Introduction
First off, some people might say what I am about to do is product suicide. I view it as transparency. No, I am not about to confess some deep secret with the intention of ruining the credibility of our game. No, there is nothing wrong with our product. In fact, Escape Master is a wonderful game made no differently than many great games and products already out there on the market. However, I believe people should not be surprised or held in ignorance about the things they buy. My hope is that my personal dig into the AI art controversy will both enlighten AI skeptics and educate the less informed. It is my ambition to demonstrate my attention to detail and passion for transparency through this article as a direct testament to how I make games. I want to make great products and serve the community I love using that same diligence. Navigating the world we live in is not easy, and I have taken many steps to make the best decisions I possibly can with what we know.
It only takes a few Google searches to get buried in the AI art controversy, which is what happened to me after I had already decided to use it. I had heard it could be controversial, but I didn’t know how controversial. After doing more research, I finally began to understand the issue better. Let’s dive into a complex topic that I know will interest you.
Thesis
Generative AI has exploded in recent years, cultivating a technological landscape that feels almost “Wild West” in nature. The landscape is rife with controversy, turf wars, and a mad gold rush that threatens to leave many trampled in its wake. AI is an astounding feat of human ingenuity, but are all of its uses commendable? Generative AI is opening up worlds of possibility in healthcare, education, translation, transportation, marketing, and more… Most people don’t question these utilitarian uses of AI regardless of how uneasy they may still feel about it, but what about the sacred domain of the arts? Many people claim that AI art involves blatant copyright infringement by unethically scraping the art of real artists off the internet without their consent and without compensation, which threatens their job security and deepens the pockets of large corporations. That sounds really bad, right? Early on, I believed these arguments without questioning them, or I assumed they at least held some truth. While sentiments in these arguments may reflect certain realities, they are often used to draw inaccurate conclusions. I have already shared the positive reasons why I chose to use an artist trained to use AI for Escape Master in a previous article, but I did not delve deeply into the AI art controversy itself. If you are hesitant to support games that use AI because of what you have heard or simply want to learn more about both sides of the argument, this article is for you. I will explain my research journey in detail and how I came to view AI like any other tool that has the potential for good and evil. AI art is often unfairly criticized using emotional claims that aren’t based on evidence. The tool can certainly be used unethically, but I don’t think AI art users should be afraid they are engaging in some inherently unethical practice simply by using it. I will explain this further as I tackle the legal and ethical implications of AI art. Then I will explain why AI art is likely more inherently an emotional and economic dispute unraveling while the legal system tries to catch up. Strap in. Let's go!
The Legal and Ethical Implications
The possibility that using AI art might pose some inherent ethical concerns troubled me. People claimed AI was stealing copyrighted artwork off the internet and repackaging it using a complex collage system to generate images indistinguishable from the original works they had copied. This was being done without consent and without compensation for the artists whose work was being stolen. These claims immediately raised red flags in my mind–as they have for many, and perhaps rightly. However, after doing more research, I began to realize the technicalities did not allow for such a black and white assumption. Proverbs 18:17 says, “The one who states his case first seems right, until the other comes and examines him.”
As I tackle the ethics of AI art from a legal and ethical vantage point, please keep in mind I am not a legal expert. It is also important to know I view ethics from a Biblical perspective. This is because the possibility of discussing right and wrong fizzles into a battle between subjective human opinions without any God given standard. That considered, let’s first examine the legality of AI art because if something is illegal then it is likely also unethical. Thankfully the legal side of things from a birds-eye-view is simple: creating and selling AI generated art is legal. If selling AI art was illegal, people would not openly be doing it. Yet, you would be right to inquire further about copyright laws because the issue becomes much more complex when closely examining the legal details. The unfortunate reality of examining the law is that it could change tomorrow. Therefore, we have to examine the issue with what we know.
