Artists, you will learn to love your AI apprentice
The initial reaction to AI art from many artists has been anger and fear. This is entirely understandable, a tool has appeared that will allow anyone, even people with no artistic skill or training, to make artwork. But worse than that — this tool has been developed by harvesting the work of artists from the internet and using it as training data, so that the output from the AI art tools often mimics the style of famous artists, a style they have worked for years to hone. Typical reactions are along these lines:
“How dare they use my art as a training tool without my consent and without compensating me”.
“It’s not fair that I had to study for years to achieve this skill, and now anyone can do it with no effort”.
I understand both these reactions, but I think they are misguided. I think in time, all artists will learn to love AI art generation tools, despite their initial revulsion.
For thirteen years I have worked as an art teacher. I have taught students how to create digital art for videogames. I have devoted a significant part of my time to creating tutorials, teaching classes, helping people become artists, so it saddens me to see artists I admire react with hostility to a new technology that will enable even more people to become artists. It seems like their fear of competition is stronger than their desire to empower others. If I had that attitude I wouldn’t have taught anyone. I wouldn’t contribute to the open-source 3D printing community. I wouldn’t make free tutorials. I wouldn’t share work-in-progress videos with my followers.
Now they are quite right to complain that existing AI art tools have trained on their work, and can replicate their style without their permission. But that has always been the case, this is nothing new. It has always been possible to hire someone to replicate another artist’s style exactly, with or without their permission.
The problem here isn’t that their style can be copied, it’s that their style can be easily copied with no effort or training on the part of the user.
Many artists feel that the use of their catalog of art in order to train these AI tools entitles them to some form of compensation.
“I wholeheartedly support the ongoing protest against AI art. Why? Because my artwork is included in the datasets used to train these image generators without my consent. I get zero compensation for the use of my art, even though these image generators cost money to use, and are a commercial product.” — Loish
I can certainly empathize with this reasoning. After all, if a musician samples an existing work to create a new song, they pay a license fee and possibly royalties to the original artist whose track was sampled. But what if I sample a very small part of someone’s song, or if I make something very similar, but not identical to an existing musical work? Copyright law varies from country to country, but it frequently takes the form of “you need permission from the copyright holder if you use a ‘substantial’ part of their work” and the exact meaning of “substantial” is left to the courts to determine. I think we can apply the same to visual arts. If a “substantial” part of your work shows up in an AI generated image, then the creator of that image (the person who directed the AI tool to make it) has violated your copyright and needs your permission to use that work. If an AI generated image does not look substantially like a work you created then it doesn’t matter how it was trained, the final result doesn’t violate your copyright. You cannot copyright your style.
I know the counter-argument to this, that ‘training’ an AI isn’t the same as training a human. Loish states it herself in her Instagram post against AI art:
“Many have compared image generators to human artists seeking out inspiration. Those two are not the same. My art is literally being fed into these generators through the datasets, and spat back out of a program that has no inherent sense of what is respectful to artists.” — Loish
Fortunately her assertion that her art is literally being fed into these generators isn’t true, AI doesn’t eat artwork (yet), however the rest is true: the program has no sense of respect. But the program didn’t generate the artwork on its own. A human prompted it to, and a human decided to keep the result. That human has the ability to decide if the result of using the AI art tool is respectful.
Here is my opinion on the difference between training a human and training an AI:
There is no difference between:
1. Hiring a human who has trained to mimic the style of an existing artist, without that artist’s permission, to produce an artwork to your desired specifications…
and
2. Using an AI art tool that was trained to mimic the style of an existing artist, to produce the same artwork.
The end result is the same. There is no legal or ethical difference.
The question of copyright arises if the resulting image is substantially similar to an existing work, and in that case the person who commissioned the work (or prompted the AI art tool) must seek permission from the copyright holder in order to use the new work commercially.
Studying the old masters is a time-honored tradition, a right of passage for any budding artist. We learn from those before us by emulating them, then growing beyond duplication into finding our own style, our own voice. If it is acceptable for a human artist to learn by mimicking the work of others, then why should it be wrong to use a tool that does the same?
