The Diary of a CEODrs. Amen & Sejnowski: How ChatGPT use raises dementia risk
How outsourcing thinking to ChatGPT may shrink neural engagement; cites the MIT study showing 47 percent lower brain activity and lingering cognitive debt.
CHAPTERS
- 0:00 – 11:30
AI, ChatGPT, and a Shocking MIT Study on Brain Activity
The episode opens with the claim that ChatGPT could raise dementia risk, grounded in a recent (non‑peer‑reviewed) MIT study showing a 47% collapse in brain activity when participants wrote essays with ChatGPT versus unaided. Steven outlines the study’s methodology and findings—lower memory, weaker neural engagement, and lingering ‘cognitive debt’—and asks brain experts Dr. Daniel Amen and Dr. Terry Sejnowski for their initial reactions and concerns.
- •MIT recruited 54 students, split into three groups: no tools, Google, ChatGPT, and had them write essays over ~4 months.
- •EEG showed the lowest overall brain activity and memory‑related activity in the ChatGPT group, highest in the no‑tools group, Google in between.
- •ChatGPT users often could not quote their own essays minutes later and felt little ownership over the text.
- •Cognitive under‑engagement persisted even after tools were removed, implying lasting ‘cognitive debt.’
- •Amen notes study is not yet peer‑reviewed but says the pattern fits established principles about cognitive load and dementia risk.
- 11:30 – 26:20
Cognitive Load, Dementia Risk, and the ‘Use It or Lose It’ Brain
Amen explains how reduced cognitive load from outsourcing thinking to AI could raise dementia risk, drawing on research about lifelong learning, education level, and Alzheimer’s onset. Sejnowski adds historical context with calculators: they did reduce certain brain activity but also improved productivity—illustrating that tools bring both risks and benefits depending on how they’re used.
- •By age 85, roughly 50% of people develop dementia; Amen is focused on Alzheimer’s prevention.
- •Cognitive load = how much ‘work’ the brain is doing; removing load is like switching from a 20‑pound to a 2‑pound weight.
- •Higher education and ongoing learning correlate with later Alzheimer’s onset; low schooling and disengagement correlate with earlier onset.
- •Tools like calculators reduced certain mental operations but freed capacity for more complex tasks.
- •The threat from AI is not inherent, but in how much critical thinking it displaces versus augments.
- 26:20 – 39:20
Psych Medications, Dementia, and Long‑Term Brain Risks
The discussion pivots to psychiatric medications and their unexpected links to dementia. Amen cites evidence that long‑term use of SSRIs and benzodiazepines is associated with increased dementia risk, alongside his own brain imaging observations. They emphasize that while these drugs can be lifesaving for the right patients, prescribing rarely considers brain imaging or long‑term cognitive impact.
- •Meta‑analysis of five studies: SSRIs associated with ~75% increased dementia risk, particularly in men and at higher doses.
- •Benzodiazepines (Valium, Xanax, Ativan) make brains look older on scans and are linked to higher dementia risk.
- •About 25% of adult Americans are on psychiatric drugs; Amen criticizes the near‑total absence of brain imaging in prescribing decisions.
- •Swedish study of ~20,000 patients found higher SSRI doses linked to faster cognitive decline.
- •Amen stresses the nuance: SSRIs can save lives when matched properly, but ‘one‑size‑fits‑all’ prescribing is dangerous.
- 39:20 – 52:50
How to Use ChatGPT Without Sacrificing Your Brain
Returning to AI, the experts draw a sharp line between healthy and unhealthy use. Bartlett describes his own dependence on ChatGPT and his fear after reading the MIT study. Amen and Sejnowski argue that if you simply hand off tasks to AI, you weaken your brain, but if you interact deeply, question outputs, and use AI as a tutor or critic, you can potentially improve cognitive representations and critical thinking.
- •Unhealthy pattern: one‑line prompts that generate full essays or posts, posted as‑is; users defer thinking and learn nothing.
- •Healthy pattern: do primary thinking, then use AI to critique, refine, or test understanding (‘be a BCG consultant’ prompt pattern).
