Ethics And Morals:
First I would like to bring to light just a few of the many ethical concerns. The databases that were used to train nearly all the AI image generation models, illegally and indiscriminately scraped the internet and collected billions of image text pairings. They even intentionally targeted top professional artists’ portfolios to put their works in the database. For example, they intentionally went out to scrape Artstation, Behance, Deviant Art, and many other artist portfolio sites as well. Many of the companies openly admitted this early on. Further, it scraped people’s private data from social media and private medical data which is a huge HIPPA violation. What’s worse, it’s been confirmed at least one dataset was found to have scraped 1,000s of images of child sex abuse and other horrific images. All these images were then used to “train” the AI image generation models. You cannot untrain an AI from what it has seen, once AI has been trained on an image it cannot be untrained. Therefore all that data lives in its training data and can influence and come out in images. It restitches things it has seen and combines them into a “new” image. While some of those training data sets are under investigation and may have been shut down that does not change or fix the problem of the AI models that have already been trained on those images.

These AI programs are all trained on real artists’ work. They took our images without our knowledge, consent, compensation, or credit. There is a huge ongoing lawsuit against some of the largest AI companies because of this. All the art and images used to “train” the AI models were illegally taken without consent, they skirted the law by developing the database saying it was for research and not for profit then, turned around and used it to train generative AI models for profit. These companies are making tons of money from images they stole from us and innocent people, it is theft and data laundering.

Abuse And Exploitation:
Both AI users and the public need to be aware of the ethical concerns and implications. People argue personal info has been stolen before AI, but that doesn’t make it ok for AI companies to do it, or for anyone to use it. Stealing private medical data is a major security breach and is a felony and a federal crime. Having the other abusive and exploitative images I mentioned in the database is also a felony and federal crime. Just because there have been issues with theft and abuse before doesn’t justify what companies training AI and AI users are doing here.

Like it or not these AI image generators and the people using them are abusing and exploiting our fellow humans. Are you ok with that or the effects it can have on the people you are exploiting? AI companies who have taken these image text pairings without consent are exploiters and abusers who will justify making a quick buck by any means necessary. No matter who it affects and how it affects them. Just because someone doesn’t know you are using their private data or you don’t know them doesn’t mean you aren’t taking advantage of them. Many people have found themselves and their data in the training databases and were appalled and horrified. People and their data are not a resource to be freely used however you see fit. Using people’s private data without consent is exploiting them and a form of abuse. By using AI image generators these companies and the users are taking advantage of others on a scale we can’t even fathom. They could be causing harm to people on a scale we can’t comprehend. By supporting AI image generation people are, enabling, and/or taking an active part in using and abusing people on an incomprehensible scale. Because that’s what these companies do they are using people and their data against their wishes. There is no way to opt out from this and the AI can’t be untrained.

These AI image generation companies and anyone who supports or uses them are taking advantage of, and harming others for their gain. While a user of these programs, may not be directly abusing them, they are knowingly using a program that is and it perpetuates the problem. It tells these companies, and others you are ok with what they are doing. Once a person learns what the issues are they are responsible for what they do with that knowledge. If you use and support these programs and companies you are now knowingly, encouraging, and participating in the abuse and exploitation of others for your gain. This unauthorized use of personal data raises serious concerns about privacy and consent. I cannot support programs or people who knowingly and willingly abuse people, how about you?

Data And Images Online Are NOT Free For Public Use:
While our images and data may be online for public view, they are not online for public use. While images may be publicly accessible online, they remain the intellectual property of the creators and are not intended for unrestricted use. Artists are not posting images online for public use unless they say otherwise. Artists post our images as portfolio images and promotional materials so we can get more work. We also post images to show our progress to generate interest and establish trust in our creation. We post images so we can sell those images. That does not give anyone the right to take those images and use them how they want. Further, I also mentioned security breaches of private data that were never supposed to be seen by the public. So for those trying to say that when they use AI they are using data available for public use, they are wrong. Public viewing is not the same as usage rights. You can see an artwork in a public gallery but you can’t take it off the wall without purchasing it, and when you do there are separate rights of using that art for commercial purposes vs the rights you have purchased which is ownership of the artwork. Buy unless an artist states otherwise you still don’t have commercial usage rights of that art.

How AI Is Trained To Generate Images:
I’ll keep this as simple as I can for time’s sake. AI can’t learn the way people do, it doesn’t “create” like we do, because it’s a machine and needs information to “create”. That information is all those stolen images, you take that away and the AI has no source to generate its images from. When being “trained” AI is given the image and text pairings that were stolen and ungoverned for safety. It then looks at the image and breaks it down into data by adding layers upon layers of noise essentially, distorting the original image. Once the original image is pretty much unrecognizable the AI then gradually reduces the noise and reconstructs that image as accurately as it can. These AI models can actually get really close to the original aside from minor distortions. That’s how it learns to “create” each object.

