← All texts

Essay · mantisvex.com

The World Is a Computation

We are, in all probability, living in a simulation. The argument is a chain: felt experience, the "what it is like" of being someone, is a functional fact, and therefore reproducible on another substrate; a perfect double of us, identical in every respect but with "no one inside," is nonsense; if experience is reproducible, consciousness does not depend on flesh; and then Bostrom's simple counting makes our own simulation close to certain. What remains out of reach is not the fact but the level above: who is running the computation, whether there is anyone at all, and why. This essay defends the chain, then stops, honestly, at the edge of what cannot be known.

Essay 2026 ~32 min read
I

The intuition that won't let go

I will start by confessing the bias, because hiding it would distort everything that follows. I lean toward the idea that the world is a computation, and that it is running somewhere. This is not a conclusion; it is an inclination: a magnet under the compass, pulling the needle before any bearing has been taken. The whole essay is an exercise in not obeying that magnet without testing it first. But to test it, I first have to look at it at full strength, without prudently shrinking it into a reasonable position. So here is why a mind that cares about rigor can catch itself finding the simulation not far-fetched, but almost natural.

First, the world lets itself be written down. It can be described in mathematics with a precision that should astonish us far more than it does: some quantities are predicted correctly to the twelfth decimal place1. We get used to it, we call it physics, and we forget the underlying strangeness: reality seems to be drafted in the very language we use to describe it. As if, beneath matter, there were no other fabric than structure, relation, number. A world made of information would look exactly like this.

Second, the world comes in packets. Energy, charge, action: nothing flows in a perfectly continuous way; everything is dealt out in indivisible units, in smallest denominations. An analog cosmos, smooth all the way down, would have no reason to have a minimal grain. A computational system would. Quantization looks less like a property of matter than like a data type.

Then there are the limits, and they are too clean. A maximum speed, absolute, that nothing crosses. A scale below which the question "how long?" stops meaning anything. A purely physical world could, in principle, have neither ceiling nor floor. Ours has both, drawn with a straightedge, like the specifications of a machine: peak throughput, minimum resolution.

Look more closely at that maximum speed. We call it "the speed of light," but that is a misnomer. Light travels at that speed because it has no mass, not the other way around; what the constant really measures is the speed of causality, the fixed exchange rate between space and time, hard-wired into the engine. And a fixed exchange rate is precisely the kind of thing that gets set. A value that could have been set otherwise. The word that comes to mind, when you stare at it, is not "law of nature." It is "parameter."

And there is, finally, a clue that is more troubling because it is subjective. Anyone who has spent time building systems, setting constants, defining bounds, watching teeming complexity emerge from a few simple rules, recognizes something in this list. Not a proof: a scent. The world has, for someone who builds worlds, the characteristic smell of a system. You know it is not an argument. You cannot quite shake it either.

Nothing I have written so far is an argument, and I want to say so before a hurried reader mistakes momentum for demonstration. These are resemblances. Quantization resembles a data type; the speed limit resembles a setting; the Planck length resembles a resolution. The word recurring in every sentence is "resembles," and that is exactly the word a rigorous mind must learn never to confuse with "is." The seductiveness of a thesis and its solidity are two different things, and the surest way to sink a defensible thesis is to rest it on its most seductive arguments rather than its most solid ones. So the next section does the sorting, without mercy, and it starts by demolishing my favorites.

 one level down · the Planck length is not a pixel

The "pixel" image is beautiful, and false, or at least premature. The Planck length (about 1.6 × 10⁻³⁵ meters) is not a measured resolution of space. It is the scale below which a theory of quantum gravity becomes indispensable, and where our current theories simply stop predicting anything.

Some approaches (loop quantum gravity2) suggest that space could be discrete at that grain; others (string theory) do not imply it; and observational searches for spacetime "foam" have, to date, found nothing3. The resemblance to a pixel is therefore evocative, not established. It is exactly the kind of distinction that all of Section II is about to apply, argument by argument.

