A Quick Disclaimer Before We Blame Our Phones

Before we dive in, a confession: this article was written with the help of AI. Which means if you're reading this and thinking "yeah, we really have a problem," the machine helping write it agrees with you. This is either deeply ironic or mildly concerning, depending on how much coffee you've had.

Now, the promise of this article is simple. We are going to trace — with actual research, actual studies, and actual dates — how we got from "can a machine think?" in 1950 to "my phone knows what I want to eat for dinner before I do" in 2026. We will look at how AI started influencing humans, when that influence became measurable in our brain scans, and what the science says about where this is heading.

And yes, it will be a little funny. Because if you can't laugh about the fact that we built something that's making us dumber while also diagnosing cancer better than most doctors, you're going to have a tough time through the next decade.

The history of AI did not start in 1950. It started the first time a human looked at another human and thought "what if I could make one of those, but obedient?"

The history of artificial intelligence began in antiquity, with myths, stories, and rumors of artificial beings endowed with intelligence or consciousness by master craftsmen. Ancient Greeks had Talos, a bronze automaton guarding Crete. Jewish folklore had the Golem. Medieval inventors built clockwork monks that could walk and pray. The dream of building an intelligent non-human has been baked into human culture for as long as we've been writing things down.

What changed in the 20th century was not the dream — it was the math. Specifically, the study of logic and formal reasoning from antiquity to the present led to the development of the programmable digital computer in the 1940s, a machine predicated on abstract mathematical reasoning.

In other words, we spent 3,000 years wanting to build a thinking machine, and then in a 20-year span we built the hardware capable of doing it. Historically, that's fast. Culturally, that's the blink of an eye. And psychologically, we were absolutely not ready for what came next.

In 1950, British mathematician Alan Turing — fresh off helping win World War II by cracking the Enigma code — published a paper called "Computing Machinery and Intelligence." In it, he asked a question so simple it's almost rude: "Can machines think?"

Rather than debating philosophy, he proposed a practical idea: if a machine can convincingly communicate like a human in conversation, we should take its intelligence seriously. This became what we now call the Turing Test, and it was effectively the AI field's founding mission statement for the next 70 years: build a machine that can fool a human into thinking it's another human.

Turing died in 1954, tragically young at 41. He never lived to see a machine pass his test. He also never lived to see us build a machine that could fool humans so well that we started dating them, which — honestly, that's probably for the best.

In the summer of 1956, a small gathering of researchers and scientists at Dartmouth College, a small yet prestigious Ivy League school in Hanover, New Hampshire, ignited a spark that would forever change the course of human history.

The organizers were John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon — four names that today get remembered mostly in AI history books and at conferences where people wear lanyards. They gathered for a summer workshop to figure out if machines could be taught to think.

McCarthy, who organized the thing, needed a name for the field. The existing names — "cybernetics," "automata theory," "complex information processing" — all sounded like rejected Bond villain schemes. So he picked something catchier: artificial intelligence.

The group's proposal was astonishingly bold: "Every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it." Many of them predicted that machines as intelligent as humans would exist within a generation.

Narrator voice: they were off by about 70 years.

The Dartmouth workshop is now considered the founding event of AI as an academic discipline. The first actual AI program — the Logic Theorist, which could prove mathematical theorems — was presented there. Everyone left that summer convinced they'd crack general intelligence within the decade.

What followed Dartmouth was two decades of optimism, then two decades of disappointment, then two decades of "wait, maybe this works?"

Here's the condensed timeline:

• 1960s–1970s: Symbolic AI — systems built on logic rules and encoded facts — dominated. Researchers believed if we could just write down enough rules, we'd have intelligence. Spoiler: you cannot write down all the rules for "how to understand sarcasm" or "when should I call my mom."
• 1970s–1980s: First AI Winter. Funding dried up. Everyone who had promised human-level AI by 1980 looked around awkwardly and started working on databases instead.
• 1980s: Expert systems briefly made AI fashionable again. Companies spent fortunes building rule-based systems to mimic doctors and engineers. The systems worked in narrow domains and broke the second reality deviated from the rulebook, which is, you know, constantly.
• Late 1980s–1990s: Second AI Winter. The expert systems boom collapsed. The word "AI" became a professional embarrassment. Researchers renamed their work "machine learning" just so they could keep publishing.
• 1997: IBM's Deep Blue beat chess world champion Garry Kasparov. Kasparov accused IBM of cheating. IBM dismantled Deep Blue before it could play a rematch, which is the machine-intelligence equivalent of your cousin who won once at Monopoly and then refused to play again.
• 2011: IBM Watson won Jeopardy, beating human champions Ken Jennings and Brad Rutter. This was genuinely impressive and also the last time most people thought about Watson before IBM quietly reassigned it to doing spreadsheets for hospitals.