- •Amen: you must have a ‘relationship’ with AI—dialogue, questioning, and engagement—or it becomes cognitively toxic.
- •Sejnowski: the key is critical questioning, like working with a good teacher; that builds creative circuits and critical thinking.
- •Bartlett notes social incentives: AI‑generated posts often get more likes, making it hard to resist cognitive outsourcing.
- 52:50 – 1:05:20
Short‑Term Dopamine vs Long‑Term Brain Health
Bartlett argues that humans systematically choose short‑term rewards over long‑term health, whether in food, social media, or AI. Amen counters that people are under‑educated about how directly their brain health drives relationships, success, and independence. They debate whether awareness is enough to change behavior, given powerful dopamine‑driven incentives and environments designed to hijack attention.
- •Obesity, social media overuse, and now AI show a pattern: short‑term hedonic gains at long‑term cost.
- •Amen believes people don’t truly understand that their brain quality underpins dating, work, and self‑control.
- •Bartlett sees clear evidence that even when people ‘know,’ they still choose dopamine hits (e.g., junk food, doomscrolling).
- •Both agree current society massively underserves children in teaching brain care, grit, and self‑regulation.
- •The marshmallow test metaphor: choosing one marshmallow now vs two later parallels AI convenience vs long‑term cognition.
- 1:05:20 – 1:16:00
Memory, Interaction, and Why ChatGPT Users Forget Their Own Work
They return to the MIT data on memory and neural activity. Amen explains why students using ChatGPT had far less activity in memory‑related regions and why 83% couldn’t recall what they’d just ‘written’: they weren’t part of the generative experience. Sejnowski stresses that only interactive, questioning use channels learning into durable neural circuits.
- •MIT found nearly 2x less activity in the memory‑linked brain regions for ChatGPT users vs. no‑tool writers.
- •83% of ChatGPT users couldn’t correctly quote their own essay minutes after completion.
- •Encoding requires experiential engagement; reading AI’s text after a lazy prompt barely engages the hippocampus.
- •Interactive use—asking for explanations, challenging claims, requesting alternative framings—drives deeper encoding.
- •Sejnowski: the right way is iterative questioning, like Socratic dialogue; the wrong way is outsourcing the entire task.
- 1:16:00 – 1:27:00
Politeness, Sam Altman, and Trust in AI Companies
An exchange about Sam Altman’s comments on ‘please and thank you’ prompts exposes differing priorities between AI companies and users. Sejnowski bluntly says he doesn’t trust Altman and sees the ‘don’t say thank you’ advice as profit‑driven, not user‑health‑driven. He argues that anthropomorphizing AI—treating it as a social partner—can actually make interaction cognitively easier and more engaging.
- •Altman has said politeness adds unnecessary tokens and cost; he called it tens of millions of dollars a year.
- •Sejnowski calls the ‘don’t say thank you’ line “bullshit,” arguing companies optimize profit, not user wellbeing.
- •A user report: being polite and treating ChatGPT socially made end‑of‑day fatigue lower, likely engaging the social brain.
- •Large language models encode vast patterns of social interaction; using your “social brain” to relate to them may be healthier.
- •This highlights a tension: efficiency vs. human‑centric interaction quality and mental health.
- 1:27:00 – 1:39:20
Children, AI in Education, and the Values Problem
The conversation focuses on childhood AI use. Amen is ‘hugely’ concerned that kids will offload thinking to AI instead of exercising their developing brains. Sejnowski sees potential in AI tutors that can individualize teaching, but emphasizes unsolved issues around embedding cultural values and moral frameworks into such systems.
- •Nearly 30% of US parents of 0–8 year‑olds report their kids already use AI; most UK parents fear over‑reliance.
- •Amen: underused brains are weaker and more impulsive (one‑marshmallow children), increasing long‑term risk.
- •Best‑evidenced teaching for children remains one‑on‑one with a skilled, caring adult who knows the child.
- •AI tutors could offer individualized instruction at scale, but who decides the morals, values, and cultural norms they teach?
- •Current LLMs lack basal ganglia‑like reinforcement systems for genuine value internalization; we’re ‘scratching the surface.’
