The consequences of AI artwork

The subject of AI art is en vogue. While theexplosion of Chat GPT into the mainstream, and the threat that that tool poses to all theemail and copywriting jobs of the world has been dominating the mainstream conversation,image generation tools like Stable Diffusion and Midjourney have become the subject of alot of conversation in my communities, among artists and creatives. Opinions are, as they areabout anything, divided, although in the circles where I run, people are generally… less thanenthusiastic about these tools, at best seeing them as intellectually and spiritually vapid spamengines, and at worst as an existential threat. Now, before I did this YouTube nonsense fora living, I did a different kind of nonsense: I was a mid- to low-tiercomic artist and illustrator,.

Making rent by taking commissions and prayingthat someone would subscribe to my Patreon. For someone who is in the position that Iused to be, AI art tools are, objectively, a direct threat to their livelihoods.The market that forms the basis for their business is being directly targeted fordestruction, and so is the skilled labor of every mid-level concept and texture artist,every in-betweener on an animation production, every storboarder and every artist who makesa living making Twitch emotes and drawing people’s D&D characters. Even the usuallypreternaturally resilient not safe for work art market is feeling the squeeze, with imagegeneration tools increasingly able to output infinite amounts of instant gratification.AI art tools are designed, explicitly,.

To eliminate the cost of labor for that kindof work, and since no human being can compete with “free,” those tools are going to driveworking artists directly out of business. So. Let’s talk about AI art. And something I want to say up front is thatI am not very interested in debating whether AI art is art. I am of the If I Walks LikeA Duck school of thought, and AI art is, for all practical intents and purposes, art. Rightnow we can usually spot it if we pay attention, but the technology is advancing and we willall fail that Turing test someday. And if our only objection to AI art is thatwe can tell that it isn’t “real” art, then that objection is going to fail the momentthe machine can make it look convincing enough.

So. AI art is art, but the moreinteresting question is what kind of art is it? What are the conditionsfor its production? And what does it do? And I think in order to talk about those things,we should first very clearly establish that… The fact that we all collectively call thisstuff “AI art” is the result of an incredibly successful marketing and branding exerciseby the people and companies who stand to get rich from selling the technology. AI artis not made by AI. There is no artificial intelligence involved with any stage ofthe production process of AI art. Despite what the “AI difficulty” settings in yourfavourite video game might have you believe, nobody has ever invented a true artificialintelligence, and certainly nobody is using.

One to generate big tiddy anime waifuswith horribly distorted nightmare hands. These tools are not run by artificialintelligences, they are run by machine learning algorithms. These algorithms areno different in nature than the one that decided whether to recommend this YouTubevideo to you, no different than the one deciding which ads to show you on everywebsite that you visit, and no different than the algorithm shoveling recommendedcontent into your face on TikTok, Instagram or Twitter. They may have different purposes,but they’re made from the same technology. It is important to establish thisfor a couple of reasons. First of all because by branding these products as“artificial intelligence,” the people who.

Seek to profit from it are engaging in what Iwould call essentially a kind of soft fraud, a blatant false advertising. It makestheir products sound more sophisticated and futuristic than they really are, andobscures the reality of what’s being sold. Secondly, conceptualizing AI art as“artificially intelligent” muddies the waters of all conversation about it. Implyingthat there is an artificial intelligence at play, a mind, opens the conversation to all sortsof philosophical debate about the true nature of consciousness and creativity. How can you saythat these machines are not creative? Who are you to define what counts as a real consciousness? Itcreates a smokescreen of false ethical dilemmas, and makes people give a leniency to AI art thatwe would never extend to the algorithm shoveling.

Four hundred million banner ads in our face whenwe’re just trying to read a goddamn article. Machine learning algorithms,in the very simplest terms, are prediction engines. They are designed toattempt to predict a desired output based on a given input. When you ask ChatGPT towrite you an essay, it cycles through an unspeakably complicated series of mathematicalcalculations to predict, one word after another, what a successful output might look like.It doesn’t think about any of those words, what they might mean, what kind of context theyhave, all it does is predict mathematically which series of words in a particular orderwill lead to a satisfactory result. I bring this up because I want to dismantle aparticular argument. I’ve seen it argued many.