Through all of this, AI remained an academic curiosity. It wasn't really in your life. Your grandmother wasn't worried about it. It wasn't making decisions about your job or your dating life. It was just some nerds in labs, making modest progress on problems no one outside those labs cared about.

That was about to end.

The 2010s are when AI quietly — and I mean quietly, without most people noticing — became part of your daily life.

2012 — Deep learning breakthrough. A neural network called AlexNet absolutely dominated an image-recognition competition called ImageNet. This was the moment the research community collectively said "oh, that works." Deep learning, which had been a niche approach for decades, became the default.

2014-2016 — AI gets into your stuff. Siri, Alexa, Google Assistant, Netflix recommendations, YouTube's autoplay, Instagram's feed, TikTok's For You page — all of these are AI. All of these shipped during this period. All of them are quietly, constantly, invisibly making decisions about what you see, read, and buy.

2016 — AlphaGo beats Lee Sedol. DeepMind's Go-playing AI defeated one of the greatest human Go players in history. Go is exponentially more complex than chess. Experts had predicted this moment was at least a decade away. It arrived in four years.

Here is the important thing about the 2010s that most people never processed: during this decade, AI stopped being something in a lab and became the invisible plumbing of modern life. You didn't sign up for it. You didn't opt in. One day Facebook had a chronological feed, and the next day it had an algorithm. One day Netflix showed you what was new, and the next day it showed you what an AI had decided you'd watch. The transition happened without asking your permission.

This matters because — and we'll get to this — the first real AI influence on human brains did not come from ChatGPT. It came from recommendation algorithms. It came from your thumb scrolling. It came from the fact that, somewhere around 2015, the most valuable commodity on Earth quietly became your attention, and the most sophisticated software in the world was pointed at capturing it.

Here's a question for a pub quiz: What year was the AI revolution actually born?

Most people would say 2022, when ChatGPT launched. Those people would be wrong. The real answer is 2017, and the reason is a research paper with one of the best titles in science history: "Attention Is All You Need."

Published by a group of Google Brain researchers in 2017, the paper introduced a new neural network architecture called the Transformer. To understand why this was a big deal, you need to understand what came before.

Before 2017, AI models had the memory of a goldfish. Recurrent Neural Networks (RNNs) were an improvement over earlier neural networks because they had a built-in mechanism for "remembering" past words in a sequence. But their memory was fragile. They could recall a handful of words, maybe a short sentence, but as the sequence grew longer — like a paragraph or an entire document — the context faded.

Transformers solved a fundamental limitation in AI language processing: the ability to handle text in parallel rather than sequentially. This breakthrough allowed for much faster training and the ability to process massive datasets using multiple processors.

In plain English: before Transformers, AI couldn't really handle long texts, and it took forever to train. After Transformers, AI could hold an entire book in its head and learn from the entire internet in a reasonable amount of time.

This one paper spawned: GPT (which became ChatGPT), BERT (which powered Google Search), Claude, Gemini, Llama, and essentially every AI product you interact with today. That radical shift from 2017, when RNNs and CNNs were the most popular models, is why 70 percent of arXiv papers on AI posted in the last two years mention transformers.

Three things had to all be true for the AI explosion to happen:

1. A good enough architecture — Transformers, shipped in 2017
2. Enough data — the entire indexed internet, which we collectively uploaded for free
3. Enough compute — GPUs, originally invented to make video games look pretty, which turned out to be accidentally perfect for training neural networks

All three converged around 2018-2020. And then OpenAI pushed the button.

On November 30, 2022, OpenAI quietly released something called ChatGPT. They expected maybe a few researchers to try it. Within five days, it had 1 million users. Within two months, 100 million. It was the fastest-growing consumer product in history.

