Are We Living in a Simulation? A Modern Look at the Simulation Argument

0:00This episode is brought to you by Progressive Insurance. Do you ever think about switching insurance companies to see if you could save some cash? Progressive makes it easy to see if you could save when you bundle your home and auto policies. Try it at Progressive.com. Progressive Casualty Insurance Company and Affiliates. Potential savings will vary. Not available in all states.

0:22The old question was whether our reality might be simulated. Today, we're asking the next one. What happens when civilizations like ours become the simulators?

0:34Suppose tomorrow morning someone proved the universe was not the base layer of reality. Not fake in the sense that your life did not happen, or that your friends and family did not matter, or that the stars were really painted on the ceiling. But fake in the sense that reality is already on something else. A deeper layer. A computer. A dreaming mind. A machine. A universe above ours. Or something so strange that words like computer and machine are only metaphors. Now imagine your first response is not fear, or denial, or even awe.

1:09It's irritation. Because your taxes are still due. Your back still hurts. Your inbox is still full. The coffee still needs brewing. And the simple reality is that most people are not really all that surprised by the idea of some higher layer of reality. While the specifics vary, that has basically been the default opinion of much of humanity since long before computers. Philosophically, it does not matter much whether we are some deity's dream, some deity's creation, or some development team's simulation.

1:41Either way, we are talking about a reality beneath another reality. That, I think, is one of the underappreciated parts of the simulation hypothesis. If we are in one, it does not a very good job of making the question feel both cosmic and weirdly irrelevant. If the floor still catches you when you fall. If food still fills your stomach. If kindness still comforts people and cruelty still hurts them. Then, real may be a more complicated word than we usually assume. We have talked about the simulation hypothesis before, along with ancestor simulations, virtual worlds, digital minds, post-biological civilizations, the Fermi Paradox, and the Anthropic Principle.

2:22But it is worth revisiting now, because this idea has changed flavor. It used to feel like philosophy wearing a science fiction costume. But now I think it feels like science fiction walking to the engineering department and asking for a desk. We are not close to simulating a universe. Well, depth and scale of our own, anyway. But we do have untold thousands of simulated realities and subverses already. In science, video games, and even pre-computing fantasy worlds like Tolkien's Middle Earth. Worlds made of symbols, rules, and shared imagination.

2:52Not conscious minds. None of those characters think they are real, because they do not think at all. And we are not close to uploading or copying a mind yet, let alone a human civilization to a perfect digital Earth. We do not have convincing evidence that we are inside someone else's software. And if anyone tells you that they prove that we are, or prove that it is impossible, they are almost certainly overselling some very specific argument. You also cannot disprove simulation by saying the rules of this universe make simulating this universe impossible. First, a simulation only needs enough fidelity to fool its occupants, not perfect atom-by-atom duplication of everything.

3:30Second, this universe would not be the one running this simulation. We do not know the rules or limits of a higher layer of reality. Only in a specific case of an ancestor simulation can we even guess that their physics probably resembles ours. Putting it bluntly, proving you are not in a simulation is probably impossible. And I don't mean beyond a shadow of a doubt proof here, either. Or even a reasonable doubt. Even getting clear and convincing evidence is probably not possible. We will explain that as we go, but in the past, folks could shrug and assume that such simulation was impossible.

4:03A nice storyline like in The Matrix, but no more. Yet now we live in an era where machines generate images, voices, faces, conversations, histories, virtual spaces, fictional people, and increasingly convincing shadows of intelligence. Ironically, this is also a counter-argument to the simulation hypothesis itself. But the key point for the moment is that humanity developing very real virtual reality is now very hard to disbelieve. We already have worlds inside worlds. Primitive ones, perhaps, but real enough that people spend money on them, make friends on them, hoard meetings on them, and sometimes build identities around them.

4:40Again, none of that proves we are in a simulation. But it does make the old question harder to dismiss of the laugh. Because a civilization does not go from carving symbols into clay tablets to simulating conscious beings in one step. It gets there by gradually making better models, better games, better memories, better artificial people, and better artificial worlds. Until one day someone asks the uncomfortable question. At what point did the model become a place? At what point did the character become a person? And at what point did the simulation become reality?

