Exponential tools, Stone Age game theory

Lex Fridman opens with a thought experiment. If aliens had watched Earth for the whole of civilization and had to file a short summary — a Hitchhiker's Guide paragraph, not a book — what would they write? Schmachtenberger first refuses the easy anthropomorphism. He has no idea whether their senses or consciousness even track what ours do. Constrain the puzzle, he says, and assume they are enough like us to have gotten here, only more advanced.

Those observers would not see a clean ladder of progress. They would have watched the Library of Alexandria burn. They would have seen that two thousand years ago Greece could produce something like the Antikythera mechanism, a clockwork computer, and then lose the technology. Every so often, he says, there is a giant self-induced fire. The record is not a dialectic of improvement. It is advance, then wreckage, then advance again.

Looking at the present — past nuclear weapons, into globalization, the Anthropocene, and exponential tech — they would notice that the decision procedure has not moved as far as the tools. Humans still allocate power through rivalry and game theory, roughly as they did in the Stone Age. His guess is that this is one of the planets that does not make it to intergalactic life, because it blows itself up in technological adolescence. If it is going to make it, it needs rapid progress in the social technologies that can guide, bind, and direct physical technologies, so that people become safe vessels for the power they are getting.

Most of the great civilizations are gone. Schmachtenberger does not find that obscure. Egypt, the Inca, the Maya, the Aztec — they had life cycles, and they ended, sometimes by war from outside, sometimes by eating themselves. Easter Island is his clean case of self-termination. If an island consumes resources faster than they replenish, and grows its population while doing so, the utilization rate itself accelerates. You do not get a gentle S-curve to a steady state. You get an exponential climb, a wall, and a collapse.

The definition he wants is simple. A self-terminating system is any system that depends on a substrate and is debasing that substrate. Fridman offers the standard optimism: human ingenuity, productivity, and innovation might outrun consumption. Schmachtenberger does not dismiss that hope. He also will not let it travel alone. Naive techno-capital optimism and naive techno-pessimism are both failure modes. Positive possibilities and catastrophic risks are rising together. For a flourishing outcome to last, every catastrophic risk has to not happen. Any one of them is enough to zero the rest.

He wants generator functions, not a catalog of last century's accidents. Joseph Tainter's work on collapsing complex societies, Baudrillard on institutional decay, Jared Diamond's popular *Collapse* — the point of reading them is to abstract why systems fail, then design against those patterns. Slow death is overconsumption: overfishing, deforestation, fossil extraction, pollution, a linear materials economy demanded by an exponentially growing money supply. Fast death is nuclear war, bioweapons, uncontrolled AI, nanotech. Neither replaces the other. They stack.

For most of history, existential risk to a group was local. An empire could fall and neighbors could fill the space. World War II and the bomb, Schmachtenberger says, were the first time humans could mess up habitability at planetary scale. It was not that people had been peaceful before. They had simply lacked the power. Superpowers could no longer fight a real kinetic war they might win. They could fight diplomatic wars, a Cold War, and proxy wars through countries that did not hold the big weapons.

Bretton Woods was the social technology built for that novelty: the United Nations, the World Bank, the IMF, globalization trade agreements, mutually assured destruction. The bet was that deep economic interdependence would make war unprofitable. He grants the result. About 75 years passed without superpower-on-superpower war. The same architecture also pulled the world into planetary boundaries faster. Rapid GDP growth meant overfished oceans, cut forests, climate change, toxic mine tailings. You cannot run an exponentially growing linear materials economy on a finite planet forever, he says, and it has to be exponential if interest and fractional-reserve banking expand the money supply, because goods and services must then grow to match.

Interdependence also created fragility. COVID showed how a problem that started in one part of China could shut travel, starve fertilizer and pesticide shipments, and produce crop failures and locust damage in North Africa and Iran. Cascade dynamics became their own catastrophic risk. Meanwhile the weapons map changed. Two superpowers with one catastrophe weapon became eight or nine nuclear states, plus drones that can hit infrastructure, biology, and other arms races the UN never bound. Mutually assured destruction is a two-body Nash equilibrium. Multiple agents make a three-body problem: if A shoots B, whom do I shoot? Nukes are hard to build. Weaponized drones are not, and non-state actors can reach them. Bretton Woods does not prepare a world of many agents, many catastrophe weapons, planetary limits, and cascade fragility.