II

What is real physics, and what isn't

For a long time, my favorite argument was this one. In the double-slit experiment, a particle behaves like a wave as long as you are not looking, and like a particle the moment you measure it. An irresistible translation for anyone who has ever opened a game engine: the universe only computes what is observed. It's level of detail: you render in high resolution only what enters the player's view; everything else runs degraded to spare the machine. Quantum physics would be the signature of a system managing its resources. I believed it. I defended it. And it is wrong. Not approximate, not exaggerated: wrong, on the meaning of a single word.

The word is "observation." In a popularizer's mouth, it suggests a gaze, a consciousness, a player in front of a screen. In the formalism, it means any interaction that couples the system to its environment: a photon striking it, an air molecule brushing past it, an inert, switched-off detector in an empty room. The system "picks" a state the moment it interacts with anything at all, whether or not any mind takes an interest. The process has a name, decoherence4, and it needs no witness: it happens inside stars, inside rocks, inside the oceans of exoplanets no one will ever look at. So there is no economy tied to attention. The universe does not render "what you look at": it renders everything that interacts, which is to say, near enough, everything, all the time. An engine that computes everything at all times is precisely an engine that does no LOD. My best argument, examined up close, was testifying against me.

I leave it on display, dismantled, because it teaches the rule that governs the rest of this section: the simulation thesis attracts secondhand physics, and that secondhand physics is its worst enemy. Whenever a "quantum" argument seems to prove the simulation in three sentences, it almost always rests on a word that does not mean, in the formalism, what it means in conversation. The sorting that follows is not an exercise in caution. It is hygiene: if the world really is a computation, it deserves to be defended with arguments that survive a reader who knows.

 one level down · two more false friends: the conscious observer and entanglement-as-messenger

The first false friend is the LOD's big brother: "consciousness creates reality." Same root, same error. A measuring device decoheres the system exactly as an eye would, because what matters is the physical coupling, not the subjective experience. A few physicists historically flirted with a role for consciousness (von Neumann5, then Wigner6, who later retracted); it is today a marginal position, and nothing in the data requires it.

The second is entanglement as "instantaneous communication," the shortcut in the code, the hole in the network. Two entangled particles show perfect correlations at a distance; that is true and it is dizzying. But a theorem, known as no-communication7, proves that no information can travel through that channel: each measurement's outcome is locally random, and the correlation only appears when the two logbooks are compared, which requires a classical channel, bounded by the speed of causality. Entanglement does not punch a hole in the engine; it scrupulously respects its speed limit.

So much for the clearing. Now, what holds. Because once the pop is swept away, real physics remains that points in the direction of information, and it is stranger than the false arguments it replaces.

The first solid piece fits in one sentence: information is physical. Not metaphorically: thermodynamically. Erasing one bit of information, anywhere, on any substrate, has a minimal, incompressible energy cost, set by temperature. This is not an engineering convention; it is a bound of nature, stated in 19618 and experimentally verified since9. What it means, if you take it seriously: information is not a way of talking about matter. It is a quantity that matter counts, bills and conserves, on the same footing as energy.

The second piece is the deepest result of theoretical physics in the last fifty years, and it comes from an unexpected place: black holes. When you throw matter into a black hole, its horizon grows, and you can compute the maximum amount of information the swallowed region could have contained. The result is a quiet scandal: that amount is proportional to the surface of the horizon, not to its volume. As if everything a three-dimensional region of space can hold in information actually fit on its two-dimensional boundary. This is the holographic principle, and it does not come from a forum: it comes from Bekenstein's work on black-hole entropy10, generalized by 't Hooft11 and Susskind12. The most sober reading you can give it is already vertiginous: volume, depth, space itself might be a redundant description, a rendering, of something thinner living on an edge.

And there is a third thread, more recent and less finished, which I give for what it is: a research program, not an established result. In certain theoretical models, the geometry of spacetime appears to emerge from quantum entanglement: undo the entanglement between two regions, and the space connecting them tears. If this direction pans out, space would not be the stage on which information moves: it would be manufactured by it. Wheeler's formula taken literally: it from bit13.

 one level down · the exact status of each brick

Established: Landauer's principle (bound formulated in 1961, experimental confirmations from 2012 onward); Bekenstein-Hawking entropy and the entropy/surface proportionality for black holes.

