Humans are the only species that skip sleep on purpose

Walker’s fascination did not start with pillows. In the United Kingdom you can begin medicine at 18, and he was at the Queen’s Medical Center, stunned by anesthesia: within seconds, a fully conscious human being could lose all mentality and awareness. He read about hypnosis, sleep, and dreams when those topics still looked like charlatan science. Almost every colleague he knows is an accidental sleep researcher. Nobody, he says, put a hand up at age five to become one.

The accident happened in his PhD. He was trying to identify different forms of dementia early with electrical brainwave recordings, and it was a disaster — no result after no result. Weekends, he sat in the doctor’s residence inside a little igloo of journals. Some dementias, he noticed, ate away sleep-related brain areas; others left them untouched. He had been measuring patients while they were awake. He started measuring them asleep, and the results arrived.

Then he wanted a bigger question: was the sleep disruption that came with dementia a symptom, or a cause? Twenty years ago, nobody could answer a simpler one: why do we sleep? He did not yet know that some of the most brilliant minds in scientific history had already failed at it. He figured he would do a couple of years of sleep research, solve it, and go back to his patients. Two decades later he is still there. Hard questions, he says, meter out their difficulty all the same, whoever asks them.

The old answer — we sleep to cure sleepiness — is as empty as saying we eat to cure hunger. It tells you nothing about what food, or sleep, actually does. The question has been upended. Is there any physiological system in the body, or any major operation of the mind, that is not wonderfully enhanced when we get sleep, or demonstrably impaired when we do not? So far, for the most part, the answer seems to be no.

Walker’s working answer for why we implode so fast without sleep is blunt. Human beings seem to be one of the few species that will deliberately deprive themselves of sleep for no apparent biological reason. A few animals do skip it, but the reasons are hard: severe starvation, caring for a newborn — a female killer whale leaving the pod to give birth — or birds flying a transoceanic migration of some 3,000 miles. For the most part, the rest of the animal kingdom does not do what we do.

Because evolution never faced sleep deprivation as a common challenge, it never built a safety net. The contrast is the fat cell. Ancestors swung between famine and feast, so nature invented a way to store caloric credit and spend it in debt: adipose tissue. Where, Walker asks, is the fat cell for sleep? Where is the banking chip you can draw on after an all-nighter? There does not seem to be one. She never had to invent it.

From an evolutionary perspective, sleep is almost idiotic. While you sleep you are not finding a mate, not reproducing, not caring for young, not foraging — and you are vulnerable to predation. On any one of those grounds it should have been strongly selected against. Yet in every species studied carefully to date, sleep is present. Even earthworms, millions of years old, have active periods and a passive sleep-like state called lethargus. Sleep, he suggests, evolved with life itself and then fought its way through every step of the pathway.

That is the origin of the famous line in the field: if sleep does not serve an absolutely vital function, it is the biggest mistake the evolutionary process has ever made. An energy-conservation hypothesis tried to explain it as a low-battery mode, and it has been blasted out of the water. The difference between lying on the couch awake and lying on the couch asleep is only about 140 to 150 calories. Sleep is an incredibly active process. Mother Nature, Walker says, did not make a spectacular blunder.

Walker used to sort the mind into three main states, anesthesia aside: wakefulness, non-rapid eye movement sleep, and rapid eye movement, or dream, sleep. He no longer believes the boxes are that hard. He thinks in a state-space diagram, more continuum than switch. A daydream is a different wakeful state from sitting across from someone, present and focused outward. Depending on how you define a dream, there is some dreaming in almost all stages of sleep. Calling non-REM unconscious, he says, is already a stretch.

The continuum gets stranger under deprivation. Individual brain cells, and clusters of them, can fall into a sleep-like state even while the animal still looks awake. Humans do this too. In a microsleep the eyelid partially closes and the brain lapses, but behaviorally you seem awake. Walker does not see a one-or-zero step function. He sees a slide.

The danger is the driver’s seat. At 65 miles an hour in a two-ton vehicle, it takes about a second to drift from one lane to the next and about two seconds to go completely off the road. Have a microsleep at the wheel, and it could be the last microsleep you ever have. Unlike alcohol, which often means you react too late, falling asleep means you do not react at all. At that point there is a two-ton missile on the street, and no one is in control.

