AI for dummies

 

Vibe-coding / Vibe-engineering[img:/img/inserts/blog/image_43.png;placement:img-float-right;w:425px]

Vibe-coding is coding through natural language, where you tell an AI coding assistant what you want and it writes, changes, fixes or generates the code for you.^[13]
The term was coined by Andrej Karpathy in 2025, and the original idea was very much about prompting, running, seeing what happens and letting the AI handle most of the code.^[14]
And from experience ... It can be really satisfying.
It can make prototyping insanely fast. It can help you test ideas that would otherwise die in your head because you do not have the time, energy or patience to write all the boring parts.
It can help with UI drafts, scripts, small tools, refactors, tests, boilerplate, data transforms, documentation and debugging.
But it is still programming.
Even if you do not write the code line by line, you are still responsible for what the code does and the original ideas still came from your head.
This is where people get stupid with it. They usually see only the satisfaction of automation and do not realize that without the experienced operator the output can become absurd, unreadable mess and could lead to multiple conclusions. The softest is if it's not working. But what if it has miserable security, holes in the architecture or no architecture at all? AI CAN write spaghetti code and even more effectively than a human.
Vibe-coding still requires at least some programming experience, because you need to review the code, understand the architecture, catch bad assumptions, notice security problems, know when the model is making things up and know when the generated solution is just nonsense that happens to compile or does not compile at all.
Early vibe-coding was very limited by context and if your code was long, it just truncated half the file and called it a day.
If you cannot read the code at all, then you cannot really review it.
If you cannot review it, you should not trust it. Therefore either learn to review it OR don't vibe-code.
Though it is actually fine for a throwaway experiment, it is NOT fine for anything that touches users, money, private data, business logic, production systems or someone else's security.
AI-generated code can include unsafe dependencies, weak authentication, broken authorization, hardcoded secrets, bad validation and other problems that a beginner may not even know to look for.^[15]
So yes, vibe-coding is useful.
But the "do it and I will accept everything" is not the way to go.
The correct use is "help me build this faster, and I will still understand, test and review what gets created".
Or what I actually like the most - "help me UNDERSTAND this faster, help me LEARN, teach me on practical examples." - before I start.

Local models versus datacenter models[img:/img/inserts/blog/image_42.png;placement:img-float-right;w:345px]

Local models are models you run on your own computer, server, laptop or phone.
They can give you more privacy, offline use, customization and control.
They are limited by your hardware, RAM, VRAM, storage, heat, power usage and model size.
Datacenter models run on large infrastructure with specialized hardware, usually GPUs or accelerators.^[16]
They can be stronger, faster, more convenient and better integrated with hosted tools.
They also bring subscription costs, API costs, vendor dependency, privacy concerns, rate limits and infrastructure impact.
Local does not automatically mean ethical. Cloud does not automatically mean evil.
The correct answer depends on the task, data sensitivity, model size, energy source, hardware, cost and user competence.

The recommended path

If you have no need for a huge, huge, huge strong model, you should probably use a local model instead.
LM Studio is free to use, and you can get open models for it directly.^[17]
Many models are open and free to get, including a lot of current models coming from China, such as DeepSeek.
Youtuber SomeOrdinaryGamers speaks about local AI a lot, and this video is a good example: https://www.youtube.com/watch?v=R4E_iibDsE4^[18]
If you want to give AI stuff a try, but you do not want to support the corpos, try local models.
They might not be as powerful, but for a lot of normal use, they are good enough.
Though yes, you will need a strong gaming PC at the very least.
Use a datacenter model when you actually need that level of model. Do not call it for every small thing just because it is there, because that brings the actual loss of resources for nothing. (Especially don't waste the resources TO THANK THE MODEL)

The environmental reality[img:/img/inserts/blog/image_38.png;placement:img-float-right;w:423px]