Many times that AI image generation tools workin a way that is comparable to the process of human creativity. AI art tools mix up a databaseof pre-loaded images to arrive at something new, and isn’t that just what we do? Don’thumans also just mix and match their experiences? How can you say it’s so different? This fundamentally misunderstands how brainswork, the human mind does not run on binary code. But it also completely misunderstands howthese machine learning algorithms work. Machine learning tools do not make decisions, they don’tthink critically, they don’t reflect, and they have no awareness of what they’re doing. Theyare simply engines of manipulated probability. If you want a visual way to conceptualize thesetools, imagine a giant pachinko machine with a.

Vast and impossibly complex network of pegs.These pegs bounce the balls back and forth in insanely complex ways until they land in aspecified pattern. You as a user can make the balls land in any position and pattern thatyou want, simply by changing how the balls go into the machine. Your prompt decides the output,but at no stage of the process is any part of the machine “thinking.” It was designed to allow ahuge range of outputs from a huge range of inputs, but the balls don’t bounce on the pegscreatively, they bounce according to math. All of this is to say that “AI art”is a misnomer. There is no artificial intelligence involved in any of stageof this process. I will not entertain credulous philosophical arguments about “buthow do you define intelligence?” because I’m.

Not eighteen anymore and this isn’t an Introto Philosophy class with a stoned TA who gives you a passing grade just for showingup. These are machine learning tools, they are probability models, and they are notcreative in any reasonable sense of that word. Of course, just because the toolsaren’t in themselves creative, human beings can still usethem in creative ways… right? No. The answer to that question is no, but first let’sstep back for a second. I just a spent a long time arguing that AI art tools are not equivalent,or even similar, to human minds, and thus you can’t compare what a machine learningalgorithm does to human creativity. But let’s assume I’m wrong about that,let’s assume that that whole section.

Of the video was just flat out wrong.Let’s assume that they are equivalent: when a machine learning algorithm generatesan image based on a prompt, it’s engaging in the exact same creative process that a humanwould, and thus its output is equally creative. If this is true… what the hell makes youthink the prompter counts as the artist? When I take a commission from a client, theytell me to various levels of specificity what they want me to draw, and my job is to followtheir direction. But when the piece is finished and I deliver it to them, everybody on earth,from children to copyright lawyers to national governments, universally understand that I amthe artist who created the art. The commissioner may have prompted me, but no matter how specifictheir instructions, they will never be the artist.

So if the creative process of a machine learningalgorithm is equivalent to the process in a human, then why should the prompter be creditedwith anything it makes? In this scenario, the machine is the artist. In thescenario, the machine is creative. The prompter is merely the commissioner.Even the most arrogant Medici patron of the Renaissance would have never havegone around Florence claiming that he sculpted the statues of Michelangelo justbecause he asked for them to get made. So if machine learning algorithms are creative,then “AI artists” definitely can’t be. But again, I don’t really believe this. Machinelearning algorithms are not creative, they are, like pencils, brushes, rulers and paints, merelytools, and all tools can be used creatively..

In theory. If we ignore all real world context,all questions about how the tools were made, how they function, ethics, copyright, the lot ofit, then yeah, absolutely. Like many of my fellow artists, I can absolutely see how these things areuseful. You can save a lot of labor with something like that, and if a creative mind applies thetool in a clever way, creativity will result. Unfortunately, we don’t live in a world withoutcontext. This is not the platonic realm of ideas, this is the dirty, complicated world ofreality, where things have nuance and circumstances outside the nature of the tool mayconspire to render the tool creatively useless. Because here’s the simple fact: AIart tools as they exist today are universally created with stolen labor.And I’m not using that as a legal term,.