Fridman spent five or six years at MIT on the human side of driving, and microsleeps were a central worry. Could computer vision, with no contact sensors, read rising drowsiness from a face? Walker thinks the eyes carry a lot. There is eyelid aperture — partial closures, full closures, and how long each lasts. There may be information in the pupil as well, because the autonomic nervous system shifts as you slide between states, and it partly controls pupillary size. Add steering angle, steering maneuvers, and pedal pressure, and he suspects a combinatorial phenotype of “you are starting to fall asleep.”

Most production cars still use a crude signal: steering-wheel reversals. If you have to keep correcting the car, the system infers sleepiness. Walker agrees it is powerful, and also very rough. How drowsiness shows up is strikingly person-dependent. Some people slow their blinks first; others start to drift the wheel; others change at the pupil. He does not think the field has to hunt only for a universal cue that covers 90 percent of drivers.

His moonshot is a Savile Row suit of sleepiness features. Imagine a sleep tracker at night whose data the car can see. First learn the signature of that individual when well slept; then watch for deviations and pattern-match them to sleep history. You no longer need one-size-fits-all. You need the well-rested baseline of this particular snowflake. Fridman adds a twist from Tesla Autopilot: decades of human-factors work say people supervising automation lose vigilance fast, yet in the studies he has seen, drivers on Autopilot were not getting less alert. He guesses residual fear of dying is still holding them. As the systems get better, that urgency will fade, and the car will need a driver-facing camera. Hands on the wheel, both men agree, is an outdated proxy for attention.

Fridman asks, from a robotics angle, how fundamental consciousness is. Walker treats it as a deeply embedded feature with a panoply of biological benefits. Dream research is his evidence. It is vanishingly rare, he says, to become someone other than who you are in a dream. You can have a third-person perspective, as if you have risen above your body, but the sense of conscious self almost never disappears. That, to him, is a strong sign that consciousness is not a nice trick sitting on top of the mind.

And yet last night, when you went into dreaming, you became flagrantly psychotic, for five good reasons. You saw things that were not there: hallucination. You believed things that could not be true: delusion. You became confused about time, place, and person: disorientation. Your emotions swung wildly: what psychiatrists call being affectively labile. Then you woke up and forgot most of it: amnesia. Experience any one of those five while awake, and you would probably be seeking help. Nested at the heart of that psychosis, the present self remains. Abandoning it, Walker says, takes a kind of biological escape velocity.

Fridman is in the minority in AI who think high general intelligence, defined as operating in a living world of humans, requires consciousness. Most of the field treats the two as separable: you can build a superintelligent chess machine with no inner life. Walker loves that the question is still embryonic. Answers, for him, are just ways to get to more questions. He also thinks it is a missed opportunity to treat the ultimate artificial mind as a replica of ours. We already have human consciousness. Why not try to build something complementary, even orthogonal?

That same humility shows up in how he talks about being wrong. It can take 20 years to build a reputation and two minutes to lose it; if he is mistaken, he does not want to stay mistaken a second longer than he has to. Science, unlike mathematics, never hands you a proof, only a smaller set of things that might still be true. He has launched a short-form podcast with two goals: democratize sleep science, and reunite humanity with the sleep it is desperately bereft of. He is more afraid of not trying than of trying. The first edition of *Why We Sleep* had errors; they were corrected in the second. Ten facts a page times 350 pages is not going to be perfect out of the gate. The duty is a humble mind that fixes what it gets wrong.

Fridman is a heavy coffee drinker who can down several cups and still knock out. Walker’s first sentence, coming from a sleep scientist, is meant to sound strange: drink coffee. The health benefits are well established. The paradox is that sleep is spectacular for health, caffeine can hurt sleep, and coffee still shows a Venn diagram of overlapping benefits. The reconciliation is antioxidants. For many people in Western diets, the humble coffee bean has been asked to carry almost all of their antioxidant load. Decaffeinated coffee has a constellation of benefits too. It is the bean, not the caffeine, doing that work.