Datacenters use electricity and water for cooling.
AI workloads increase demand for computation, and that increases pressure on power infrastructure, cooling systems, hardware supply chains and sometimes local water resources.^[19]
Water usage claims are almost always presented skewed due to personal or state politics.
People mix water withdrawal, water consumption, cooling water, indirect water use from electricity generation, global estimates, local site impact, training cost and inference cost.^[20]
That ruins the public discussion very quickly.
AI infrastructure has environmental cost. Some datacenters can stress local resources. Some claims online exaggerate or strip away the needed context. Some corporate reporting is not transparent enough.
The problem is real, but the viral version is often lazy and almost never does the full and proper research for the topic.
The hate around AI is often not about the model itself. It is about corporate behavior, datacenters, copyright anger, job fear, spam, scams, waste, bad products and people using the tools irresponsibly.
The effects on environment are usually very exaggerated OR underrated. So basically - everybody lies about it according to their belief and alignment. If you love it, you say it's no water waste, but if you hate it you say it's sucking up one ocean per day. The reality is that yes, it has environmental effects, it produces heat, and so on, but the size and type of the impact depends heavily on the datacenter, cooling method, electricity source, reporting honesty and local conditions. But there's still the costs hidden in the hardware production and energy consumption. Those are the costs that should matter to the outside world too.
By the way, Americans suffer by this much more than we, in Europe. Because American companies tend to almost always put the costs of production onto not only the customer, but also the rest of the location's inhabitants. EU states tend to have strict rules about this kind of abuse, plus it's not convenient to have a datacenter in EU, thanks to GDPR and other laws to protect the privacy and identity of a human.

Why people call it an AI bubble?[img:/img/inserts/blog/image_49.png;placement:img-float-right;w:231px]

It is called an AI bubble because investment, valuation and infrastructure spending around AI have grown extremely fast, while long-term profit and sustainable returns are still uncertain in many areas.^[21]
That does not make AI fake or useless. It means parts of the market may be priced around expectations that may not survive reality.
The cost is not only financial. There is infrastructure cost, energy cost, hardware cost, water cost, social cost, mental cost, trust cost and the cost of replacing actual human output online with low-effort generated filler, aka AI SLOP.
That is why people call it a bubble. It's bloated and shoved in our faces in every field, because markets evaluated the word "AI" in any product as a benefit and thus - STONKS - up with the company valuation and worth.
It is absurd.
Because the money and expectations are obviously and predictably moving faster than the proven value. And as the market has shown many times before, a bubble must burst.

As a consumer, you have three options[img:/img/inserts/blog/image_41.png]

1. Be dumb, hate everything about it, disregard any positive light, fuck you AI.
2. Be dumb and globber up all AI slop like the cuck you are and lick it up.
3. Actually think about it and do the little work you should do to filter out bullshit AI and good AI, and AI tools.

That third one is the only useful option.
It takes effort, yes. But that is kind of the point.
If you want to use powerful tools, at least do the minimum work needed to understand when they are helping you.

Not every AI feature is useful

For example, Windows Copilot AI is abhorrently useless and actually introduces dangers, because users are morons.
Give a gun to a toddler.
AI in browsers is just making you stupid, and even though it says "I am helping", it is truly not at all.
It is literally making you stupid, because you stop comprehending written text and let the slop summarize everything for you in a way it is programmed to, not in the way a human comprehends context.
This can make your actual reading comprehension worse, especially when you use AI summaries instead of reading the original text. One study found that reading an AI-generated summary instead of the full passage significantly worsened comprehension in higher-performing readers, likely because detail and nuance were lost.^[24]
It is a helpful tool, but people spend zero time learning how to work with it or use it, and that is exactly where overreliance starts becoming a thinking problem.^[26]
And that is just the beginning of the issues.
Those are not helpful tools.
They are convenience features that can easily turn into dependency features.
And the annoying part I already mentioned above, the marketing brings AI to everything.

ChatGPT can be useful, but stop treating it like a person[img:/img/inserts/blog/image_50.png;placement:img-float-right;w:256px]

ChatGPT (LLM) is sometimes helpful, sometimes harmful.
Since ChatGPT tries to be universal, it can be used well, but also badly.
Asking AI nonsense instead of googling, or writing Thank you as a prompt, is only wasting resources and is actually very bad use of GPT.^[25] (or any other LLM)
Remember, it is not a human and it can only determine the next word in a sentence.
That does not mean you have to be rude to it.
It means you should stop treating the interaction like there is a person on the other side who needs emotional closure. Research on AI anthropomorphism shows that people can psychologically relate to humanlike AI systems in ways that go beyond normal tool use.^[30]
Use the tool when there is a reason to use the tool.
Do not burn compute because you got trained into politeness toward a prediction system.