I think the reason many people feel that the AI training is somehow different to human training is because it isn’t fair. It allows someone to benefit from other artists’ work and to bypass years of study and effort to instantly gain an ability that should take time to acquire.
But here I think artists are underestimating themselves. Your ability do draw isn’t what makes you an artist. Give an AI art tool to an untrained person and they don’t suddenly become an artist — they will only be capable of outputting derivative, uninspired, bland content that looks like AI art.
What makes you an artist is your imagination, not your manual dexterity. The AI art tool bypasses the manual-dexterity aspect of visual art production. Instead of having to draw the picture, you need only describe it. However, using an AI art tool doesn’t instantly give you a sense of style, taste, an understanding of composition, light and contrast, line weight, balance, pathos. Your years spent studying the old masters, painting with physical media, learning Photoshop, whatever you did to reach your current ability level — those years weren’t just about training your hand, they were training your mind. Those years gave you the ability to imagine something wonderful, and then make it exist. If you suddenly lost the use of your hands, would you no longer think of yourself as an artist?
If you agree with this idea, that an artist is more than their drawing or sculpting ability, then you should start to realize that AI art tools are far more of a gift to talented artists than they are to the untrained.
Many artists have realized this already.
Daytoner is a concept artist with a clearly defined style of mechanical/robotic creations.
He downloaded the Stable Diffusion AI art tool (which is free and open source) and re-trained it. Instead of learning from other people’s work, it trained exclusively on his own mechanical designs. It then produced this output (which Daytoner painted-over to improve the quality):
His response on Instagram is very different to the negativity many artists are expressing towards this technology:
When my very inexperienced AI apprentice threw this back at me yesterday I was shocked for a moment and was about to throw my pen against the wall. At the same time I was stunned by the pure audacity that was coming at me. It was like it was saying:
“Hey man, what you do is cool and all but why are you holding back so much?” (making a super unemotional face saying it)
Holy shit!Seeing this makes me want to push myself more and I am literally wondering why my stuff is not better and even more crazy weird and out there. Thank you AI for this eye opening moment!
I can only say if you have a certain style of your own, try it out! It is scary, yes, but this one was very inspiring. It is so amplified :D It had me sitting there like I was the small human that needed to be taught a lesson.
I hope his enthusiasm will be contagious. I hope other artists will realize they can now train their own AI apprentice, for free, and it will help you push your style further. It will speed up your workflow, create reference images for you, assemble mood-boards, remix existing work. It’s not going to replace drawing. You still need a large collection of work to train the AI and get it to work in your style. And you still need skill to paint-over the flaws in the AI’s work (the fingers, it’s always the fingers).
A tool has appeared that reduces the skill level required to make art. It opens visual art to a new audience. How you react to this new tool says a lot about who you are as a person.
Is your reaction “I don’t like this, I don’t want other people to use it. It’s not fair. I should be compensated”?
Or is your reaction “This is amazing, how can I use it? How can I help other people use it? How can I make it easier for AI to learn from my previous work — where do I opt in?”
As an artist, a teacher and an advocate of the open-source movement, the second reaction is a lot more appealing to me.
In closing I want to address one additional aspect of AI art, which is the fear that it will put people out of work. If a single person can use AI tools to match the output of 5 artists, then 4 people are going to lose their jobs, right? Not necessarily. As technology improves, the number of artists working in game studios and visual effects/animation studios doesn’t decrease — instead the quality and realism of their output increases. In fact it takes a lot more artists to make a game or a movie now than it did 30 years ago. When 3D animation software appeared, did that mean all traditional 2D animators were out of work? No, for most of them their skills were directly transferable to 3D animation. Their knowledge of timing, weight, motion, anticipation, deformation — these were all more important than their illustrating skills. 3D animation software just allowed them to produce more content in a shorter time. I believe that the only artists who are in danger of losing work to AI tools are those that refuse to learn. If you’re already working in a digital art field, I think in a short time you will learn how to use, and eventually to love your AI apprentice.