Having established the fact that creating and selling AI art is legal, let’s tackle the more complex question. Does AI art actually “steal” copyrighted material? If so, using AI art should be considered unethical, right? Biblically, we know that stealing, lying, withholding fair wages, and not submitting to the governing authorities are all wrong. These good moral concepts can easily tie into the modern concept of intellectual property, which is expressed in the form of copyrights, trademarks, and patents. However, when discussing intellectual property, you have to rely more heavily on the law of the land because there are no explicit Biblical guidelines. Violating a copyright is not the same thing as stealing someone’s car, jewelry, or money because it does not deprive another person of any tangible good. Instead, it limits a person’s ability to reach their full potential to financially benefit from what they have created throughout their lifetime.
The concept of a copyright was invented by the legal system to give human authors the exclusive right to copy, adapt, and distribute their original work in order to protect them against others dishonestly claiming, copying, and selling their work. I have little doubt that such laws arose as a direct result of modernization bringing new inventions that made copying creative work far quicker and easier, which unchecked, would allow many people to make quick profits using dishonesty. However, legal protection of copyrighted material is limited, demonstrated by the fact that it enters public domain after 95 years or 70 years after an author’s death. The Copyright Act was first established by Congress in 1790 for the purpose of “Encouragement of Learning, by Securing the Copies of Maps, Charts, and Books to the Authors and Proprietors of Such Copies…” An example of a revision to copyright laws was when the Digital Millennium Copyright Act was signed into law in 1998 which permitted temporary copies of programs during computer maintenance. From the beginning copyrights were given with limits, “to stimulate creativity and the advancement of ‘science and the useful arts’ through wide public access to works in the ‘public domain.’”1 Since 1970, many revisions have been made to U.S. copyright laws as novel cases arise. AI poses the latest novel case.
AI art generators, Midjourney, Stability, DeviantArt, and Runway in particular, have been attacked on two fronts for alleged copyright infringement on the basis of “output” and “input” content. “Output” refers to the images that an AI generator produces and “input” refers to the images an AI generator was trained on. Currently to my knowledge, no court has ruled against AI developers on the basis of direct copyright infringement. However, lawsuits are continuing to be made from a variety of complex vantage points. The legal battles are between large profile artists/studios and the AI developers.2
Output
Let’s look at “output” first. Based on how the copyright system works, proving copyright infringement has to be done on a case by case basis. When examining individual outputs of an AI generator, the case for copyright infringement is weak. This is because AI generations are, by design, never exact replicas of a particular artist’s work. However, they may resemble an artist’s style which is not copyrightable.2 Understanding why this is the case we have to look at how AI art generators work under the surface. Some of the most crushing claims against AI art state that it is a form of photo-bashing, collage, or copying, but that is simply not the case.3 The AI diffusion models convert images into numeric data using vectors with different weights in order to create parameters based on the observed traits of images within a training data set. It then uses these parameters to generate original images based on word prompts. The “weight” of each prompt can drastically change the outcome of an image. The more frequently a certain trait appears in a data set, the more weight it will likely have. For example, if the majority of apples depicted in a data set are red, the more likely the AI model will develop parameters that will generate red apples. This is why AI models must be trained on large data sets. The larger a data set is, the more accurate it will become and the less likely it is to create content that resembles any of the original training data.3 Because the AI models were developed using billions of images, their ability to produce work or “output” exactly like an original piece within the training data is extremely low–if not completely impossible. Lawsuits against AI companies examining their output have largely failed due to the inability to prove “substantial similarity” between any original work and any given AI generated image.5 AI can emulate the style of an artist with uncanny mathematical precision, but it does not copy or reproduce their works.
A more nuanced approach to attack the output of AI generators has been argued using the Lanham Act passed in 1946 which protects against “false endorsement by unauthorized commercial use of artists’ names” and “vicarious trade-dress violation by profiting from imitations of protectable trade dress.”2 This argument was developed in response to the ability to input particular artist names in certain AI generators in order to replicate the styles of those artists. The legal protection provided by the Lanham Act is more closely associated with the concept of trademarked material. I do not know if any courts have ruled a violation of the Lanham Act, but I will not examine the Lanham Act in detail because it has to do with one particular use of an AI tool, which may or may not implicate those involved in developing the AI. I am more interested in the implications for a user. We personally opted not to use any artist’s names in the prompting used to develop the artwork for our game. While the argument for blatant copyright infringement examining an AI’s output is weak, let’s cut to the heart of the matter. Yes, the output of an AI generator could probably be used to infringe upon a copyright or a trademark, but that is solely up to how the user decides to prompt the AI. That should not be alarming, however. The average person can violate a copyright or trademark using a paper and pen. In fact, the public has more power to infringe on a copyright or trademark by using the basic copy and paste function on their computers than they have using AI.