I’m sure some slimeball lawyer in a fancysuit can point to a technicality in some badly written copyright statute fromthe 1830s that makes it not a crime to steal a hundred million pieces of artwork for adatabase, somehow. I’m using it as a moral term. The ONLY reason AI art tools work is becausethey have enormous databases of previous art to draw upon. This is the fuel that powers them,and those databases were compiled from millions of images that were used without permission. AIart tools as they currently exist are a means by which you can extract the benefits of millionsof man-hours’ of other people’s labor, without their permission, without having to pay them,and without adding an ounce of effort yourself. AI art tools, at their core, cannot createanything new – they are converters. They.

Convert the collected effort and labor ofpast work into new forms. That past work, those millions of images, those are thefuel. They are the thing that powers the whole operation. Without those images to drawon, these algorithms cannot produce anything. Without the labor of artists to createtraining data, there can be no AI tools. Without the training data of work doneby traditional artists, there is no prompt in the universe that an AI artist can write thatwill cause the images they want to appear. All the creativity that is possible with AI arttools is possible due to the creativity of the artists whose work form the basis for the tool’sfunctionality. No training data, no algorithm. With the currently availabletools, AI artists cannot be.

Creative. No matter how good their ideas,no matter how creative their prompts, the only possible output from afuel tank of stolen creativity… is plagiarism. And I think we generally allagree that plagiarism isn’t creativity. Even if we imagine an AI generation toolentirely based on public domain data, however, there are intrinsic limitationsto these machines. As I said, they are converters – the most an AI generation tool canever do is rearrange the data it has access to into a different pattern based on a prompt.The absolute highest quality of work it can produce will therefore always be just a littleless than the quality of what is fed into it, but more importantly, this means it inherits everysingle bias already present in its training data.

So if, for example, its training data contains asubstantial majority of light-skinned characters, the model will always assume that the most likelydesired result from a prompt is lighter skin, because that trait is disproportionately weightedin the data. Some people have already found that even if they ask image generation tools fordark-skinned characters, the tools will make the skin as light as possible, because that biasis present in the training data. This problem extends to everything – anime art for example hasa tendency to default to larger eyes for female characters and smaller eyes for male characters,so if you ask an image generation tool trained on anime art for male and female characters,these are the tropes it will reproduce. The end result of that is homogenization, image generationtools, simply due to the way they are constructed,.

Will regress towards the lowest commondenominator, always. In simplified terms, whatever is most common is assumed to be most desirable.Whatever is most popular will be most copied. I hope we can all agree that regressing tothe lowest common denominator is the opposite of creativity. Creativity is supposed to pushand expand the boundaries of human imagination, and image generation tools by their natureare designed to produce means and averages. They're designed to copy what already exists. And this isn’t a problem you cansolve by simply giving the algorithm more training data – that trainingdata has to come from somewhere, and if the sources of training data have abias, if they have limitations or problems,.

It doesn’t matter how much of it you feed into thealgorithm, those problems will persist, always. Those inherent limitations to creativity will exist, always. Anyway. We’ve spoken a lot about creativity, butin doing so I feel like I have been giving these tools a little too much credit. The marketingaround AI art certainly emphasizes creativity, but that’s not really what these tools are for.The purpose of AI art is a lot more practical, and a lot more pedestrian, because the truth is… This is a power loom, or an automaticloom. They were first invented around 1785, and are one of the big and iconic creations ofthe industrial revolution. The textile industry was a major hotbed of invention and advancementin production machinery, and the development of.

Automatic looms formed the basis for manyother machine automation projects to come. Weaving was traditionally a very laborintensive job, and a very skilled one. Memorizing the steps and movements necessaryto turn wool and fibers into woven cloth was a process of learning that could take alifetime to master, and there were huge deltas in quality and value between the workof a good weaver and the work of a novice. With the introduction of the automaticloom, the quality of work was standardized, and the amount of labor required to turn a givenvolume of wool into a given volume of cloth plummeted. As a result, not only did weaversand textile workers rapidly lose their work, those who were able to retain jobs working in thenew textile factories were subjected to brutal.

