Not on My Grave

My barber Tommy was standing behind me with a pair of clippers when he asked whether I had seen the news about the artificial-intelligence researcher who thinks AI might kill everybody. This is apparently what barbershops are for now. There was a time when a man sat in a barber chair and was asked whether he wanted the sides tapered. Perhaps someone argued about baseball. Maybe an elderly man complained about taxes. Now you sit beneath fluorescent lights wearing a black polyester cape while a man holding a razor near your carotid artery asks whether you have emotionally prepared for the extermination of the human species by self-improving software. “No,” I told him. “I try not to read the news anymore.” Tommy stopped for half a second. Not dramatically. Tommy is a barber. Barbers understand that civilizations rise and fall but the fade must remain symmetrical. I explained that the news had begun to seem less like information and more like a subscription service for problems over which I possessed no jurisdiction. Every morning, somewhere between coffee and brushing your teeth, you are invited to contemplate a new mechanism by which everything you love may disappear. War. Fascism. Climate collapse. Financial collapse. Pandemic. Artificial intelligence. Nuclear weapons. Microplastics. Apparently your cutting board is killing you too. The extraordinary thing is that after informing you of civilization’s imminent destruction, the article usually ends by asking whether you would like to enable push notifications. Tommy laughed. Then I went home and, naturally, looked it up. This is one of the peculiarities of modern man. He says he has escaped the news and then spends forty minutes researching the apocalypse his barber mentioned. There he was. Young. Brilliant. Cambridge. Serious eyes. The sort of face on which the beard has arrived before the authority has. A researcher from one of the most sophisticated artificial-intelligence laboratories in the world explaining, with unmistakable sincerity, that the machines being built by people like him might conceivably kill all of us. Not metaphorically. Not “destroy traditional retail.” Not “disrupt the advertising ecosystem.” Kill everybody. The species. Homo sapiens: 300,000 years, good run, unfortunate quarterly guidance. And I stared at this young man and experienced an emotion I had not expected. It was not fear. It was irritation. I wanted to reach through the screen, straighten his collar, hand him a juice box and say: My dear boy, who exactly plugged the thing in? Because something extraordinary has happened to the language surrounding artificial intelligence. The machine is spoken of as though it arrived from another galaxy. AI will become more intelligent than us. AI will take our jobs. AI will manipulate governments. AI may escape. AI may kill us. Listen carefully to the grammar. AI is always the subject. Human beings have disappeared from the sentence. Apparently the machines financed themselves. They incorporated in Delaware. They raised billions of dollars. They negotiated the electricity contracts. They purchased warehouses full of GPUs. They hired the engineers. They built the cooling systems. They crawled out of the Pacific Ocean one moonless night carrying extension cords and said, Gentlemen, we require three gigawatts. At some point, one must admire the achievement. We have created the first industrial revolution whose architects narrate themselves as bystanders. And this, I began to realize, was why the young man’s frightened face bothered me. Not because he was young. Not because he was frightened. He may have excellent reasons to be frightened. Perhaps he understands the danger much better than I do. The strange thing is that nobody older seems to be standing behind him. Once, societies had people whose function was not merely to know things, but to have survived things. Every civilization arranged this differently. There were elders, priests, grandmothers, judges, village councils, teachers, rabbis, imams, tribal leaders, statesmen. Some were wise. Some were idiots. Some undoubtedly gave terrible advice while smelling faintly of onions. But the social function existed. These people possessed something the young did not yet possess: time. They remembered the famine. They remembered the war. They remembered the winter when the river froze. They remembered which neighbor betrayed whom and which feud nearly destroyed the village. They knew that the world had ended before. Several times. And somehow breakfast had still been served the following morning. Their authority therefore rested not merely on information but on memory. A frightened young person could come to them and say, What happens if the wolves come? And the elder did not open a PowerPoint presentation. He said: We build a fence. What happens if the harvest fails? We ration grain. What happens if the neighboring kingdom invades? We send word to the valley and prepare the walls. Modern civilization has improved this process considerably. Today the child says, “What if the wolves come?” A twenty-seven-year-old computational researcher adjusts his microphone and says, “Our current models suggest a nontrivial probability of total village extinction.” This is what we now call expertise. I am not mocking expertise. Expertise is indispensable. If my kidney fails, please do not summon the village elder. Take me to the nephrologist immediately. But expertise and wisdom are not the same faculty. Expertise tells you how frightened you should be. Wisdom tells you what fear requires of you. And somewhere along the road to becoming the most technically sophisticated civilization in human history, we misplaced the second half. Everyone is frightened now. Nobody is the adult. The strange thing is that the most powerful people are often the most helpless-sounding. The politician explains that the system gives him no choice. The executive explains that shareholders give her no choice. The investor explains