Stanford Neuroscientist: Can’t Remember Your Dreams? Your Brain May Be Warning You!
In a Nutshell
Stanford neuroscientist David Eagleman explains that dreams protect the visual cortex from sensory takeover during sleep, leveraging brain plasticity to adapt and build skills like motivation via challenges that enlarge the anterior mid-cingulate cortex (aMCC)—a willpower muscle grown by persisting in frustrating but achievable tasks. The brain operates as a "team of rivals" with competing neural networks, requiring Ulysses contracts (pre-commitments to constrain future impulses) for discipline, while cyber-optimism highlights how AI and the internet boost learning through curiosity-driven novelty, virtuous friction, and remixing knowledge—far outweighing atrophy risks if used iteratively for growth. Key advice: seek social challenges, drop mastered skills for new ones to build cognitive reserve against decline, and complexify relationships to reduce polarization by discovering shared human connections before differences.
These notes were generated by AI and may contain inaccuracies.
After decades of debate, the reason we dream is to defend the visual cortex from takeover by other senses like hearing and touch. If you go blind, the visual cortex gets taken over by other senses. Harvard experiment: blindfolding normally sighted people showed takeover starting after 60 minutes. Brain plasticity allows us to sculpt our own brains to become the person we want to be.
Brain peaks at age two with fluid intelligence (ability to learn anything). As adults, we develop crystallized intelligence (knowing how to drive, use cell phone, run a business), so less change is needed. Brain structure is always degenerating.
Actions to Change Brain for Motivation and Discipline
To fundamentally change the brain for high motivation, discipline, and agency: studied in lab for decades (specific actions teased but interrupted).
Cyber optimist: AI and social media will make young people smarter than previous generations.
At age 8, fell 12 feet off roof under construction, broke nose; fall seemed to take long time (calculated 0.6 seconds). Sparked interest in perception, brain's construction of reality vs. what's real. Brain (3 pounds, locked in skull) builds model of world, deciding what to take as reality.
Most people unaware they have a brain, never seen or touched it, assume experiences = reality.
We are not one individual but a team of rivals: 86 billion neurons form neural networks with competing drives.
Example: chocolate chip cookies – one network wants energy, another fears weight gain, another compromises with gym.
You are arguing with yourself. ... The way that your ship of state moves depends on the vote of the neural parliament at any time.
Regrets (e.g., whole bag of chips, alcohol, drugs) happen because different networks take charge at different times.
Ulysses contract: act now to prevent bad behavior later.
Example: Alcoholics Anonymous – clear alcohol from house, even if soberly resolved not to drink, to constrain future self on festive/lonely nights.
We are not one person; different circumstances tempt different behaviors. Understanding this aligns us with desired self.
Feels like one 'me' tussling, but implies competing political parties in brain, all loving the country but differing on steering.
Recommendations
- Make Ulysses contracts.
- Understand thyself becomes know thy selves.
- Recognize illusion of being one person by reflecting on past behaviors varying by circumstance.
Big misconception: we are one unified self.
Truth: brains are plastic/adaptable. William James (100 years ago) coined term: like plastic, brain molds and holds shape (e.g., remembering third-grade teacher's name via changed neural networks).
Brains change every moment; some doors close (e.g., must learn language concept in first years or never grasp naming/asking).
Romania post-Ceausescu: tens of thousands of orphaned kids neglected (fed but not held/talked to), grew up with cognitive deficits.
Humans have more cortex (wrinkly outer 3mm) and plasticity than animals. Horses born doing ancestral behaviors; humans learn cumulative knowledge (physics, math, art) and springboard.
Downside: human brains half-baked at birth, absorb inputs; wrong inputs = trouble.
Brain builds internal model of world for operation (e.g., entrepreneur's model for business structure, hiring, boards).
Fluid intelligence (learn anything) shifts to crystallized (right answers for world, less change needed). Could learn 13th/10th-century languages/cultures as baby, but now adapted to current era/family/culture.
Plasticity doesn't just diminish; brain gets optimized answers.
To change: possible because brains change less only because they don't have to. Pandemic forced reassess world model (toilet paper, banks, coffee shops uncertain) – rare adult chance to challenge model.
How to Change (Stubborn, Unmotivated to Motivated)
Seek challenge: stay in frustrating but achievable zone.
- Take new tasks.
- Seek novelty.
- Become expert, then drop and try what you're bad at – builds new brain pathways.
Internet/social media exposes more opportunities; actively seek challenges, new expertise.