Caffeine still hits sleep in at least three ways. For most people it makes it harder to fall asleep and harder to stay asleep. Even if you are the person who can have three or four espressos with dinner, fall asleep fine, and stay asleep, the deep sleep you get will not be as deep. You lose somewhere between 10 and 30 percent of it. To drop deep sleep by about 20 percent, Walker says, he would probably have to age you 15 years — or you could do it every night with a cup of coffee. Past about three cups a day, the health benefits turn down the other side of a U-shaped curve. Dose and timing make the poison.

Caffeine’s half-life is about five to six hours, so a quarter-life of 10 to 12. A noon coffee still has a quarter of its caffeine circulating in the brain at midnight: tucking yourself in and swigging a quarter cup before the light goes out. Clearance is run by cytochrome P450 liver enzymes, and a gene called CYP1A2 sets the speed. Some people clear it in as little as two hours; Walker, hand up, can sit at eight or nine. Tolerance is a different layer. The body internalizes adenosine receptors, so you need two cups to get what one used to give you, and going cold turkey produces withdrawal. For the average adult, he would cut caffeine maybe 10 hours before bed — before 2 p.m. If you are struggling with sleep, dial it down and see.

Sleep, Walker says, is profoundly tied to memory, in four ways. First you need it before learning. A well-slept brain is a dry sponge, ready to soak up new information; a sleep-deprived one is waterlogged, and the memory circuits themselves are in worse shape. Then you need it after learning, to take those freshly minted traces and hit save. It is nowhere near as quick as a digital system. The change is physical and biological, at the level of brain cells, and it takes hours. Sleep cements the memory into neural architecture so it is less likely to be forgotten. He thinks of that as future-proofing information.

The third benefit is the one he is most enchanted by. Sleep does not simply strengthen individual memories. It intelligently integrates, cross-links, and associates them — informational alchemy. You wake with a revised mind-wide web of associations. That, he suspects, is why you have never been told to stay awake on a problem. Some version of “sleep on it” exists across languages, which tells him the creative benefit crosses cultural boundaries. The French, he notes, are closer to sleeping *with* a problem. The British sleep *on* one. He thinks that says a great deal about the romantic difference between the two.

The fourth layer, which grabbed him about eight years ago, is forgetting. Forgetting may be the price we pay for remembering, and letting go has enormous benefits. The brain, he suspects, has finite storage; we cannot spend a lifetime as pure aggregators. It is not useful, evolutionarily, to remember where you parked three years ago. The neurologist Alexander Luria, in *The Mind of a Mnemonist*, studied people who seemed to forget nothing. He spent the first half of the book looking for a chink in their armor and never found it. Then he inverted the question: not what is the benefit of always remembering, but what is the detriment of never forgetting. The lives he described are not ones you would want.

Fridman hears an AI problem in that. Lifelong continuous learning is still open, and the right way to formulate memory is actually forgetting: you have to be exceptionally selective. He remembers ideas and quotes and cannot hold names and titles, which makes academic conversation with someone like Andrew Huberman painful. Walker is the same with names, and thinks anxiety at first meeting is often the culprit, not a broken memory system. What he does not forget are faces, textures, the height of a trouser hem over a shoe. He can still see Fridman in a hotel lobby two years on: the headphones, the weave of the shirt, the boots the other person has already forgotten. “What I remember well is the feeling we shared,” Fridman says. The labels are what drop away.

Walker splits associative processing into two modes. Wakefulness is a Google search gone right: you see the obvious connections, page one. Dream sleep is group therapy for memories. Everyone gets a name badge. Sleep gathers the pieces of the day and forces you to talk not to the people at the front of the room, but to the ones at the back you thought you had no connection with. Chat long enough and a distant, non-obvious link appears — page 20 of the search, a field hockey game in Utah that still, somehow, is about you. Fusing things that should not go together, when they do, can produce marked jumps in evolutionary fitness. That, to him, is the biological basis of creativity, and the algorithm REM sleep was designed to run. Memory pinball: you shoot the day’s information into the attic of cortex and watch where it lights up. You could argue, he says, that we dream therefore we are.

