Blog written by Diego Gisbert Llorens
AI is here! And its uses are many: from coding, to text generation, to more creative endeavors such as composing music and creating art. Some hail it as a tool to democratize the access to art and creativity, a new medium with endless possibilities, while others denounce the lack of ethics behind it and fear for their professional future. Gloves are off in the art world, and both sides are getting ready to touch questions that are very close and personal. Let's take a look at the situation.
Only Rembrandt can paint a Rembrandt, or so Harry Donovan (Jason Patric) says in the 1997 movie Incognito. That movie is a reflection on the artistic identity of a creator, beyond technical excellence, and what makes an art piece unique and significant. Well, to be fair, it also touches on the hypocrisy and corruption of the art world, but those are topics for another day.
With the dawn of Artificial Intelligence generated images among us, and the debate raging like wildfire among digital artists, animators, illustrators and other professionals of the field, the reevaluation of what defines the value of art and artist alike seems to be dramatically appropriate. This very week, many relevant figures in the concept art and illustration world deleted their accounts on Artstation, the main hub for the industry at the moment, or left their galleries empty in protest against the platform's less than ideal response to the problematic of AI-generated “art” which, in case you haven't been following the news, involves very shady practices that the industry is currently ill-prepared to deal with.
AI art generators such as Dall-E, MidJourney and Stable Diffusion offer the chance for anyone to create an image by writing down a string of words or “prompt”, which their Generative Adversarial Networks (GAN) will translate into what they consider to be the most accurate version, based on a score applied to the immense database of images they have access to. Said databases can be expanded by users “feeding” them with pieces from desired artists, so the AI is trained to recognize the patterns upon those styles are based, and replicate them after a fashion. Stable Diffusion was trained on a core set of over 2.3 billion pairs of images and text tags scraped from the internet.
“People sometimes think these programs take existing art, put it into a bucket, stitch bits of it together, and then people sell the results as an NFT”, says Black Label Art Cult (BLAC), an AI art advocate, in regards to the very common accusation of art theft. However, the databases said generators are trained on include pieces created by many named authors and taken without their consent. Access to those billions of images was granted under special permission, under the premise (or disguise) of being destined for scientific research and exclusively for non-profit, although said generators are linked to companies which are, at the moment, generating consistent profit off of them, such as LENSA, which has made 16.2 million USD in 2022 and has not paid a single penny to the artists whose work it owes its own existence. The most utilized databased, LAION 5B, contains 5.8 billion text and image data, including copyrighted and private data, gathered without its rightful owners consent or knowledge. And, obviously, without compensation.
Artists are outraged and even afraid that this new technology could spell their doom and take over many job options, specially the less specialized, lesser paid jobs that are often the entry points to the industry and the base of many artists' livelihood but, could it be so?
San Francisco's ballet used AI generated images to promote one of their plays earlier this year. Tor Books, one of the biggest and most famous science-fiction and fantasy publishers, recently published the cover art for the upcoming Christopher Paolini's novel Fractal Noise; after some backlash, the publishing house, reputed for the high quality of its covers, issued a half-hearted and pretty vague apology on its Twitter, but will move on to publish the novel with its current cover piece.
Possible legal issues aside, the fact that companies and entities with more than enough budget to cover their artistic needs would resort to such ways to do so, presumably to cut overall costs of production, is shameful. Independent authors and small companies with close to no budget could find AI art generators to be a blessing, allowing them to generate visual material above their possibilities and perhaps empowering them to bring their ideas to life with a surprising level of quality, whereas they would have struggled or been unable to successfully publish their projects at all otherwise. Bigger companies, however, can and should afford the relative luxury of hiring skilled professionals to deliver above-average quality work tailored to the specific needs of their projects, something the AI art is currently unable to do.
But, why the outrage? Progress is inevitable, and new technologies have always led the way. Some jobs just go obsolete, and that's life. Besides, how is AI different than using, let's say, Photoshop or Blender? How is the AI using of references any different from the way artists get inspired by other artists all the time? Are stablished artists simply gatekeeping by trying to declare who is and artist and who is not? What is art and what is not?
Those above are the most commonly mentioned points on the pro-AI art camp. And they are all wrong. Let me tell you why.
