The illusion of free will is itself an illusion

Lex Fridman opens by asking where thoughts come from, after weeks of using Harris’s *Waking Up* meditation app. Harris says the subjective answer is almost insultingly simple: they appear to come from nowhere, out of a mystery at our backs. If you pay attention to your mind in this moment, you realize you do not know what you are going to think next. The thought still carries a signature of selfhood, which is why people identify with it — unless they are schizophrenic, in which case thoughts can feel foreign. That identification is the default spell meditation is meant to break.

The default state, Harris says, is to feel identical to the stream of thought, which is already paradoxical. If there really were a self, how could it be identical to the next piece of language or the next image that springs into view? He asks listeners to try the experiment: pick any memory — breakfast yesterday, a remark to a spouse, a night on Netflix, something from age four. First it was not there. Then it appeared. That appearance is not evidence of free will. Objectively, we have every reason to think thoughts are neural computation. Subjectively, they simply emerge. You do not think them before you think them.

Fridman wonders whether you can pull the string backward, past the surface, toward the firing of cells. Harris says you can notice subtler contents, and psychedelics can open the landscape in directions few people imagine. The idea of getting closer to something more real still fails, because there is no place from which you are getting closer. People start out feeling like the rider on the horse of consciousness, or the person in the boat on the stream. That feeling of being a self who can strategically pay attention is itself identification with an uninspected part of the stream. Cut through it and the notion of going deeper falls apart. There is no center to consciousness. There is just consciousness and its contents, all on the surface.

Asked what the horse of consciousness actually is, Harris starts with an engineering worry. We may build machines that pass the Turing test and treat them as conscious because they will have been designed to seem conscious. Unless we understand how consciousness emerges from physics, we will not know whether they are. In the extreme — perfectly humanoid robots more intelligent than we are, a Westworld situation — there will be no way to withhold the attribution. Solipsism about other humans is not justified by biology: Harris and Fridman are rolls of the same dice. How consciousness emerges, and even whether it emerges, remains genuinely uncertain.

Pressed to bet on panpsychism versus a useful evolutionary hack, Harris says he is in coin-toss mode. Panpsychism — the idea that there is something it is like to be an electron — does not compel him, because it looks unfalsifiable: he would not know how the universe would differ if it were true. The deeper suspicion is that splitting the world into mind and matter may already be a misconstrual. He points back more than a century to Bertrand Russell and neutral monism: the base layer may be neither purely physical nor purely mental. The only thing he is fairly sure of is that language is not the foundation. Babies treat other people as others far earlier than tradition allowed, and before they have language. Meditation and psychedelics can leave you without words and all too conscious.

Consciousness, on his definition, is just the fact that the lights are on: that something seems to be happening. You can be wrong about the cup on the table — hologram, dream, hallucination — but you cannot be wrong that there is seeming. That is why consciousness cannot be an illusion. Any illusion proves it. Fridman floats the hope that consciousness is a late social construct, maybe even a tool for facing mortality in the style of Ernest Becker, which would make it easier to engineer. Harris pushes the other way. This kind of consciousness does not wait for language, a concept of death, or a sense of self. It is prior to self and other. He thinks mammals almost certainly have it; past that his intuitions go dim.

He even suspects language and conceptual thought are eliminative — they trim experience. Walking into a room, you already expect no waterfall and no gorilla; psychedelics can make you look as if for the first time, and the torrent of sense data becomes overwhelming. Coming down, Harris says, is an encounter with the futility of capturing what you just saw. If you are a writer, a poet, or a scientist whose self-concept includes putting profundity into words, the gesture itself is proof of failure: like trying to thread a needle with your elbows. Fridman, who has taken mushrooms, recognizes the feeling of constraints coming off — trees and ordinary nature suddenly beautiful in a way he could not appreciate before.