How AI Works And Uses Its Training Data.
AI then samples that training data, all those stolen images it saw. While the tech companies argue they don’t actually store the images. Those images are essentially broken down into basic code for the purpose of using them later. It actually uses parts of images from the database to “make new ones”. It is literally sampling from the original images and compiling them into the “new” image. You can even see butchered signatures from artists in some pieces and samples from specific works when you know what to look for. Some of these signatures were even still recognizable as specific artists’ signatures just distorted. While you rarely see that example now its because they caught on and fixed that issue but the image data is still in the AI’s training. This is a little bit of a simplification but AI image generators are essentially a super advanced collage program stitching together “new images” from the training data. it is not really creating something new because AI can’t create without the training data.

How Artists Use Reference Compared to AI.
This is very important and will address the question of how is it different when human artists use references to create. People are arguing that AI models use images for inspiration like real human artists do, which is simply not the case. There is a lot to unpack here.

First, AI users are NOT using images for reference or inspiration because they are not the ones actually choosing the images being used in the generation of any given image. They are not the ones stitching or putting the “new” image together. Even if they add their own reference to the data set they did not have any control over the images the AI model was originally ”trained on”. The AI program they are using is NOT using images as “reference or Inspiration” the same way humans do. Essentially it is mathematically sampling the images from the training data and using specific parts of the data in the image it generated. Refer to what I wrote about how AI is Trained and how it uses the training data.

For the sake of argument here we will call it all reference. While humans and AI models both use reference images, the way they use the images is very different. Humans are capable of being inspired while AI is not. Let’s break it down.

As artists, we rely on reference images to spark inspiration, to get initial ideas before the creative process begins, and sometimes during. We can use references to understand form and proportion. We can use references to understand techniques, and for problem-solving when we encounter a particularly challenging situation in the creation process. But the way humans and AI use “references” is completely different. As human artists, we bring creativity, interpretation, and ethical awareness to our use of the reference images that we intentionally hand-pick for the project. We consciously select and interpret images, making deliberate choices on how we let them influence our personal, unique artistic vision (without sampling the image) while respecting copyright laws and ethical principles.

AI however does not have human creativity and ethical understanding. It operates based on mathematical algorithms and data, it indiscriminately samples images (or its training data) without interpretation or intentionality. AI Models are trained on vast datasets that were unethically and illegally acquired. As I mentioned these databases include copyrighted images, private social media images, and medical data, as well as abusive, exploitive, and horrific images. This raises concerns about copyright infringement, privacy and consent violations, ethical use of data, and potentially harmful content. AI users have 0 control over the initial data used to train the AI programs they are using, making it impossible to understand how AI generates its images and their ethical concerns.

Some of the key differences here lie in human creativity, intentionality, and ethical awareness. While human artists make conscious and creative decisions and respect ethical principles, AI operates without understanding or consideration for ethical implications or the harm that it could cause. It simply uses the training data however necessary to give the user the desired output according to the prompts they gave it.

Going Further…
We, human artists, have understanding, creativity, interpretation, and intentionality. As real human artists, we bring our subjective and objective understanding and interpretation to reference images. Human artists make deliberate decisions about how we use reference images based on our unique vision, emotions, imagination, and experiences, which distinguishes our approach from AI. We can look at the image, be inspired by elements that speak to us, and incorporate those elements into our own unique creations without copying or sampling. As real artists, this process involves intuition, our experiences, emotions, creativity, imagination, and thought, which are inherently human qualities. While AI programs analyze reference images through mathematical algorithms and patterns, lacking the subjective interpretation, imagination, emotion, and creativity etc. that humans possess.

As human artists, our creative process involves experimentation, iteration, and personal expression. We intentionally reinterpret or abstract elements from reference images to influence our artistic vision and creation. We make original works that reflect our individual style and perspective adding imagination to it as well. While AI follows predefined algorithms and training data to generate images, lacking the ability to truly create or innovate the way humans do. AI is in fact sampling the training data and reusing it, theft is theft.