Working conditions and systemic abuse. Factoryowners were safe in the knowledge that because tens of thousands of people had lost theirjobs, if any worker started complaining they could simply be fired and replaced by the nextstarving bastard willing to work for lower pay. In response to these intolerable conditions,workers began to organize and engage in both political activism, petitioningthe government for regulations, aid and legal remedy… and they began toengage in strike actions, including breaking into factories and physically destroying themachines that threatened their livelihoods. In England, these striking workers began to callthemselves by the name of Luddites. If you know that term today you know it as a mocking insultthat describes a “backwards, technophobic idiot,”.

An obstinately primitive reactionary who hatesscience and progress and resents all technology. The real history of the Luddites,though, is very different. The Luddites didn’t smash machines becausethey “hated technology,” they didn’t do it because they were upset about how kids beon their phone nowadays. They did it because the way that factory owners and the wealthyclasses USED the machines, put their actual lives at risk. These were people who spent years,entire lifetimes, learning to be good at weaving, good at treating textiles, good at doingtheir craft. They established communities, culture, shared histories, folk songs,and traditions, and the livelihoods and survival of their families were based on theexpectation that if they put in hard work,.

If they learned a difficult and valuable skill,and made that skill available in the market, then they in turn would be rewarded witha fair day’s pay for a fair day’s work. And then suddenly a machine could doit cheaper, and their human labor, their human craft, their human lives,became surplus to requirements. There isn’t anything inherently immoral or evilabout automation and machinery. They’re just tools. But in the hands of capital, thesetools were used to destroy communities, to destroy livelihoods, destroy cultureand traditions. They were used to throw human beings into precarity andpoverty, for no other reason than some dickhead factory owner wanted to geta little bit richer a little bit faster.

Maybe you think the benefits are worth it, inhindsight. Like yes a lot of people suffered, but it was temporary, and future generationsreaped the benefits. It’s a valid position to take, and I’m wearing a cheap t-shirt fromChina too, so I don’t exactly have a high horse from which to be sanctimonious about it.But if you were there, back then, and you saw it happening in front of you, you saw the desperationand the impoverishment and you saw this part of society being wiped out in front of you… would youhave turned up your nose at them? Would you have said, “this is gonna be worth it in 200 yearswhen people can buy a 5-pack of t-shirts for 20 dollars on Amazon”? Or would you have thought,hey, maybe it would be good if we came up with a way to do all this technologicaladvancement without hurting people?.

Wouldn't that be better? I bring all of this up becauselike the automatic loom before it, AI art is automation. This story has playedout before, and we know how it’s going to end. As I’ve said, AI art is a morally neutral tool.But so is a hammer, until you use it to bash in someone’s head. Corporations will use machineart as an excuse to destroy the value of creative labor, to drive down the costs of production,and to lay off people by the thousands the moment it is viable to do so. And they will do it for no otherreason except that some dickhead CEO wants to get a little bitricher a little bit faster.

If you want to imagine the benefits of this newand amazing tool, you also have to remember the people whose livelihoods are going to getcrushed by it. People are going to be forced to give up their businesses, give up theircraft and give up their dreams. People are going to go into debt because of this, peopleare going to lose their homes because of this. Not because AI art is evil,but because it creates a set of incentives which the amoralworld of business will follow. Genies generally don’t go back in their bottles.AI art is here, it comes with a host of problems, and we’re gonna have to deal with that. Sowhat are the solutions? Well first of all, I don’t fucking know. I’m a YouTuber with a historydegree, not a policy expert or a copyright lawyer..