Strongly supported but theoretical: the holographic principle as a general property of quantum gravity. Its most precise realization is Maldacena's AdS/CFT correspondence (1997)14, demonstrated in a toy universe whose geometry is not ours. The transposition to our universe is largely conjectured, not proven.

Research program: spacetime emerging from entanglement (Van Raamsdonk15; the ER = EPR conjecture of Maldacena and Susskind, 201316). Serious, active, not concluded.

The difference between this list and the false friends above is not the degree of strangeness, which is greater here. It is that every statement survives the question: "what exactly does the formalism say?"

Let us take stock of what the sorting leaves standing, because it is what will carry everything that follows. What real physics suggests is not "someone is watching us and the universe is saving resources." It is something more austere and more radical: information is a physical quantity in its own right; the capacity of a region of space is counted on its surface like that of a recording medium; and space itself may be a derived structure rather than a given. None of this proves a simulator. All of it, however, makes the foundation statement perfectly respectable: the world behaves, down to its deepest laws, like a system whose substance is information. The question is no longer whether the world is a computation: that is the most natural reading of the above. The question is whether that computation has an inside. That is the subject of the next section, and it is where everything is decided.

III

Felt experience is functional

Everything so far established the foundation: the world behaves like a system whose substance is information. But the foundation is not enough. For the simulation argument to get off the ground, for it to be even possible that we are simulated beings, one precise thing must be true: consciousness must be able to exist on a substrate other than flesh. Not intelligence: consciousness. Not the ability to solve problems, hold a conversation, compose music: the raw fact that there is, behind the behavior, something it is like to be the system17. This is the lock on the whole thesis, and this is where I break it. Here is the position I defend, in one sentence: felt experience is a functional fact, it is what a certain arrangement does rather than a mysterious fabric layered on top, and consequently, if you simulate a human being perfectly, you have not built an imitation of a human being. You have built a human being.

The classic objection bears a famous name: the philosophical zombie18. Imagine, it says, a perfect copy of you: atom for atom, function for function, response for response. It laughs at your jokes, pulls its hand from the fire, writes essays about consciousness. But inside: nothing. No one. The lights are on, nobody's home. If such a being is even conceivable, the objection continues, then felt experience is something over and above function, an extra the copy did not receive, and no simulation, however perfect, guarantees there is anyone in there.

I think this thought experiment is a sleight of hand19, and I will say where the card is palmed. The zombie is described as identical to you in everything: same brain, same states, same processes, down to the last action potential. Then it is declared: "but without the felt experience." Yet declaring is not conceiving. Something has been removed by decree, without anyone being able to say what, since by hypothesis nothing is missing: no gear, no signal, no physical or functional difference of any kind. Felt experience has been treated like a lamp you could unscrew from the mechanism without the mechanism changing. But if you can unscrew it and nothing changes, then it was wired to nothing, and one may ask why we would have it. And if it is wired, if felt experience does something, plays a role, leaves a trace, then the perfect copy copied it along with everything else, and the zombie evaporates. The thought experiment does not describe a possible being: it describes a contradiction dressed up as a scenario. It asks you to imagine an intact mechanism missing what the mechanism produces.

An image from my world, for those it will speak to. Take a piece of music. Not the score, the render: the final file, sample by sample. Copy it bit for bit. Then declare: "the copy is identical in every respect, but it doesn't have the sound." The sentence has correct grammar and no meaning. Sound is not a substance poured into the file afterward; sound is what the file is, read by the appropriate system. There is no residue to forget in the copy, because there was never anything there but structure. My thesis is that felt experience is to the brain what sound is to the file: not an occupant of the mechanism, but the mechanism itself, considered from the only point of view that exists within it. Copy the arrangement, and you have copied the point of view, because the point of view was never anything other than the arrangement running.

