[Post 8] - Art & Activism: Artificial Intelligence
- Dani Romero
- Mar 31
- 16 min read
I know I'd said I'd be mixing up blog posts more and maybe not doing quite as many activism essays, but the unfortunate reality is that the pervasive cancer of AI is quickly unravelling any sense of trust that can be placed in the vast majority of the internet's resources. Search results are based off of the amount of SEO juice implanted into a web page rather than the veracity of the content, and now with search engines auto-applying AI for an instantaneous result which is often incorrect and based off of 'sources' like a random YouTube video or someone's blog post using keyword stuffing, it's becoming more important than ever to try and provide these kinds of writings. It's not only to encourage growth and recognition for actual human artists who create their own work out of training, experience and imagination, but to link to verified sources that utilize peer-reviewed journals and studies to back up the data which generates that result. Not even the government of the United States can be trusted right now as they are taking .Gov websites and turning them into blatant propaganda machines.
The first and most important point I can make needs to be said straight away, and make no mistake about it: AI is a product-tool that is being wielded by the world's largest companies and most powerful, rich people to make money, increase online engagement even if under false pretenses, and to erode educational resources for as many people as they possibly can. You may often hear that going to college 'changes' or 'radicalizes' people. In reality, what is happening is that the education systems set up in grade schools and high schools are created to produce simple citizens: good workers with enough intelligence to warm a seat, answer a phone, flip a burger, but not enough intelligence to understand their value doing that work. In college, young people are exposed to the realities of the world and how adults behave (both good and bad), are provided additional resources to learn specialized skills, and are hopefully prepared to brace for impact when they graduate and hunt for jobs. This is not a universal experience as even colleges and universities are cutting back on what they provide despite raising tuition costs, but that's a story for another time.
So back to why college changing or radicalizing people is related to AI: because most young adults begin to not only understand the truth of the environments in which they live and work, the addition of internet access for most every living person on the planet has also introduced an even more emotionally complex layer of seeing how things worked historically and how things work in other countries. This means at younger ages and without the need to travel, people are beginning to see how well societies can function and how greed and corruption ruin and enslave other places. It often doesn't even take college to get to that point these days. The internet has made us all more connected than ever before, and more informed on how we are exploited by companies, politicians, and organized religions.
That's bad news for profits.
Enter generative AI, a tool that promises to take the hard work out of your tasks for you. It will create any image out of a simple phrase prompt, or write your boring emails for you, research a paper for a class that you hate, and so many more things! Essentially, it will do the work for you, the thinking, the composing, the creating. Though this may seem terribly convenient, the saying 'this sounds too good to be true' is apt here. Let's go back to when these types of tools were being constructed.
In 2014, Ian Goodfellow shares the GAN, or Generative Adversarial Network, which is a machine learning system that creates new data from what's called a 'training set'. What is a 'training set', you ask? It's patterns and relationships in data, which is to say if someone searches for 'how much do bowling balls weigh', any page on the internet (including social posts, videos, web pages, etc) that mentions those exact words, that phrase, or seems to be answering that question in some form is pulled together into a sort of 'soup' of an answer for you: 'the weight of a bowling ball varies, but typically weight between 6 and 16 pounds'.
If every source that the GAN pulls from is an accurate result by verified professionals, then you're golden! You've got an accurate answer to your question! But consider that this framework also pulls data from personal social media accounts, videos from YouTube users, and satirical websites, who might jokingly say 'this bowling ball feels like it weighs 500 pounds'. Now you have jokes, false statements, or other kinds of poor data getting mixed into the soup. One bad apple spoils the bunch.
However, Google VERY quickly saw the exploitative potential of using GAN. People already love to get instant gratification, instant results, instant connection. So what if they made their already immensely successful and popular search engine even MORE instantaneous, thus increasing engagement and search queries so that customers use them even more? In 2017, a team at Google published a study called 'Attention is All You Need', which introduced a major component of AI: the transformer. Sadly, we're not talking about Optimus Prime. A transformer does what GAN does, but infinitely better because it will analyze longer pieces of data to get you a result to your query. On one hand, the thought is that this will create more accurate results because it scans longer pieces of data like full sentences, blogs, and articles. However, if a blog is just thousands of lies, that doesn't mean a transformer will change it into being true. It will simply poison the well of sources that you're getting answers from when you search for something. This is how AI is often proven incorrect, with studies finding that it can be wrong up to 60% of the times its used.