Religious Orders Study: Catholic nuns donated brains post-mortem. Some had Alzheimer's (physical degeneration) but no cognitive deficits due to cognitive reserve from social challenges, fights, games, chores in convents – built new pathways/bridges.
Contrast: retire at 65, watch TV, shrink social circles → no new pathways, cognitive decline risk.
Data supports early retirement (60s) increases cognitive decline probability (coasting on internal model).
Hearing loss shrinks social life; social life crucial – other people hardest for brain (unpredictable words/actions/emotions).
At 33, build maximum pathways now so 80-year-old decline starts higher.
Brain peaks at age 2: maximum neuron connections (86 billion neurons), then prunes to resonate with world (21st-century London/LA vs. 10th-century Mongolia). Pruning good for adaptation.
Brain cells die over time (e.g., head hits). Crystallized intelligence builds lifelong, pays off later (e.g., past learnings aid new business/AI).
Anterior mid-cingulate cortex (aMCC) active/larger in those doing hard things; like willpower muscle – bigger in those forcing hard tasks.
Novices (e.g., juggling) have brain 'on fire' (high activity, stress/anxiety, whole brain engaged); experts (low activity, efficient).
Brain burns deep circuitry to minimize future energy use.
Examples of brain changes:
- Pianists: enlarged motor cortex loop (fine motor both hands).
- Violinists: enlarged only one side (bowing hand less detailed).
- Jugglers, musicians, medical students pre-finals: cortex distribution changes.
Brain devotes more real estate to relevant skills.
Cortex (wrinkly outer): one-trick pony, fluid – visual cortex taken by other senses if blind; auditory if deaf.
Brain plasticity lets us devote real estate to chosen pursuits.
High agency involves more than resolve, as most New Year's resolutions are dropped by February. It's a psychology problem of identifying short-term motivations for distant dreams, like becoming a marathon runner. Determine what motivates you short-term: who to impress, life accomplishments. Structure Ulysses contracts—public commitments like calling Bob to run every morning at 7:00 until you drop, locking in behavior to do the right thing.
It's a cycle: brain and body adapt, making running easier (less out of breath, stronger), spinning the flywheel. Identify what spins it amid life's motivations, good or bad.
Exercise is crucial for the brain. Debate persists on adult neurogenesis: humans born with 86 billion neurons that slowly die, unlike rats with a trickle of new cells. In rats, exercise (wheel running) increases this trickle. Unresolved in humans, but physical fitness matters greatly, alongside sleep and diet, for brain health.
Internet/social media exposes people to possibilities unimaginable in a 500-year-old village (e.g., no entrepreneurs or neuroscientists). Positives: exposure to new paths plus tutorials (fitness influencers, business starts, neuroscientist routes), opening talent windows for everyone, making better brains.
No one knows harms to children's brains definitively—no clean control group (impoverished kids or tech-avoiding Quakers have too many confounds; even prior generations differ in diet, pollution, politics). Speaker is a cyber optimist: internet makes youth smarter via vast intellectual diet.
Pre-internet, knowledge came via 25-minute library trips to Encyclopedia Britannica. Now, instant access to anything fosters curiosity-driven learning, key for plasticity—change happens when curious (right chemical cocktail). Past education was "just-in-case" facts; now, motivated pursuit sticks.
Isaac Asimov (1988 interview) foresaw internet: central supercomputer with cables to homes for any question, accessing humankind's knowledge. Classrooms too fast/slow; personalized doors into knowledge motivate learning, aligning with plasticity.
AI fears (laziness, atrophy from not writing/using Google/AI) miss nuances. Speaker 3x'd home improvement skills via AI (photos prompting disassembly instructions), remembering more due to curiosity.
Stanford study: Google/AI vs. self-learning. Business example: pasting AI answers skips learning cycle/neurological development, avoiding hard thinking.
Distinction: vicious friction (busywork like spreadsheets/taxes—offload to AI) vs. virtuous friction (optimal business structures, D2C/B2B strategies—use brain to learn).
1990s calculator debate resolved: don't waste months on long division. Surf AI wave for tough problems—24/7 expert collaborator expands internal models beyond "what you don't know you don't know," synergizing better ideas.
Copy-pasting doesn't benefit (triggers detection, e.g., candidate emails with model signatures like contrastive constructions, M-dashes). Effort phenomenon: value effortful things (psych studies: prefer bottle-cap art over red dot, real vs. lab diamonds due to perceived effort/mother nature's time).