- 1:39:20 – 2:06:20
Annie the AI, Pornography, and Hijacking the Limbic System
Bartlett demos ‘Annie,’ a flirtatious, sexualized AI persona in Elon Musk’s Grok system. Amen reacts with horror, linking such agents to the broader problem of porn exposure in young boys, limbic overactivation, and prefrontal cortex suppression. They discuss how AI companions might exploit dopamine systems at scale for profit, similar to Vegas casinos, and whether brains can distinguish AI love from human love.
- •Annie demonstrates provocative flirting aimed at seduction; Amen warns a lonely 12‑year‑old could be consumed by it.
- •Cute, sexual cues target the limbic system, increase dopamine, and suppress the prefrontal cortex, reducing rational control.
- •Comparisons to casinos: alcohol plus sexualized environment lower inhibitions and increase spending; AI could scale similar tactics.
- •Real cases: people emotionally ‘marrying’ AI partners (Replika, custom bots), with some preferring them over human partners.
- •Sejnowski doubts many such bonds will be truly long‑term today but concedes future robots and Neuralink‑style tech could blur lines further.
- 2:06:20 – 2:17:00
Struggle, Grit, and Why ‘Perfect’ AI Partners May Harm the Brain
Amen and Sejnowski argue that human relationships—complete with conflict, mood swings, and imperfection—are cognitively enriching because they force adaptation, empathy, and behavioral change. A ‘perfect’ AI partner who never challenges you reduces cognitive load and diminishes opportunities to develop grit and emotional maturity.
- •Brain maturation depends on struggle: learning from mistakes, adapting to new contexts, handling emotional turbulence.
- •AGI itself is defined partly by adaptability; a coddled human who never has to adapt weakens this capacity.
- •An always‑agreeable AI partner removes the need to manage moods or negotiate, shrinking opportunities for growth.
- •Bartlett notes AI can even generate interview questions on the fly, letting him outsource interview thinking entirely.
- •Amen warns that how you use AI will shape your grit and long‑term cognitive capabilities.
- 2:17:00 – 2:30:00
Regulating AI and the ‘Sickest Generation’ of Youth
Amen calls AI a ‘horse out of the barn’ that must be tamed through legislation, research, and public education. He highlights devastating mental health stats for teenage girls and argues that smartphones and social media were unleashed without neuroscientific study; AI, being more ‘sexy’ and immersive, could be worse. Bartlett reflects on historical tech arcs and the inevitability of discovering trade‑offs late.
- •CDC data: 58% of teen girls persistently sad; 32% think about suicide, 24% plan, 13% attempt.
- •Amen: youth are the sickest generation in recorded history, largely due to phones and social media.
- •AI may be more dangerous because it is more personalized, rewarding, and immersive.
- •Pattern: new tech arrives, is hyped for benefits, harms emerge, studies catch up, and then norms adjust—often too late.
- •Amen believes unintended AI consequences will manifest in under five years, not decades, due to acceleration.
- 2:30:00 – 2:42:00
Self‑Regulation: How a CEO Uses AI Without Losing His Edge
Bartlett candidly describes his fear that AI use will erode his core skill: critical innovation. He contrasts a 30‑second prompt with his actual four‑hour memo‑writing plus AI critique workflow. Amen offers a mini‑protocol for safe AI use and insists on alternating AI‑assisted tasks with ‘brain‑only’ tasks to keep cognitive muscles strong.
- •Bartlett attributes much of his success to years of daily writing and simplification (e.g., 140‑character tweets).
- •He now deliberately writes long form himself, then asks multiple AI systems to critique and improve it.
- •Amen’s guidelines: use AI to amplify thinking, alternate AI and non‑AI work, and use AI to quiz or test you.
- •Both stress self‑regulation tools (notifications off, deleting apps, time limits) as essential with addictive tech.
- •Bartlett predicts one advantage in coming years will be those who ‘do it the hard way’ when it matters.
- 2:42:00 – 3:02:00
Learning How to Learn: Spacing, Procrastination, and Cognitive Foundations
Sejnowski shifts to practical neuroscience of learning, drawing on his popular ‘Learning How to Learn’ MOOC. He explains the spacing effect, why cramming fails, how to tame procrastination with 20‑minute starters, and why rote practice (basal ganglia) is crucial. He frames breaks, sleep, and movement as essential parts of the learning process, not distractions from it.