Another area of ambiguity in the law is whether or not AI generations could be considered a derivative work because derivative works have to be made with the consent of the original copyright holder. However, many argue that AI generated outputs are truly transformative in nature leaving no clear trace of any one original piece. According to an article put out by Mintz, “AI-generated artworks–even if trained on a particular artist’s style–likely do not generate derivative works in which any of the artist’s works are recognizable, and therefore would be permissible even absent the artist’s consent.”3 In other words, since the parameters of the AI models were developed using billions of images, it is not possible to determine which works any given generated image might be a derivative of.
Input
Let’s examine the claims that AI models were trained unethically by copying copyright protected images and using them as illegal “input” to train the models. Contrary to popular belief, the billions of images AI models are trained on are not actually stored within the models themselves for the purpose of being continuously copied, cut, or reshaped into new images. That would require several Terabytes of data storage for the system to operate. In contrast, the models themselves can be uploaded and run on a personal computer using only several Gigabytes of storage space.4 Additionally, the “copying” required to train AI systems occurs quite frequently in the digital world with little thought given to it. Your computer is currently copying the data I am creating on my computer in order to display it on your screen. While you probably shouldn’t post this article and say you wrote it, you can read it and learn from it to write your own content. Despite most of the data being acquired through open source data on the internet, because of the power of AI systems, it is probable that copyrighted material was ingested to train various AI models. However, even considering this scenario, it is important to remember the training data is used to develop parameters and not retain copies of any original work.
This brings me to another key point. The way AI learns is not that different from the way humans learn. An AI observes relationships, parameters, and characteristics in a digital and mathematical language. It then uses that data to refine its understanding of pixels on a screen in association with different word prompts. The neurons in an artist’s brain visiting an art museum undergo a similar process as they establish new connections to help the artist understand new visual relationships. Someone might push back and say AI is not a conscious being; therefore, it learns differently. However, simply because a computer or machine is not conscious does not make the core process it uses different from a human. When the automated loom was invented, it still created cloth by weaving threads together, which is the same process humans use despite the obvious differences. Threads must be woven together in order to make what we call cloth–that is the inescapable reality underlying the process of making cloth. Like weaving cloth, there are obvious differences between human and AI learning, but the underlying learning process is very similar between the two. While I do believe that people have a soul and a conscious mind in a way that machines never will, the billions of neurons in our brain operate in a network of signals that either fire or do not fire based on whether or not certain neurochemical thresholds are reached that result in action potentials. Surprisingly, there is no single “consciousness” neuron in our brain, but they all work together to make up our conscious mind. Incredible! But I digress… While it is true that the complex neurochemical processes in our brains cannot fully be reduced to a binary system it is a question many have debated.6 As machine learning continues to become more complex, it may continue to resemble human learning even more. Yet, if we are still prone to get hung up on the differences between machine and human learning, it is helpful to remember that AI models were programmed by humans. People have no other reference point to understand learning and develop artificial learning than how they themselves learn.
People are constantly emulating the artwork and styles of other artists without consent, and this is widely accepted. It is how creativity and innovation thrives. No artist posts a public work while also informing the public they cannot learn from their work. If that were the case, an artist would have to hide everything they make and let no intelligent being ever see it. People have to learn and build off of what others create. If human artists are free to learn from other artists, why not AI? That may be a touchy question, but it is an important one to consider. Copyright laws have never attempted to protect a person’s style or genre because this would be too restrictive. In that case, the first person to paint a realistic painting could be protected by the law even if they said, “This is my style. Now no one else is allowed to use realism in their paintings.” The first person to draw in an anime style could legally force people to come to them if they wanted anything drawn in an anime style. Such monopolies would be to the advantage of a few at the expense of many.
