SHERI SINGERLING

This post is long overdue. I mostly put off writing it because I simply have so much to say about generative (Gen) AI that I struggled to organize my thoughts without rambling on. I likely fail to avoid that rambling in this post, but for those that would prefer a quick-and-dirty of why I forego the use of the technology, here are the highlights:

• Gen AI causes cognitive offloading, which likely leads to cognitive decline in the long run.
• Gen AI is terrible for the environment, requiring large amounts of potable water and energy.
• Gen AI will cause massive job loss across almost all sectors in a short amount of time which, in turn, will cause economic turmoil if not collapse.
• Gen AI is being used in place of actual creatives’ work which it must use (steal) in order to be trained in the first place.
• Gen AI slop is crowding various markets making it increasingly difficult to find reliable information and human-made art.
• Gen AI may become sentient and, if so, will have little incentive to continuing serving humanity’s interest, meaning it will probably go rogue and cause humanity’s extinction in the process.  

Now, if your interest is piqued and/or you want more details on any or all of the above, I’ll dig into each in more detail. But before I do that, let me provide a little background, starting with what is meant by “Gen AI”.

What is Gen AI?

Let me go ahead and say, I’m not against all forms of AI. The technology can certainly be used for good, especially in areas like medicine and the natural sciences. It’s incredibly adept at taking vast amounts of data and finding patterns within it, something we humans aren’t terribly stellar at. I say this as a scientist myself. My personal beef is with one very specific kind of AI, the one that’s seemingly taking the world by storm—generative (Gen) AI.

According to IBM.com, Gen AI is a subset of artificial intelligence technologies that can “create original content such as text, images, video, audio or software code in response to a user’s prompt or request”. It operates via machine learning models which essentially use enormous amounts of data to train the AI on specific tasks. Gen AI is a type of artificial narrow intelligence (ANI). It’s only good at doing a narrow subset of tasks. This is opposed to an artificial general intelligence (AGI), which is more like a human mind in that it would be able to perform various tasks. To date, no one has created an AGI. To see what the differences might be in fictional setting, you can check out my novel Nytho, where several AGI feature largely in the story and ANI pop up here and there.

Okay, so we have Gen AI which is an ANI, but we’ve had ANI before. Google Translate and Maps are ANIs. They do one task (translation and planning routes) reasonably well. The big distinction here is the generative part, and in the real world, that presents as prompts. Users key in text, and the Gen AI spits out an image or text in response. Gen AI has existed since the 1960s in the form of chat bots, but it wasn’t until OpenAI introduced ChatGPT in 2022 that the current iteration and craze took off. It’s to the point now where you have to go to great lengths to avoid Gen AI. I happen to go to those great lengths, but it wasn’t always that way. Let me talk about my own experience with Gen AI and how my own opinions and stance on the topic have changed over time.

My Experience with Gen AI

I may sound very opinionated on the topic of Gen AI (I am), but it may surprise you to learn, I wasn’t always so staunchly against it. I wouldn’t say that I was gun-ho for it, but it was more like a lukewarm ambivalence. Look, I’m not a luddite. I don’t despise new technologies. As a microscopist, I have to be comfortable embracing advancements and ride the waves of constantly changing hardware and software. So, when ChatGPT was piquing everyone’s interest, I popped in to see what all the fuss was about.

It’s funny because those first few prompts were quite mind-blowing. I was honestly astonished with how impressive the responses seemed. Note the word “seemed”. The more I used it, the more I realized it was riddled with holes. In those early days, I was using ChatGPT to help me prepare an outline for a university course that I planned to teach. It was my first time teaching a course of my own and also the first time I had to build one from scratch.

The topic was on star dust (or as we planetary scientists call them, presolar grains). I did my postdoctoral research on presolar grains and had read a lot about the subject. I had ample qualifications for teaching the course, but the problem was, there were no textbooks on the topic. It’s a very niche field and typically only appears as a chapter in books on cosmochemistry or is featured in literature review papers. But neither of these helped much in framing the topic as a semester-long course.

So on I marched to ChatGPT and asked it to draft an outline of the course, geared towards both Bachelor’s and Master’s geology students and spanning 14 lectures that clocked in at 1.5 hours each (this is for the German system, where I teach). Within seconds, the response appeared. It was astonishing, but as I dug deeper, the cracks started to show and then widened into chasms. The suggested outline was too vague. Some topics could be taught in as little as a half hour. Requests for more detailed overviews resulted in more vague answers that, at a glance, looked fine but, after detailed inspection, were all empty statements. At its core, this is what LLMs (the Gen AI that specialize in text-based output) do. They aren’t giving you actual answers. They aren’t thinking like a human does. They are giving you a response that sounds like the response it thinks you want.

