting and refining artwork by adjusting the parameters — very similar to that of the photographer.
For example, if the mood of a project needs to be darker and tenser then you can add more emphasis on certain colour palettes, change how the scene is lit, and take inspiration from artists who specialise in similar themes. Even in these early stages, you can wield a remarkable amount of control over specific parts of the image if you know the correct commands — and this will only get better with time. In future iterations, an AI assistant trained with a specific data set (Icelandic landscapes, for example) might be able to offer live improvements as you draw, filling in tedious details like grass or leaves in your own style, according to the mood you want the scene to portray.
This is very different from our current understanding of ‘artist’, but that’s what makes it so exciting. The new possibilities of being able to produce high-quality work so much easier means a team of full-time game artists can produce entire worlds for video games, rather than just a handful of explorable areas. Individuals wanting to create their own games will have all the tools they need to create immersive experiences without quitting their job or spending 3 years on the project. People will have more time to explore more ideas because it takes a fraction of the time to test if a ruined abbey, aqueduct, or altar fits your vision best. This is especially true for companies with budgets and deadlines — time is a real constraint in the number of things you can try, limiting the process of creativity.
Given the time savings and efficiency of AI art, concerns have been voiced over the number of artists that will be needed in the future, especially in the game industry. Would we not simply employ fewer people to do the same amount of work? I strongly believe that if companies implement AI into their creative process, they’ll want to train their own custom AI trained on specific datasets for specific art styles, which is a process that will require artists — and using this trained AI will require even more artists. The boundaries of industries employing artists will also be pushed — it’s unlikely (for example) that big game development companies will take the savings of employing fewer people and having a mediocre game, while others spend the same or more and outcompete them with more expansive worlds and incredibly stylistic art generation.
To illustrate further, the Techdirt podcast explains that when games first started out, very few people were devoted to developing game art — you simply couldn’t do that much with the technology. Once Computer-Generated Imagery (CGI) became mainstream, the number of people working on game art exploded as it evolved from basic 2D scenes to small 3D worlds to entire galaxies you could explore at your leisure. Similarly, AI art may increase the demand for artists, rather than diminish it, due to the sheer leap in potential that the technology will support.
Finally, it’s worth noting that while AI offers many promises traditional art is still likely to stick around for a while — after all, the digital camera didn’t render traditional artists obsolete. AI art technology is far from perfect, and many are not allowing things such as explicit prompts or explicit images due to very real problems concerning deepfakes, as well as issues of non-explicit (but still very damaging) content like fake news. Once we can generate realistic AI images — or even videos — of any scene, the extent to which we can trust media will be put into serious doubt. If you haven’t already been convinced how hard it is to distinguish AI art from human art from this article, a survey shows that 75% of the time people thought images generated by an autonomous AI called AICAN (not even prompted by humans) were made by a human artist. We have some time yet to consider how we might tackle these problems, but the window is rapidly closing as AI makes leaps and bounds in progress.
Further issues arise regarding ownership and copyright. Is it fair to generate art in a specific artist’s style, without them receiving credit for it? Considering how AI creates something completely new, and the fact that anyone can choose to emulate another artist’s style, I would argue copyright is a lesser problem. Ownership, on the other hand, is more complicated. AI art is an inherently collaborative process — there are the prompts of the user, but then also the engineers that created the code, and the authors of the entire system. No one party can reasonably have been said to create the piece of art, so how should authorship be distributed? For now, it is given to the user that used the AI software — but we’re only at the infancy of the debate.
Ethical questions and sinister consequences aside, I believe AI art offers the possibility to revolutionise art and open its doors to more creative minds. Whether or not you support it, the technology is already here, and likely here to stay. Google has their own version called ‘Imagen’ which is in the works, as well as Facebook’s own ‘Make-A-Scene’. AI art has sold for hundreds of thousands of dollars at auctions, and some industries like stock imagery have already recognised the potential of these tools and banned them from their site, which is a testament to their current and future capabilities.
I’m looking forward to how AI art helps people discover, create and express themselves like never before. The game development industry could make great use of the tech, and there’s no telling which other fields may be impacted — if there’s a way you can harness AI art generation, then there’s no better time to start learning than the present.
Sources
As well as having a personal interest in AI art as an amateur artist and philosopher, I studied various podcasts to assimilate my ideas, the most comprehensive of which is undoubtedly Techdirt’s which I highly recommend. I also assist in managing a large Instagram account (@digiartique) promoting digital artists, and have seen first-hand the widespread and rapid embracing of the technology as artists transition from traditional techniques. All other sources are linked where mentioned.
James is an undergraduate passionate about social issues, climate change, and philosophy. If you enjoyed what you read, subscribe here to be the first to be updated when he posts new content.