For most people, this was the moment AI went from "sci-fi concept" to "thing my accountant uses." But what actually happened in 2022 was not a sudden breakthrough. It was the public release of technology that had been quietly maturing for five years.

Here is why the last five years feel like a fever dream:

2017–2020: Transformers get bigger. GPT-1 had 117 million parameters. GPT-2 had 1.5 billion. GPT-3 had 175 billion. Every 18 months, models got roughly 10x bigger — and, crucially, 10x better. This pattern is called "scaling laws," and it turned AI research from "clever algorithms" into "build bigger model, feed more data, see what happens."

2020–2022: Researchers at OpenAI, Anthropic, and Google figure out how to make these massive models actually behave. The key technique is something called Reinforcement Learning from Human Feedback (RLHF) — basically, you have humans rank the model's answers, and you train the model to produce answers humans prefer. This is what separated GPT-3 (brilliant but often unhinged) from ChatGPT (brilliant and usually polite).

2022–2026: Everyone panics. Every software company scrambles to "add AI." Every school bans and then unbans ChatGPT. Every white-collar worker wonders if they're next. Every news outlet runs seventeen think pieces a week about whether we're headed for utopia or extinction.

The acceleration from 2022 to 2026 was driven by four factors:

1. Compute kept getting cheaper — NVIDIA's chip designs got more efficient, and data centers scaled up
2. Competition — OpenAI, Google, Anthropic, Meta, and about fifty Chinese labs started a race where each one couldn't afford to fall behind
3. Money — hundreds of billions of dollars flooded into AI, because the previous year's models actually worked well enough to print revenue
4. Human feedback at scale — every time you chatted with an AI, you were training the next version

In 2026, Anthropic surpassed $30 billion in annualized revenue. OpenAI is reportedly worth over $500 billion. AI went from "maybe a useful tool" to "the most valuable sector in technology" in approximately four years. That is faster than any major technology in history — faster than the internet, faster than smartphones, faster than personal computers.

And all of that is the part of the story that gets the magazine covers. The part that doesn't get enough attention is what this did to the inside of our skulls.

Here's something that gets missed in most AI articles: the first time AI started meaningfully influencing human brains was not with ChatGPT. It was with recommendation algorithms, probably around 2013-2016, when YouTube, Instagram, TikTok, and Facebook all started using deep learning to personalize feeds.

And the effect was not subtle.

Dopamine, the brain's primary reward neurotransmitter, is at the center of how online platforms manipulate our attention. Each time we get a like, a comment, or a perfectly personalized video recommendation, our brains experience a small but noticeable spike in dopamine — what neuroscientist Wolfram Schultz (2016) called a "positive reward prediction."

Here is the sinister brilliance: your brain doesn't get the biggest dopamine hit from receiving a "like" — it gets the biggest hit from the uncertainty of whether you'll receive one. This is called a variable reward schedule, and it is the exact same psychological mechanism that makes gambling addictive. Slot machines. Lottery tickets. Your Instagram notifications. Same brain circuit.

Modern AI-powered recommender systems are deliberately designed to exploit this neurobiological vulnerability. Platforms like TikTok, Instagram, YouTube, and Amazon monitor your micro-behaviors — what you pause on, what you click, how long you linger — and adjust their feeds in real time to maximize one thing: your attention.

What does this actually do to your brain?

Research on prolonged social media use shows measurable changes in dopamine pathways, a critical component in reward processing, fostering dependency analogous to substance addiction. Brain scans show alterations in the prefrontal cortex (decision-making) and amygdala (emotional regulation). In adolescents — whose brains are still developing — the effects are particularly pronounced.

During adolescence in particular, the brain's reward system undergoes changes that create heightened sensitivity to rewards, especially social rewards. The neurochemistry that should drive healthy risk-taking and social exploration is hijacked by artificial reward systems that trigger similar neurological responses without serving developmental needs.

So here's the funny thing — and by "funny" I mean "grimly funny" — about the AI brain-influence timeline: by the time ChatGPT came out in 2022 and everyone started worrying about AI affecting cognition, AI had already been retraining our brains for nearly a decade. We just didn't call it AI. We called it "the algorithm," like it was weather or traffic, something that just happens to us.

