Rise of the Robots – AI Art

Normally, I write about controversial topics from a moderate and reason-heavy standpoint, but this topic, perhaps ironically, is one I have an undeniable emotional attachment with. In short, I am fascinated with the idea of AI. Ever since I was a kid and first encountered computers, I had a kind of faith of imagination that all the science fiction predictions of AI and Robots were not only likely but essentially inevitable. I didn’t think we would ever meet alien intelligence, I believed we would create them. Over the years, I’ve had many debates where I claim what computers will someday be able to do, and I am told, “no, it’s impossible.” And over the years, my expectations have proven correct time and again.

So naturally, every time a story comes out about an AI doing something amazing, I feel both vindicated and incredibly excited as the world is one step closer to the wild imagination of my childhood and the dreams/nightmares of the science fiction I so passionately enjoyed as a kid. That said, AI art was not one of the popular spots on the Machines replicate mankind bingo card, partly because the crossover of talented artists and AI researchers isn’t broad. While the idea of AI art has been around quite a while, I think few of us thought it would master the arts as quickly as it has. Most futurists have egg on our faces about how long its taken AI to tackle language so the act of making paintings kind of snuck up on us.

My enthusiasm is also owed to my innate love of change and disruption. While I’m no chaos god, I do have a bit of the trickster spirit in me and I get very stimulated and excited by change, especially unexpected change. Something in my psyche just responds with excitement and imagination to events that upend what we know, and pose new opportunities and challenges. To me, that’s the siren call of adventure and learning.

Recap for the uninitiated

Really quickly, in case you don’t know, in 2022, AI art took the internet by storm. Numerous machine learning neural network AIs became publicly available. These programs sweep the internet, analyze existing artwork that is publicly available and can then produce new artwork based on what they have learned and the direction of a user. The user enters some text prompts, and the AI quickly creates images based on those prompts. The system basically combines its visual understanding of the words entered with the style and implementation of art associated with those terms and creates an image.

As of this moment, these images often have curious artifacts in them, things that don’t make much sense to us as humans because the AI doesn’t have the same conceptual frame we do. A figure might have seven fingers on its hands, or it may have a twisted face on a piece of jewelry, or buildings may be half floating in the sky. But very quickly, these are disappearing as users “train” the AI by judging its work and adjusting their prompts. Every day, it gets harder to identify the idiosyncrasies, and the quality of the art is good enough that it’s won art contests, unbeknownst to the judges of the show. The programs can generate this kind of art in seconds and can make hundreds of variations. And for the most part, they are either free or incredibly cheap to access and use.

Stopping to think if we should

The real debate of the moment is whether the use of this art is ethical. From out of nowhere, this technology threatens the livelihood of commercial and fine artists everywhere. Commercial consumers of art are clearly motivated to use this incredibly cheap and increasingly high-quality source of art. Not only do they need to pay only pennies for the art, but it can also be produced incredibly quickly. Even in the fine arts field, the absolute flood of AI art that is coming can easily drown out human artists’ work and even call into question their authenticity as human-produced work.

A human artist could spend a lifetime and never approach the technical quality or flexibility of the AI systems. More critically, they could never produce the kind of volume that an AI can. Whatever non-intrinsic advantages a human artist could claim to have, the AI is going to quickly surpass them, just as AI can now defeat the best human chess masters. The two intrinsic qualities human artists have to their advantage are their physical skills and their inherent rarity. While robots are real, they aren’t cheap, and the fine motor skills of most artists have proven to be one of the more challenging areas of robotics. That advantage won’t last forever. Human rarity, the fact each artist is a unique individual, is something that truly matters in the fine art world, but again, proving that authenticity of individuality may become difficult as AIs become adept at mimicking every individual style.

So basically, without some kind of social constraint, the vast majority of artists working in a capitalist framework are doomed by this technology. Their commercial value is under incredible threat.

Well, should we?

I think we need to do a lot of reflection before we can answer this question with confidence. I think the place to start is to enumerate some of the pros and cons of AI art generation, starting with the most pressing and obvious issues and moving into the more esoteric ones. Once we have a grasp of those, we can start looking at courses of action, their consequences, and the practicality of those choices.