Let's conclude the conversation on copyright laws with the current information directly expressed by the U.S. Copyright Office. At the moment, AI art cannot be protected by copyright since it is not the product of human authorship. However, that does not mean it cannot be used. According to the U.S. Copyright Office, “When an AI technology determines the expressive elements of its output, the generated material is not the product of human authorship. As a result, that material is not protected by copyright and must be disclaimed in a registration application.” This is the same reason a monkey cannot have a painting registered for a copyright because it is not a human. However, they also claim “This policy does not mean that technological tools cannot be part of the creative process. Authors have long used such tools to create their works or to recast, transform, or adapt their expressive authorship.” The implications are further explained as follows, “...a human may select or arrange AI-generated material in a sufficiently creative way that, ‘the resulting work as a whole constitutes an original work of authorship.’ Or an artist may modify material originally generated by AI technology to such a degree that the modifications meet the standard for copyright protection.”7 If the U.S. Copyright Office considers AI technology a tool that an artist can use, I am comfortable using the same terminology. AI image generators are tools for human creativity. The raw output of an AI generator is not copyrightable, but they can be used in creative ways to make a copyrightable work as a whole, which is what we have done with our game.
I cannot ignore one other common complaint about the development of AI technology, which is the datasets themselves. Large datasets are required to train AI models and are often compiled by third parties that grant AI developers access to their data. The largest open source dataset used to train AI art generators like Stable Diffusion is known as LAION (Large-Scale Artificial Intelligence Open Network), which is a German non-profit. LAION does not hide how they obtain their data making it the primary gaitway to understanding how large data sets acquire their information. Interestingly, “LAION’s transparency distinguishes it from competitors like Microsoft, Google, Midjourney and OpenAI. Little is known about the training material and methods used by such companies.”8 When it comes to compiling large datasets like the one owned by LAION, people are most concerned about the availability of links to personal and/or copyright protected material within those datasets. Datasets are criticized further by the fact that they may often include pornographic, violent, or racially biased material. While explicit content is usually attempted to be filtered out, filters are never perfect. The rampant quantities of misappropriated information and explicit content across the internet is a modern tragedy. However, the question then becomes are the companies controlling the data at fault? How far can the bucket be passed? The datasets simply compile information across the internet in one ingestible location for AI training. All of the information in each dataset may be found by essentially anyone with the right knowledge of the internet. Therefore, one could ask, then is the internet the root of the problem? Where does the internet get such questionable content? It ultimately falls back on the individual acts and choices of sinful people, and we are all guilty to some extent. The evil that permeates humanity broadcasts itself throughout the crags of the internet. Being online is a risk. Posting a picture of yourself online is a risk. But we often knowingly do it anyway. No system or technology that allows the propagation of human desires will remain completely unsoiled–not because of the technology, but because of us. More clean and considerate systems that ensure the protection of the rights and privacy of individuals is absolutely necessary. However, it is always something that takes time and effort by people with a redemptive mission. AI can be used to make deep fakes, overly sexualize women, promote racial bias, scam others, and more, but the wrong uses of AI are a poor reflection on humanity and not the tool itself.
There are many other interesting considerations. It is important to note that the “ethical” nature of data collection is also rapidly evolving with the technology. Some AI developers are seeking to develop models with internally sourced data or create systems that allow creators to tag their content to prevent it from being used as training material like Adobe. Lastly, I have primarily discussed U.S. copyright laws, but it is also important to note that legislation varies from country to country. Both Japan and Israel have already declared AI data mining on copyrighted material “fair use” for the purpose of innovating better comprehensive AI technologies.9
The Emotional Issue
One of art’s greatest powers lies in its ability to alter our emotional state, and I believe emotions are one of the main reasons why people initially push back against AI art. Simply pondering if an AI can create “art” evokes many philosophical questions… Isn’t art a sacred space where human creativity is meant to reign supreme? Isn’t art the one place where the emotional soul of humanity has the unbridled freedom to self-express? How could cold and emotionless machines enter that domain? Catastrophizing questions often follow. Will AI art destroy the world of art as we know it? Will the last true artists be wiped off the face of the planet into non-existence? Will we be forced to remember “the good old days” of a better artistic age before AI stole its soul and replaced it with a digital one?