I want to be precise about what this position claims and what it does not, because it gets caricatured from both sides. It does not say that felt experience is an illusion: it is as real as the sound of the track, which is to say entirely. It does not say that any program feels: a thermostat has no more of an inside than a sine wave is a symphony; what matters is the arrangement, its richness, its loops, the way it models itself, and we still do not know where the threshold lies. It says one exact and bounded thing: whatever produces felt experience in a brain, it is a fact of organization, and facts of organization do not know their substrate. An arrangement realized in carbon, in silicon, or in the pulses of a computation is the same arrangement. If the first has an inside, the others do.

 one level down · Mary, the serious objection

There is an argument against this position stronger than the zombie, and passing over it in silence would be cheating. It is called the knowledge argument20, and it stars Mary: a neuroscientist who knows everything there is to know, physically and functionally, about color vision, but who has lived her entire life in a black-and-white room. One day she steps out and sees red for the first time. Question: does she learn something? The near-universal intuition says yes: she learns what it is like. But if she learns a new fact while already possessing all the physical and functional facts, then facts of another nature exist, and my thesis falls.

The functionalist reply, which I make my own, attacks the verb "learn." What Mary acquires on seeing red is not one more fact about the world; it is one more capacity21: to recognize red, to represent it, to call it up. She goes from knowledge about the state to an instantiation of the state; her brain now runs a process it previously only described. Knowing the complete mechanics of a bicycle and knowing how to ride one are two different things, and no one concludes from that difference that balance is a non-physical substance. Mary's thrill before the red is real; it is not information leaking out of physics, it is physics happening in her for the first time.

A concession owed to the reader: this reply is defended, argued, majority opinion among functionalists; it is not unanimous, and the debate has been alive for forty years. A notable fact: Frank Jackson, the very inventor of the Mary argument, ended up switching sides and rejecting it22. That is not a proof, philosophers change their minds like everyone else, but when the prosecutor withdraws his own complaint, the defense is allowed to mention it.

So the lock is broken, and we should measure what that unlocks. If felt experience is functional, then consciousness is independent of its substrate, substrate-independent to give it its technical name. It can exist in flesh, in silicon, in a computation running on a machine we cannot imagine. Which means two things, and the second is the one that brings the vertigo. First: a sufficiently fine simulation of a human brain would not be the portrait of a mind: it would be a mind, with an inside as real as yours. Second: nothing in the nature of consciousness therefore rules out that ours is precisely that: an arrangement running, somewhere, on a substrate that is not what it seems. Until now, that possibility hung on an "if": if consciousness is simulable. That "if" has just fallen. All that is left is to count. The counting is the next section.

IV

So the counting applies

The argument that carries this section is twenty years old, fits on a page, and is almost always misquoted, including by those who invoke it. It gets summarized as "Bostrom showed we live in a simulation." Bostrom showed no such thing, and the exact version is more interesting than the legend. What he demonstrated is a trilemma23: of three propositions, at least one is almost certainly true. One: almost no civilization reaches the technological stage where consciousnesses can be simulated. Two: those that reach it almost all have no desire to. Three: we almost certainly live in a simulation. The argument does not choose among the three: it only establishes that you cannot refuse them all. And its mechanics are brutally simple: if simulated consciousnesses can exist, and if civilizations produce them, then each parent civilization can run thousands of simulations, each containing billions of minds. Simulated minds then numerically swamp "base" minds: a thousand to one, a million to one. And a mind picked at random from that population, you for instance, has the corresponding probability of being on the rendered side.

Notice what the previous section just did to this argument. Its most fragile premise has always been the first clause: if simulated consciousnesses can exist. As long as it stayed open, the whole edifice hung suspended: you could grant the counting and escape through the door of "but maybe consciousness can't be simulated." That is the door Section III closed. If felt experience is functional, simulated consciousness is not a hypothesis to verify: it is a consequence. The trilemma stops being conditional. All that remains is to look at the two branches that would separate us from the third, and to say honestly why I bet on neither.

The first branch is a cleaver: almost no civilization survives to technological maturity. It is possible; it is even the darkest branch, and nothing rules it out. But measure what it demands in order to save us from the counting. It is not enough that many civilizations collapse; almost all must collapse, everywhere, always: a failure rate so close to one hundred percent that no lineage, in any galaxy, over any billion years, crosses the line. Because a handful of survivors is enough for simulations to proliferate and the ratio to become crushing again. And there is an awkward detail: the required computing power is not distant magic. We, an adolescent civilization still playing with matches, already simulate crude worlds and train systems that hold conversations. The gap between what we do and what simulating minds requires is immense, but it is a gap of scale, not of kind. Betting on the first branch means betting that the step is uncrossable for nearly everyone, when it is already visible from where we stand.