The anecdotes are famous, and he flags them as anecdotes. Dmitri Mendeleev spent months trying to fit the known elements into a logical grid, even making playing cards of them and dealing them on long European train journeys. He fell asleep, the elements danced, and they snapped into place. Mary Shelley’s *Frankenstein* came in a dream at Lord Byron’s home. Paul McCartney received the melody of “Yesterday” in sleep, and “Let It Be” from a dream of his mother, Mary McCartney — not the biblical figure people assume. Keith Richards went to bed with a guitar and a tape recorder and woke to find the tape run to the end: the opening chords of “Satisfaction,” then 43 minutes of snoring. Laboratory work, Walker says, now reliably shows that sleep inspires creativity and problem-solving. He does not need the rock-and-roll stories to make the scientific case.

Engineer Jim Keller has tried to turn that into a protocol: for an hour before bed, think hard about a technical problem, then sleep. Walker does not object to a light hand on the tiller. He also will not pretend it is free. Mother Nature has had 3.6 million years to decide what to upload at night, and her algorithm is probably pretty good. Direct your sleeping cognition at a chip-design problem, and something else may not make it onto the buffet — perhaps the emotional events of the day. You may wake more problem-resolved and less emotionally resolved. Fridman counters that nature never designed us to think about neural networks, and that left to itself his body would like him to be lazy. Walker’s conservative question remains: at what cost?

Fridman tries, half-jokingly, to justify his life. He loves almost everything he does, and that love includes struggle, apparent stress, disappointment, and self-doubt. Sometimes he stays up all night on a problem, sleeps in the day, or runs on very little plus an hour or two of napping. He has looked up Winston Churchill’s schedule and found it uncomfortably familiar. Nikola Tesla was said to sleep two hours; Edison three, though Edison was a habitual napper with cots around the house. Walker knows the feeling. Being a scientist, he says, is not what he does, it is who he is. Sleep, working out, showering, and eating are what he does in between a 20-year love affair with the topic. Experimental ideas have kept him up. He still cannot advocate the schedule from a scientific perspective. It will not come without a blast radius.

Margaret Thatcher and Ronald Reagan were proud of sleeping around four hours. Knowing the links between sleep and Alzheimer’s, Walker has wondered whether it was coincidence that both died of that disease. The elastic band of deprivation can stretch a long way before it snaps — maybe not a microsleep at 32, a heart attack at 42, or cancer at 52, but something in the 70s. Tesla died of a coronary thrombosis. A Harvard study followed people with no signs of cardiovascular disease and found that those getting less than six hours of sleep had a 300 percent increased risk of coronary-artery calcification. Churchill fought depression he called the black dog. A more recent Harvard paper, Walker says, associated polyphasic sleep with worse physical, cognitive, and especially mood outcomes. Sleeping like a baby is not perfect for adults.

Fridman is bothered that people who care about him always say sleep more, and almost never say work harder on the dream. He thinks Silicon Valley over-indexes on work-life balance, and that reaching Olympic heights sometimes requires a chaotic sacrifice of health you would never calculate on paper. Walker splits the question. Is it wrong to do that to yourself? He disagrees. Is it unhealthy? He agrees. He has not found anyone resilient in every domain after sleep loss; you can look fine on attention and still be vulnerable in mood, blood pressure, or gene expression. As long as you are not hurting anyone else, it is that person’s choice. He will salute the flag of great heights. The only real danger is never asking what you are optimizing for. Plenty of people look back at the end of life and do not wish they had worked harder. Others, including people in medicine, shaved years off their own lives to solve a problem the rest of us now benefit from. If his appendix ruptures at 4 a.m., he is grateful for the team that is awake.

His job, as he defines it, is to turn unknown unknowns about sleep into knowns, then let you run the algorithm. He will not tell you the output. The scene he does not want is someone starting a job at 32, waking up at 65, and asking where the life went. Fridman comes out with it: he enjoys the madness, including passing out on the carpet after a night of programming, and he wants to be embraced rather than scolded. Walker’s reply is gentle. He wants Fridman to have as much high-quality life as possible. If that means five decades on the planet, thrilled every day by the work, that is a 50-year journey of delight. Both men use mortality as a lens, not a terror: stand at the future point of death, look back, and ask what you would regret in the life you have now.