Pro: Cheap Art

Honestly, art isn’t all that expensive, to begin with. We live in an age of incredible access to art, and it suffuses nearly every consumer product in existence. Yet, there is still a cost to having original art created to your specifications and desires. That remains something of a luxury good that mostly the middle class and upper class have access to. AI art promise to make high-quality custom artwork at a near-zero price point. A custom portrait will cost you less than a candy bar so long as you have at least a smartphone, and almost 80% of Americans living in poverty have one.

The technology is even more attractive for content creators, people who are engaged in some kind of creative endeavor online but who normally would have to pay others to create graphics for them. These people have become a big market for freelance artists and stock art services, but they are very cost-conscious, and the lure of nearly free, high-quality art is incredibly strong. Anything that lowers their production cost, especially for those without an established audience, is very attractive.

Any technology that not only lowers the cost of production but also decreases the time to produce and the complexity of the production process is going to be a godsend for the consumers of visual art, and there are an incredible number of consumers of visual art. Not only that, but lowers the cost of entry into any activity involving artwork, which opens the doors for people who would otherwise not have the opportunity. Even those who normally have no budget for art for their blogs or other hobbies and activities can afford full-color custom artwork in exactly the style they want for next to nothing. That’s just plain fun and enjoyable for millions of people.

Con: Starving Artists

Artists have never had it easy. While great artists can become wealthy and widely celebrated, they are far and away the exception. Most of the few that can actually manage to make a living at it will have lower than average incomes while working very long hours. They may well love this work, but it is nonetheless a very risky and challenging way to make a living, and always has been. But now, all bets are off. What they can spend a lifetime trying to master and produce, an AI can learn in a year and produce in seconds. An AI can make millions of unique art pieces a day, incorporating the genius of 1000 years of humanity’s most talented artists. Every year there will be fewer and fewer ways for humans to compete with these programs in an open market.

This con can have further downstream consequences. AI art itself can suffer from the loss of the individual input of millions of creative humans who have to divert their energies from creative endeavors to more profitable activities. The richness we can gain from cheap AI art can ultimately be undermined by losing access to the creative output of professional human artists. Whether AI creativity can eventually feed upon itself is hard to say at this point, but we know that what makes AI creativity possible is all the human creativity it has to learn from.

Pro: New Creativity

AI art can increase human creativity in two significant ways. Firstly, it gives vastly more people the ability to see their creative ideas realized. With such a low cost, anyone and everyone can turn their conceptual ideas into fully realized images. I’m fortunate enough to sometimes work in a field where I can direct artists to create what I want to see, and it’s a wonderful feeling, especially for someone who lacks the skills and discipline to do it for myself. To be able to ask a machine to create anything I can imagine is an incredible extension of my own creative thoughts, and now it’s something millions of people can do, which they could never do before.

The second source of new creativity is the AI itself. These are essentially alien intelligence that, while they learn from humans, don’t have the same human context we do. They do things we find weird and strange because they don’t have the same grounding of experience or the breadth of thought we do. Seeing this machine’s imagination can spark new ideas and inspiration in us. Just like nature provides a wealth of non-human inspiration, the web of logic and information inside these neural networks produces wholly original and alien artifacts we can now experience. Chess and Go programs have not only defeated the greatest masters, but they have also taught humans how to play better than ever before and opened our eyes to new strategies and ideas even generations of human players were not able to invent for themselves but are now at our disposal.

Con: Lost Authenticity

AI image generation generally has the ability to create authentic looking images that are not authentic at all. The line between reality and fantasy blurs and the idea photo evidence goes from a rock solid proof to being almost meaningless. The critics will see AI art in everything, leading to false claims of fabrication, while many AI images will creep into places they are unwanted through both error and deception. Digital artists will need to submit proof of work and there will be a constant battle of detection vs deception that calls everything into question. The true nightmare scenario is that people simply lose interest in art, artists, and anything truly human, simply accepting everything as essentially fake and meaningless.

So What Do We Do?

I’ve never been a proponent of banning technology that people can use for positive purposes. Much of my career has been about creating software that effectively does work other people used to do. I even programmed myself out of one of my early jobs, writing software that did what I was initially hired to do by hand. I’ve watched technology both wipe out existing jobs and create new ones. Ultimately, unemployment doesn’t change all that much, but the kinds of jobs that exist do. That said, this is the fist time technology has come for jobs as enjoyable and spiritual as being a visual artist.