Fridman wants to know why dystopia is easier to imagine than a future that works. Schmachtenberger answers with combinatorics, not morbidity. There are vastly more ways to rearrange the tens of trillions of cells in a body that do not make a person than ways that do. You can kill an adult in a second and you cannot grow one in a second. A building comes down in minutes and took years to raise. Entropy is the easy gradient. A protopian path has to avoid every catastrophic branch at once, which is why it is harder to picture than explosion.

He is not only talking about Genghis Khan. Rockefeller and Standard Oil looked like a constructive project: bring a new energy resource to the world. They did. The second-order effects, in his telling, include climate change, oil spills, Middle East wars, and a long political wreckage. Facebook, building a dating app and then a social app for tagging pictures, was not trying to build a democracy-destroying product. Maximize time-on-site for an ad model, let AI optimize a feed for whatever holds attention, and you get limbic hijack, cognitive bias, group identity, and no shared reality. "They weren't trying to do that," he says, "but it was a second order effect" — and a fast one, because exponential tech distributes farther, quicker, with a bigger jump in capacity.

War, for him, is harm we are trying to cause, or at least know we are causing. An externality is harm we supposedly did not intend, or that we accept as a side effect. Catastrophic risk comes from both: an exponential-tech arms race between in-groups and out-groups, and the application of more powerful tech to some ordinary goal whose second- and third-order effects on the commons, other people, or the environment can exceed the original problem. Innovating our way out of each new externality, he argues, ignores that we cannot keep up with exponential curves, and that finite spaces do not allow exponential externalities forever.

GDP is a rotten index of a good civilization, Schmachtenberger says: it rises in war, with healthcare spending on sicker people, and with addiction. He does not believe there is a single right metric. Any frozen metric set becomes a paperclip maximizer, a false idol of the map over the territory. Wisdom, in his phrasing, is discernment that leads to right choice beyond metric optimization. Still, he will play the thought experiment. One index he likes is the inverse of addiction, defined as compulsive behavior destructive of what people themselves value. The more a society produces that compulsion, the less healthy he thinks it is.

Comfort and addiction sit close together. The way off that path is hormesis: exposure to supernormal stimuli without desensitizing to everything else. He thinks a future civilization has to create ritualized discomfort — sweat lodges, vision quests, solo wilderness, ayahuasca, the Sun Dance, even yoga asana — so people become resilient, can face fear, and become trustworthy without being thrown into war or raw trauma. It should not be imposed by states. It should emerge from cultures that understand the value, as a voluntary practice people undergo together so that nobody bails when it gets hard.

The metric he likes even more is the coupling of emotional states. In a healthy bond, your joy and my joy, your pain and my pain, tend to move together: a compassion–compersion axis. Compersion, an uncommon English word he wishes were common, is feeling happiness at someone else's happiness. Decoupling is sociopathy: I do not want to hurt you, and I also do not care. Inverse coupling is worse and, he says, extremely common — a jealousy–sadism axis we have normalized as low-grade psychopathology. Jealousy is feeling worse when you do well; sadism is feeling better when you lose. If he had to summarize, a civilization should optimize for compersion. That rules out self-sacrificing misery dressed as love, and sociopathic or sadistic happiness. It asks for a coupling: my own drive to be happier wants to support you.

Fridman asks for the meaning of life. Schmachtenberger starts with semantics. People fuse meaning and purpose, then ask for the purpose of the universe. Purpose, he says, is a relative, utilitarian idea: the purpose of one thing is defined against other things already assumed to have value. Ask for the purpose of everything, and there is nothing outside it to point to. That does not make existence purposeless. It means the concept is too small. Daoism, in his gloss, talks about the nature of the thing where the why cannot go deeper.