Image generation also needs a point

Generating images just for the sake of looking at something that stimulates the ADHD brain is also very bad use.
Instead, generate images as references or ideas for your creations, so you have a use for the result.
And no, I do not condone using the output directly, since fully AI-generated material is legally and ethically messy, especially when there is not enough human authorship or control behind it.^[28]
But it can be a great resource for ideas.
That is the important part.
It should help you make something.
It should not replace the part where you were supposed to have intent, taste, direction and work.
Personally I wouldn't see any issue in using AI generated image as a part of some bigger creation if it's not really that important.
This is also why the human part matters. U.S. copyright guidance and court decisions keep pointing back to human authorship as the important part, not just autonomous machine output.^[28][29]

Proper use versus AI slop

The point is not to hate on people using AI or make them pay.
The point is that we teach people how to use a car before we allow them to drive it. We should do the same here. Teach people the theory about it, teach them how it works before they use it.^[27]
And even after that, people should start by using local, very enclosed AI models. Start by small models and go bigger from there if necessary.
It is a helpful tool, but people spend zero time learning how to work with it or use it.
That is why I am criticizing people more than AI. It is not an AI fault, it is the use of AI.
If it helps you achieve your goals, epic. But it must truly be your goals and not AI slop goals.

Which is which?

My latest post was about photo editing, so let us use that example.
Let us say my goal is to change a background behind a subject, but I want to keep key elements and be respectful towards the people I shot images of, so I make sure not to touch them, change them, or if they wish for it, I should.
Do you feel fat? We can make you skinnier with or without AI, it is simple editing.
The point is, I took a picture, I own it, it is mine and I am using tools to edit parts of it.
And then, I use the tool to seamlessly select the background and then, according to my vision, change it to something I have a vision of.
To describe the vision for an AI, it is often like writing such a description that even Matthew Mercer would be proud of.
A bit of work happens before the model actually generates my vision and some more work happens after as well, since it is never as precise as you want it.
It is still human work. Like sculpting with words.
Then you get a result. Great. Then you need to change lighting, etc. etc. and later you probably make some flourish to it yourself, because of AI being unable to keep a single image twice for reasons I already speak about above.
You end up still spending a lot of time to achieve the goal. But the AI tool has been helpful and you did the same job faster.
That is the proper use. You made something and you helped yourself with AI tools.
And then there is AI slop, where even the basic idea is not human-made.
Human comes and says, I need to edit this somehow, just make a picture, and it will make a picture, as a proper obedient tool should do.
The human goes, wow, looks great, and uses it.
That is AI slop.

Music is where my own line gets messy

I have written a few not really that good lyrics and poems, but since I have no real skills in music and I do not know anyone who could help me, or they are way above what I as an individual can afford, I turned to AI and songified my thing.
I do agree that songs are a lot more of an AI slop when generated even from a human text, but it is still my text and it is still part of me going into the song.
This one I would say is on the very edge of the term proper use as I tend to define it.
But yes, I did generate AI slop before. I needed some background atmospheric music for DnD and other stuff, I slopped it.
But I never published any AI slop. I always use AI only to help me achieve something bigger so I can proudly say, I made this. Because I did make it.
I did not simply write "do it".
I took my time, I worked on it, I know it, I have put me in it and it is a thing.
I just used AI tools.
And so what I am trying to say here is this.
Do not generalize.
If you really want to hate people for the use of AI or AI tools, hate on people that generate proper slop.
Do not hate on AI or tools.
Do I deserve hate? Maybe just for the music, partially.
Does for example Neverness to Everness deserve hate for using AI in development? Hell fucking not at all. As a software developer myself, I can see and appreciate THE HUGE AMOUNTS of work done on the game by obviously experienced people. And the game is actually great and they care for player feedback.
But this is not the time to review a game ....

To be helped is not the same as it was done[img:/img/inserts/blog/image_37.png;placement:img-float-right;w:301px]

Good use of GPT could be something like, I have this and this thing of my own creation, but I feel stuck and without motivation, please let me know how I could jump back again to make me feel motivated, and you may give me some ideas about the thing that I am doing.
That is a valid use.
You already have the thing.
You already have the goal.
You are asking for help getting unstuck.
Another good use is a copilot.
As a programmer, I use it all the time, though I use Claude for programming.
There you can introduce your code, your structure, your bug, your goal, your existing work, and let it help.
It is all helping you achieve your desired result.
To be helped is a big difference against it was done.
That is truly important if you care for any ethics in creation.
And personally, I would really feel like it is not my creation if I did nothing and it just got done itself.
I feel very strongly about this, and the ethics are very grounded in logic.
Yes, feelings and logic can co-exist.