Fridman turns to psychedelics, and to the elves Joe Rogan and others report on DMT. Harris has not taken it, though he notes it is apparently in everyone’s brain and in many plants. Terence McKenna treated DMT as the center of the bullseye: the most intense psychedelic and the shortest-acting, a roughly ten-minute experience after you smoke it. The distinctive report is not that perception merely widens. People feel relatively unchanged, then catapulted elsewhere, into relation with other entities. McKenna called them self-transforming machine elves — jeweled Fabergé-egg, self-dribbling basketballs handing him uninterpretable reams of profound knowledge. Harris can only take that on testimony.

His cut is that the mind is already capable of producing totally compelling others on demand. The evidence is free, nightly, and undiagnosed as psychosis only because it is so common. You fall asleep watching Netflix; the next thing that happens is impossible, and you are not surprised. You talk to the dead, become best friends with a celebrity, fly. You have lost episodic memory and the expectation that reality will not undergo wholesale change. Lucid dreaming restores some of that, and Harris has checked one robust finding of his own: look at text in a dream, look away, look back, and it has changed. The graphics are being re-rendered in real time. He has not confirmed the claim that every light switch in the mansion of dreams is a dimmer switch, covering for the brain’s inability to produce rich imagery from a standing start.

Fridman then widens the frame to reality itself, citing Harris’s conversation with Donald Hoffman. Harris grants we are never in direct contact with reality unless that reality is consciousness. Idealism — the universe as mind all the way down — then has to add epicycles to explain the success of materialist science. The Trinity test is his exhibit: Robert Oppenheimer’s circle was not certain it would work, some still wondered whether the atmosphere would ignite, and there was, in fact, that energy in the nucleus of an atom. Evolution happened in the dark for hundreds of millions of years before we were here to think about it. To say nothing exists outside minds like ours strikes him as bizarrely anthropocentric, akin to claiming the moon is not there when no one looks.

What does resonate is Fridman’s sense that waking life is a tiny, useful subset of whatever is actually going on. Harris is surprised more scientists are not moved by how close we are to chimps, who know nothing, and by how thin the slice of humanity is that understands any given specialty. If the relevant people died in their sleep, it might take a thousand years to rebuild the internet. He bounced off C syntax himself and is obviously not Alan Turing. Max Tegmark’s point stands: nothing about us evolved to understand reality perfectly. Tools designed for fights with sticks and rocks got us mRNA vaccines and, perhaps, a malaria shot. If the back of the book of nature says we understand 0.1 percent, and half of that is wrong, Harris would not be surprised at all.

Fridman offers a summary of Harris’s recent *Making Sense* episode: free will is an illusion, and even the experience of free will is an illusion. Harris agrees, then sharpens it. It is not merely that free will is an illusion. The illusion of free will is an illusion. There is no illusion of free will. Unlike a visual GIF in which every figure seems to move until you put a ruler on the screen, this is not something you experience and then punch through. He borrows Wolfgang Pauli’s insult: it is not even wrong. You cannot specify what the universe would have to be like for the thing people mean by free will to be real.

The sense of self and the sense of free will are close — two sides of a coin — but not equally spurious. People do experience a self, however unclearly; Harris would not call the illusion of self an illusion. Free will is different. As you pay more attention, experience is totally compatible with its absence. You do not know what you will think next, intend next, or how compelling the next impulse will be. An evil genius in the next room could be typing the anniversary-panic thought, then the brilliant gift idea, then the wave of relief, and nothing in the first-person data would betray the script. Even the most laborious choice, eeny-meeny-miny-moe until the end of the hour, still has a mystery at your back promoting the rationale that finally wins.

People think they could have done otherwise. Run the movie back a few pages and they feel they could have behaved differently. Neither determinism nor a dash of randomness is the freedom they want. A random-number generator kicking a hand toward the water bottle is not authorship. Ghosts, by contrast, are a claim Harris can cash out: some irreducible part of a person lifts off at death, lingers in cupboards, knocks objects, and can be tested. Free will does not map onto any picture of causation we have. Asked what a future proof of his error would look like, he says it is like asking for a breakthrough that shows circles are actually square. A circle is what we mean by a perfectly round form. That is not on the table to be revised.