Being a sleep researcher, Walker says, does not make you immune. He has had three or four severe bouts of insomnia, usually reactive, triggered by stress. Insomnia is the condition in which sleep controls you rather than the other way around. On a bad night the knowledge becomes a trap. He lies there thinking his dorsolateral prefrontal cortex is not shutting down, noradrenaline is not ramping down, and the sympathetic nervous system is not giving way to the parasympathetic. At that point he is dead in the water for the next two hours. The Woody Allen neurotic of the sleep world, he calls it.

The thing he is least proud of is years spent railing against his chronotype — whether you are a morning type, an evening type, or in between. Society is desperately biased toward morning people and the early bird catching the worm. Maybe, he says, but the second mouse gets the cheese. People tell him they have terrible insomnia; he asks whether the problem is falling asleep or staying asleep, then: on a desert island with no responsibilities, when would you sleep? Midnight to eight, they say. They currently get into bed at ten because work starts early. That is not insomnia. That is a mismatch between biology and schedule. Align the two, and it often disappears. He is “desperately vanilla,” neither strong morning nor evening, naturally up around 6:45, alarm at 7:04 because why not be idiosyncratic. For a while he thought that if he was not in the gym by seven something was wrong with him. He abandoned the shame. Fridman is happiest going to bed at 5 a.m. — an extreme evening type, which has the bonus of sleeping through the morning’s email drama. Fight biology, Walker says, and you normally lose. The way you know you have lost is disease.

The best treatment on offer is cognitive behavioral therapy for insomnia, or CBT-I: about six weeks with a therapist, including online. In the short term it matches sleeping pills; after you stop, the benefits last for years, whereas pills typically bring rebound insomnia. Walker’s own four tools: first, actually address the stressful thing, and remember that night thoughts are bigger and more catastrophic than the same thoughts in daylight. Second, keep regularity — and after a bad night, do nothing. Do not sleep in, do not go to bed earlier, do not nap, do not drink extra coffee. Sleep in and you will not be tired at the usual time; go to bed early and your rhythm will not want you asleep; a nap is a valve on a pressure cooker, taking the edge off the sleepiness you need at night, like snacking before the main meal. Third, meditation, which he dismissed as woo-woo until the data were compelling; he has not stopped in six years, and meditation apps discovered a hidden evening peak of people self-medicating insomnia. Fourth, a wind-down routine. Sleep is not a light switch. It is landing a plane. Adults abandoned the bedtime ritual they still give children, and they need it back.

Fridman’s days sometimes run 28 or 30 hours, averaging perhaps six hours of sleep per 24, and he has gotten good at the chaos. The closest evidence, Walker says, is rotating shift work, and the news is not good: higher incidence of depression, diabetes, cardiovascular disease, obesity, and stroke. The circadian clock, the suprachiasmatic nucleus, would have the personality of a creature of habit. Sleep, Walker keeps saying, is an investment, not a cost. Effectiveness when well slept typically exceeds effectiveness when not: why boil water on medium when you could do it in half the time on high? After about 20 hours awake, a human being is as cognitively impaired as someone legally drunk. He does not know a CEO who would brag about a team that is drunk all the time, yet the airport warrior — three time zones, email at 2 a.m., in the office at six — still gets lauded. That sleep machismo is not Fridman’s problem. Fridman is driven by passion. If you skip sleep because you think sleep is stupid, you are wrong. If you skip it because you are in love with the work, that is a gray area — just know what you are trading, and that the thing you get is worth it.

Fridman does intermittent fasting, sometimes one meal a day, sometimes none. On time-restricted eating, Walker would like more data. Animal studies looked promising; human replications have been more disappointing, including work nearby at UCSF. Three decent studies he knows — two in obese people, one in healthy-weight people — found no advantageous benefit to sleep. Time-restricted eating did not necessarily harm sleep. It did not seem to improve it either.

True fasting is a different state. The best evidence comes from religious practice, especially Ramadan, 29 to 30 days from sunrise to sunset. Walker lists five changes, none of them particularly good for sleep. Melatonin release decreased and came later. People found it harder to fall asleep. Total sleep decreased. A wake-promoting chemical, orexin, increased — which is why fasters often feel they can stay awake longer. Deep sleep seemed unaffected; REM, or dream sleep, decreased. REM matters for emotional first aid, memory, creativity, and hormone function, including testosterone, which peaks just before and during it. When orexin is deficient, the result can be narcolepsy. When it is high, you stay awake.