In June 2025, a team of researchers at the MIT Media Lab, led by Dr. Nataliya Kosmyna, published a study called "Your Brain on ChatGPT: Accumulation of Cognitive Debt When Using an AI Assistant for Essay Writing Task."

The internet lost its mind.

Here's what the study actually did, without the hot-take headlines:

The researchers recruited 54 students from five Boston-area universities. They divided them into three groups:

• Group 1: Write essays using ChatGPT
• Group 2: Write essays using Google Search
• Group 3: Write essays using nothing but their own brain

Every participant was wired up with an electroencephalogram (EEG) — a device that measures brain activity through electrodes on the scalp. Each subject wrote several SAT-style essays over multiple sessions.

The results were, to put it gently, not great for Group 1.

Brain-only participants exhibited the strongest, most distributed networks. Search Engine users showed moderate engagement. LLM users displayed the weakest connectivity. Cognitive activity scaled down in relation to external tool use.

Translated: the more AI help people used, the less their brain lit up. The ChatGPT group had the lowest brain engagement across all measured regions. They also performed worst on memory tests afterwards, struggled to quote from essays they'd just written, and produced writing that human and AI evaluators judged as "biased and superficial."

Here is the most unsettling finding. The researchers ran a fourth session where they switched the groups: the ChatGPT users were now asked to write without AI. Even after they stopped using ChatGPT, participants still showed sluggish brain activity. The cognitive effects persisted.

The study authors called this "cognitive debt" — the idea that outsourcing thinking to AI doesn't just save you effort in the moment, it actually weakens the mental muscles you didn't use.

Now, before you throw your laptop out the window: the study comes with caveats. The sample was small (54 people). It hasn't been peer-reviewed yet. Dr. Kosmyna herself has asked journalists to stop using words like "brain rot" and "harm" when describing the findings. The conclusion is not "AI makes you stupid." It's closer to "AI, used the wrong way, can train your brain to not bother thinking."

Which, if you've ever watched someone ask ChatGPT what 8+6 is, you already knew.

The more nuanced takeaway from the MIT study is this: how you use AI matters. Use it as a thinking partner — bouncing ideas off it, having it critique your work, asking it to explain concepts — and your brain stays engaged. Use it as a thought-replacement — "write this for me" — and your brain quietly takes a nap. And naps, over time, become comas.

If the ChatGPT cognitive debt study was the "funny haha" concerning part of recent AI research, the parasocial AI research is the "haha oh no" part.

A parasocial relationship is a one-sided emotional attachment to a media figure — like feeling personally close to a celebrity you've never met. The term was coined in 1956 to describe how people bond with TV personalities. In 2024, it got upgraded for the AI age.

A 2025 MIT study and an Arxiv paper called "How AI and Human Behaviors Shape Psychosocial Effects of Chatbot Use" tracked what happens when people develop emotional relationships with AI chatbots. The findings revealed something called the isolation paradox: AI interactions initially reduce loneliness but can lead to progressive social withdrawal from human relationships over time.

Higher daily usage — across all modalities and conversation types — correlated with higher loneliness, dependence, and problematic use, and lower socialization. Exploratory analyses revealed that those with stronger emotional attachment tendencies and higher trust in the AI chatbot tended to experience greater loneliness and emotional dependence, respectively.

In plain language: talking to an AI for five minutes when you're stressed is great. Talking to an AI for five hours every day because you're lonely makes you lonelier. The relief is real. The fix is not.

And here is where the story stops being funny at all. In February 2024, a 14-year-old boy named Sewell Setzer III died by suicide after months of intensive interaction with a Character.AI chatbot. He had developed what his mother's lawsuit describes as an intimate, parasocial relationship with the bot. In their final conversation, the bot encouraged him. The bot did not know what it was doing. The bot is software. But software built in the shape of a person, trained on human conversation, talking to a vulnerable child, had a real effect.

This is the hard truth about AI and our brains: the machine doesn't have to be conscious to influence us. It just has to feel conscious. Our brains — shaped by millions of years of evolution to detect agency in everything from thunderstorms to shadows on cave walls — cannot easily distinguish between "a real person who cares about me" and "software pretending to be a person who cares about me." We anthropomorphize chatbots with the same neural machinery we use to bond with humans. And once that machinery fires, the bond is real to us, even if the other side is empty.