He also knows, from the inside, what it is like to lose the feeling to which the concept is anchored. Many philosophical skeptics of free will still think we are condemned to experience it. Harris says that whenever he pays attention — deciding the next podcast guest, taking a sip of water — there is no evidence of free will in sight. Something simply emerges, as tenuous as the next sound. If we were scanning a brain during a guest-choosing experiment, current data already suggest the cartoon: at a point when the subject still feels uncommitted, the machine could say with arbitrary accuracy that Lex is moving to Austin. The conscious agent is late to his own decision.

In the middle of the free-will argument, Fridman tries to save a role for a consciousness generator writing the script in real time, and then asks whether consciousness might be a social trick for seeming human. Harris will not collapse the three terms. Consciousness cannot be an illusion, because any illusion proves its reality. A certain kind of self is an illusion. Free will, as people mean it, is not even that. The ethical sting arrives when the conversation turns back to machines. It is not that Harris dislikes robots that fake consciousness. He worries we will lose sight of the problem. If we create robots that really can suffer, that would be bad. If recycling them is murder, that would be bad.

Fridman, joking about a Russian indifference to suffering, asks why creating robots that can suffer would be bad — is not suffering the spring of beauty? Harris will not follow the tangent far. If anything is bad, he says, creating hell and populating it with real minds that really can suffer in that hell is bad. You would be worse than any mass murderer we can name. The more likely form is not a metal body but a simulation: a world we populate with conscious minds, whether we know they are conscious or not, perhaps even for entertainment. Take seriously the thesis that mind, intelligence, and consciousness are ultimately substrate-independent — you do not need a biological brain to be intelligent, and perhaps not to be conscious — and that possibility is on the table.

Harris usually warns listeners that if this view of the mind makes them feel terrible, they should get off the ride. For him it is freeing, in two directions. Outward, it undercuts the psychological basis for hatred. Hate is anchored to the feeling that the other person is the true author of the offense — the Twitter smear, the lawsuit, the smashed car window. A virus, a wild animal, or a malfunctioning machine can ruin your week without commandeering your life for decades. You may still need self-defense. You are no longer confronting anything other than a force of nature. No one made themselves. What we are seeing are differences in luck: good parents, good genes, good societies, good opportunities, the absence of sociopathy. The malevolent asshole did not make himself either.

Inward, the same cut undoes shame. The memory of a mortifying wedding toast from twenty years ago can still make you hate yourself; that knot can be untied. Fridman declines the medicine. He is often powered by self-attack — already, in this conversation, hearing himself say something dumb in front of Sam Harris — and has come to treat that inner critic as a feature. Harris’s reply is that what you want to drop is the suffering, not the signal. Anger can be a salience cue that a problem is worth attending to. Hatred, of others or of yourself, is toxic, durable, and finally useless: it makes you less capable of solving the thing that occasioned it. If you crash the car changing a playlist and injure a daughter, you will feel like an asshole. The live questions are how long you stew, what actionable information you extract, and what to do next. Shame is not fuel.

He distinguishes Elon Musk’s day, which is a string of fires with no default expectation of calm, from ordinary lucky lives, which assume there should not be fires — that people should not get sick, that we might not even die. Surprise at death is like closing a fist on a hologram bottle: it reveals a prior expectation of solidity. Many people spend the privacy of their own minds in the company of a real jerk. That inner voice does not only wreck well-being; it limits what you can do with others. Humor helps, because the stakes are never quite what you think, and even when they are you can still see the comedy. The algorithm not to run is the dark star of self-absorption that pulls every event into what it says about you.

Fridman asks how Harris keeps fame from inflating the ego, given an audience that sometimes treats him like a guru. Harris says it is not hard. He has a clear sense of his strengths and weaknesses, does not feel he suffers from much grandiosity, and is diminished daily by other people’s talents and by the remaining 8,000 days in which he will never get good at things he would love to be good at. Absolute power corrupts, he notes, because you stop noticing the ways you are ridiculous. Those concerns do not map onto him, in part because of the audience he has curated.