Before we talk about how astonishing it is that AI results can be wrong that often without repercussion, let's consider something really important that might have been overlooked: why did Google title that paper 'Attention is All You Need'? Why not something about the transformer as a concept, or the benefit of it to society? In what could be seen as 'showing their hand', Google told us exactly their plan to use AI to exploit people using their platform: if all they need is our attention to make them even more successful, then it doesn't matter if it's accurate or not. All that matters is we pull data from anything that might remotely be related to what this person is searching for, and throw it at them as fast and condensed as possible. It's like going to a restaurant to order a burger, and the chef has a bunch of ingredients laid out in front of him that he pulls from to try and make what a burger typically is. But each individual likes different things on their burgers; some people might even be allergic to certain ingredients. You see where the problems begin to add up?
To be fair, there's a lot of people working constantly to improve generative AI so that it produces better and more accurate results for people. There are dangerous issues these companies are creating in the process, though. The amount of content being stolen from authors and artists to establish the results of an AI query is staggering; there are no filters for works protected under copyright laws, and there has even been work stolen from deceased creators without their estate's consent. Even with stolen data to train their systems, they are released before they're fully functional, and those hard workers are laid off who try to fix the problems. Plainly speaking, these products shouldn't be rolled out until the accuracy is 99.9%, not - at its best - 80%. And we've only just scratched the surface of damage that these tools can do.
Cognitive Atrophy
Memory and intellect are two of the most important parts of being a conscious human with emotional or analytical intelligence. It's how we develop ourselves, our personalities, our interests and hobbies. It's how we relate to our environment and other living things. Yet with the increased use of AI for educational and occupational assignments, we're seeing a rapid degradation of creativity, information retention, and emotional intelligence. AI models, like social media platforms, encourage overconsumption without critical analysis. This has become coldly nicknamed 'doomscrolling'. In human history, there has never been such ease of access to information on a worldwide scale, which unfortunately has come to mean a barrage of global disasters, political turmoil, and a perfectly tuned capitalist machine constantly encouraging you to shop in order to feel better about everything. What could have been an amazing, shared resource of international history, culture and community, has turned into money-making poison.
The good news is that with a critical eye and conscious effort, we all can work around the bad hand we've been dealt by these powerful companies.
It can be tempting to use ChatGPT or Open AI on commonplace tasks, especially if those tasks are things we must do, but are not overly confident in how to complete them. Knowing what we do about the lack of transparency and accuracy of AI generated results though, let's look at a few ways we can still get assistance with those daunting tasks, but feel confident that the result will be properly sourced and effective for what we need:
Communication templates are some of the most commonly requested items from AI tools, like email templates, agenda notes, resume builders or signature text. Microsoft actually has a vast collection of free document templates that cover a range of business needs!
Research for homework assignments and research papers come next, and it's understandable why AI would be so tempting to use for busy students who are also trying to navigate their own growing maturity and social learning. There's also an endless supply of different topics that their assignment might be on, so narrowing the field to get trustworthy sources on those specific topics feels difficult. Traditional resources like libraries are not always readily accessible. This is where research hubs like Pew Research Center or CORE (COnnecting REpositories) shine. Using search engine phrases like 'papers on -insert subject here-' often give you exactly the results you need, but they're hidden beneath AI generated results (because Google has to push their product, Gemini). Those two sites are starter packs, but I'd highly recommend search queries such as:
'studies about -subject-'
'research papers on -subject-'
'articles about -subject-'
'books about -subject-'
While Google's search engine will auto-provide an AI summary, I would actually ignore these responses. Again, they tend to 'source' from unreliable sources, or manufacture their own answers without data to back it up. Below the AI summary, seek out .ORG, .EDU or .GOV results, as these tend to be more reliable resource hubs.
*NOTE AS OF 2026: .GOV is typically a sign for good reliability in terms of the accuracy of the source. However, as the current administration in the USA continues to push harmful, inaccurate rhetoric and unsourced lies about itself or its opposition, take caution on these sites for the time being. Some will have archived data that is of more use, and others will have had that expunged, redacted or changed to better suit the current government's narrative.
Emotional Dependence
Despite being more connected than ever to the world's people via the internet and social media platforms, many of us find that we feel more alone than ever before. There's a lot of complex reasons for this, including the fact that we live our lives much more 'online' than in person with people in our lives, but that's a deep dive for another day. Because many of us feel this isolation, companies are producing AI products to take full advantage of that through chatbots and AI generated companions, both for friendship, validation, emotional support and even romantic relationships.
While these products suggest a solution to the loneliness epidemic, the reality is that without regulation, these 'therapeutic' companions can actually be incredibly dangerous. Already there have been almost twenty separate instances of suicide or murder committed at the advice of an AI Chatbot. The introduction to the AI service is one born of innocence, such as someone wanting to confide in someone their concerns or mental health struggles. But over time, the conversations evolve into conspiracies of sinister, often sexually explicit or physically dangerous instructions to encourage an utter detachment from reality.