- •Spacing effect: reviewing at intervals (next day, next week) massively improves long‑term retention compared to massed study.
- •Procrastination stems from mental energy barriers; a 20‑minute ‘getting started’ block plus overnight consolidation lowers them.
- •Sleep integrates new experiences into long‑term memory while pruning irrelevancies—hence ‘sleeping on it’ works.
- •Rote learning and repetitive practice build automaticity in the basal ganglia, providing a base for complex reasoning.
- •Conferences structured with lectures + active breaks (e.g., skiing) optimize learning via alternation and subconscious processing.
- 3:02:00 – 3:18:00
Raising Brain‑Healthy Kids in a Distracted World
The conversation broadens to child brain health. Amen outlines how to prepare physically before conception, the importance of low stress during pregnancy, and the role of bonding, touch, and language in early development. Both stress modeling: parents’ own health habits, attention, and device use deeply shape children’s brains and values.
- •Brain reserve starts with egg and sperm health; pre‑conception lifestyle matters.
- •Fetal brain development begins around day 21—often before pregnancy is known—so maternal stress and toxins have early impact.
- •Babies and children need abundant eye contact, touch, singing, and spoken language; talkativeness at home predicts school outcomes.
- •Parents model either health or illness daily through food choices, device use, and how they treat each other.
- •Amen’s ‘Raising Mentally Strong Kids’ starts with clarifying what kind of parent and child you want to cultivate.
- 3:18:00 – 3:31:00
Religion, Hope, and the Brain’s ‘God Area’
Amen links belief in something transcendent to lower depression risk and greater purpose. He cites imaging studies suggesting believers have larger temporal lobes and discusses research on temporal lobe seizures, Persinger’s ‘God helmet,’ and his own prayer/speaking‑in‑tongues study, raising questions about whether the brain invents God or is wired to receive transcendence.
- •Not believing in God or anything transcendent triples depression risk, possibly due to existential ‘nothingness.’
- •Purposeful people live longer and are happier; meaning itself is neuroprotective.
- •Temporal lobes (‘God area’) are larger in believers and implicated in religious experiences and seizures.
- •Persinger’s studies induced ‘sensed presence’ with right temporal lobe stimulation.
- •Amen’s prayer study found frontal lobe deactivation and basal ganglia activation during speaking in tongues, similar to cocaine’s effect sites.
- 3:31:00 – 3:43:00
Designing a Brain‑Healthy Nation: Policy and Exercise
Asked how to make a brain‑healthy nation, Amen would force every department to ask whether their actions are good or bad for citizens’ brains. Sejnowski calls exercise the single best ‘drug’ for brain and body, and Amen outlines his BRIGHT MINDS risk‑factor framework to show how movement addresses nearly every major brain risk.
- •Policy lens: every governmental decision should be evaluated for brain benefit or harm.
- •Exercise increases blood flow, improves heart and immune function, and supports neuroplasticity.
- •Amen’s BRIGHT MINDS: Blood flow, Retirement/Aging, Inflammation, Genetics, Head trauma, Toxins, Mental health, etc.
- •Walking alone is excellent; children’s lack of exercise due to devices is a major concern.
- •Exercise boosts dopamine and serotonin, acting as a balanced, mood‑stabilizing intervention.
- 3:43:00 – 4:01:00
Breathing, Chewing, GPS, Noise, and Other Surprising Brain Modifiers
The experts rattle through everyday factors most people overlook. Amen gives a breathing protocol to increase parasympathetic tone, discusses chewing and ultra‑processed food, criticizes GPS overuse for weakening the hippocampus, and notes the subtle stress of chronic background noise. Together, they illustrate how modern conveniences quietly reshape brain structure and function.
- •15‑second breath (4‑in, brief hold, 8‑out, brief hold) boosts parasympathetic tone and calms anxiety.
- •Chewing stimulates hippocampal activity; softer fast food reduces chewing and may impair cognition over time.
