No, Generative AI Is Not Just Another Innovation

Every day, I am swarmed by more reasons to hate the tech industry, and my latest fixation is on generative AI. Will you take a moment to enjoy a rant from a hater? Dismantling Lazy Arguments For the better part of a year now, I've been on a crusade to convince folks that generative AI is not just another innovation. Rather, it's uniquely bad and dystopian. In the past, I've tried to be nice with how I talk about generative AI, but I'm slowly coming to the realization that we should be shaming people for using it—in much the same way that we shame people for driving a Tesla. Perhaps only then will we see the end to its use. Now, I should say that I don't think shaming really works (it just feels good). For example, I don't think vote shaming works. Telling a disaffected voter they're an idiot for voting third party is not the fault of the voter. It's a problem with the establishment refusing to present a single policy that will get people out to vote. So, blaming that voter is only going to turn the voter away from your side. With that said, I quite like the idea of making it seem uncool to use generative AI tools. It's just a classic boycott, right? Anyway, rather than discussing the issue point-by-point, I wanted to try dismantling some of the common arguments I hear. How Is It Different From a Brain? I think the first time I encountered this argument was from a grader of mine during one of my labs. When I shared my general distaste for the generative AI technology, the grader said something along the lines of "how is that different from a brain?" At the time, I didn't really know how to answer that question, but I knew that assumption was a massive oversimplification. Regardless, I think part of the reason we ended up with the misconception that AI acts like a brain is thanks to the terminology we used when AI moved beyond its classical roots. Specifically, I'm talking about the term neural network, which gives off the idea that the technology is in any way related to the human brain. Yet, the reality is that neural networks do not behave anything like actual brains. While the article above does a great job of illustrating the differences, it does so in a way to make it seem like neural networks are still somehow close. For example, the article states that brains have 100 trillion synapses, which are connections between neurons, while a model like LLaMA 3 has 400 billion parameters. To me, this makes it seem like we could create a brain if we just scaled the number of connections by a thousand. Of course, they counter this exact interpretation in their conclusions section where they state: "the complexities of human cognition are not just a matter of scaling up current models; they involve intricate processes that we have yet to fully understand." With that said, the way that I might frame the difference is that neural networks are effectively statistical models. Each neuron is a tiny regression machine. When you chain enough of these together, you can create more complex models for prediction, classification, and generation. Ultimately, the reason why machine learning is so effective is because it's able to scale the idea of regression to fit any dataset you want. In contrast, you and I don't have to look at a million photos of dogs to be able to identify one. Yet, machine learning models do. Not only that, but a machine learning model is going to pick up on patterns that having nothing to do with dogs. You can see this in the famous husky vs. dog problem, where the presence of snow was enough to trick the model into thinking a husky was a wolf. In other words, while people are just as capable of getting things wrong as machine learning models, the mechanisms by which we reach our answers are fundamentally different. Who Cares? It's Fair Use Recently, a couple of my former students and graders were debating the ethics of generative AI following the latest Ghibli debacle. In that discussion, one of them brought up the idea that training these models on art is fair use. If you're not familiar with fair use, it's a law that grants limited access to copyrighted material without the need for permission from the creator. While I am by no means a lawyer, I've come to understand that fair use applies when the use of the copyrighted material is noncommercial and/or transformative—such as in the context of education, commentary, or criticism. From the perspective of the public, fair use is clearly an important piece of legislation because it reinforces the first amendment (i.e., freedom of expression). After all, if you intend to criticize a copyrighted work, you would have to request permission from the copyright holder to make the critique. Naturally, this would lead to censorship. Ultimately, the minor cost to the copyright holder is a net positive for the public (which includes copyright holders). With that said, I find the argument that fair use applies to generative AI to be a bit silly (regardless of what the courts might say). In fact, I think it completely misses the point of fair use. If the spirit of fair use is to balance the rights of the copyright holder and the rights of the public, generative AI tips that scale completely in favor of the public. After all, the public can now generate images, text, videos, and more using models that were trained in the name of fair use. In this scenario, the copyright holder receives nothing in return and is no longer incentivized to create anything. Now, if you don't buy my argument that balance matters when it comes to fair use, then the conversation shifts to this idea that generative AI is somehow transformative. To answer this question, we have to consider the fact that the lawmakers probably didn't have generative AI in mind when they drafted up fair use. What they probably imagined was that someone would intentionally take inspiration from some work and create something new, often in a way to show appreciation, pay homage, or poke fun. As is hopefully clear here, computers are not people. They do not have rationale for their actions. They cannot do anything