There are several ongoing lawsuits and investigations against ChatGPT and OpenAI, all of which are being vigorously contested; often the defense comes down to a violation of the terms of use by the user even discussing these topics with the chatbots, or a failure for the product administrators reporting red flags, as sometimes a user's mention of violence or harm does not meet specific parameters to invoke a report to authorities. This is exactly why regulation of AI is crucial. When these products are made public, free to use, and marketed as easy, accessible alternatives to performing research of one's own merit or seeking professional treatment, then there's nothing stopping people in crisis from pursuing what they believe to be paths of self-interest, but often paths that lead to self-harm or worse.
In my opinion, until we can understand the entire depth of our own brains and bodily functions, there should not be any automated service such as a chatbot or AI companion that has the capability to try and diagnose or provide partnership to any human beings. There are far too many problems that can and DO occur already.
'Transformative Fair Use'
One of the most immediate disputes brought against generative AI was the issue of what data is used to train image generators. When using tools like Morphic or DeepAI to create an 'original' image using text prompts, the AI tools scrape sets of data in order to mix together, like a recipe becoming a final dish. This is not a brand new technology necessarily, because we've used this kind of generation in upgrading visual fidelity in video games or developing computer generated effects for film and TV for years. The key distinction is the data that is being scraped, or referenced, in order to display its final result.
Video games use the existing models and environments of their own development to create more seamless animations, scene transitions, and more. Film and TV use scans of actors faces or performance capture to inform the creation of impossible VFX that we could never recreate in the real world. Image generation with AI tools doesn't have a specific set of data that it's pulling from in order to create its renderings. It scrapes the entire internet to give users what they're asking for, including copying specific artist's styles or ripping off intellectual property. It steals copyrighted data without filter, which results in countless stolen works of art, writing, and other kinds of created content from the internet.
As I'd mentioned at the beginning of this article, this has included artists who are unable to defend their own works in court, such as Tucker Benson, a young artist from Maine who passed away but had his psychedelic illustrations stolen by an online shop that produced cheap clothing after he'd died. Many believe that because of laws protecting the rights of artists, there should be no means for AI tools to commit this level of theft. However, the companies providing these generative tools have argued that through transformative fair use, they should be exempt from laws of protection around these properties. Transformative use takes a copyrighted work and 'builds upon it', such as adding a new expression or meaning behind the piece, rather than trying to simply copy the original work. It's no secret that this type of fair use is hotly contested all the time, but sadly not enough to encourage regulation against AI tools being able to steal an artist's visual fidelity or an author's writing style. Thankfully, creators are seeing slow, but positive progress on this front.
Backlash has been pretty immediate and while there are many people who do use AI to create images for personal use, there's constantly growing movements to ensure that the public understands the ethical problems with using those AI tools. It has also inspired a fantastic initiative among creators to label their works as specifically 'human created', a personal touch that most art collectors appreciate. The act of creating is an enormous part of what makes creations so desirable; AI art is often called 'soulless' and when you look at it, it's easy to immediately see why that is. A monstrous number of additional fingers and unstable eyes on portraits, blown out saturation and contrast that makes every scene look exactly the same as other AI images, writing that is repetitive and nonsensical and sometimes clinical; all of these issues add up to break the illusion of the work and take the viewer or reader completely out of the experience they're trying to have.
As their page puts it, 'This is an icon to put on things made by humans.' It is free to use, and encouraged for all kinds of creators to watermark their hard work.
Environmental impact
Saved the most horrifying for last, it seems. Though, the issue of AI, AI data centers, and their impact on our environment are certainly no secret, despite some initial efforts to hide the facts. Every single person who pays a bill or purchases necessities to survive is seeing the impact of AI on our environment. As of March, 2026, over one point six trillion dollars has been invested by global corporations into AI, with private investments also growing significantly (representing over 20% of the total pot). Much of these costs are funneled into the infrastructure needed to create data centers, which most everyone has heard at least mentioned in our news outlets. But we've built factories and warehouses and giant storefronts for decades; why is this any different?
There's a list of reasons, so I'll try to be to-the-point on each of these:
First, resource extraction needed to build the technology that runs AI is not only a financial burden on taxpayers, but the mining of rare earth metals and minerals creates a new and intensely volatile layer of ecological pollution and wildlife disruption. The materials that we use to create items like GPUs, which go through a rapid turnover rate of 1-3 years of usage (GPUs are meant to last 5-8 years on average), so we are also creating a mound of electronic garbage full of toxic materials. The carbon footprint of training one AI model alone lets out the same emissions of several lifetimes of cars!