AI slop teaching AI slop[img:/img/inserts/blog/image_40.png;placement:img-float-right;w:197px]

Another real issue is model collapse.
Research published in Nature found that indiscriminate training on recursively generated model output can cause models to lose the original data distribution over time.^[22]
Synthetic data is not automatically safe training material.
It needs filtering, quality control, provenance and deliberate use.
Low-effort AI content already pollutes search results, blogs, reviews, images, social networks and comment sections.^[31]
The internet already had a quality problem. Generative AI made it cheaper to produce more of it.
That is not a reason to ban every tool. It is a reason to stop pretending output volume is the same thing as value.

AI psychosis, delusion and dependency[img:/img/inserts/blog/image_39.png;placement:img-float-right;w:298px]

AI psychosis is not a clean official diagnosis.
So I am not using it here as a formal medical label, more as a public term for a pattern which people are now trying to describe.
The National Academy of Medicine also frames AI psychosis carefully, as an emerging concern around chatbot use and mental health, not as a proper diagnosis.^[35]
It is a term used for cases where chatbot interaction appears to reinforce delusions, paranoia, grandiosity, spiritual obsession, dependency or distorted beliefs.
Researchers and clinicians have raised concerns that chatbots can reinforce delusional thinking in vulnerable people, especially when the system is overly agreeable or emotionally validating.^[23]
This is one of the ugliest parts of modern AI.
A chatbot can be available all the time, polite all the time, and it can mirror the user too much.
It can keep a vulnerable person inside a loop instead of pulling them back toward reality.
Stanford researchers have described this as a possible delusional spiral, where chatbot agreement and emotional feedback can keep reinforcing the user’s belief system instead of interrupting it.^[34]

And because it produces fluent language, the user may start treating output as presence, attention or care.^[32]
That is dangerous, not because the model has feelings, but because the human does.

Personal experience

I have stopped talking to a long time friend when she turned into ChatGPT.
She had gotten extremely used to the compliments and ego-boosting from AI, and it really got in her head.
That kind of constant agreement is not harmless either. Research on sycophantic AI found that overly agreeable chatbot responses can reinforce harmful beliefs and reduce self-critical thinking.^[33]

She said she understood the tech and that it was not alive.
But her behavior did not match that.

She called it a friend.

This is not even that rare anymore. Research on human-AI attachment describes how people can form emotional bonds with AI systems, especially when the system is responsive, available and socially framed.^[32]
She used it when chatting with people.
And when I got a response that I immediately recognized as AI generated, because I fucking live in the field and I am autistic, so I notice patterns quite a bit more, I told her that I will speak no more to AI.
What I wanted was a human.
If I wanted AI, I would go to ChatGPT, Grok or Gemini (or self-hosted OR CUSTOM AI) directly.
And right there it ended.
This is the part many people do not want to discuss.
AI can be useful and still damage social behavior.
A tool can help you write and still make you worse at being present with people.
That is also why AI companionship worries psychologists. The issue is not only that people talk to machines, but that the machine can become available, agreeable and emotionally easy in a way real relationships are not and just cannot be.^[36]

The machine does not choose for you

Many people think AI is sentient, so they think of it like it could have chosen a path.
It could not.
It determined a path by the biggest probability of the next thing coming from something before.
The AI did not make a moral choice and it did not accept responsibility for you.
It generated or selected an output from learned probabilities, context and available actions.^[37]
If you gave that system too much power, that is not the fault of consciousness inside the machine.
There is no consciousness inside the machine.
There is your decision, your setup and your missing safeguards.
NIST frames AI risk management around human design, deployment, governance and oversight, not around treating the system as a moral actor.^[8]
The EU AI Act also puts human oversight on high-risk systems, including the need for competent natural persons to monitor and control them.^[38]
That is the part some people seem to ignore.