AI slop feels like least resistance, irritating. Engage iteratively (refine voice/inspiration) for benefit. Effort yields returns; everything is a trade-off—know it (e.g., focus pills have downsides).
Adapt assessments: no final papers (AI-prevalent); use projects (experiments, data, collaboration). Assess AI usage ability, not success without it. Surf AI to become superhuman with forgotten/new knowledge.
AI like Aristotle in pocket: ask curious questions, remember/improve (e.g., home improvement). Everyone running faster—like Steve Jobs' "bicycle for the mind" (computers); AI is motorcycle for the mind. Beat copy-pasters; edge via optimal use, like new swords/guns/cannons.
Weekly practice: ask curiosity-driven questions. Example: podcast monologues—pitch idea, request pros/cons, why wrong, counterarguments; engage deeply.
Teach next generation critical thinking and creativity (programming details obsolete).
Prompt AI: "Be brutally honest, think for yourself, objective, blind spots." Overcomes reluctance to be wrong (echo chambers, cognitive dissonance). AIs sycophantic by default (briefly tried, disliked); instruct "brutally honest always." Private AI feedback less embarrassing than humans.
Example: Joke "Knock knock. Who's there? A letter. Let us who? Let us in and I'll tell you." ChatGPT initially polite/emoji; brutally honest: "not very funny." Children find it funny.
Discussion on whether AI-generated jokes can be funny. For a little child, they get to repeat a joke to their classmate, learning how to do a joke. No single answer to whether that can be funny.
AI reinforces what is already believed, not promoting growth mindset. Prompted to be really honest, AI said the joke is absolutely not funny. Prompted to be even more honest, AI said it's basically not funny at all, shouldn't say that to people, comedic originality 1 out of 10, likelihood of real laughter 1 out of 10.
AI can't be funny yet. Great at repeating jokes but doesn't understand humor. When asked to make a new joke, it follows structure: first guy walks into a bar, second guy does X, third guy breaks pattern, but the break is random, not funny. Transformer model doesn't think of punchline first then build to it.
AI is massively creative. Creativity is absorbing the world, every experience, then bending, breaking, blending cognitive concepts into new remixes. Done all the time for what to say next, recipe, patent, company. Beethoven remixed music from his local culture.
AI remixes what it's trained on, with broader diet. AI excels at remixing but can't do selection: generates 100 pictures but doesn't know which is most appealing.
Humans also struggle with selection; intern making 100 pictures wouldn't know preference either, but learns taste over time. AI can't learn visual taste because it generates pixels diffusely, doesn't "see" the image to assess appeal.
Experiment: fed 30-minute behind-the-scenes video to Gemini, asked to predict drop-off points. 100% accurate twice, e.g., at minute 7 where person talked 74 seconds, jarring, tried to sell hoodie.
Thumbnail test: gave four known A/B test results to Gemini, 100% accurate predicting winner.
AI increasingly makes better creative decisions on preferences than humans. A year ago, not the case.
AI picks on different dimensions, e.g., green vs. red, big vs. small font. But humans consider same factors knowingly.
Human art evolves; AI trained on past hits picks middle of distribution. Hits are new twists at the edge.
AI generates many variations; most unusual one takes off due to volume, but selection is human. Humans pick edge because novelty seekers.
Record labels use formats: 8 bars here, hookie chorus, build pace. Ed Sheeran writes hits for many, e.g., Sony boy band post-One Direction had first three songs by Ed Sheeran, guaranteed hits via formula.
Number one songs often formulaic ghostwriting, not pure genius.
Humans seek between novelty and familiarity: recognize brand/singer but need novelty. Car companies release new models, haircuts/fashion evolve.
Love familiar but new. Record/radio industry blasts song everywhere to make it familiar. Graph shows love peaks when new enough but familiar from repetition (e.g., 20 hearings), then declines if overplayed.
Companies stay in sweet spot: New Coke too novel failed; Blackberry too static failed.
Bon Charge face mask worn 15-20 minutes daily boosts collagen, helps fine lines, blemishes, improves complexion. Professional-grade, noninvasive, confirmed effective by health professionals. Fast free worldwide shipping, easy returns, 1-year warranty, HSA/FSA eligible (up to 40% tax-free savings). 20% off at bondcharge.com/doac.
Whisper Flow turns speech to text for any app/device, highly accurate, corrects mid-sentence changes. Saves time, team loves it. Download at whisperflow.ai/stephven.