The usage of a new tool, such as photography, digital painting, 3D modelling, etc., not only requires technical knowledge and skill to be used, but it also cannot enhance the creative abilities of an artist beyond their sense of aesthetics or artistic criteria, their taste, if you will.
Using references is often a must, something that I encourage my students to do as a preliminary stage of any piece. However, I also make sure to understand the references are a guideline, an inspiration, not something to rigidly stick to, in which case they hinder the process rather than enhancing it. Human decision-making process is imperfect by definition, and this results in the usage of references as very personalized filtering process in which different artists will look for, and find, immensely different bits of useful information in a particular image, due to their specific goals ,their current project, their actual level of skill, emotional and intellectual experience, and so on. We simply cannot help it, to take what we see and make it slightly ours through our own take, result of our own unique baggage of vital experiences, desires, limitations, etc.
And, above this all there's the one specific trait that sets human artists apart: intentionality. Any artist worth their salt knows how to follow the fundamental guidelines of visual language, composition, color theory, symbolism, etc. But we also know why, and when not to do it, when to bend or ignore the rules to create unexpected tension, to enhance the narrative, to provoke an emotional response and, in any case, deliver the message to the audience in the way we and we alone can do so. Or in the way most suited to the needs of a specific client or project.
This is why our job cannot be automated and will never go obsolete. Because the value of what we do goes beyond technical skill, flashy styles or conventional visual appeal: it is based on connecting through the share experience of existing as a human being.
Ok, this all sounds very pretty, but let's get down to the business side of things again. The reaction of the comic industry has been clear, with top-level artists like Dave Rapoza and Lois van Baarle expressing their rejection and concerns against AI art for many of the reasons listed above, whereas Jon Moisan (Acquisition Editor at BOOM Studios!) wrote, “If you submit AI art to me in an attempt to get work and I find out, I'll do everything in my power to make sure you're blackballed from the comics industry. There's no room for frauds in this industry.”, opinion seconded by Heather Antos, Senior Editor at IDW. Conventions such as New Zealand's pop culture Armageddon Expo have announced a total ban on AI art in their events, whereas Guillermo del Toro called AI art “an insult to life itself”.
Now, let's say a videogames studio decides to use AI to generate the bulk of its visual output in record time, and close to zero cost. The art itself, freshly out of the generator, is almost unusable and would require human supervision and post-production in order to suit the specifics of the project, which could be done by a small team of experienced artists. Now, this feast would eventually lead to famine, as the one quality necessary to make any proper use of that AI generated material is taste earned through hands-on experience would be largely out of reach for the next generation of artists trying to break through. Entry-level jobs could easily be replaced, whereas high level jobs would remain mostly untouched, destroying the artistic ecosystem. That would mean sacrificing a whole generation of potential talent for a few years of hypothetical short-term gains.
Is this scenario possible? Yes, I am sure it is already happening. It is nonsensical and short-sighted? Also yes.
As we speak, initiatives launched in the US and Europe by artists' associations aim to set a legal framework that limits the usage of AI art while respecting and protecting the rights of human creators and their work. In the meantime, many people feed pictures of their own faces to AI generator apps to create funny profile pictures, not aware of the metadata said files include and which could jeopardize their privacy once fed to the algorithm.
To summarize, at the moment the usage of AI generated art is unethical and implies the risks of potential plagiarism and copyright infringement. Most companies would not risk using it until a safe legal framework is in place, neither will they hire someone based on their prompt-typing skills and AI-based portfolio. As a learning tool, the AI art generation provides little value, since all useful learning as an artist happens during the creation process.
This leads me to the last point, that I purposefully omitted until now: is anti AI art simply another form of gatekeeping? No, it certainly is not. It is a warning to all those young and aspiring creative people who seek a shortcut where there is none. Regardless of your field, be it design, concept art, music or gallery painting, the joy and the beauty of creation is in the process, and it is the only way to truly learn the skills and develop your taste, but also to find the questions that will lead you to what you truly have to say as an artists.
As Picasso said, “computers are just boring. The only thing they provide is answers.”
Take a look at these links for more information on this topic, as well as another Gofundme campaign to help protect artists:
Gofundme: Protecting artists from AI generated content
KOrtizblog: Why AI models are not inspired like humans
Crew Machine: Pros and cons of AI generated content
BBC News: The rise of AI artists
NFT Now: Fear vs ethics: Where AI art critics go wrong