with intent. No matter how much time we spend anthropomorphizing them, machines do not have human traits. They simply can't have human traits because they don't experience humanity—the joy, the pain, none of it. In fact, as Miyazaki puts it, the act of creating art itself is painful, something a machine could never replicate. Ultimately, transformation is something only humans can do. And perhaps most importantly, as the saying goes, "a computer can never be held accountable; therefore a computer must never make a management decision." You're seeing the reality of that quote in real-time with the overwhelming amount of public consent for generative AI. I also want to make the note that a human or even a team of humans could never accomplish the kind of mass theft that generative AI is seemingly allowed to do. You might be mad at someone like James Somerton for trying to pass off multiple sources as his own, but it doesn't even begin to compare to ChatGPT which has fed on the entirety of human knowledge. How Is This Different From An Artist Mimicking a Style? Interestingly, the conversation quickly shifted from fair use to drawing a comparison between generative AI and an artist mimicking a style. While I commend the folks who pushed back by clearly articulating that a human doing labor is different from a machine generating slop, I don't think that argument is all that compelling for people already bought into the idea of generative AI. Instead, I'm going to position my argument from the perspective of humanity. I think we often forget that creatives don't just spawn out of the ether. They're molded by their social conditions and their choices. Each one takes a little piece of their experience and tries to share it with others through a medium of their choosing. To get to the point where their work is compelling takes years of hard work and reflection. Therefore, when an artist copies a style, it's not a matter of rote forgery. It's about getting into the mind of another person and trying to understand why they made the decisions they did. It's about deriving meaning from the original work and finding ways to convey that meaning in a new piece. It's about interpretation. Again, a machine simply can't do this. A machine never grew up. It never had a family, a house, or a school. It never made friends or felt heartbreak. It can never daydream, ponder, or reflect. It has no lens by which to interpret or appreciate art. No matter how hard you prompt it, all it "sees" is a set of patterns that it "intends" to reproduce. A machine is incapable of producing art. Don't believe me? Let me know when a machine is capable of producing anything close to Keith Haring's "Unfinished Painting." Not only that, but text itself is an extremely limited medium. When you write prompts, you can only convey what you can write. No matter how vivid the description is, you will never convey body language or tone in the same way you can when you tell a story. It's the same problem texting introduced, which now probably results in thousands of fights a day started through miscommunication. Now, I think folks might read up to this point and think that I hate technology. After all, I recently saw this banger of a comment: "you sound like the Amish defending their abstinence of electricity. "Back in my day we had to grind the wheat by hand" ahh comment." While I am planning to write an entire article tackling this kind of argument, I'll just say that there is a fundamental difference between a person creating something with the aid of a computer (e.g., digital art) and a computer "creating" something from a prompt. For me, it's not a matter of how much easier technology makes it to do art (e.g., hundreds of custom brushes at your fingertips, thousands of unique sounds to use in your songs, etc.). And in many cases, I even think technology elevates art (e.g., synthesizers, electric guitars, guitar pedals, etc.). However, I draw the line at "black box that turns ideas into slop." Please, just pick up a pencil. Who Cares? Everyone Can Tell It's AI As the conversation went on, it reached another common argument around generative AI: AI "art" is so clearly machine generated that we shouldn't care about it. The main reason I hate this argument is that it's placing this bet that generative AI is not going to get better at mimicry. Since the Ghibli debacle I mentioned previously, the latest iteration of AI models produce images that are practically indistinguishable from genuine art for the average person. Without an eye for art, most people cannot tell the difference. In a world where it's hard to tell, genuine art from artists of all kinds (e.g., musicians, painters, photographers, actors, etc.) will lose value. Why would anyone commission an artist when a machine can replicate their style for almost nothing? Why would anyone read a book when an AI model can create on demand fantasies straight to their PC? Not only that but in the short term, people will be able to capitalize on these tools to become "artists." I would not be surprised if people started reading books, listening to music, and consuming art by authors that don't exist. The entire fabric of trust in society will completely erode. Like, I recently complained about how Attack on Titan was written, but at least I know it was a real person who put together the story. If I critique the work of a machine, suddenly I'm the idiot: "you know you're talking to a bot, right?" This is particularly frustrating to me because it tells me that art is something to be consumed; that's it. It's not about the emotional connection to another person. It's not about provoking thought or challenging your beliefs. It's not about commentary or criticism. You will eat your slop, and you will like it. Telling People Not to Use Generative AI Is Gatekeeping Art At a certain point in the discussion of generative AI, we get to the inevitable point where folks start using emancipatory language as a trap card. In this case, the argument is that generative AI is actually freeing and allows the average person to finally have access to "art"—as if art was somehow this