At the scale of a human life he offers three modes that have to stay in a virtuous relationship. Being is taking in and appreciating the beauty of life that is here now — sunset, music, breeze — a mode in which you are not asking what it all means because you are full of experiencing it. Doing is adding to that beauty: art, technology, science, protection. Becoming is getting better at both, able to be more moved and more able to contribute. Any mode alone fails. Pure being slides into hedonism. Pure doing becomes the person who cannot stop and burns out. Pure becoming is the hole of never being enough.

Where the doing comes from matters more than the task list. If it comes from a fullness of being, it is more likely to notice externalities. If it comes from an inner emptiness trying to get full, it will sacrifice other places and keep a chunk of attention locked inward. He rereads the Buddhist claim that desire is the cause of suffering. There is a desire that is lack seeking fulfillment in the world, and it perpetuates suffering. There is also a desire that is already full of the beauty of life and wants to add to it — closer to a bodhisattva vow than to craving. Eastern traditions, he says, leaned into unconditional presence; Western ones treated desire as the source of creativity. Creating from wholeness in service of the whole is a different act than creating from deficit.

Fridman asks whether death is fundamental to a meaningful life — terror in the Ernest Becker sense, or at least the melancholy of an ending. Schmachtenberger has felt fear of death. He has also had moments of thinking he was about to die that felt peaceful and beautiful. Much of what people call fear of death, he suspects, is fear of a story about afterward: hell, a bad reincarnation, the bardo. Fear of non-experience itself looks weaker in ordinary life. Stay up late enough and you long for deep sleep, for experience to stop. That longing is not morbid.

Finite time does more than create urgency. In his observation, most people start to emotionally mature shortly before they die. If you lived forever, you could stay focused on what is in it for you forever. If life with a capital L continues — consciousness, people falling in love, people adding to beauty — and your particular life does not, then your life can still have effects beyond it. Old men plant trees whose shade they will never sit in. Death forces the widening that compersion and compassion describe.

He remembers a poem by Hafez in which the Sufi writer tells a lonely reader to think of him, leaning his spirit across a millennium to offer comfort. Someone centuries ago was already caring about a stranger's loneliness. That, for Schmachtenberger, is among the most beautiful human moves: something that transcends what's in it for me. Einstein's line about the optical delusion of separate things sits nearby. When he imagines the world after he is gone — young people in love for the first time, people stunned by sunsets — he cannot not feel connected. Asked whether it saddens him to be forgotten, he says he already depends on unnamed people without whom he would not be here. Gratitude toward ancestors beyond names, and the question of whether we are being good ancestors, seems to him a healthy idea.

Fridman tries to force a choice: big democratic state, tight parental tribe, or Michael Malice-style anarchy. Schmachtenberger will not take the A-or-B test. He would not like the United States federal government to vanish tomorrow, and he also does not want it to continue as it is. Humans spent a few hundred thousand years in groups under Dunbar's number, often taken as about 150, though Fridman notes the original paper is a range. Below that scale everyone can know everyone, harm is not anonymous, lying is hard to cash, and a council can still hold the big decisions. Past it, you can hide from the group and feel less bound to strangers. A civilization lasts, he says, when binding energy exceeds cleaving energy.

The body is his model for what should replace the binary. Cells, tissues, organs, organ systems — you would not want tens of trillions of cells as a sea of protoplasm and call it pure democracy. Each higher layer is based on the lower ones and also influences them. He wants governance at the individual, the family, the community, the city, the region, the nation, and the globe, with checks, balances, and two-way information flow. Global governance in that sense is not a top-down imposition with no brakes. It is governance at the scale we actually affect, because anyone ruining the ocean ruins it for everybody. He uses the word governance on purpose, as a process, not government as an imposed thing.

Cities matter more to him than nation-states. Nation-states, in his blunt phrasing, are largely fictitious, defined by wars and agreements to stop wars. Cities sit on real proximity: putting certain things next to each other still creates value. He cites Jeffrey West's work on scale: companies and nation-states, built on complicated agreement structures, get diminishing production per capita as they grow past tribal size, while cities get increasing productivity per capita. People can feel agency in a city. Neighborhoods, and non-geographic networks of affinity, should have their own layers. The live question is how to get emergent order — not chaos, and not order via oppression.