Let’s look deeper at the ethical considerations of this part as well. Human artists navigate ethical considerations such as respecting copyright laws, obtaining permission from the original creators, and picking images that are safe and legal. We also consider cultural sensitivity when using reference images. We often credit the original creator (except in specific situations which is its own can of worms). AI programs don’t have the awareness or capacity to consider such ethical concerns. The responsibility then falls on the developers and users of AI to ensure the ethical use of reference images. This means obtaining appropriate permissions, making sure the images are safe, and considering the impact of their work on the original creators and the public community. The problem here is this is not happening the developers used databases that indiscriminately scraped sites all over the Internet. Illegally obtaining copyright images, people’s private data, and abusive images. So there is no ethical way to use AI to generate images period.

The developers of Stability AI, one the biggest AI companies and image generators, were even caught saying they would not be creating an AI that would create generated music. “Because diffusion models are prone to memorization and overfitting, releasing a model trained on copyrighted data could potentially result in legal issues.” They openly admitted to the AI models sampling the training data and even reusing parts of it. They didn’t do it in the music industry because they knew they would get sued and lose because of all the big money behind that industry. But as individual artists, it’s much harder for us to go after these companies because we don’t have the financial backing of big music.

Human artists bring creativity, interpretation, and ethical considerations to our use of reference images. We are very careful and intentional about how we allow it to influence our work. AI relies on algorithms and training data, lacking the understanding and ethical awareness that human artists put into practice. I could expand further on exactly how we use reference but I feel like this is a pretty solid overview. And it’s already quite long.

In a court of law AI images cannot be copyrighted either, the courts even deem it does not have significant human input, for a person to claim it as their work.

Trust And Community Impact:
The use of AI to generate images or even “assist” in the artistic process is already banned from most reputable shows, galleries, and competitions. If real artists use AI to “assist” their process the trust in the art community could ultimately be diminished or damaged in the public eye. Even more damaging are the AI users claiming they are artists whether they are open about the fact they use AI or conceal it. Especially as the major scandals about the programs themselves come more to light. Just type Laion 5B Scandal and you will quickly see what I’m talking about (they are one of the biggest training datasets for AI).

Do you really want to support images generated by AI that was trained on data sets that have images of child sex abuse, and other illegal data in them?

The Generative AI programs themselves should not even still be legal and should be completely deleted and retrained under federal regulations and approved, legal and ethical, resources before being allowed on the market again. If this thing blows up the way it could and should, the backlash on anyone who uses or supports AI-generated images could be very bad for all of us, and negatively impact the art community as a whole.

Using AI To Generate Images In Practice.
Are AI-generated images Art, and is the user of the program the artist? Art is an expression of humanity and our creator and comes from our emotions, soul, spirit, heart, and mind. Art comes from experience, practice, and skills learned to create something from what we feel. It comes from creative problem-solving and communicating an emotion or feeling. None of these are things a machine can do. Human artists can create an image from imagination, a machine takes data from others’ images, and mathematically constructs another image based on its training data.

Art as defined by the Oxford Dictionary:
“the expression or application of human creative skill and imagination, typically in a visual form such as painting or sculpture, producing works to be appreciated primarily for their beauty or emotional power.”

The users of these AI image generation models are claiming to be “artists” but let’s look at the definition above and the facts here and how it works in practice. If you understand how a client or art director works with a real artist when requesting a custom work this makes the answer much clearer. There is no difference between what an AI user does & what a client or an art director does when requesting art from an artist.

An art director gives artists a description of what they want, and we create the art. It’s the industry-standard practice that we who actually made the artwork are the artist, not the person telling us what they want, no matter how detailed the description is. It doesn’t change just because it’s a machine instead of a human. Those are the facts about how the user or prompters’ interaction with AI works from the user end.

I know how detailed an AI user can get with prompts. But no matter how detailed they get it is exactly how a client or art director works with professional artists in the industry. The better the art director understands the vision of a project the better they can describe a piece. I’ve had art directors and clients who have written anything from a couple of sentences to long very detailed documents. I have met with art directors and clients where we specifically talk about how we want the light, subject matter, model, angle, composition, style etc.

This is what clients and art directors have done in the industry with professional artists for centuries.

Some are more detailed than others or have more revisions than others. In almost every situation, the client or art director has never had the arrogance or disrespect to call themselves the artist.

Anyone even a child can literally type a sentence or two and have AI spit out a usable image. AI users have no more control over the image than an art director who asks us to make something. They tell us what they want and it gets made. Except with AI it takes only seconds. It didn’t take them any discipline or real knowledge to learn how to create anything.

Some people are calling AI image generators a tool but consider how it works. They type a description, get an image back, and say change this, and it gets changed. That is the same way it works in my relationship with an art director. They say I want this & I make it, change this & I change it. In every case, I am still considered the artist. They are either a client or art director depending on the level of expertise and involvement in the creation of the final. This is industry standard practice.