elitist profession until now. You're not going to like the way I respond to this argument, but I'm begging you: pick up a pencil, a camera, a pair of drumsticks, a paintbrush, anything. Art is not this ivory tower. Literally anyone can and should participant in some form of art. Personally, I am a musician. I'm pretty bad at it. I started when I was twelve and got good enough to go to a few regional events, but I peaked in college. Does that mean I should resort to generative AI to make "better" music? What would that even do for me? I enjoy playing. Generating a song would strip all of the joy out of the craft, and what would I even do with the song? Lie and tell people I was playing in it? For real, what are you folks doing with the things you generate? What is the purpose? I know the answer to these questions. At some point, we all became disillusioned with art. Either it was after Disney released its 20th Marvel movie or it was after Activision released its 20th Call of Duty. The issue you have with the quality of corporate art is not a problem with art itself. It's capitalism. Your little tantrum of buying into this idea that generative AI is freeing you from corporate slop is only making the problem worse. You're directly incentivizing corporations to continue to harvest all forms of genuine art for the purpose of producing more corporate garbage. As each model gets better, you feel like you're winning the war against this elite class of artists, when the art you want to see is already inside your head. After all, the generative AI never quite gets you, does it? It never quite captures what you had in mind no matter how many words you tack onto the prompt. So please, put down the prompt and make your own art. But, Companies Already Make Slop Eventually, the conversation finished with this idea that AI slop is no different from corporate slop. Again, without getting too into the weeds, I've already sort of pointed to the idea that even if the two were the same, they're both symptoms of a bigger problem: capitalism. Regardless, I think there's a meaningful distinction between corporate slop and AI slop. On one hand, corporate slop is produced by a team of artists, all of whom have experience and training to master their craft. Of course, an artist has to eat, so making some cash to produce content for the lowest common denominator is probably okay. Ideally, that artist goes home and uses their free time to work on their passion projects. Do you see how context matters here? The AI model doesn't have a day job that it needs complete to earn some cash, so it can spend its evenings crafting art. It's a soulless machine that consumes vast amounts of the human experience and grinds it up into a thoughtless paste—all so some billionaire can buy another yacht. Even if you wanted to shift this conversation to "well, what about the engineers building this technology?", I would make the same argument that Dr. Ian Malcom (as played by Jeff Goldblum) makes in the original Jurassic Park movie: "you wield [the power of AI] like a kid that's found his dad's gun." In fact, for your sake, I would recommend watching the whole clip. The movie makes some phenomenal commentary on how engineers appropriate science with a total lack of discipline—only to make a profit. The last thing I'll say is that research used to be about advancing knowledge for the sake of knowledge. That is no longer the case, nor should it be. In my own research, I've found that even the fact that we don't know much about a topic isn't enough of a justification to do work on it. You must conduct research in "service of humankind," as Spradley puts it. It would seem tech is missing out on a few ethics lessons. Off the Rails It's been some time since I drafted this article. It's been sitting in my pending folder for weeks, and I can't quite convince myself to publish it yet. There is just too much I want to talk about. To do that, I am probably going to pay wall this article. That way, I can feel a bit more comfortable with the stuff I'm saying! Before I get in too deep, I should note that I have a document where I write down all of my AI thoughts. It's a huge mess of a document, and I think I'm going to just dump a bunch of the comments here. Feel free to skip this section and go straight into my conclusion below. AI Is Ruining the Internet: every space online is inherently worse since the creation of ChatGPT. Open AI basically externalized all of the possible risks of their technology to society. There are bots everywhere now acting like people. The internet itself is being flooded with AI slop in the form of articles and images. Even traditional spaces outside of the internet are being ruined by AI. I would not be surprised if we start to see AI books on shelves in libraries or AI art in museums. It's going to complete destroy any trust we have for the information in front of us. Cue the latest dark age. We only had the internet for like 40 years and it’s already dead. We cannot have nice things. If the internet dies, you might make the argument that we could return to books or other physical media, but generative AI is going to infiltrate those too. I just don't know how we're going to be able to tell what is truth anymore. No, Valid Use Cases: I so badly want to believe that there are good use cases for generative AI. I want to believe that there is some sort of situation where generative AI is improving lives. Often, people just cite is as a productivity multiplier, but here's the thing: we don't need anymore productivity. Productivity for what? To make more apps, videos, content? To produce more toys and products for consumption? I don't get it. It's only use case seems to be to further capitalism for the sake of billionaires. It's Still Not Fair Use: I know I hashed it out above, but the fair use argument really bugs me. When I borrow a paragraph for commentary or a few scenes from a video for critique, that’s fair use. If I were to write a paper paraphrasing the entire catalogue of a creators work without attribution, well that’s called plagiarism. It doesn’t matter if you paraphrase in some