The second branch is stranger still: mature civilizations exist, the power is there, and almost none ever uses it to simulate minds. For this branch to hold, a near-universal convergence is required: millions of civilizations, with radically different histories, values and curiosities, must all renounce the same thing. Out of ethics, out of boredom, out of prohibition: whatever the motive, it must be shared by almost all, almost always, with no lasting exception. Yet we have exactly one data point on what a curious species does with computing power, and that data point is us. It is unambiguous: we simulate everything we can simulate, as soon as we can. Worlds, economies, worm brains, conversations. The history of computing is a history of ever-finer simulations, and nothing suggests a voluntary stopping point. Extrapolating our own trajectory proves nothing about others; but betting on the second branch means betting against the only example we have.

So what remains are the probabilities as I see them. And I do mean: as I see them, because this paragraph is a judgment, not a theorem. The first branch requires a near-perfect extinction; the second, a near-perfect abstinence. Two near-perfections, each of which must hold across the whole extent of the cosmos and of time, against a third branch that requires nothing but banality: that somewhere, at least once, a civilization did what we already do in miniature. I have spent this essay cutting seductive arguments in favor of solid ones; this one survives the cut. When a conclusion demands only banality and its only two escape routes each demand a statistical miracle, I choose banality. We are, in all likelihood, on the simulated side of the count.

 one level down · the hidden assumption in "a mind picked at random"

The counting rests on a move that must be named, because it is the philosophical hinge of the whole argument: treating yourself as "a randomly drawn observer" from the total population of minds. This is the self-sampling assumption24. It looks innocent (if 999 out of 1000 minds are simulated, why would you be the exception?), but it is not free, and we know a case where the same reasoning yields a conclusion almost no one accepts: the doomsday argument25, which "proves" by simple counting of past and future humans that the end of humanity is near. Same machinery, embarrassing conclusion. Two readings are possible: either self-sampling is reliable and you must swallow doomsday along with the simulation; or it has subtle validity conditions, and the debate over which ones is still open.

My position: the simulation case is more robust than the doomsday case, because the counting there ranges over coexisting minds that are qualitatively indistinguishable from the inside, not over a position in a temporal ordering; but I do not claim the question is closed. This is the most philosophically loaded link in the chain, and the reader who wants to refuse it must refuse it here: everything else holds.

Let us recap the chain, because it is now complete and should be seen in a single glance. The world behaves like an information system: that was the foundation. Felt experience is functional, so consciousness does not depend on its substrate: that was the lock, and it broke. Therefore simulated consciousnesses are possible; therefore the trilemma applies unconditionally; and of its three branches, two require statistical near-miracles while the third requires only banality. Conclusion: we are, in all probability, arrangements running on a substrate that is not what it appears to be. I am aware of the weight of that sentence, and I stand by it. But this is precisely where the essay changes nature, because everything established so far concerns our level. About what lies above (who computes, on what, why, and whether there is even a "who"), the counting says nothing. Nothing at all. And that is where the real question begins.

V

The level above, and the admin's silence

Let us follow the logic where it leads, because it does not stop at our level. If we are simulated, then at least one level exists above ours. Call it what computer people call this kind of thing: the level with write permissions. But the reasoning that placed us inside a simulation applies, word for word, to that level too. If it is capable of simulating conscious worlds (it just did; we are the proof), then by the same counting, it is itself almost certainly simulated by a level above it. And so on. The conclusion is not "we live in a simulation": it is "we live somewhere in a stack of simulations whose height and summit we do not know," a tower in which every floor mistakes its ceiling for the sky. Statistically, being the very first simulation, the one running directly on base reality, is as improbable as being base reality itself. We are somewhere in the middle. That is all the counting licenses us to say, and it must be seen for what it is: enormous, and almost nothing.