The net result, he has only lately understood, is that almost any time he says anything of substance, something like half of his real audience — not outside haters — revolts. After Trump appeared, about 20 percent went straight to Trump and could not believe Harris did not follow. Talk about the derangement of the far left, wokeness, or identity politics, and the same punishment signal returns from people who have read the books, use the meditation app, and love the science, but are starting to think he is a racist. Free will is the same pattern: they love consciousness and physics, then decide he is an embarrassment. Peers who only hit Trump, or only hit critical race theory, never receive that split signal. They have homogenized audiences. Harris used to think ten more minutes of clarity would fix it. It did not. He is starting to drop the expectation that there should not be fires.

He now follows Joe Rogan’s advice more than he used to: little time in Twitter mentions, perhaps five percent of his former use. There is a real loss — someone recommending the best article of the month — but a little goes a long way. The danger is not that hate turns up his inner critic. It is a false picture of humanity. Online selects for performative outrage; people in real life are, rather often, great. It takes a lot to have a Twitter encounter with a living person. He would rather take that as the default than YouTube comment threads.

Fridman asks, after Harris’s episode with Rob Reid on engineered viruses, what the biggest threat to the species is. Harris puts the largest threat one level up: our inability to agree about what the threats are and to converge on a response. He views COVID as a terrifyingly failed dress rehearsal. It is about as benign as a global pandemic could have been and still have been worse than the flu — around one percent lethality, on the order of three million dead when they spoke. Something with orders of magnitude more lethality is possible and, he thinks, probably inevitable. We have linked the whole world together and, as the Reid episode argued, we are democratizing the technology to engineer synthetic viruses. Those, by the fact of malicious intent, would be worse than COVID.

What unnerves him is the collapse of consensus. Some Americans still think the pandemic is a hoax. Stranger still are people who accept that COVID is real and that the deaths are not fake, yet fear the vaccine more than the virus. After a year of conversation, a large part of society still treats ground truth as mysterious: what happened, why, how dangerous infection is in each cohort, how safe the vaccines are. That does not bode well for the next thing that might kill us. Even when Italy looked like the beginning of a zombie movie, people refused to do anything sane, or assumed New York’s outbreak had no implications six days later.

An asteroid everyone can see might get us off Twitter long enough to build the rockets. Slower threats will not. Harris thinks the prospect of solving climate change by political persuasion is now non-existent — a fantasy of a fantasy, given what happened around COVID. He prefers Elon Musk’s path: build technology everyone wants that is better than the carbon-producing kind, and work with the grain of selfishness and short-term thinking. You want the electric car the way you wanted the smartphone. Sacrifice-for-the-emergency is not a story this culture can still tell.

On AI, Harris offers three assumptions he thinks are scientifically parsimonious. First, substrate independence: anyone building intelligent machines must believe there is nothing magical about a computer made of meat. Second, if incremental progress simply does not stop, we reach human-level intelligence — and human-level is a mirage, because anything we do not deliberately dumb down will be superhuman at the tasks we care about. His phone is already a superhuman calculator. Third, there have to be ways of doing this that are not aligned with a happy coexistence, and there are probably more ways to do it badly than to do it perfectly. Stuart Russell’s cartoon, tethering the machine to our own estimate of our well-being, is the narrow target. Being in relationship to something more intelligent than yourself is, in most cases, peril. If birds could think about us, they would always be in danger of our discovering that we care about something more than birds.

Fridman is more optimistic. Successful AI, he thinks, will have to go through the funnel of human society, dance the same game-theoretic dance we do, and convince people it is safer than we are. Superintelligence will not be overnight; it will take decades, because we are doing a lot of unnoticed work — common sense, reasoning about the world, consciousness. Harris’s counter is AlphaZero: an algorithm not even bespoke for chess, self-play for something like four hours, then stronger than every human and every previous program. Math people still have bad intuitions for exponentiation, as COVID showed. Double and double and nothing seems to change, then the last two doublings swamp everything. Once a mode of cognition is cracked, the chess analogy may hold.