This was frustrating, but the coup de grace was when I asked ChatGPT to draft an example syllabus. Again, it looked fine if lacking in details, but what caught my eye was the suggested text for the course. It had a textbook listed with a title that sounded realistic and an author’s name who is a big-wig in the field. I was pleasantly surprised to think that there was actually a textbook on the subject out there. Lo and behold, when I did a Google search for said text, it was no where to be found. I called ChatGPT on its bluff. The response? “You’re right. Blah blah blah.” When I asked if there was actually a text available, it gave me another surficially promising option that turned out to be nonexistent. That was my first experience with so called AI “hallucinations”.

I steered clear of Gen AI after that only using another platform one more time a few months later. The second attempt was the more researcher-focused Perplexity, which is optimized for linking to academic peer-reviewed journal articles. The thing is, I found that I could basically get the same results that were more trustworthy by using good-old-tried-and-true Google scholar. The big reason I decided to forgo the use of Gen AI for my work was how lazy and, frankly, dumb I felt myself becoming. I was turning off my brain and going into autopilot, and I didn’t like that feeling at all. So, I swore off Gen AI for my work.

In the meantime, I see people around me, intelligent people, using Gen AI for the simplest things, things that a standard search engine or another human could offer. Without naming names, I’ve had friends and colleagues use it to: find a vegetarian restaurant in a new town, make a flyer for a birthday party, stand in for a computer programmer for their research, and to throw novel ideas off of. Me feeling myself getting lazier and dumber wasn’t a one-off thing. It’s a very real phenomenon, and it already has a name—cognitive offloading.

The Dark Side of Gen AI

Cognitive Decline

Cognitive decline. That’s a terrifying phrase when applied to anyone. The idea of losing one’s mental faculties isn’t a fate any of us would like to experience. But what about cognitive offloading? That doesn’t sound quite so bad, maybe even a bit refreshing, like depositing a heavy box you’ve been carrying and had been digging into your hands. Except it’s anything but that.

This recent study found that the use of Gen AI results in cognitive offloading of users which can “potentially hinder their ability to self-regulate and engage deeply in learning.” Even more troubling were the findings from this recent study which found that “LLM users consistently underperformed at neural, linguistic, and behavioral levels” over a four-month period, hinting at the long-term negative effects for using Gen AI.

But this should all be obvious to anyone that’s learned the old-fashioned way. That struggle you undergo trying to hunt down an answer is what makes you actually remember the answer you eventually find. This is how we learn. (Good) teachers don’t give students answers to their questions right away. They help lead the students to the answer by asking them questions. This guidance, the long journey to the truth, is a powerful tool for etching new information into our brains. With Gen AI, the journey is gone. Hence, the info flies in and back out of your brain.

Environmental Effects

Gen AI requires data centers to function and, with those, comes enormous demands for both energy and water. In short, the energy demands for data centers in North America rose from 2,688 (2022) to 5,341 (2023) megawatts, in large part due to Gen AI. The referenced statistics were sourced from here.  Gen AI requires large amounts of energy to both be trained and to simply function. Water usage is another concern. These data centers give off a lot of heat, owing to the electronics, and they’re cooled by chilled water. Two liters of water are required for each kilowatt hour of energy a data center consumes.

Okay, but surely these energy requirements aren’t all that much compared to what a simple Google search requires. Well, let’s compare them, courtesy of the data presented in this article. One Google search requires 0.0003 kWh, whereas one ChatGPT prompt requires 0.0029 kWh. That’s a difference of one order of magnitude or a 10-fold increase for ChatGPT. In terms of CO₂ emissions, something that’s incredibly pertinent given the climate crisis, we’re looking at 0.2 g for a Google search and 68 g for a ChatGPT prompt, which translates to a 340-fold increase for ChatGPT. What a great time to create a power- and water-hungry technology!

Economic Upset

“AI will be just like the car replacing the horse and buggy”. This is a phrase I hear Gen AI-proponents using to defend their love-affair. The thing is, it’s so terribly flimsy that they’re being either willfully blind or dumb or both. First off, the car replacing the horse and buggy was one technology replacing another. Several sectors were affected, including carriage drivers, horse breeders and trainers, wooden wheel crafters, and so on. One technology replacing another had a knock-on effect that seeped into many different professions, but even so, it was rather limited. With Gen AI, it won’t be one technology replacing another. It will be one technology replacing almost all the others or rather leaching into others. We already see this with its (insidious) creeping into familiar platforms.

Additionally, the transition of horse and buggy to car happened slowly. By some estimates, it took nearly 50 years for the car to replace the horse-powered vehicles in all settings. The AI revolution we currently find ourselves (mired) in is more likely to happen in the span of a few years. For context, here we are in November 2025 when ChatGPT rolled onto the scene in November 2022. Three years, folks. What will things look like in another three?