It’s gonna take a lot of effort by a lot of smartpeople to figure out this shit, and whatever solutions are implemented are necessarily goingto be imperfect. That’s just the way of the world. But if I have to give some suggestions… okay. Thefirst and most obvious is for the training data to simply be licensed. The creators of the tools canreach out to the human beings whose labor forms the core of the functionality for their algorithm,and pay them fairly for the labor it took to make them. That would at the very least make theuse and deployment of these tools somewhat ethical. That would at least compensate people fortheir work, and respect their right to consent. The problem of course is that under thosecircumstances, the only entities with the money and IP library to create truly powerful AIart tools are going to be… corporations. Disney,.

Universal, Fox, Netflix, Tencent… all thecompanies that already monopolize culture will be granted an even more powerful tool tocontinue that cultural monopoly. Copyright law is already the personal hobby horse of theDisney Corporation in the United States, and there is no reason to expect that they willabuse the concept of copyright any less once they become able to infinitely replicate the workof every person who has ever worked for them. With robust regulation, they might be madeto pay royalties and licensing fees to the artists whose work they exploit, which is betterthan nothing, but anyone who imagines that these tools will spur the vanguard of some kind ofpopular revolution of radically free creativity is absolutely fucking dreaming. Broke collegestudents with big ideas and creative hearts.

Are not suddenly going to be taking on Hollywood,because even if these tools theoretically allowed one person to do the creative output of a hundredpeople, any studio could hire a hundred people and get the equivalent output of ten thousand, plusa marketing department, plus an international publishing arm, plus an army of lawyers to crushall competition before it even gets started. These tools will not “democratize” anything.They aren’t fucking magic. They will be subject to the same market forces and the samematerial conditions that everything else is. If left unregulated, thesetools are and will always be based on theft. All of the economicincentives guarantee that they must, and the corporations will be the biggestthieves of all. WITH regulation, well,.

They’re as vulnerable to corporate corruptionas anything else is, but we might have some hope that they don’t catastrophically destabilizethe entire creative labor force. We might have some hope that they don’t shove the value ofcreative labor off a cliff and into a landfill. But okay, I am being generally very cynical here.Do I see any positive potential in these tools? I can try. Let’s say a hundred artists get together in aunion, and all agree to pool their collective work into an image generation tool. Everyone inthe group can use the tool to help them work, and thus everyone benefits from the advantagesof everyone’s collective labor. In return, everyone pays a nominal union due, the money fromwhich is used to do maintenance, provide member.

Services, and retain whatever lawyers are goingto be required to keep on top of legal issues. Now, that’s just an imaginary idea I pulled outof my ass, but the point I’m trying to get at is that if these tools have a positive use, thenit has to be based on the consent and mutual benefit of the people whose work gets usedto create them. It cannot be based on theft, and it cannot be based on the exploitationof the many for the benefit of the few. More than anything else, this is what Iwant you to take away from this video: All the blue-sky promises and tech-utopiangarbage vomited up by Silicon Valley jackoffs is noise designed to distract you from avery simple truth: nothing gets made for free. AI art tools do not summon as if bymagic an infinite spring of new creativity;.

They simply repackage the work of other peopleand pretend it came from nothing. They are based on the stolen work of thousands of people who werenot asked, did not consent and have not been paid for the value they now help to create. OpenAI didnot fashion ChatGPT as a naturalistic language chatbot with the sheer magic of programming, theyexploited the brutally underpaid manual labor of real workers, to filter their languagemodel into a product they could sell. It is always a mechanical turk. There isalways exploited labor hidden under the table. If we are going to live in a world of AI art andimage generation, if we are going to somehow make these tools into a net positive for the world,we must stop thinking of them in the terms sold to us by people who stand to be billionaires ifthey become normal. We must always, always ask:.

Whose labor was used to build thistool? And who could it be used to hurt? Think hard, be compassionate, and havesolidarity. The only means we all have to protect ourselves from abuseis the support of the person standing next to us. Anyway, I don’t want to advertise anything today.Support the channel if you want to, please don’t get into stupid fights in the comments, and maythe tides of history wash gently over us all.

Sharing is caring!