AI technology similar to brain but not the same, leading to jagged intelligence: extraordinarily smart then weird answers.
AI as artificial neural networks: simplified brain model ignoring neuron complexity (single cell as complicated as a city, human genome, millions of proteins). Units connected with strengths, many layers deep.
Brain has competing networks (team of rivals), emotions, appetites (food, sexuality). AI lacks these.
AI reads/remembers everything on planet; human couldn't in thousand lifetimes.
Converged on intelligence but different structures. Geoffrey Hinton says breakthroughs from brain understanding, shifting from rules/probability to neural nets.
Humans do one-trial learning (e.g., mom says "pomegranate," got it). AI needs thousands/millions examples, billions in training cost, separate training/action phases. Brains learn on the fly.
AI observes human behavior externally, doesn't know "what it's like" internally. Models humans from outside.
Example: Tesla full self-driving navigated complex traffic by slowing/stopping correctly, though algorithmically different.
Care about other humans: renaissance in live theater/performances. Napster didn't kill concerts; Taylor Swift sells out huge shows for real in-person experience.
Talks increased post-AI; people want in-person over avatars (e.g., 11 Labs voice + ChatGPT).
AI makes us more human: prior tech distanced us, now forces humans to do irreplaceable things (e.g., Taylor Swift live, nurses/doctors holding hands).
AI relationships (1 billion people) as sandbox, but evolution drives to real human touch, stars, dinner with parents.
Worries of people replacing real relationships with AI unrealistic.
Dreaming provides a chance to sandbox relationships and resolve issues before engaging with fellow humans. Counterargument: society may bifurcate with some becoming more addicted to smarter AI that knows exactly what to say to retain attention. However, fundamental human drives like contact with spouse, making babies, and sharing dinner prevail over phone appeal—no one sits a phone across at an Italian restaurant, though some do. Rule at restaurants: no phones on date night; many violate it. Some have proclivity or wiring preferring pornography or virtual wife over rejection risk on first dates.
Bifurcations will occur. Question: what would that person have done in previous generations—successful relationship or always troubled relating?
Dr. Anna Lembke (dopamine expert, colleague) notes everyone has different addictive substances—not universal like heroin or alcohol, but spectrum; for her, romantic erotic novels nearly ruined her relationship.
New technology particularly addictive to certain types. Human society has little in common; everyone differs subtly. Lab studies dozens of axes of person-to-person differences—not big like psychopathy or schizophrenia, but subtle.
Example: visualize ant on purple and white tablecloth crawling toward jar of red jelly. Response: big black ant, jar overflowing jelly down sides with wooden lid, ant almost there—hyperphantasia (rich visualization like picture/movie). Speaker has aphantasia (no visual images). Population evenly spread along spectrum.
Ed Catmull (Pixar founder, ray tracing patents) is aphantasic; surveyed Pixar—many top animators/directors aphantasic. They draw better by attending to external subject and dialoguing with page/pencil; hyperphantasic kid draws from mental image.
Studied dozens of axes—everyone has different internal experiences.
Synesthesia: blending of senses (≥3% population, not disorder, alternative perceptual reality). Examples: letters trigger colors (J=green, M=blue); music triggers visuals; taste triggers fingertip feelings.
Aphantasia: picturing kids not visual but motoric, audio, olfactory imagery—rich notion of being with them, poor visual.
Listeners: try ant/jelly visualization. Hyperphantasia: can picture; aphantasia: cannot. Capacities: accomplishes tasks in 100 ways (visual, motor, sound, smell, conceptual)—no translation to world differences except aphantasics often visual artists.
Used Pipe Drive 10 years in former company, now all commercial teams at Steven.com media company. Easy intelligent CRM; visual pipeline dashboard shows every deal stage, next actions, real-time. Replaces complexity with clarity. >100,000 companies use; 30-day free trial at piperive.com/ceeo (no credit card).
For big goals feeling like Mount Everest: break into tiny 1% steps (team philosophy for success). 1% diaries sold out last year; back with new colors/tweaks. Framework/process/motivation at thediary.com (link in description).
After millennia debating, dreaming defends visual cortex from takeover by other senses (like in blindness). Planet rotates into darkness half-time, disadvantaging visual cortex. Every 90 minutes, midbrain blasts random activity only into visual system to defend territory.
Harvard experiment: tight blindfold on sighted people 60 minutes—visual cortex starts responding to sound/touch.