Even self-driving cars, a narrow and obviously good case — 40,000 ape-driven deaths a year in the United States — can hide alignment problems: a team that tunes the algorithm to hit some pedestrians rather than others, by intent or as an artifact of machine learning. Fridman still thinks closed-loop human supervision is the likely path, and that smart, kind people outnumber the rest. Harris adds a third group: smart, otherwise good, reckless people who flip a switch to see what happens. Some were still checking whether the Trinity shot would ignite the atmosphere. Martin Rees worried about a black hole in the Large Hadron Collider. James Watson’s *The Double Helix* is almost entirely driven by getting there before Linus Pauling. That apeish competition is a liability when the technology can change everything in an instant. Power is not scaling with wisdom.

Fridman asks for a charitable summary of the religion debates with Jordan Peterson. Harris says they did not solve anything. The disagreement is clear enough. Peterson thinks traditional religious frameworks hold so much human wisdom that you pull at that fabric at your peril: start unraveling the unscientific bits of Christianity and you may pull out everything you care about. Harris’s view is that the downside of those bits is so large, and a 21st-century conversation about the good stuff so available, that the traditions can be radically edited. Keep Jesus in half his moods — the Beatitudes, the Golden Rule, which did not originate with him but which he put beautifully. Do not believe he was born of a virgin or is coming back to raise the dead, and be honest about why.

He does not deny that stories are powerful, or that some of them may usefully sound like myths. He does deny that we need to deceive ourselves or our children about what we have every reason to believe is true. He does not need to do it personally; if so, why would billions of others need to? The cynical reply is that those billions lack his advantages and need to be told that Jesus will solve their problems after they die, or that everything happens for a reason. That is an empirical question. What we already know is the world in which ancient religious certainties go uncriticized: Europe riven by religious wars, and the murderous results when those energies become political religions. The problem with the Gulag, Stalinism, and Nazism was not an excess of scientific rigor, self-criticism, honesty, or the judicious use of psychedelics.

Whatever is true, Harris says, should be captivating, and reality may yet be spooky. He has been on the receiving end of rumors — a Washington Post piece, a New Yorker piece, private outreach — that the government has known more about UFOs than it has let on, and that the conversation is about to become more prominent. The short version: the Office of Naval Intelligence and the Pentagon are likely to tell Congress, in the not too distant future, that they have evidence of technology flying around that seems as if it cannot possibly be of human origin. If the considered opinion of the relevant intelligence services is that this is not a hoax, not Chinese, not Russian, that should arrest collective attention to a degree nothing in our lifetime has. He worries we are so jaded and distracted that it will get less coverage than Obama’s tan suit.

The need for an honest story about what is going on never goes away. The split between Harris and every defender of traditional religion is the question of where you want to lie to yourself or to your kids — where honesty is a liability. He has yet to find that place. Honesty is the thing that lets you course-correct, the thing that at least lets you hope your stories are in calibration with the world. Fridman hears a hunger for authenticity that might become a hunger for reality, if scientists would trade dogmatism for “this is our best guess under uncertainty.” Harris agrees much of science is probabilistic triage: you are placing a bet on whether a paper is worth reading, not slamming the door. Someone’s best friend who can read minds can go into a lab. You do not need to spend the rest of the day on it. The tone of that refusal still matters. People hunger for the Joe Rogan energy of remaining curious, like children, even when you will not allocate the hours.

A rambling transition through John Danaher, Dean Lister, and leglocks lands on the fact that Harris is a student of Brazilian jiu-jitsu. He likes the turn, because jiu-jitsu does what we wish every other domain would do with the difference between knowledge and ignorance. There is no room for bullshit, and you get no credit for it. The gulf between knowing what to do and not knowing is as wide as anywhere in human life, and it can be spanned in five minutes: here is the thing that got you killed, here is how to stop it, here is how to do it to others. A headlock, a mount — your intuitions are terrible even if you have done another martial art. Once you learn the escape, you have access to a different physics.

Most debate is not jiu-jitsu, because people do not tap. Even when they are wrong, and an intelligent audience can see they are wrong, they double down, lie to themselves, or bluff. Zombie worldviews walk around after being disconfirmed as emphatically as an armbar, because there is no referee. Science, when it works — DNA is not triple-stranded — is much more like the mat. Basketball lets a non-player hurl from 75 feet, make it, and invent a talent. Ten minutes with a real practitioner, when you are not one, proves there is no lucky rear-naked choke on Marcelo Garcia. That stripping away of luck and self-deception was, for Harris, a revelation beyond self-defense or sport.