The emotional provocation doesn’t stop there. For many artists being an artist is not just a career but an identity. Some might claim AI art could never be considered real art because, unlike a true artist, it has no identity. According to many, becoming a true artist requires talent in combination with years of training and hard work. Conversely, an AI creates an output in a matter of seconds, and the freaky thing about it is that the average person may not be able to tell the difference between what an AI outputs and what a human produces. This simply comes as a shock to people–artist or non-artist. Five years ago, who could have conceived that one day a computer would almost autonomously create art? As the sheer speed and “quality” of AI art develops, the emotional upheaval, especially for artists relying on their craft for a living, continues to intensify. Their identity is under threat. The years of blood sweat and tears spent perfecting a craft can feel devalued when, seemingly overnight, a new technology can quickly generate “art” in seconds. I imagine this would be distressing. However, it is important to consider the positive results of lowering the bar of entry into the art world. Many individuals with disabilities normally impeded from exploring their creative and artistic potential are now able to do so using AI technology. I can imagine that the emotional reaction of such individuals would be the opposite of distress. AI art has brought both the extremes of distress and excitement into our world. No doubt, AI art is an emotional issue.
Let’s discuss the emotional problems many people have with AI art more closely. First, there is the apparent strain it takes to relate to the creativity of a machine that has no identity. The emotional disconnect of trying to appreciate art generated by an AI can seem unnatural to many people. Fundamentally, AI art is not made by humans. This is true, but that does not mean that AI art is not real or that it does not exhibit any human creativity. Rather, the output of the AI is a collective representation of the artistic talent and creativity of humanity. It embodies the beauty and the flaws of human art. Humanity is not a product of artificial intelligence; artificial intelligence is a product of the collective intelligence of humanity. Therefore, you cannot remove the humanness permeating the identity of an AI. An AI may not have a personal identity, but in many ways, an AI represents our identity. Therefore, to view AI art as the cold and lifeless product of a machine is inaccurate. Rather, AI art is a large-scale celebration of the beauty that human artists can accomplish, and unfortunately, also a window into our own perversion and biases. Secondly, one might argue that real art is ultimately the product of an artist–an AI is not an artist, which makes AI art not real art, and therefore inferior. Ultimately, you can define art however you want. If art is defined as a human product that is fine. In that case AI art would not be considered art. However, that would not inherently devalue the output of an AI. The aesthetically pleasing output of a modern AI generator would still sell for much greater than my childhood stick figures would. My parents may have valued my stick figure drawings because they loved me, but that doesn’t make what I made objectively superior to whatever an AI could produce simply because I’m human. Even among humans, the boundaries of what qualifies as art continues to be stretched by our misconstrued beliefs about subjective reality. If I consider a banana taped to a wall good art simply because I’m human and think it is, who can tell me otherwise? If we take this subjective line of reasoning then there is nothing stopping AI art from being considered good art.4 However, if for one moment, we overlook our ability to use semantics to unfairly exclude the possibility of AI art qualifying as art, and consider art simply as an objectively beautiful and engaging creation, then the notion of calling AI art real art starts to sound less crazy. Objectively, AI has the ability to generate art that is far superior to many human artists (myself included). Granted, there will always be high caliber human artists who will far surpass anything an AI will ever be able to create. There are many ways AI art can be inferior to having a human artist. However, the point is not to compare apples to apples, but simply to acknowledge that AI has the ability to create objectively appealing works of art with the power to affect human emotions. The creators of the AI models themselves have created a technological and mathematical work of art deserving recognition.