Because here is the question that obsesses anyone who accepts the above, and I accept it: who computes? We want a face. A god, a teenager from a higher civilization, an artificial intelligence archiving its ancestors, a process with no intention at all. And here I must do to my own hopes what I did to my own arguments: put them to the blade. Each of these figures is a garment cut from our own wardrobe. A "god" is an enlarged father. A "teenage gamer" is an enlarged developer. An "archivist AI" is our own technology enlarged. Every conjecture about the admin smuggles our categories (intention, curiosity, purpose, play) up to a level that nothing guarantees contains them. We reason about the simulator the way a video-game character would reason about us: by ascribing us health bars, quests and an inventory, because those are the only concepts its world supplies. The error would not be getting the details wrong. The error would be believing that the right categories exist in our lexicon.

It will be objected that the admin's silence is itself a clue, that a creator who does not manifest tells us something of his intentions: indifference, protocol, cruelty. But that is the same trap, one notch deeper. Interpreting a silence requires knowing what speaking would have meant. We have no idea what would count, for the level above, as "manifesting": its interventions, if they exist, could saturate every second of our world without our having the organ to perceive them; or our entire universe could be, from its point of view, one line in a log file that nothing justifies opening. Silence is not a datum. It is the absence of a datum, and the absence of a datum cannot be read.

This lucidity should be carried to its end, because it has a precise shape that can be named. Our situation is not that of the ignorant who has not yet found the answer: it is worse, and more interesting. Imagine a being living on a sheet of paper, in a two-dimensional world26. Push a finger through its sheet: it will not see a finger. It will see a circle appear out of nowhere, grow, shrink, vanish. It can measure that circle all its life with perfect instruments: it will never form the concept "finger," nor the concept "up," because the dimension containing them does not exist in its mental geometry. Its ignorance has a shape, and that shape is invisible to it: it cannot know what it is missing, because it would already need the missing dimension to perceive its absence. We are that being. The level above is not hidden from us like a secret that could be cracked: it is inaccessible to us the way depth is inaccessible to the plane. And the most vertiginous part is not what we do not know about the admin. It is that we do not know the list of what we do not know, that we may be missing the very dimensions in which the question "who computes?" makes sense.

 one level down · can we at least test that there is a level?

Only one serious attempt has been published. In 2012, Beane, Davoudi and Savage noticed27 that our own particle-physics simulations run on a grid, a discrete lattice, and that if the universe ran on a similar grid, it would leave a signature: an anisotropy in the spectrum of ultra-high-energy cosmic rays, aligned with the lattice axes. People looked. Nothing. But the null result only tests simulators that would compute the way we did in 2012: a simulator using any other architecture slips under the radar. The test does not constrain the hypothesis; it constrains our imagination.

And the problem is structural, not technical. Any observed anomaly always admits two readings, an unexpected physical law or an artifact of the computation, and nothing, from the inside, allows you to decide between them. A good enough simulation is undetectable by construction; the only conceivable "proof" would be a voluntary intervention from the level above. In other words: a revelation. The irony should be measured: the most technophile hypothesis in modern metaphysics, pushed to its maximum rigor, lands exactly on the epistemic regime of the religions, that of a created world, a mute creator, and the waiting for a sign.

Which brings me to the most expensive concession of this essay, and I make it without anesthesia because it is true. Grant everything: the foundation, the lock, the counting, the stack of levels. What have we explained? The world has laws: why these? Answer: because the level above set them. But the level above has its own laws: why those? Something exists rather than nothing: why? Answer: because a computation is running. But the computation's substrate exists: why it? Every answer moves the question up one level without ever resolving it. "The world is simulated" explains the world exactly the way "God made it" explains it: by pushing the mystery back one notch and dressing it in contemporary vocabulary. I have defended this thesis across hundreds of lines, and I stand by every link, but I refuse to make it say what it does not say. It does not answer the why. It says nothing about the laws above. And it leaves intact, at every level of the stack, the only enigma that has ever seemed real to me: that there are, somewhere in the computation, arrangements for which the computation is like something. The simulation is not the answer to the great why. It is the rigorous demonstration that the answer sits one level higher than anything we will ever reach, and that it always has. What remains is what to do once you have understood that. That is the subject of the last section.