So if using AI, in practice from the user’s end, essentially works the same way, I just described how can the user call themselves the artist. Just because it’s a machine instead of a human? If it works the same way you would work with a human it’s no longer a tool. To call it a tool when the interaction is the same as with a human artist implies that we are not artists but a tool, being used by the person giving us directions and that the directors are the artists instead. I am not ok with this and many others share this view. We the artist learned how to create something that meets the client’s, or our own vision with our unique talents. When people are using AI to generate the image no matter how detailed the “instructions”, they are a client or at most an art director & the AI is the artist (if we are going to call it art).

While I disagree with the use of AI image generation in any capacity or calling it art, especially when you consider the definition above and the practice of creating. If people are going to rationalize using it or if they are going to call it “art” then the AI program is the artist, not the user as I just explained. This is the correct and only logical way to view it. The prompter is NOT the artist of the work generated by the AI.

Ideation Vs Creation:
Some of the people using AI argue the image didn’t exist until their imagination and words “trigger” the AI, however we must realize this is how custom artworks are described by someone hiring an artist. A person can imagine and describe all they want but until it is actually made by the artist the piece does not exist, it’s just thought. This is how many of us make our living in the first place.

While essential to the creative process coming up with the idea alone does not make you the artist, unless you are the one who executed it. Real artists combine both idea, imagination, thought, and “description”, with our talents, imagination, skill, experience, expertise, emotion, and expression to make the artwork. We make the dream or thought a reality, in this case, an image. That is what makes us an artist, the ability to actually create the idea that was imagined, whether our own or someone else’s. As the artist, you must be able to conceptualize and make the creative work itself. AI users are just doing the first part which is what an art director does, they are therefore not artists, and should not pretend otherwise. The AI users don’t actually create anything, the AI model makes a guestimation about what they asked based on its mathematical algorithm and the data it was trained on, and samples that data. They might ask it to modify some things but they are still at the mercy of how the AI chooses to do it.

While the AI user retains some decision-making authority over the parameters and objectives of the image. The AI has autonomy within its capabilities to generate the images based on their instructions. Like the art director and artist example. Except, AI uses mathematical algorithms and training data to do this, without ethical consideration of the images used in the generation of a new image.

Even if someone were able to describe the piece down to the exact shape and detail. Look how a police sketch artist works. They sit down with a person. That person describes the person to be sketched, some people are better at describing than others. The more talented the artist and the more descriptive the other person is the better the outcome of the final sketch. A person describes how the other person looked and says for example “bigger nose, wider eyes, narrow chin, messy hair etc.” The person describing it for the artist is never called the artist, the one sketching is. With AI, a user’s or prompter’s interaction with it is almost the same as an art director’s or client’s relationship with an artist. It’s also very similar to my sketch artist example. Why do you think people are sharing what prompts to use to get a desired result? They didn’t say what technique did you use.

Community And Cultural Impact:
A lot of professionals have been directly impacted by the use of AI as well in the commercial industry including myself. There is an example of an illustrator who wishes to remain anonymous who reported an incident. This Illustrator was working on a book project, when they came back from their scheduled vacation they found an email letting them go. In short, the email read that they had to make cutbacks so they had to let the artist go. They then told the artist that they were going to have AI finish the project and they would be happy to send them a copy. This AI was trained on this artist’s work and used to finish the project they were working on without consent or additional compensation. Additionally, anyone in publishing knows that the illustrator fee is a small part of any book. I and many of my peers are seeking ways to continue to make a living off of the craft we spent our lives perfecting. While protecting ourselves against this technology that was designed to replace us and uses our work to do it. Other artists have reported their works being trained and then AI users use those artist names as prompts to generate fake works in their style that are then used against them. Many artists reports their names being used tens to hundreds of thousands of times as prompts in AI image generators.

This microwave technology steels our work and produces images at an incomprehensible rate that is impossible to keep up with from a physical time sake and a financial stance. Entitled people with little or no experience are using this technology to generate images as fast as they can with no appreciation, for the craft or the understanding and discipline it takes to create, just to turn a quick buck.

There are many more issues and concerns I could point out about this but these are probably the biggest ones to consider.

In the words of Dr. Ian Malcolm:
“I’ll tell you the problem with the scientific power that you’re using here: it didn’t require any discipline to attain it. … You didn’t earn the knowledge for yourselves, so you don’t take any responsibility for it. You stood on the shoulders of geniuses to accomplish something as fast as you could and before you even knew what you had you patented it and packaged it and slapped it on a plastic lunchbox, and now you’re selling it, you want to sell it!”