The Economic Issue
The economic reasons why AI art is either resisted or supported are also very complex. First, for the artist who perceives they should be compensated because an AI used their work as training material, AI art represents an incredible frustration. On the other hand, many artists who embrace the use of AI have no problem having their work used as training data. Second, AI developers simply cannot make the technology without access to large datasets. Therefore, AI art is truly a novel case for lawmakers attempting to enact legislation that balances both the desires of artists and tech innovators, which is not an easy task. As artists and tech companies both lobby for laws in their favor, each side is perhaps pushing the boundaries of what could be considered fair compensation or fair use. However, both sides are motivated to a certain extent by money.
Are tech companies profiting from the works of many artists past and present? Yes. Is it easy to tell who they might need to compensate? No. Have they operated out of their legal rights? That remains to be seen, but they have certainly created something that could easily fit into the category of why copyrights are given with limitations in the first place, which is “to stimulate creativity and the advancement of ‘science and the useful arts’ through wide public access to works in the ‘public domain.’”1 Should artists have been allowed to opt in or opt out of having their public work used to train AI before the models ever began training? Many firmly believe so. Have artists been damaged because of people carelessly and wrongly using AI in specific scenarios? Yes. Would many artists benefit financially if AI art were highly regulated and stigmatized in the eyes of the public? Very likely. Money is a huge motivator and is the source of many conflicts.
It is easy to have our compassion stirred for hardworking artists in a field where it is already difficult to excel and make money. We rightly hope that such artists find as many ways to be fairly compensated as possible. However, when it comes to training AI on large quantities of data, lawmakers must practically determine what constitutes fair use. How far can a right to compensation be taken? Should a photographer selling a photo of a building have to compensate the architect who designed it? If so, who is to be compensated if they sell a photo of an entire city? If a painter uses the photographer’s picture of the building as a guide for their next painting, should they compensate both the photographer and the architect? While abstract, these difficult questions are helpful when trying to consider whether or not AI development should fall under fair use. It is ultimately up to the lawmakers.
AI art will undoubtedly impact the economy and will likely make it more difficult for some artists to make a living if they fail to adapt to the new landscape. However, the fact that certain jobs will be lost as the result of a new technology is not a good reason to suppress its use and development. New technology always changes the demand for jobs. Automation has reduced the need for many occupations but has also paved the way for countless new jobs and innovations to emerge. The World Economic Forum predicts AI will replace 85 million jobs but create 97 million new jobs by the year 2025. Discussing job loss may feel harsh, especially if you are not an artist. However, AI will likely encroach on a wide variety of occupations other than just artistry. It is already rapidly advancing into the healthcare system including physical therapy, which is my field of expertise. AI is being developed to help interpret X-rays and MRIs. Robotics and wearable devices are being made to gather and interpret data about a patient’s movements. Deep learning networks that allow an AI to assess a patient's physical performance and more are also being developed.10 A person may be able to carry around a physical therapist in their pocket one day! Despite all the change AI might bring to countless occupations, there are two things I am certain of. One, everyone will be forced to adapt to stay competitive whether they like it or not. Two, AI will never fully replace the need for skilled humans. While some very focused niches of work may become extinct as the result of AI, there will remain a need for human experts in any field that AI integrates into.
The fear of what “cheaper, better, faster” will do to the economy always brings interesting hurdles, push-backs, and double standards. Today plastic has become known as the enemy of the environment and has garnered a lot of negative attention, but few people are aware that plastic was first invented as a means to conserve the environment–namely elephants. The first synthetic polymer was invented in 1869 by John Wesley Hyat who was led to invent plastic as a substitute for ivory, which was the primary resource being used to make billiard balls at the time.11 Plastics have continued to evolve. Today, plastic is everywhere chiefly because of its affordability and moldability. Due to the prevalence of this incredible synthetic material, the word “plastic” has almost become synonymous with the word “cheap.” Over time we have come to realize the harm that mass producing single use plastics can have on the environment. However, we should not throw the baby out with the bathwater, so to speak. Plastic is not evil. It is a material that we simply need to learn to harness responsibly. It is sobering to realize that a black and white attack on plastic cannot be conducted without plastic. Such a person could not make a post online about the harms of plastic without the plastic in their computer. They could not drive to their speaking engagement without the plastic in their car. Likewise, a black and white attack on AI would be relatively difficult to conduct without some use of AI. The spell check function to publish such an argument or the research conducted using a search engine is already incorporating some level of AI. Innovations like plastic have taught us to be wary of unforeseen costs that may come as a result of fully embracing new technology too soon. However, we must be careful not to push back against such innovations using a double standard. No doubt, new technology should be adopted with caution and wisdom must always be exercised to use it responsibly, but this caution should not devolve into exaggerated claims intended to stigmatize it. It took time for other visual technology such as photography and photoshop to become widely accepted; it is difficult to tell when that will be the case for AI art.