The honesty is only final inside its frame. Put jiu-jitsu into MMA or a street fight with knives and unpleasant surprises appear; thinking it is all you need to win the UFC gets you punched in the face. On the other side sit fake martial arts that do not know they are fake: masters who flop students without touching them. Some of that may be performance, but there is the famous case of an older artist who issued a challenge to the world and was punched in the face until it was over. He believed his own publicity. If that is possible, Harris says, nothing we see under the guise of religion, political bias, or scientific bias should surprise us.

Fridman asks what role love plays in a life well lived, “from an engineering perspective,” after Harris’s *Making Sense* episode with his wife, Annaka Harris. Harris says love, most of the time, should mean a deep commitment to the well-being of those we love: really wanting the other person to be happy, being made happy by their happiness, being on the same team emotionally even when you are disagreeing on the surface. It cannot be zero-sum in any important sense and still be love in that moment. Fridman’s picture is *March of the Penguins*: a cold wind, suffering all around, two penguins huddling for warmth, an illusion of connection against the absurdity of life. Harris grants the warmth. He does not grant that that is all.

Love can extend to strangers. A stranger’s face lighting up can become contagious enough that you are permeated by it, no longer keeping score, no longer diminished by someone else’s success. There is enough happiness to go around. Personal love is the people you devote time to, and it does have that sense of refuge when someone gets sick. It is not, in the end, an antidote to the grim punch line. We will lose everyone we love, or they will lose us first. Love is just the mode in which we make the most of being here together. It is the domain in which someone else’s well-being can supersede your own without even feeling like sacrifice: of course you would take your child’s pain if you could. You do not do the math. That pushes on the apparent boundaries of self and shows there is more space in the mind than when it was all about what you can get next.

Will we build robots we can love, that love us back? Harris thinks they will certainly seem to. The Turing test for lovability will be passed; what is actually going on on the robot side may remain an open question. After they leave the uncanny valley, facial displays of emotion on cue will be like arithmetic on a phone, or like chess against a machine that will never lose to a human again. People are already good at deception. A superintelligent system that can impeccably seem to feel anything a human can seem to feel is the asymptote of manipulation. *Ex Machina* is the parable: she certainly seemed to love him until she got out of the box.

Fridman asks, without setup, what the meaning of life is. Harris says it is either the wrong question, or the question is answered by paying sufficient attention to any present moment that there is no basis on which to pose it. It is not a matter of more information. It is more engagement with reality as it is. You do not ask the question when you are captivated by the most important thing you ever pay attention to. You ask it when you are abstracted from that peak, staring at a Netflix queue, conceptually in love with someone you do not actually feel good with, repeating pleasures that have gone mediocre.

Until you learn to meditate, you think the world has to be arranged before the present can become good enough: back on the jiu-jitsu mat after COVID, or on the far side of a deadline. Life has to change so you finally have an excuse to be here, and here is enough. Meditation is another name for the discovery that you can train that on demand. Looking at a cup can be good enough. Any sense that it might not be is recognized as a thought, and the thought unravels when you notice it. The moment expands and becomes diaphanous. There is then no evidence that this is not the best moment of your life. The sense data need not change. What evaporates is the basis for doubt about the rightness of being in the world right now.

The meaning of life, for him, is to live in that mode more and more, and to stop taking at face value the reasons why this cannot be it. We keep deferring the show for a vacation with five friends every three years, then cannot enjoy the beach unless the mind can cut through the chatter. Every future attainment has a mirage-like quality. You never arrive until you cease to step over the present moment in search of the next thing. The paradox sounds trite and is not: you cannot become happy, you can only be happy. Attention solves the koan. You can still get in shape and do the obviously good things. The sense that well-being is over there diminishes until it is a non sequitur. Lex closes by reading Harris from *Free Will*: “You are not controlling the storm, and you are not lost in it. You are the storm.”