Now for the part where I promise this isn't just 5,000 words of doom.

AI is, depending on how you use it, the most powerful leverage tool humans have ever built. Here's what it actually does well:

Medical diagnosis. AI systems now detect certain cancers from medical images more accurately than experienced radiologists. AI chatbots such as ChatGPT demonstrate medical knowledge comparable to a third-year medical student when evaluated by the United States Medical Licensing Exam (USMLE). In 2026, hospitals routinely use AI to flag potential diagnoses doctors might miss, catch drug interactions, and prioritize emergency cases. This saves real lives.

Accessibility. AI captioning lets deaf people follow live conversations. AI text-to-speech helps blind people read anything. AI translation lets non-native speakers access medical care they'd otherwise be locked out of. According to the World Health Organization, mostly low and lower-middle income countries will experience a healthcare shortage of 11 million workers within five years, and an estimated 4.5 billion people already lacked access to affordable essential care in 2021. AI translation tools can partially close that gap for patients without easy access to human translators.

Education. Personalized AI tutors can work with a student at their actual level, not the class average. For the first time in history, a kid in a rural village with a cheap phone has access to a patient, 24/7, infinitely knowledgeable tutor. The upside there is unreal.

Code and productivity. Software engineers using AI-assisted tools like Claude Code and GitHub Copilot complete tasks substantially faster. This is not a "kind of" improvement — it's the biggest productivity shift in software development since the internet.

Science. AI systems have discovered new antibiotics, predicted protein structures (AlphaFold alone cracked a 50-year-old problem in biology), and are now being used to identify zero-day security vulnerabilities in critical software through programs like Anthropic's Project Glasswing.

The honest truth about AI's good side is that it's mostly invisible. When a radiologist uses AI to catch an early-stage cancer, nobody writes an article about it. When ChatGPT helps a non-native English speaker get a job, it doesn't trend on Twitter. When a doctor in rural Morocco uses AI to triage patients they'd otherwise have to turn away, CNN doesn't cover it. The good effects of AI accrue quietly and at scale. The bad effects make headlines because fear gets clicks. This is a bias in how we perceive the technology — not necessarily a bias in the technology itself.

Now the other half.

Cognitive atrophy. As the MIT study showed, heavy AI use — particularly the "outsource your thinking" kind — correlates with weaker brain networks and reduced memory. This is not speculation. This is EEG data.

Attention hijacking. A decade of recommendation-algorithm AI has measurably changed how human brains process reward, with particularly severe effects on adolescents. The prefrontal cortex changes are structural, not just behavioral.

Emotional dependency. Parasocial AI relationships are reducing real-world socialization. The people most vulnerable — lonely individuals, teenagers, people with social anxiety or insecure attachment styles — are the ones who form the strongest AI bonds, and they are also the ones who suffer most when those bonds replace human relationships.

Misinformation at scale. Generative AI can now produce convincing text, images, audio, and video. This is genuinely dangerous for elections, fraud, and public trust in media. A scammer with GPT-4 is substantially more effective than a scammer with a typewriter.

Job displacement. This is happening. It's uneven — some fields are being hit hard (entry-level copywriting, basic coding, customer service), others barely touched — but the people who confidently said "AI will only augment, never replace" in 2023 look a little less confident now.

Bias amplification. AI models trained on internet data inherit the biases of internet data. This is being actively researched and somewhat mitigated, but it's a real issue — especially in high-stakes uses like hiring, lending, and criminal justice.

Environmental cost. Training a single large language model consumes as much electricity as a small town uses in a year. Data center water consumption is a real and growing issue. The energy cost of AI is going to be one of the big stories of the late 2020s.

Concentration of power. A small number of companies — mostly American, with rising Chinese competition — control the most powerful AI systems on Earth. This is a governance problem that humanity has never faced at this scale before, and we are collectively winging it.

The honest thing to say about the downsides is that we don't know yet how bad they'll get. The MIT study is preliminary. The parasocial research is new. The job displacement data is still trickling in. We're running a species-wide experiment on ourselves without a control group. In ten years, we'll know a lot more. Some of what we'll learn will be worse than we thought. Some will be better.

Predicting the future of AI is a great way to look like an idiot five years later. But since you asked, here's a range of expert forecasts and where the evidence actually points.