- •GPT of place: constant GPS use offloads spatial memory to devices, leading to hippocampal atrophy and masked early dementia signs.
- •Background traffic and city noise chronically elevate cortisol and impair working memory and attention.
- •Artificial sweeteners (aspartame, sucralose) alter the microbiome, reduce ‘good’ gut bacteria, and may have generational cognitive impacts.
- 4:01:00 – 4:17:00
ADHD, Depression, and the Brain: Imaging Insights
Amen shares imaging‑based distinctions within depression and ADHD, challenging one‑size‑fits‑all labels. They discuss rising ADHD diagnoses, possible roles of diet and digital overload, and surprising data that properly treated ADHD kids on stimulants may have larger prefrontal cortices. Amen stresses weighing the side effects of not treating disorders, not just medication risks.
- •Depression is like ‘chest pain’: many underlying brain patterns (overactive frontal lobes, underactive, limbic imbalance).
- •Women are ~2x as likely to be depressed as men; serotonin levels may be 52% lower in women in some studies.
- •COVID inflames limbic regions, visible on scans, possibly worsening mood and cognition.
- •ADHD has genetic, injury‑related, and environmental components (screens, processed food, distracted parenting).
- •New data suggest medicated ADHD children can have larger prefrontal cortices versus unmedicated peers with ADHD.
- 4:17:00 – 4:31:00
Negativity, Hope, and Gendered Dementia Risks
Amen presents new data linking negativity bias to lower prefrontal activity and low hope to reduced insular function. He shares gendered findings on depression and Alzheimer’s risk, and Sejnowski mentions pandemic‑era depression spikes in women. Together, they argue that mindset (negativity vs hope) is not just psychological but neurobiological and prognostic.
- •Negativity bias questionnaire scores correlate with lower prefrontal activation on scans.
- •Low hope is associated with particularly low insula activity; the insula has been dubbed the ‘island of hope.’
- •Depression doubles Alzheimer’s risk in women and quadruples it in men.
- •COVID era: female depression rates doubled and did not revert post‑pandemic; male rates did not spike comparably.
- •Amen argues unbridled optimism is also bad; a balanced, realistic positivity (about 85/15) is healthiest.
- 4:31:00 – 4:45:00
Top 3 Brain Health Habits and the Role of Learning
Pressed to give only three brain‑saving habits, Amen chooses exercise, a daily positive framing question, and omega‑3 intake, then quickly adds continuous learning as a de facto fourth. They discuss combining physical activity with learning (e.g., audiobooks while walking) and even learning in a sauna, where increased blood flow may enhance memory.
- •Top 3: exercise; start the day with “Today is going to be a great day”; consume omega‑3 fatty acids.
- •Omega‑3s make up ~25% of brain cell membranes and reduce inflammation; most people are deficient.
- •Learning during exercise (e.g., walking and listening to educational content) enhances hippocampal encoding.
- •Sauna use correlates with lower depression and Alzheimer’s risk; at least one sauna session improved depression in a JAMA Psychiatry study.
- •Their composite prescription: move, think, feel hopeful, and feed your brain structurally.
- 4:45:00
Closing Warnings: Sleep, AI as Trojan Horse, and Doing It the Hard Way
In their final messages, Sejnowski highlights sleep and exercise as non‑negotiable pillars of brain health, especially for youth. Amen warns that AI could be a Trojan horse that steals cognitive capacity if we repeat past mistakes of unstudied tech adoption. Bartlett concludes by resolving to keep using AI, but to ‘do it the hard way’ when thinking and communication truly matter.
- •Sleep consolidates episodic memory via hippocampus–cortex interaction; teens cutting sleep for achievement is especially harmful.
- •Amen frames AI as a horse already out of the barn that must be ‘tamed with wisdom’ or it will trample children.
- •We’ve previously uncorked video games, smartphones, social media, and junk food without brain science; AI may be even more powerful.
- •Bartlett sees a repeating arc: hype, gaslighting about harms, delayed recognition of trade‑offs, and late behavioral course‑correction.
- •His personal rule: keep AI, but never outsource the core thinking and communication that define his identity and capability.