that markets give him no choice. The general explains that security gives him no choice. The technologist explains that technological progress gives him no choice. The newspaper explains that the algorithm gives it no choice. The billionaire explains that competition gives him no choice. Everyone appears to command something. Nobody appears responsible for anything. Power has learned to speak in the grammar of victimhood. This may be the defining rhetorical invention of our age. The wolf no longer says, I am hungry. The wolf says, You have no idea how difficult predation has been for me. The wolf has a podcast now. The wolf has trauma. The wolf has retained counsel. The wolf would very much appreciate some empathy while he eats the sheep. And so the people funding the machines tell us the race cannot be stopped because somebody else might win. The people building the machines tell us the machines may become uncontrollable. The people governing the country tell us technology moves too quickly for government. The government tells us corporations have too much power. The corporations tell us markets are beyond their control. The markets, having wisely declined all interview requests, remain unavailable for comment. This is how responsibility evaporates. Not through conspiracy. Through grammar. The most consequential political sentence of the twenty-first century may turn out to be: “There is no alternative.” Whenever you hear it, check your wallet. Somebody is about to do something he very much prefers to describe as physics. There is no alternative to globalization. There is no alternative to austerity. There is no alternative to war. There is no alternative to automation. There is no alternative to the AI race. There is no alternative because China. There is no alternative because markets. There is no alternative because innovation. There is no alternative because history. History has become an extraordinarily convenient employee. She works twenty-four hours a day and accepts blame for everything. But artificial intelligence, for all the theological language now gathering around it, remains humiliatingly physical. It lives somewhere. This seems worth remembering. The god has an address. The god requires cooling. The god requires chips fabricated in factories. The god requires copper. The god requires substations. The god requires transformers. The god has zoning disputes. The god occasionally trips a circuit breaker. The god, for all its rumored omnipotence, receives an electricity bill. Perhaps someday that changes. Perhaps genuinely autonomous systems become extraordinarily difficult to contain. Perhaps they distribute themselves, deceive us, replicate, manipulate infrastructure and make the problem vastly harder. Fine. Then say that. But do not leap from “this may become extraordinarily difficult” to “human beings are spectators before destiny.” There is no destiny here. There are decisions. That does not mean there is one enormous red OFF button in Washington labeled ARTIFICIAL INTELLIGENCE which the president has simply forgotten to press because someone stacked mail on top of it. It means something more important. These systems remain embedded inside human institutions. And institutions can be governed. Electricity can be governed. Chips can be governed. Data centers can be governed. Capital can be governed. Training runs can be governed. Corporations can be governed. Liability can be imposed. Standards can be created. Treaties can be negotiated. Inspection regimes can be invented. International institutions can be strengthened. None of this is easy. But difficult is not the same word as impossible. The disappearance of that distinction may be one of the great pathologies of our age. We have somehow arrived at a civilization capable of putting telescopes in space that can look billions of years backward into the birth of galaxies, yet when somebody suggests regulating a corporation we shrug and say the logistics seem challenging. We can split the atom. We can edit genes. We can land machines on Mars. But apparently a billionaire with excellent lawyers represents the outer boundary of human sovereignty. This is why the deterioration of international institutions matters far beyond diplomacy. After the catastrophes of the twentieth century, human beings attempted something astonishing. They tried to build rooms in which enemies would speak before killing one another. The United Nations was never pure. It was never democratic enough. Great powers protected their interests from the beginning. Hypocrisy was built into the walls. Of course. It was built by humans. But the idea was immense. There are problems too large for tribes. There are problems too large for nations. Therefore humanity must occasionally sit at one table. Then we weakened the table. We ignored it when inconvenient. Powerful states treated international law as binding upon their enemies and advisory for themselves. Governments bypassed institutions, hollowed out agreements, withdrew from obligations and then pointed toward the resulting impotence as proof that international cooperation had always been childish. It is an elegant trick. Break two legs off the table. Piss on the carpet. Then announce that furniture has failed. Artificial intelligence may eventually become one of those questions that no country can solve alone. Not because humanity needs a world government run by humorless Belgian bureaucrats determining your toaster settings. Because the incentive structure is obvious. If one laboratory slows while another accelerates, the first fears losing. If one country regulates while another does not, the first fears losing. Everyone may privately prefer restraint while publicly racing. Human beings have encountered this problem before. It is why treaties exist. It is why arms-control agreements exist. It is why inspection exists. Diplomacy is not what we do when everyone trusts everyone. If everyone trusted everyone, diplomacy would be unnecessary. Diplomacy