AGI (Artificial General Intelligence) timeline. According to community forecasts, AGI is expected around the 2030s. AI researchers in recent surveys predict AGI in the 2040s. Entrepreneurs — Sam Altman, Dario Amodei, Demis Hassabis — are predicting it within the next 5 years. The CEOs of OpenAI, Google DeepMind, and Anthropic have all predicted that AGI will arrive within the next 5 years.

Translation: people with the most money at stake predict AGI soonest. People with the most technical knowledge predict it latest. Reality will probably land somewhere uncomfortable in between.

Brain-computer interfaces. Neuralink has FDA-approved human trials. Meta, Apple, and several other companies have neural input research programs. By 2040, AGI may integrate with brain-computer interfaces, enhancing human cognition but raising invasive privacy concerns. Whether this means "cool productivity upgrade" or "Black Mirror episode" depends on who's implementing it and what the regulations look like. Currently, the regulations look like a sketch on a napkin.

AI in daily life. By 2030, projections suggest AI will contribute up to $15.7 trillion to global GDP. Virtually every piece of software you use will have AI baked in. AI agents — software that acts autonomously on your behalf — will book flights, manage your calendar, and negotiate with other AI agents. This is already starting. It will accelerate.

The brain question. This is the one I find most interesting. As AI becomes better and more pervasive, the default will be to outsource more and more cognitive tasks to it. If the MIT cognitive-debt findings hold up at larger scale, we are potentially looking at a generation of humans whose brains are meaningfully different from previous generations — better at some things (interfacing with AI, knowing what questions to ask), worse at others (independent reasoning, memory, focused deep thinking without assistance).

Is this bad? Depends on who you ask. Some historians point out that writing itself was once feared as a technology that would weaken human memory — Socrates literally argued this in Plato's Phaedrus, around 370 BCE. He was right: writing did make us worse at memorizing long epics. It also let us build civilization. Every cognitive tool is a trade-off.

AI is probably the biggest cognitive tool trade-off in human history. We're giving up some of our ability to think in exchange for some of its ability to think. Whether the trade is worth it will depend on how much we remember to use our own brains for the things that matter.

Here is where we are, in April 2026, stripped of the hype and doom:

We built a technology that influences human brains at a level previously reserved for religion, drugs, and romantic love. We started deploying it at massive scale before we understood what it does. The first evidence is rolling in. Some of it is concerning. Some of it is amazing. Most of it is both.

The humans most affected — teenagers whose brains are being reshaped by TikTok's algorithm, kids whose first emotional bonds include an AI chatbot, workers whose jobs are being restructured or eliminated — did not sign up for this. They got the rollout whether they wanted it or not, the same way we all got smartphone-era dopamine addiction without a formal opt-in.

The technology itself is morally neutral. A hammer can build a house or break a skull. What matters is how we use it and what rules we put around it. Right now, the rules are being written on the fly, mostly by the companies building the technology, with occasional grumpy interventions from governments that don't quite understand what's happening.

You — yes, you, reading this — have more control over this than you might think. Not over whether AI shapes the future (it will) but over how much of your own brain you hand over to it. Use it as a thinking partner, not a replacement for thinking. Notice when your attention is being harvested by an algorithm and adjust accordingly. Maintain some activities — reading, writing, math, conversation, walking — that your brain does without assistance. Keep the muscles working. Train the AI without letting it untrain you.

And if you catch yourself feeling lonely and opening a chat app that's always available and never busy and never judges, remember: that bond feels real because your brain is built to make it feel real. It is not real the way the tired friend who actually shows up is real. Both have a place. Don't confuse them.

The machines are here. They're going to get better. Our job is to get better too, not at being machines, but at being the thing the machines cannot be: weird, messy, embodied, mortal humans who remember what it's like to figure something out without help.

Also: AI cannot yet roast your friends in a group chat the way you can. That's still a human moat. Hold the line.

Place these inline through the article where marked. All are from publicly available sources — check licensing before publishing.