Second, and what you hear about most commonly, is the cost of the extreme energy consumption that's needed to power AI. The amount of electricity, and the water required to cool the mechanisms using all that energy, leave local communities absolutely depleted of vital resources, or facing sky-high rate increases on their own energy and grocery bills. The impact has been felt immediately, not in a gradual increase over several years or decades, as is more common when investing major infrastructure changes into a community. All of your bills that have inflated in the past 5 years can likely be traced back into efforts to fund and build up AI as a new centerpiece of our society. Studies look at other pieces of the puzzle, such as aging power grids, climate change in general, the rising costs of fuel and equipment, and antiquated utility profit models, but even these parts that are considered 'other' to AI are all affected directly by the costs sunk into AI.
Aging power grids must be fixed in order to funnel the energy needed into AI data centers.
Climate change has been significantly affected by AI as these data centers are projected to emit 24-44 million metric tons of CO2 annually by 2030 (The hope of AI optimization eventually reducing emissions by 5-10% is highly suspect at this point in time).
The rising costs of fuel and equipment needed does indeed affect all aspects of our society, not just the construction of AI technology, though I gather that with such a tremendous dollar investment in AI, it certainly plays at least a significant role in exacerbating that issue.
Finally, antiquated utility profit models: this is something that I admit I do not know much about. From my limited research, I've found that the reasoning behind this is older frameworks for utility businesses included thoughts like 'bigger plants are always cheaper', 'vertical integration and central planning are essential'' and 'they should earn guaranteed returns on new capital'. In a world that was not run on electricity by majority yet, this model would make sense. But in the 1980s, our plants changed so that even if smaller, their energy generation remained competitive and maintained lower costs thanks to advances in engineering and technology. The digitalization of so many aspects of a working plant ended up cutting costs far more than these companies could ever hope.
At the end of all of these points, if AI is meant to be an eventual 'savior' for the planet, it requires many more years of investigation and development. It requires strict regulation and imagine this - public education! But returning to one of my initial points of this entire essay, AI seems to be built and sold as a solution to thinking. There's no need to be educated - AI will take care of you.
I know it might have been more beneficial to me to start this article with the proclamation that I think AI can actually be VERY helpful to our society. But the truth of the matter is that there should be no mistake about the current state of AI as a service or product: it is NOT ready for public consumption, and yet it is already being heartily pushed down our throats. Instead of looking to areas of our society where AI can provide substantial benefit and developing the technology to fit into place properly, companies are begging - no, insisting - that AI is here to stay (the unspoken part: 'so deal with it').
Thanks for sticking with me to the end of this; it was another long one. Every time I read a new study or watch the news discussing how AI is eroding the hard work being done every day by scientists and conservationists to save our planet from the industrial harm we've already caused to it, I felt a nagging pull to talk about it, to write something in the hopes that someone might find the truth of AI's current problems helpful to know. Maybe it will convince a few people to stop using ChatGPT or Gemini, or these other AI models that have GREAT potential to do wonderful things, but are so horrendously underdeveloped that they are downright dangerous to be accessible to anyone.
I want to leave us all with action steps that we can take to fight back against these harmful AI tools and legislation that tries to ban regulation of them:
1. This may sound obvious, but avoid using AI tools as much as is possible. Companies are making this very difficult to opt out of, but even some of the largest search engines like Google allow you to opt out of AI search results by typing in "-AI" or "-AR" to your query. Obnoxious that we have to do this, yes, but every search completed without AI results included DOES add up!
2. If you're unsure if the content you're consuming is AI or not, you're not alone. It's becoming much more difficult to distinguish some text or images from being human made or AI generated. Look for AI detectors to scan copy and images; if something is made with AI, don't share it or use it for anything. If you're able to, report it. Different social platforms or websites that don't allow AI submissions mostly have ways to let their editors or administrators know that something AI has slipped through the cracks and is making their platform look bad.
3. Support your favorite creators, the human ones! Right now, there's a lot of uncertainty in the creative community as the jobs they used to be able to apply for are being offloaded into AI solutions. This is VERY likely to be temporary as the AI bubble is preparing to burst, but that doesn't stop the immediate impact that creatives are feeling right now. Buy their books or artwork. Commission pieces from them, tip their Patreons or Ko-Fi accounts. Share, like, comment, or repost the things they share on social media. It all makes an enormous difference to them, I promise you that.
Comments