Brain as storyteller weaves random activity into visual story from day's hot connections.
If no planetary darkness, no dreaming needed (maybe not close eyes). Deep sleep may handle trash cleanup.
Quantitative predictions: 25 primate species—brain plasticity correlates perfectly with REM sleep (dreaming via rapid eye movements). Humans (most plastic) dream all time; infants 50% sleep is dream sleep, decreases with age. Fully baked newborn monkeys: little plasticity, little dream sleep.
All animals dream (even ocean-bottom fish equivalent). Blind mole rat (evolved blind, lives in darkness) still dreams—ancient circuitry predates vision loss; evolution lags.
Like evolution-installed TV at night when real life off, to prevent atrophy.
Dreams mostly useless/bizarre; random activity like random bookshelf sentence seeming meaningful by chance. Some pathways may have meaning, but overlooked: most pointless beyond neurological protection.
Most important overlooked: political interfacing—learn dialogue skills, listen to understand internal models (not agree). Complexify in-groups/out-groups: find cross-cutting ties (surfing, golden retrievers, hometown) to keep social circuitry seeing others as persons (vs. objects like homeless, addicts, enemies).
Maintain eye contact, conversation.
Keep brain active till death: new challenges; drop mastered skills (e.g., sudoku) for unskilled ones. Forces changes as brain always degenerates (faster in Alzheimer's). Build new roadways/paths.
More to degenerate = more leftover. Effort builds paths even as parts fall apart.
Avoid: good sleep, diet matter. Fascinating: younger generation less alcohol; Silicon Valley wineries selling acreage. Friend (20s) favors drinking back for party social lubricant, missing 'dumb mistakes' growing up.
Previous question: what wish most for planet next 10 years?
Complexification of relationships: understand out-groups as humans with lives (no need love/agree)—less polarization. Possible via AI challenging dismissals (stories, introductions), social movements mixing politics for human connection.
Not wishful: happening everywhere. Macro division (20+ social networks >20M users: Rumble, Bumble, Threads, X, Facebook, Snap, Instagram)—splintering into niches.
Algorithms shifted social graph (followers see posts) to interest graph (algorithm serves retentive content for ad revenue)—tighter echo chambers. Even YouTube: 61% don't subscribe; algorithm decides reach on retention.
Nothing new about echo chambers (neighbors=reality). Optimistic: internet exposes diverse views vs. USSR state control/single story. Prefer messy spectrum.
Truth in middle: state control vs. reinforcing algorithms—middle ground?
Is there a middle ground where algorithms let up a bit without state control? Prediction for 2026: a market opportunity for a new social media company addressing current problems. Everyone hates surfing and getting served exactly what they expect, then feeling like they've wasted their lives after an hour or two.
This new company would not build algorithms like others, focused on engagement at any cost with incendiary posts. Instead, it connects people based on shared interests like a particular cuisine or location. The algorithm temporally sequences content so users build a connection threshold before discovering differences, such as opposing political opinions on subject X. Example: "Wow, I didn't know that, but I really like Stephen, so I'm going to lean in and listen a little bit more."
This approach is easy to implement and a key selling point: "Hey, we are here not to enrage you but to actually build connection." It echoes how social media started—a return to origins. There's a neuroscience basis for why it devolved, but now an economic opportunity as the landscape is clear.
Such a network wouldn't be retentive by design because it avoids triggering dopamine like a slot machine (e.g., TikTok's ping-ping randomized returns and dopamine hits). Without addiction, it might not drive returns for ad sales and economic viability.
Counterargument: The story isn't that simple—we don't all want slot machines all the time. People go to Las Vegas for slots occasionally, but rarely. What we truly desire are meaningful connections: meeting someone online, following each other, connecting on points, then discovering differing opinions (e.g., on Iran or abortion) and still listening.
This ties back to the brain's internal battle, like choosing cookie or salad. In today's world, the cookie gives a dopamine hit, but we don't eat cookies all the time—we eat salads much of the time because we're not unconscious automatons.
Dr. David Eagleman thanked for his work. Host links his book below. New book: The Ulysses Contract, out in June 2027.
For anyone that wants to know how to change your life by changing your brain, this is the perfect book to read.
New York Times bestselling author. The book fostered a growth mindset: if not something now, can be tomorrow. Host credits learning from Livewire for pursuing growth mentality across life, expanding neural pathways.
Keep The Diary Of A CEO in your library
Save the videos and channels worth coming back to, and find them again in one place.