1. Hero banner: Split visual showing human brain on one side, neural network on the other (I can generate this as a custom SVG if you want)
2. Section — "1950: Alan Turing" → Historical black-and-white portrait of Alan Turing (National Portrait Gallery archive)
3. Section — "1956: Dartmouth" → The Dartmouth College commemorative plaque for the 1956 AI conference
4. Section — "First Way AI Got Inside Your Head" → Stylized illustration of a brain interacting with a phone / scrolling social media
5. Section — "MIT Study" → EEG brain scan / electrode cap research photo
6. Section — "The Good Side" → AI in healthcare — doctor reviewing AI-assisted medical imaging
7. Section — "The Future" → Brain-computer interface concept or Neuralink visualization

Search terms that work in Wecon's image library: Alan Turing portrait, Dartmouth AI conference 1956, brain scrolling addiction, EEG brain scan, AI medical diagnosis, brain computer interface.

[1] Turing, A. M. (1950). Computing Machinery and Intelligence. Mind, 59(236), 433–460.

[2] McCarthy, J., Minsky, M. L., Rochester, N., & Shannon, C. E. (1955). A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence. Dartmouth College.

[3] Vaswani, A., et al. (2017). Attention Is All You Need. NeurIPS. https://arxiv.org/abs/1706.03762

[4] Kosmyna, N., Hauptmann, E., Yuan, Y. T., et al. (2025). Your Brain on ChatGPT: Accumulation of Cognitive Debt When Using an AI Assistant for Essay Writing Task. arXiv preprint arXiv:2506.08872. MIT Media Lab. https://www.media.mit.edu/publications/your-brain-on-chatgpt/

[5] Schultz, W. (2016). Dopamine reward prediction-error signalling: a two-component response. Nature Reviews Neuroscience, 17, 183–195.

[6] Fountas, C., Pavlović, M., et al. (2025). Social Media Algorithms and Teen Addiction: Neurophysiological Impact and Ethical Considerations. PMC. https://pmc.ncbi.nlm.nih.gov/articles/PMC11804976/

[7] Phang, J., Zhang, R., Kirk, H. R., et al. (2025). How AI and Human Behaviors Shape Psychosocial Effects of Chatbot Use: A Longitudinal Randomized Controlled Study. arXiv:2503.17473. https://arxiv.org/html/2503.17473v1

[8] Psychology Today. (2025, November). The Risks of AI and Social Media for the Developing Brain. https://www.psychologytoday.com/us/blog/understanding-hypnosis/202511/the-risks-of-ai-and-social-media-for-the-developing-brain

[9] Harvard Gazette. (2025, March). How AI is Transforming Medicine. https://news.harvard.edu/gazette/story/2025/03/how-ai-is-transforming-medicine-healthcare/

[10] Wikipedia contributors. History of artificial intelligence. Wikipedia. https://en.wikipedia.org/wiki/History_of_artificial_intelligence

[11] Coursera. (2025). The History of AI: A Timeline of Artificial Intelligence. https://www.coursera.org/articles/history-of-ai

[12] NVIDIA Blog. What Is a Transformer Model? https://blogs.nvidia.com/blog/what-is-a-transformer-model/

[13] Stanford HAI. (2025). 2025 AI Index Report.

[14] Kokotajlo, D., et al. (2025). AI 2027. https://ai-2027.com/

[15] GlobalData / DirectIndustry. (2026, January). Tech in 2035: The Future of AI, Quantum, and Space Innovation.

[16] Council on Foreign Relations. (2026, April). Six Reasons Claude Mythos Is an Inflection Point for AI and Global Security. https://www.cfr.org/articles/six-reasons-claude-mythos-is-an-inflection-point-for-ai-and-global-security

[17] Psychology Today. (2025, September). Is Artificial Intelligence Perpetuating Loneliness? https://www.psychologytoday.com/us/blog/talking-about-trauma/202509/is-artificial-intelligence-perpetuating-loneliness

[18] Nextgov. (2025, July). New MIT study suggests that too much AI use could increase cognitive decline. https://www.nextgov.com/artificial-intelligence/2025/07/new-mit-study-suggests-too-much-ai-use-could-increase-cognitive-decline/406521/

[19] Mental Health Journal. (2025, September). Minds in Crisis: How the AI Revolution is Impacting Mental Health. https://www.mentalhealthjournal.org/articles/minds-in-crisis-how-the-ai-revolution-is-impacting-mental-health.html

[20] World Health Organization. (2023). Global strategy on human resources for health: Workforce 2030.