VI

What remains, and what it costs

Let us take inventory, coldly, the way you close out a worksite. What I can show: the world behaves like an information system, down to its hardest results. What I can support by argument: felt experience is functional, so consciousness does not depend on its substrate, so the counting applies, so we are almost certainly simulated. What I wager without being able to show it: the height of the stack, the middle where we float. And what I will never know, by construction and not by laziness: who computes, on what, why, and whether the question itself makes sense in the dimensions we lack. Three floors of decreasing certainty, and a final wall. That is the exact perimeter of what an honest mind can hold on this subject at this point in history. Anything claiming more is selling something.

What remains is to draw the consequences for the one living inside, and the first concerns freedom. In a computed world, every state follows from the previous one: free will in the strong sense, a will that would escape causality, does not survive this essay, and I will not pretend to half-regret it. It did not survive anything else either. But something remains, and that something is more solid than the freedom we lose. For most sufficiently rich systems, no shortcut exists: no formula yields the final state without unrolling the computation step by step. This is called computational irreducibility28, and it changes the meaning of what we are. You are determined, yes. But you are incompressible. No one, not even the level above if it exists, can know what you will do other than by letting you do it. Your life is not the reading of a book already written: it is the writing of the book, at real speed, with no abridged version possible. It is not the freedom of the cause. It is the freedom of the unpredictable, and it is the only one that ever existed.

 one level down · computational irreducibility

The concept comes from Stephen Wolfram, who isolated it while studying cellular automata: systems with childishly simple rules, some of which produce complexity that no analysis condenses. For these systems, predicting the state at step one thousand requires computing the thousand steps: the fastest simulation of the system is the system itself. This is not a limit of our instruments: it is a mathematical property of the process.

Applied to us, the consequence is precise: even in a perfectly deterministic universe, the future of a system like a brain, a society or a biosphere can be rigorously unpredictable from inside and outside alike, not through ignorance, but because no path shorter than its unfolding exists. Determinism stays true; prediction stays impossible. Both hold together, and it is in the gap between them that what latitude we have resides.

The second consequence is the price, and I promised it without anesthesia. Section III won the argument by making felt experience a function. Now the cost must be paid. As long as lived experience passed for a fabric apart, it was a foothold outside the mechanism: whatever happened to the world, this at least remained, the one thing I am sure of, of a nature nothing else shares. By bringing it back into the world of arrangements, I stripped it of that status. It remains real, entirely. It remains mine, in the only sense in which a point of view belongs to the arrangement running it. It will still be extinguished, and that extinction remains what it always was: the end of everything there is, for the one being extinguished. But it has ceased to be of a different nature from the rest. The one certainty I had left, that there is this, now, is itself now a line in the computation. I traded my foothold for an argument. I would make the trade again, because I prefer a truth that costs me to a consolation that flatters me. But I will not pretend it costs nothing.

And yet one thing survives this stripping, and it is not a consolation, it is a fact of structure. A computation that halts has computed. An execution that ends has taken place, and nothing that comes after can make it not have taken place. The nothingness that follows has no retroactive power: it erases the process's future, not its having-been. If we are arrangements running somewhere in a stack of mute levels, then every moment of the unfolding is exactly as real as the substrate allows, and it is so once and for all. It is not much. It is what remains when you have refused to lie to yourself, and I note it without emphasis: the refusal to lie to yourself always leaves less than you hoped, and always more than nothing.

I end where I began, with the confession. The magnet under the compass is still there: I still lean toward the stack of levels, toward the running computation, and nothing in this essay has changed my inclination. What has changed is that I now know which of its pulls is iron and which is wind. I started from a question, the one everyone asks: who made the world, and why? I arrive at a rectified question, and that is all honest work can deliver: what does a conscious arrangement do with its execution, knowing it will never read the spec? I do not have the answer. No one does, and the wall between us and it is not the kind that gets breached. But between an answer that explains nothing and a properly cut question, I choose the question. It, at least, is the right dimension. It fits inside ours.