Authoritarian states, he notes, are already coordinating more effectively on some long-horizon tasks. China has built high-speed rail at home and abroad; the United States has built none. China brought 300 million people out of poverty in a period of rising inequality elsewhere, and would be the closest country to a closed loop if global supply chains failed — Belt and Road, rare earths, transistor manufacturing. Term limits, built to prevent forever monarchs, have a second-order effect: nobody plans past the next election, and left-versus-right spends its energy undoing the last four years. That system dissipates as heat. The imposed-order alternative has its own dystopia. The goal is long-term planning without a dictator, through emergence rather than a podium.

Fridman asks whether software platforms could be that emergence, and whether technology is values-neutral. Schmachtenberger calls the neutrality claim silly. Whole schools of anthropology — Marvin Harris's cultural materialism, Marshall McLuhan on information technology — exist because tools change minds, values, and social systems. The ox-drawn plow is his illustration. Cultures that used it could grow storable grain, survive famine, and expand; others had to adopt it or lose. Animism disappeared wherever the plow went, because you cannot speak of the spirit of the buffalo while beating a cow all day. Upper-body strength moved caloric supply onto men; pregnant women miscarried at the plow; the ratio of gods tilted male. A related, reductionist line of thought even says feminism followed the tractor, once an internal combustion engine erased the strength gap. Surplus became inheritance, then capital, then class. He admits single-cause stories are too neat. The claim he will not drop is that technology codes patterns of behavior, which code the values that then rationalize the tool.

The same loop, later, is why a major leap in physical technology breaks the previous social system: the plow broke the tribe, the Industrial Revolution broke feudalism, the nuke broke nation-state war as an adequate institution and forced an international apparatus. Exponential tech following current market incentives is not, for him, a solution. The market is itself a previous social technology. Wikipedia is the contrast case he respects. It tends an information commons, is not a for-profit corporation, and does not serve two masters. Facebook and Google harvest behavioral data that can be more specific than what a therapist, doctor, lawyer, or priest holds. That is privileged information. In ordinary fiduciary law, an agent does not have a game-theoretic relationship with a principal. Platforms do. Users think they are holding a tool. He says they are dealing with an agent whose goal — maximize time on site — rivals their own.

Modern democracy, Schmachtenberger says, grew from an Enlightenment bet that huge numbers of anonymous people, unbonded, raised differently, could still make collective decisions and even compromise for strangers. That is a wild idea. It depended on shared sense-making — what is happening, what would a policy do — and shared meaning-making, the values we are trying to move toward. If you think systemic racism is rampant and I think it is not a thing, or you think climate change is almost existential and I think it is nothing, coordination fails. The work is to inhabit the other view hard enough that you can steal-man it, then look for a proposition that serves more of the values than a proposal built to win.

He paraphrases Jefferson: given a perfect newspaper and a broken government, take the newspaper, because people who understand what is going on can rebuild a government, and people who do not cannot choose well. He paraphrases Washington: the number one aim of the federal government should be the comprehensive education of every citizen in the science of government — game theory, coordination, history, sociology, economics. Not rule of law, not protecting the border. A dictatorship can do those. If the first aim is anything other than that education, he says, it will not stay a democracy long. Education and a functioning fourth estate are prerequisite institutions.

The original practice, at the scale the Founders could imagine, was the town hall. A proposition was not delivered by a special interest for a yes/no vote. It could be crafted in real time. He cites a Founding Father line that voting is the death of democracy, because voting polarizes a population into sublimated war. Vote last. First ask how the thing you care about, which this draft seems to damage, could make the proposition better. Scale killed that process: representatives riding from one hall to another. The internet could restore proposition co-creation across geography. That it mostly has not, he says, sits somewhere between insane and aligned with class interests.

Fridman makes a case Schmachtenberger is happy to join. Wikipedia's second-order effects are underestimated. It is not only that people read it. YouTubers who do not know a subject well read an article thoroughly, turn it into a compelling video, and millions learn that World War II happened, that slavery was a scale, that injustices have names. Knowledge, Fridman thinks, cannot help but lead to empathy: a breakup looks different once you know millions were tortured. That leaks into how police, lawyers, judges, juries, and ordinary citizens talk, even when the surrounding systems are flawed. Schmachtenberger is a huge fan. "You should donate to Wikipedia." There are very few systems like it, which he finds sad.