exists because we do not. And perhaps this is where I finally understood that I had been unfair to the young researcher with the frightened face. He is not the real child in this story. He is simply the child who noticed the house smells like smoke. The real children may be standing behind him: the executives old enough to be his father, the investors old enough to be his grandfather, the political leaders whose offices were created precisely so that somebody would possess the authority to say no, the institutions that have spent decades confusing authority with branding and leadership with communications strategy. They send the young scientist onto television to explain that civilization might be destroyed, and then everyone nods gravely at how complicated the situation has become. That is the diaper. Not his youth. Ours. A civilization in diapers is not one inhabited by young people. It is one in which old people refuse responsibility. We have mistaken intelligence for wisdom because intelligence is easier to measure. You can test intelligence. You can credential it. You can put letters after its name. You can publish it. You can optimize it. You can hire it for seven hundred thousand dollars a year and give it kombucha. Wisdom is more irritating. Wisdom asks what you are willing to sacrifice. Wisdom asks whether your desires should be limited. Wisdom asks what belongs to people not yet born. Wisdom asks whether the thing you can build is also a thing you should build. Wisdom asks whether victory in the race is meaningful if the race itself is insane. Most offensively, wisdom occasionally says: No. And “no” is the word modern power finds almost impossible to pronounce when money is present. The tragedy of our age may therefore not be that our machines become smarter than us. It may be that we become too immature to govern our own intelligence. We know more than any humans who have ever lived. We can summon nearly the entire archive of civilization from a slab of glass while sitting on the toilet. We can sequence a genome in hours. We can model proteins. We can translate languages instantly. We can calculate the probability of extinction to two decimal places. And still somebody must eventually stand up and say: Enough. Not because he knows the future. Because he understands responsibility. Perhaps AI really could kill us. I do not know. Neither, finally, does anyone else. The uncertainty is real. That should make us more serious, not more superstitious. If the danger is small, govern it proportionately. If the danger is enormous, govern it enormously. If the companies object, let them object. If investors lose money, investors lose money. If countries must negotiate, send diplomats. If institutions are too weak, strengthen them. If political systems have been captured by wealth, then confront the capture. If that is expensive, pay. If that is difficult, work. If that requires courage, discover whether we still possess any. This is what adulthood means. Not fearlessness. A child believes adults are people who are never frightened. Then the child grows up and discovers that adults are merely the frightened people who still have to decide what happens next. That afternoon I thought again about Tommy. The clippers moving around my ears. Hair collecting on the black cape. The ordinary intimacy of a barbershop. Two men talking about the extinction of humanity because apparently weather and sports no longer generate sufficient anxiety. And I thought perhaps this is where civilization always actually resides. Not in the laboratories. Not in Washington. Not in corporate boardrooms. Not at the United Nations. Here, in the millions of ordinary rooms where people wake up expecting there to be a tomorrow: a barber raising his children, someone driving home from work, a woman buying groceries, a father waiting outside a school, an old man watering tomatoes, a young couple arguing about rent. People whose lives almost never appear in grand theories of history, although they are supposedly the reason history matters. The future belongs to them. Not to the machine. Not to the billionaire. Not to the politician. Not even to the scientist. And maybe what I wanted someone to say to that frightened young researcher was not Shut up. Certainly not. Warn us. Tell us everything. Show us the danger. Make the charts. Run the models. Sound the alarm. That is your job. But once the alarm rings, somebody else must enter the room. Someone older, perhaps. Or perhaps merely someone who remembers what adulthood is. Someone who does not promise that catastrophe is impossible. Someone who does not pretend courage means certainty. Someone who looks directly at the possibility of failure and says: Then we will do everything within our power to prevent it. If the machine becomes dangerous, we govern the machine. If the company becomes dangerous, we govern the company. If money prevents government from governing, we govern the money. If nations cannot act alone, we build institutions through which they can act together. And if those institutions have been hollowed out, then we repair them. Not because success is guaranteed. Because responsibility does not depend on guarantees. You may frighten me. You may overwhelm me. You may be more intelligent than me. You may calculate faster than me, remember more than me and predict things I cannot predict. You may even defeat me. But you do not get to convert my fear into my surrender. Not while I am alive. Not while there is still a hand capable of reaching for the switch. Not while human beings remain responsible for the world they build. Not on my grave. Then Tommy spun the chair around. He held up the mirror behind my head. “How’s that?” I studied the fade. Civilization, for the moment, remained intact. “Little shorter on the sides.” —Elias WinterAuthor of Language Matters, a space for reflection on language, power, and decline. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit eliaswinter.substack.com