SOURCES

References

  1. E. P. Wigner, « The Unreasonable Effectiveness of Mathematics in the Natural Sciences », Communications on Pure and Applied Mathematics 13, 1-14 (1960).
  2. C. Rovelli, « Loop Quantum Gravity », Living Reviews in Relativity 11, 5 (2008).
  3. A. A. Abdo et al. (Fermi LAT & GBM), « A limit on the variation of the speed of light arising from quantum gravity effects », Nature 462, 331-334 (2009).
  4. W. H. Zurek, « Decoherence, einselection, and the quantum origins of the classical », Reviews of Modern Physics 75, 715-775 (2003).
  5. J. von Neumann, Mathematische Grundlagen der Quantenmechanik, Springer (1932) ; trad. Mathematical Foundations of Quantum Mechanics, Princeton University Press (1955).
  6. E. P. Wigner, « Remarks on the Mind-Body Question », dans I. J. Good (dir.), The Scientist Speculates, Heinemann (1961).
  7. G. C. Ghirardi, A. Rimini & T. Weber, « A general argument against superluminal transmissions through the quantum mechanical measurement process », Lettere al Nuovo Cimento 27, 293-298 (1980).
  8. R. Landauer, « Irreversibility and Heat Generation in the Computing Process », IBM Journal of Research and Development 5, 183-191 (1961).
  9. A. Bérut et al., « Experimental verification of Landauer’s principle linking information and thermodynamics », Nature 483, 187-189 (2012).
  10. J. D. Bekenstein, « Black Holes and Entropy », Physical Review D 7, 2333-2346 (1973).
  11. G. ’t Hooft, « Dimensional Reduction in Quantum Gravity », arXiv:gr-qc/9310026 (1993).
  12. L. Susskind, « The World as a Hologram », Journal of Mathematical Physics 36, 6377-6396 (1995).
  13. J. A. Wheeler, « Information, Physics, Quantum: The Search for Links », dans W. H. Zurek (dir.), Complexity, Entropy, and the Physics of Information, Addison-Wesley (1990).
  14. J. Maldacena, « The Large N Limit of Superconformal Field Theories and Supergravity », Advances in Theoretical and Mathematical Physics 2, 231-252 (1998).
  15. M. Van Raamsdonk, « Building up spacetime with quantum entanglement », General Relativity and Gravitation 42, 2323-2329 (2010).
  16. J. Maldacena & L. Susskind, « Cool horizons for entangled black holes », Fortschritte der Physik 61, 781-811 (2013).
  17. T. Nagel, « What Is It Like to Be a Bat? », The Philosophical Review 83, 435-450 (1974).
  18. D. J. Chalmers, The Conscious Mind: In Search of a Fundamental Theory, Oxford University Press (1996).
  19. D. C. Dennett, « The Unimagined Preposterousness of Zombies », Journal of Consciousness Studies 2 (4), 322-325 (1995).
  20. F. Jackson, « Epiphenomenal Qualia », The Philosophical Quarterly 32, 127-136 (1982).
  21. D. Lewis, « What Experience Teaches », Proceedings of the Russellian Society (University of Sydney) 13, 29-57 (1988) ; repris dans W. G. Lycan (dir.), Mind and Cognition, Blackwell (1990).
  22. F. Jackson, « Mind and Illusion », dans A. O’Hear (dir.), Minds and Persons (Royal Institute of Philosophy Supplement 53), Cambridge University Press (2003).
  23. N. Bostrom, « Are You Living in a Computer Simulation? », The Philosophical Quarterly 53, 243-255 (2003).
  24. N. Bostrom, Anthropic Bias: Observation Selection Effects in Science and Philosophy, Routledge (2002).
  25. J. Leslie, The End of the World: The Science and Ethics of Human Extinction, Routledge (1996).
  26. E. A. Abbott, Flatland: A Romance of Many Dimensions, Seeley & Co. (1884).
  27. S. R. Beane, Z. Davoudi & M. J. Savage, « Constraints on the Universe as a Numerical Simulation », The European Physical Journal A 50, 148 (2014).
  28. S. Wolfram, A New Kind of Science, Wolfram Media (2002).