Why would a fasting brain release a wake-promoting chemical? One of the few times animals undergo prolonged sleep deprivation is starvation. The signal says you cannot find food in your usual perimeter, so stay awake longer and travel farther. The brain does not know about Safeway. Fridman has fasted 48 and 72 hours and felt he needed less sleep, with a suspicious amount of free focus. Walker’s translation: your sleep need did not drop. Orexin hit the overdrive switch. Subjective wide-awakeness and a reduced need for sleep are not the same thing. Fasting and time-restricted eating have other health benefits, including hormesis — what David Sinclair has boiled down to “what doesn’t kill you makes you stronger.” Sleep just does not appear to be one of the things they improve.

One camp in sleep research treats dreams as meaningless heat from the light bulb of REM: you built the bulb for light, heat is just what happens. Walker does not believe it for a second. Dreaming as a conscious experience is probably more metabolically costly than REM without the experience. Whenever nature burns ATP, there is usually a reason. Dreams, they have since found, have functions. One is creativity. Another is overnight therapy. Dreaming is emotional first aid. During dream sleep you take difficult, sometimes traumatic experiences from the day, and the state acts like a nocturnal soothing balm, taking the sharp edges off so you come back feeling better. It is not time that heals all wounds. It is time during dream sleep. He calls it emotional windscreen wipers.

Content matters, not just the fact of dreaming. People who learned a virtual maze and then dreamed of the maze itself were the ones who navigated it better on waking; people who dreamed, but not of the maze, were no better. Necessary, not sufficient. The same pattern shows up in mental health. After a painful divorce, those who dreamed of the difficult event itself went on to more clinical resolution of depression a year later; equally frequent dreamers who did not dream of the trauma did not. From an engineering view, Fridman hears a virtual-reality test space that is largely consequence-free, a crude cousin of self-play in reinforcement learning and of data augmentation in computer vision. Anatomically, the prefrontal cortex — the CEO of the brain — shuts down in REM, while emotional, memory, visual, and motor centers become more active, some more than in waking. Chemically, REM is the only time in 24 hours when the brain is devoid of noradrenaline, which normally narrows attention to a spotlight. Noise gets injected into the architecture. The prison guards are away, and the prisoners are running amok in a delightful way.

Sleep also changes whether you can see beauty. Split emotion into positive and negative, and sleep deprivation does raise the negative. The larger hit, by something like an order of magnitude, is to positive emotion. You stop gaining pleasure from normally pleasurable things: anhedonia, the state often called depression, which surprises people by being more about lost pleasure than wall-to-wall gloom. It happens quickly. Fridman recognizes the stroll in that direction — meaning drains out of the things that usually give it — and he uses the state as information, especially combined with fasting, to notice what is going wrong, without acting on it until he has slept. Walker adds a harder fact. Insufficient sleep is strongly linked to suicide ideation, attempts, and, tragically, completion. In 20 years of studying sleep, they have not discovered a single psychiatric condition in which sleep is normal. That tells him sleep is a potential causal agent, and a tool for prevention. He quotes the entrepreneur E. Joseph Cossman: the best bridge between despair and hope is a good night’s sleep. He spilled hundreds of pages trying to say the same thing. Cossman did it in one line.

Asked for the meaning of the whole ride, Walker answers in kind. For him personally, the meaning of life is to eat, to sleep, to fall in love, to cry, and then to die — and probably race cars in between. Sex, they agree, is a topic for another three hours. Fridman closes with Nikola Tesla, the man they had just discussed as a two-hour sleeper: all that was great in the past was ridiculed, condemned, combated, and suppressed, only to emerge more powerfully from the struggle. Both things can be true at once. Tesla’s line about struggle, and the Harvard finding that sleeping under six hours triples the risk of calcifying the artery that keeps the heart alive, sit in the same conversation. Walker’s offer is still only information. What you do with the night is yours.