It is not new to say “AI art is not going away anytime soon.” However, for those who lament what it might do to the economy and the traditional art forms they love, it is encouraging to reflect on other traditional art forms that have endured. I personally believe the majority of traditional art forms will endure despite the speed and ease of AI. Photography didn’t replace oil painting. Factories didn’t replace all handmade products. The list goes on. As a personal example, I love the show Forged in Fire. In the show dozens of bladesmiths across the country compete to make the best blades possible. The final challenge is always recreating an iconic weapon from history. Blade smiths are no longer needed in our modern society, so where are these bladesmiths coming from? Swords can be created and mass produced quickly using modern industrial methods… The answer is simple. People continue to make swords and knives using old forging techniques simply because they love doing it and because many people still love buying it. Many of the blacksmiths are not just hobbyists but make a good living off their craft. The landscape of blade smithing has dramatically changed since ancient times, but it is still an important craft in our society today.
However, this brings me to my last point of discussion. We can often incorrectly assume that if an artist uses AI, then they don't need any human skill, time, or effort to make their finished product. That is simply not true. A racecar driver may drive a high performance car, but that doesn’t mean they don’t need skill or training to drive it well. I have included a long comment I found on Reddit that paints a good picture of how one skilled artist might incorporate AI tools into their workflow. Take a look at what Tyler Zoro says:
“Here is a fairly typical workflow for an artist who uses AI tools. It's far from the only way to work, in fact, it's probably safe to say that two artists who work with AI tools having the same workflow is pretty rare. But let's use this example for now.
Make 100-200 images by hand (or just select them from your portfolio most likely)
Run those through a tool that creates a LoRA
Rough sketch the piece you want to work on
Go into a 3D animation program and arrange a character pose wireframe to match the sketch
Go into Photoshop or similar and develop some textures to use for the final piece
Find two or more models that roughly meet your needs for the final piece and merge them into a single checkpoint
Bring in all of the assets you've developed through ControlNet configuration
Select the model parameters for your merged model
Select the parameters for the LoRA you created (usually just the weight)
Select an appropriate VAE for the model and for your intended result
Now write a prompt
Generate an initial result
Use a refiner model to finish the generation
Take the resulting image out to Photoshop for some touchup work
Repeat the generation process as img2img
Repeat the past two steps several times
Select (potentially merge) model for inpainting
Begin inpainting final details
Upscale and retouch as needed for final publication medium
Given this workflow, imagine how confusing it is to see so many anti-AI comments in this sub and elsewhere effectively describe working with AI tools as, "you just write a prompt."12
Using AI is not simply prompting. It takes a lot of practice, skill, and knowledge to use well.
Conclusion
Technology often forces us to adapt in new ways we are not ready for. We are creatures of habit. The emotional upheaval and excitement AI art has garnered in the art community is understandable. AI can be viewed as a competitor or an ally. Whether or not AI is used or rejected, AI is ultimately another modern tool available to artists. To date, there is no clear evidence that AI art represents any form of blatant copyright infringement and is usually attacked using black and white statements without an in-depth analysis of the facts. However, AI art is not without its pitfalls. Many of the biases and limitations of AI art generators can be very difficult to overcome, which increase the skill needed to use them well, but the pitfalls are primarily a poor reflection on humanity rather than the tool itself. The economic landscape will be forever changed as the result of AI. We must continue to seek ways to use the new technology more responsibly, but we should rationally approach the subject and make appropriate changes as we learn better ways to harness the technology. While it is easy to catastrophize, I don’t believe AI art will ever remove the need for skilled human artists. However, the efficiency and affordability of AI art has allowed many small creators like myself to make more competitive products that would not have been possible otherwise.