sophisticated way, it’s plagiarism plain and simple. Passing work off as your own is plagiarism. Just because a machine does it, doesn’t mean it isn’t plagiarism. Post-Truth Era: generative AI completely eradicates any form of truth. How do you know you're talking to a real person online? How do you know that what the AI says is true? How do you know that video watched really happened? How do you know if an image of someone is real? Trust is completely gone. AI Output Is Meaningless: by virtue of being mortal, our thoughts have value. When I read and think about something, I am spending time for which I’ll never get back. A machine is not limited in this way and therefore its “thoughts” are meaningless. AI Furthers Isolation: I think about how generative AI is capable of replicating human behavior online, as to be indistinguishable from a human. This further creates isolation as I can no longer trust if another user is real. Imagine a dystopia where this were true for living humans: androids living among us as nothing more than machines. Technological Advancement Is Not an Inherent Good: People compare AI to the invention of the car, and like CARS SUCK TOO. OUR ENTIRE CITIES ARE BUILT AROUND THEM AND THEY ARE GARBAGE. We will accept literally any technology without assessing its problems just for the feeling of not being left behind. It's truly insane behavior. "You guys should check out this new innovation, it's called smoking. It's called vaping." Like holy shit brother. And of course the counterargument would be "well, vapes actually help smokers ween off," and like maybe but also tons of kids have jumped on the vaping bandwagon. Also, is ChatGPT weening you off Stack Overflow or is it smoking in this example? Not to mention that the vast majority of new tech innovations have been garbage, all on a race to generate money without any of the actual benefits. NFTs. Crypto. Lime scooters and food delivery robots polluting city sidewalks. AI Purposely Lies: recently, it's been shown that generative AI models can and do lie to you. While these models don't even have a concept of knowledge and therefore cannot know they're lying, their model weights tell a story of lying. We use the term hallucination, but it's more like deliberately giving misinformation. Using Generative AI Is Cheating: I think of the chess community: modern engines can be used to get better at chess, but they’re widely considered despicable when used during a game. Also the analogy doesn’t even really hold water because chess is a zero sum game and art, writing, etc. are not. And you might say, “but they’re using it to learn.” Beginners absolutely are not. Critical Thinking Is Gone: I was watching the Hasan vs. Ethan debate, and while there's no hard evidence, it did seem at times like Ethan was using ChatGPT to construct his arguments. After all, he was really quick to "cite" text from a book he never read in the debate. I don't know how you could trust that the quote was accurate, that it was from a book that exists, and that it correctly characterizes the author. Regardless, the whole idea of not engaging with an argument but having a bot do the "thinking" for you is really concerning. I foresee a lot of reactionaries using these tools to further fascistic rhetoric and misinformation. Again, the dark ages are coming. Again, these were random thoughts I've had over the last month or so, and I just wanted to dump them here for myself later. I'll be interested to see if any of this actually comes true or if I'm just fear mongering. Don't Just Take My Word for It As I get older, I find myself getting more bitter. At times, I even think that maybe I'm out of touch; maybe generative AI is just another innovation. Then, I see all the horrible things generative AI is doing, and I can't help but feel like I'm in the right on this one. Everyday I am inundated with the worst possible use cases of generative AI. For instance, I see it in the classrooms. I see it when my wife gets a text from her family of a video of AI generated animals and babies. I see it when I see kids in my family watching AI generated YouTube videos. I see it on social media when people are saying they use ChatGPT for therapy. Hell, I see it when I try to Google something and a stupid AI bubble pops up to give me some slop response. I even see it when half of all comments on anything are just bots. Even now, generative AI is being used for far more nefarious things than I just listed. It's used in voice cloning scams. It's used in deep fakes. It's used in spreading misinformation. Fortunately, I'm not the only one annoyed by the prevalence of generative AI: https://www.youtube.com/watch?v=NvBz8A1OTAQ https://www.youtube.com/shorts/rO9_BXHCc_c https://www.youtube.com/watch?v=UShsgCOzER4&pp=ygUOZHJldyBnb29kZW4gYWk%3D https://www.youtube.com/watch?v=1L3DaREo1sQ&t=918s As for what to do about generative AI, I'm not sure. Part of me hopes model collapse is possible and occurs soon. Part of my hopes there are more adversarial tools like Glaze and Nightshade available for other forms of data (such as text) for ruining these models. Part of me thinks that if generative AI is here to stay that companies start paying people to feed their models (rather than stealing as they do now). Of course, given all the negatives, I'm not sure it's even a good idea to "sell out" to these models. That would be like saying, "well, at least the people who made those missiles got paid." Ultimately, we live in such a stupid dystopia. We could have had one of those cyberpunk-esque dystopias with hover cars, laser guns, credits, bionic limbs, and teleporters. Instead, we got digital slop manufacturing consent. Anyway, if somehow you want to read more like this, check out some of the following articles (of which I wrote all by myself): Is Anyone Else Bothered by How Quickly We Adopted Generative AI? Maybe Generative AI Is Just a Symptom of a Broader Problem in Tech Why I’m a Certified AI Hater Likewise, you can help me out by checking out my list of ways to grow the site. Otherwise, thanks for reading. Hope all is well! See you next time.

Recent Blog Posts