He wants listeners to run a full dialectic on information tech: take Tristan Harris's techno-concern all the way, including exponential curves of harm, then take the techno-optimist case a Kurzweil or a Diamandis might make, and hold both. On some metrics things get worse exponentially and better exponentially at the same time. That is destabilization of a previous system, from which either a better order or a collapse can emerge. He does not want optimism in the assessment. "I want my assessment to be just as fucking clear as it can be." Optimism belongs in the search for solutions on top of that clarity. Pessimism belongs in red-teaming those solutions for second-order effects, without collapsing into despair.

Authoritarian states, he continues, are applying exponential tech on purpose to become better authoritarian states: Internet of Things surveillance, machine learning, social-credit systems. Open societies are mostly not directing tech toward better emergent order. They are letting corporations do it under corporate law that never imagined fiduciary duty over attention or antitrust against network-effect monopolies — YouTube larger than every other video platform together, Amazon larger than the other markets together. The biggest firms of the Scottish Enlightenment were tiny beside today's. Supply learned to manufacture demand, and addiction is a superb way to do it. These corporations are in some ways more powerful than nation-states, and they debase open societies from inside: a tech feudalism, because a company is not a democracy. Two attractors, both bad: authoritarian states and feudal corporations steering exponential tech. The path he wants is the reverse — use exponential tech to build social tech that makes healthy emergent order possible, so that order can then bind the tech.

Existing problem-solving processes are not matching the landscape. More countries have nukes; nuclear disarmament did not happen; the UN sustainable development goals were not achieved; every new tech category spawned an arms race. Markets and states as solvers are too slow. DDT and cigarettes were regulated after they had already killed huge numbers. Elon Musk's point on AI, as Schmachtenberger relays it, is that after-the-fact regulation fails for self-replicating or exponentially fast systems: once you notice an autopoietic AI, you may not get to reverse it. New processes have to internalize externalities at the right speed and geographic scale, and they cannot be imposed. They have to emerge from people who want them and can help build them — a cultural Enlightenment or Renaissance around participatory governance at scale.

Fridman turns from societies to fathers. Schmachtenberger had an unusually present dad: homeschooling, work-from-home, fathering treated as a vocation. The high-level lesson he keeps is the intersection of self-empowerment with a virtuous purpose beyond the self. Wiring a house before the drywall went up, his father made the hidden runs beautiful — holes at the same height, wires twisted a certain way — because if a job is worth doing it is worth doing well, and excellence is its own reward. On miserable jobs the line was: see the job, do the job, stay out of the misery. Asked how he could have been a better son, he is embarrassed by an early-twenties phase of Dawkins-and-Hitchens certainty, sanctimonious and dismissively reductionist toward his father's less standard beliefs. He got to apologize later.

Fridman reads a passage Schmachtenberger had written: countless times, making love, watching a sunset, listening to music, feeling the breeze, he would sign up for the whole of life, pains included, just to have that exact moment. Experience itself is infinitely meaningful; pain is temporary; seen clearly, even the suffering is filled with beauty. He will not take credit for the philosophies that taught him to see that way, only responsibility for practicing them. Looking at someone he loves, he is overwhelmed that there has never been anyone like them and never will be, and that they might be gone tomorrow.

The first time he took a large dose of mushrooms, he stared at a tree in front of his house and cried at how beautiful it was. The tree was not purple. It was still green, with leaves, describable in the ordinary way, and far more capturing than it had been on a thousand unseeing walks. What the mushrooms seemed to do was shut off the narrative that would have taken him elsewhere, so he could actually see it. He decided that when he was off the mushrooms he would practice seeing the tree, because it is always that beautiful and he had been missing it. Fridman has had the same hours-long, idiotic, grateful stare. They skip vodka. Schmachtenberger had bought ginger-turmeric-cayenne shots. They toast the unique moment they get to share. They met for the first time. He will never be the same.