I think that the AI, as it stands, is more of a threat to new and developing artists than well established ones. The market for fine art strongly values the humanity of artists, their message, their mystique, and their individuality. In the commercial world, artists do a lot more than just draw and paint. The truly talented ones can understand the needs of clients and create art that is going to connect with an audience and move them to buy or remember a product. Unfortunately in the more mass market world, sometimes people just need any old picture of X, Y, or Z and for that, AI art is fast and cheap and good enough.

Ideally, I’d like to see solutions that preserve the value people can get from the technology while also protecting artists and creative people wherever we can. I’m not for an outright government ban on such technology, nor do I think that would likely work. I am for some regulation and for social and ethical standards that work towards good goals.

Copyright Concerns

One of the most important areas to resolve are the copyright concerns surrounding the training of these AI Art systems. These programs are developed by having a computer program transform digital images, along with text describing them, into a kind of interlaced set of related data which is then amalgamated with all the other image and text combinations into an enormous dataset. The AI then uses all this information to build a matrix of inferences between the words and the minute elements of the images so that when you say “I want a picture of a clown on a beach at sunset.” It walks through all this data in semi-random ways building a picture from an initial fuzzy cloud of possibilities.

The challenge from many artists is that they did not give permission and were not paid for the images that were used in this process. The tech companies and organizations behind these systems say that the images are not copied into the database, but are instead analyzed and transformed into something distinctly different. There are multiple court cases already underway to test these theories. The artists have the more intuitive argument, but the outcome of law depends on technical legal analysis and these are largely uncharted legal waters.

However these cases come down, I think there is a strong cast to be made that the copyright owners of the data that is used to create these systems are meaningfully contributing to the ability these systems display and are not generally compensated in any way. The hardware and software makers are also contributing but have indeed been paid for their work. Right now, much of the AI art programs are essentially giving away the service, or making it available at very low rates. So low that even if artists were paid a share, it would be fairly tiny given the vast number of works included in these models. Still, whatever the economics of it, some way needs to be found so that AI art systems can support and pay the artists that have trained them so effectively.

My initial thought is that some system where some portion of proceeds from AI systems should be put into a fund that can be paid to artists who contributed to the system or some subset of them. Or at least be used to subsidize art students or commission art. Ideally companies should only take art from sources they have valid licenses for, but it may well be too late to do that.

Professional Standards

This is the area most pressing to me, as AI is a big area of concern in the Tabletop Games industry. It is a business that exists because of and for the benefit of creative people. Its often as much a labor of love as a business. That said, it can still be a model for other businesses that care about creatives.

I think in a creative industry, and really any industry that uses art commercially, we should seek to support and preserve artists. AI art is ultimately a product of human artists and it wouldn’t be possible without their efforts. For that reason, I think anyone hoping to make a living, or a substantial portion of their living from creative work, should seek to employ skilled artists for their projects. Those artists in turn should use good professional judgement as to what degree of AI assistance they use in their own creative efforts, though never more than the employer specifies as acceptable.

Final products should indicate whether or not AI image generation (or other types of AI-gen) were used in creating the product so that consumers and other producers can make informed decisions about paying for the product.

Go Easy On The Witch Hunting

I’ve seen a fair amount of aggressive posturing on the issue of AI. I totally understand the passion, but I’ve seen many instances where people have come out torches in hand to castigate art when it is not AI-gen. They leap to conclusions and flame throw the target online leading to all kinds of reputational damage and distress from the target. I don’t think there is anything wrong with community policing, but like real policing, it should come with a good investigation and a measured response.

People also need to understand that companies consist of layers and multiple individuals and its easy to have both miscommunications and deception at any point in a chain of production. Companies that use AI gen should have the opportunity to address any accusations of misrepresentation or unethical or unprofessional use of technology.

Final Thoughts

I think a good number of people I know will be somewhat disappointed that I’m not 100% against the use of AI-Gen technology. I often take a middle road on controversial issues and I often take some heat for it. I’m OK with that. I’ve had some fun looking at AI generated images and have had some fun making a few. It’s a highly entertaining tool just as a thing to play around with. Its also incredibly useful for creating quick visualizations of ideas to share with others. I like that it provides utility and enjoyment to a huge number of people.

I’ve decided I won’t use it for any commercial product that I produce for sale. I’m happy to argue that its a bad idea for creative companies to use it, but I’m not willing to castigate those who do. I don’t begrudge anyone else having a different view, but I am willing to argue why I think that it is a technology that is both a threat to artists and a boon to everyone who can enjoy art.