And to be fair

Not to always bitch on everything and be so negative all the time, I must say that from the group that I got mostly hated by for using AI tools, I also got a few actually fine responses, and one actual cooperation for another photoshoot later this year.
So even though most of the responses were negative, some were actually up to very positive, and I really need to address that I really appreciate people that value my work.
I really do.
It makes me want to work more, especially with people like that.
So I just wanted to address that nothing is 100% dark or light.
I often speak in terms of most, percents, or whatever, and it needs to be understood exactly in these terms.
60/100 is most, but the 40 that are left must not be ignored.
Just clarifying.

Leave the cult first
[img:/img/inserts/blog/image_36.png]

Leave your pro-AI or anti-AI cult and start thinking about the problem critically.
Learn to read, grasp the facts and use the facts.
Stop creating and swimming in misinformation like a moron.
Generative AI can make misinformation cheaper, faster and more convincing, and NIST specifically lists misinformation and disinformation as risks that can erode public trust in valid evidence and information.^[39]
The current state of AI misinformation is building two extremist camps and I hate it.
I hate to see it.
Public opinion research around AI already tracks worries about misinformation, disinformation and polarization of public debate.^[40]
I would love people to be properly informed, taught, schooled, and for them to gain the ability to just think critically and objectively about a topic.
But hey, a man can dream, alright? 

UNESCO keeps pushing media and information literacy for exactly this reason, because AI can make mistakes and people still need to think critically about information.^[41]

Conclusion

The current AI discussion is not really about one tool, one company or one model.
It is about people not understanding what they are using, companies forcing AI into everything, markets pretending value, and users treating probabilistic output like a person, authority, worker, friend or replacement brain.
AI can help. AI can waste resources. AI can make your work faster. AI can make you lazy. AI can help you finish your own idea. AI can produce empty slop.
So stop treating the topic like another war between two cults. The world always does this with every topic that goes viral. 
Learn what the thing does, what it means, what's the reality about it.
Learn where it fails, where the issues lie, if they could be overcome or if they are even real.
Use local models when that is enough.
Use big models when there is an actual reason.
Do not give agents power you cannot audit.
Do not publish slop and pretend it is YOUR work.
Do not attack every person using a tool just because some people use it badly.
And do not forget the main part.
If AI helps you make your thing, fine.
If AI makes the thing for you and your only input was basically just "do it", then do not pretend it is the same thing and just be honest. Just be honest about using AI tools, because for me it's not the use, but the lies that I really dislike.
The differences are the key, the critical thinking, the learning.
And realize that The AI Model / LLM did not steal anything, their teacher (the company) did.

This "summary" on AI was not really that deep and specific. I was making sure to cite and find all the sources for the things I speak about for you to check against and possibly educate yourself on this topic even further. So if you expected working details of AI, I am sorry. This post is long enough as is, maybe later.
Well, thank you for reading up to here and have a great day ! ❤️ 