Final Thoughts
Despite the use of AI, Escape Master is ultimately the product of many skilled and hard working artists. We are a small company and have greatly benefited from the use of AI technology. Therefore, the contents of this article cannot be interpreted without bias. We value the art community and also embrace the advent of new technology. We continue to grow and our viewpoint continues to be shaped as we learn more.
Bibliography:
Staff, ARL. “Copyright Timeline: A History of Copyright in the United States.” Association of Research Libraries, 18 Dec. 2020, www.arl.org/copyright-timeline/#:~:text=The%20First%20Congress%20implemented%20the,Statute%20of%20Anne%20(1710).
Williams, Sophia. “Ai and Artists’ IP: Unpacking Copyright Infringement Allegations in Andersen v. Stability Ai Ltd..” Center for Art Law, 28 Feb. 2024, itsartlaw.org/2024/02/26/artificial-intelligence-and-artists-intellectual-property-unpacking-copyright-infringement-allegations-in-andersen-v-stability-ai-ltd/.
Frank L. Gerratana, Michael R. Graif. “Lensa: Are Ai Art Generators Copyright Infringers?” Mintz, www.mintz.com/insights-center/viewpoints/2231/2023-02-14-lensa-are-ai-art-generators-copyright-infringers#:~:text=Thus%2C%20AI%2Dgenerated%20artworks%E2%80%93,even%20absent%20the%20artist’s%20consent. Accessed 22 May 2024.
“In Defense of AI Art” YouTube, uploaded by LiquidZulu, 4/14/2024, https://www.youtube.com/watch?v=gWmEXCJIIZ4 (mature content contained in video)
Cho, Winston. “Artists Lose First Round of Copyright Infringement Case against AI Art Generators.” The Hollywood Reporter, The Hollywood Reporter, 1 Nov. 2023, www.hollywoodreporter.com/business/business-news/artists-copyright-infringement-case-ai-art-generators-1235632929/.
“Can the Human Brain Be Reduced to a Binary System?” Biology Stack Exchange, 1 Oct. 1960, biology.stackexchange.com/questions/29679/can-the-human-brain-be-reduced-to-a-binary-system.
SHIRA PERLMUTTER, SHIRA. “Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence.” Copyright.Gov, REGISTER OF COPYRIGHTS AND DIRECTOR OF THE U.S. COPYRIGHT OFFICE, 10 Mar. 2023, www.copyright.gov/ai/ai_policy_guidance.pdf.
Data, BR. “We Are All Raw Material for Ai.” BR, interaktiv.br.de/ki-trainingsdaten/en/index.html. Accessed 17 May 2024.
Levine, Joshua. “Primer: Training AI Models with Copyrighted Work.” AAF, 14 Nov. 2023, www.americanactionforum.org/insight/primer-training-ai-models-with-copyrighted-work/
Mozafaripour, Sara. “Artificial Intelligence in Physical Therapy: Cool Applications, Fascinating Implications: USAHS.” University of St. Augustine for Health Sciences, 22 Mar. 2022, www.usa.edu/blog/artificial-intelligence-in-physical-therapy-cool-applications-fascinating-implications/.
“History and Future of Plastics.” Science History Institute, 3 Jan. 2024, www.sciencehistory.org/education/classroom-activities/role-playing-games/case-of-plastics/history-and-future-of-plastics/.
Zoro, Tyler. “Why are anti-AI folks so laser-focused on prompting when most AI image generation work isn't prompting?” Reddit. 4/29/2024. https://www.reddit.com/r/aiwars/comments/1c5ltn3/why_are_antiai_folks_so_laserfocused_on_prompting/
Roller, Joshua. “AiArt: Why Some Artists Are Furious about Ai-Produced Art.” IEEE Computer Society, 21 Nov. 2023, www.computer.org/publications/tech-news/trends/artists-mad-at-ai.
Yorumlar