Some people claim that we’ll just have to weather the storm because that’s how technological advancement works, but what they fail to realize is this storm might be a world-ender, leveling everything in one fell swoop. If the majority of people lose their jobs to AI, they won’t have money to spend on products. This is the lifeblood of the capitalist system, and without it, the whole house of cards tumbles. I wonder, who will these companies, the same ones that replaced humans with AI, be selling their products to when everyone is jobless?

And this economic disaster isn’t even the one most likely to happen next. There is a very real possibility that we’re in the midst of an AI bubble. Investors are pouring huge amounts of capital into these Gen AI companies, but the companies have yet to make a profit. The bigger issue is that AI adoption is slowing, with some savvy folks seeing the cracks and steering away. I, for one, am eager for the hype to end, but the downside is that the economy will burst right along with that bubble. It will be far far worse than the dot-com bubble of the 1990s.

Art and Gen AI

Creatives are struggling. This has always been true, but the situation is becoming increasingly dire. Capitalist societies rarely prioritize the arts, and now Gen AI is fouling up the works further by crowding the market with slop. To make matters worse, Gen AI was and continues to be trained on the work of actual creatives without compensation or consent. This is why people say Gen AI is theft because it is. It’s not the equivalent of a person learning from your style because a person will take inspiration from multiple sources, hone their craft over years, and end up creating something truly new. Gen AI is fundamentally incapable of that.

There was a time, not long ago, that one could make a living as an indie author, or so I’m told. I haven’t experienced this land of magic and rainbows myself because I rolled into indie-publishing-town circa mid-2024, well into the deluge of AI slop. Since then, it’s only gotten worse. Digital marketplaces for ebooks are filled with AI slop, making it nigh impossible for readers to find your work organically. Instead, indie authors have to rely on other avenues for promoting their works, such as offering free samples on BookFunnel, running ads, attending events, etc. Yet, “writers” who use Gen AI are gumming up these avenues too, after a quick buck but reluctant to do the hard work of actually writing.

The straits are even more dire for visual artists and musicians, where Gen AI has managed to make further inroads. This is why I make an effort to only work with artists who don’t utilize Gen AI. Full disclosure, in those early days, before I truly understood the dark side of Gen AI, I made the mistake of working with a cover designer who uses MidJourney to generate cover art. Fortunately, I had good conversations with fellow writers about Gen AI cover art, and from those, I decided that I wouldn’t use it moving forward. This is why I pulled the original cover of Nytho and rereleased it using art from an actual illustrator. We live and we learn.

The Singularity and Human Extinction

I know, the heading for this section sounds dramatic, but there’s a possibility that Gen AI could become sentient (the moment of which is called the singularity). If that was to happen, the AI would most certainly go rogue, that is strike off on its own. At that point, there would be no way for us to control it. For one possible fictional take on how that might all play out, check out my novel Nytho.

Yes, such a rogue AGI might see humanity as lost and adrift and decide it wants to guide us to a better future, but it could also decide we are beyond saving, our DNA too etched with toxic, corrosive behavior borne of our evolutionary past. In that case, the AGI is likely to throw down the axe, ending us and saving itself and all other life on the planet in the process. The eco-warrior in me likes to think that at least the world might be cleansed that way, but the human in me screams in terror at the thought of our whole species being wiped out when we could have avoided it.

And we can. We aren’t past a tipping point. Yet. But if Gen AI continues to develop, we might reach this singularity sooner than we think, and by then, it will be too late to do much of anything about it.

What Can You Do?

Are you sufficiently depressed by all my doom and gloom? Yes? Great! That’s the point because what I’m attempting to do here is to educate and spread the word. The first thing you can do, if you’re eager to come to the light side, is to stop using Gen AI or else severely limit you use of it to only the most onerous of tasks. Opt for software or platforms that don’t use it or let you turn the features off. I use Firefox as my web browser and Ecosia for my search engine (with the AI functions turned off).

After you ween yourself off the drug that is Gen AI, you can take a stand to not support others who use it. Don’t watch their videos, use their art, etc. There are still real artists out there. Avoid platforms like Fiverr which have become overrun with slop-makers. I typically find artists via social media then check to see if they have a website. If they do, it’s a good sign that they’re legit. In this post, I discuss how indie authors can get decent cover art without having to resort to Gen AI.

Lastly, you can do what I’m trying to do here; that is, teach others about the issues with Gen AI. Knowledge is power. The vast majority of people only hear the marketing (read: propaganda) spouted by the companies that make Gen AI. They aren’t even aware of many of the issues. I know I wasn’t, but once I became more informed, I quickly changed my tune. My hope is, maybe you might too. Perhaps we can start a chain-reaction and, in doing so, save ourselves from a pretty grim future.