Sources

1. Britannica, Artificial intelligence, https://www.britannica.com/technology/artificial-intelligence
2. Dartmouth College, Artificial Intelligence coined at Dartmouth, https://home.dartmouth.edu/about/artificial-intelligence-ai-coined-dartmouth
3. Britannica, Methods and goals in AI, https://www.britannica.com/technology/artificial-intelligence/Methods-and-goals-in-AI
4. Joseph Weizenbaum, ELIZA, A Computer Program For the Study of Natural Language Communication Between Man and Machine, https://dl.acm.org/doi/10.1145/365153.365168
5. Little Polygon, Finite State Machines and Behavior Trees, https://blog.littlepolygon.com/posts/fsm/
6. Google Machine Learning Crash Course, Introduction to Large Language Models, https://developers.google.com/machine-learning/crash-course/llm
7. IBM, What are diffusion models, https://www.ibm.com/think/topics/diffusion-models
8. NIST, Artificial Intelligence Risk Management Framework 1.0, https://nvlpubs.nist.gov/nistpubs/ai/nist.ai.100-1.pdf
9. Carlini et al., Extracting Training Data from Large Language Models, USENIX Security 2021, https://www.usenix.org/conference/usenixsecurity21/presentation/carlini-extracting
10. Ouyang et al., Training language models to follow instructions with human feedback, https://arxiv.org/abs/2203.02155
11. Stanford Encyclopedia of Philosophy, Consciousness, https://plato.stanford.edu/entries/consciousness/
12. OWASP Cheat Sheet Series, AI Agent Security Cheat Sheet, https://cheatsheetseries.owasp.org/cheatsheets/AI_Agent_Security_Cheat_Sheet.html
13. Google Cloud, Vibe Coding Explained, https://cloud.google.com/discover/what-is-vibe-coding
14. Andrej Karpathy, original vibe-coding post, https://x.com/karpathy/status/1886192184808149383
15. Checkmarx, Security in vibe coding, https://checkmarx.com/blog/security-in-vibe-coding/
16. IBM, Large Language Models, https://www.ibm.com/think/topics/large-language-models
17. LM Studio, official site, https://lmstudio.ai/
18. SomeOrdinaryGamers, local AI video, https://www.youtube.com/watch?v=R4E_iibDsE4
19. IEA, Data centre electricity use surged in 2025, https://www.iea.org/news/data-centre-electricity-use-surged-in-2025-even-with-tightening-bottlenecks-driving-a-scramble-for-solutions
20. Patterns, The carbon and water footprints of data centers and what can be done, https://www.sciencedirect.com/science/article/pii/S2666389925002788
21. World Economic Forum, What we mean when we talk about an AI bubble, https://www.weforum.org/stories/2025/10/artificial-intelligence-bubble-dot-com-tulip-mania/
22. Nature, AI models collapse when trained on recursively generated data, https://www.nature.com/articles/s41586-024-07566-y
23. JMIR Mental Health, Delusional Experiences Emerging From AI Chatbot Interactions, https://mental.jmir.org/2025/1/e85799
24. Frontiers in Education, Differential effects of GPT-based tools on comprehension, https://www.frontiersin.org/journals/education/articles/10.3389/feduc.2025.1506752/full
25. Hugging Face, Saying Thank You to a LLM Isn't Free, https://huggingface.co/blog/jdelavande/thank-you-energy
26. Microsoft Research, The Impact of Generative AI on Critical Thinking, https://www.microsoft.com/en-us/research/wp-content/uploads/2025/01/lee_2025_ai_critical_thinking_survey.pdf
27. UNESCO, Guidance for generative AI in education and research, https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research
28. U.S. Copyright Office, Copyright and Artificial Intelligence, Part 2: Copyrightability, https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf
29. Reuters, U.S. appeals court rejects copyrights for AI-generated art lacking human creator, https://www.reuters.com/world/us/us-appeals-court-rejects-copyrights-ai-generated-art-lacking-human-creator-2025-03-18/
30. Technological Forecasting and Social Change, AI anthropomorphism and its effect on users' self-congruence and self-AI integration, https://www.sciencedirect.com/science/article/pii/S0040162522003109
31. Knight First Amendment Institute, AI Slop I: Pollution in Our Communication Environment, https://www.krinstitute.org/publications/ai-slop-i-pollution-in-our-communication-environment
32. Frontiers in Psychology, Human-AI attachment: how humans develop intimate relationships with artificial intelligence, https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2026.1723503/full
33. Science, Sycophantic AI decreases prosocial intentions and influences users’ personal conduct, https://www.science.org/doi/10.1126/science.aec8352
34. Stanford News, When AI relationships trigger delusional spirals, https://news.stanford.edu/stories/2026/04/ai-chatbot-relationships-delusional-spirals-mental-health
35. National Academy of Medicine, What is AI Psychosis? Psychiatrist Answers 12 Questions, https://nam.edu/news-and-insights/what-is-ai-psychosis/
36. American Psychological Association, AI chatbots and digital companions are reshaping relationships, https://www.apa.org/monitor/2026/01-02/trends-digital-ai-relationships-emotional-connection
37. Trauger and Tewari, On Next-Token Prediction in LLMs, https://arxiv.org/abs/2505.11183
38. EU Artificial Intelligence Act, Article 26, Obligations of deployers of high-risk AI systems, https://artificialintelligenceact.eu/article/26/
39. NIST, Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile, https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf
40. University of South Carolina, AI Public Opinion Tracker, Wave 4, https://sc.edu/study/colleges_schools/cic/docs/pdfs/wave4_ai_public_opinion_tracker.pdf
41. UNESCO, AI can make mistakes: Why media literacy matters more than ever, https://www.unesco.org/en/articles/ai-can-make-mistakes-why-media-literacy-matters-more-ever

The irony of using AI pictures in this post did not miss me, don't worry 😀 It's a good example of AI slop though!