5:11So today we are going to take a modern look at the simulation argument. Not just, are we in the matrix? What the argument actually is? Why ancestral simulations are a special case? Why modern AI changes the mood, but not the proof? Why ethics might stop advanced civilizations from doing this? How hacking a simulation might work or fail? And what all this means for the Fremi Paradox? And I want you to leave a comment on this one, because it helps appease the algorithm. And because I'm curious where folks land, and whether it differs from when we first discussed this over a decade ago.

5:45Do you think we are probably in a simulation? Probably not. Or the question is unanswerable. And more importantly, why? And give the like, subscribe, and notification buttons a smack too, so the algorithmic spirits of this simulated realm know you approve our continued existence. Then, grab a drink and a snack, and let's ask ourselves whether reality is real, and whether it is even the right question. The Simulation Argument The first thing we need to do is separate the simulation hypothesis from the much older idea that life is a dream.

6:18There is some philosophical overlap, but it is a different question. Human beings have probably been asking some version of that question since before we had writing. Dreams feel real while we are inside them. Myths often picture the world as the thought or dream of some higher being, and philosophers have long wondered whether our senses deceive us. The modern simulation argument is related to all of that, but it is not quite the same. The version most people mean today is associated with Nick Bostrom, and while it often widens to other discussions, it is specifically about our ancestor simulation scenarios.

6:52Simulations of a civilization's own past, or something close enough to it, run by a later civilization that arose from it. That distinction matters. If a being from a universe with totally different physical laws creates our universe as an art project, then we may not be able to infer much about them. Their reality might not have atoms, stars, entropy, three dimensions, or anything resembling computers as we understand them. Think about old video games, where your character could run off one side of the screen and re-emerge on another, although a world map was an actual rectangle,

7:24not a rectangular projection of a sphere. But if this is an ancestor simulation, then the simulators are in some meaningful sense downstream of us. Their civilization resembles ours enough that they are simulating something like their own history. They might not make it a close match socially, same as we might run one where Rome never fell, or early Mesopotamia invented steam engines. Their physics probably resemble ours, and so might their motives. Education, entertainment, research, resurrection, nostalgia, ethical training,

7:56legal reconstruction, or curiosity. So it doesn't mean it's an exact match, but it probably matches our reality except if they thought it irrelevant or impractical. Or if they specifically wanted to test a counterfactual, like how a civilization would develop if planets were light years apart instead of light minutes apart. That might be absurd in all solar systems and galaxy, but normal in theirs. The classic simulation argument goes roughly like this. Three broad scenarios exist. One, civilizations like ours usually go extinct

8:26before they become capable of running ancestor simulations, either through self-destruction, stagnation, or because such simulations turn out to be genuinely impossible. Two, civilizations do reach the point where they could run ancestor simulations, but choose not to run many of them. Maybe they are too impractical, perhaps they are pointless, or they could be considered unethical. And three, civilizations survive, gain the ability, and run lots of them. Now, we do not have evidence that strongly favors any one of those three.

8:56Every day we gain more technology and do not die, options two and three look a lot more plausible from one perspective, but it's not the same as proof. So in all discussions, this option gets paired with the principle of indifference, also called the principle of insufficient reason. If you have several plausible options, and no evidence favoring one of them over the others, you temporarily treat them as roughly equally likely until you get more evidence. Three options, about one in three each. And channel regulars may have heard me say in episodes, live streams, or interviews,

9:27that when asked about the odds of us being a simulation, I usually put it as about one in three, or just under a third. This is the complete and total reason for that number. It's not a measured probability, it's not a secret equation. It is a placeholder for ignorance. The just under part matters too, because even in option three, at least some people are still the original biological observers in the civilization that eventually makes the simulations. So the chance of being simulated is not exactly one third under this reasoning.

9:57It is one third multiplied by the chance that, if option three is true, you are one of the simulated observers rather than one of the originals. Yet they run huge numbers of ancestor simulations, that second probability may be very close to one, but is not literally one. If the third option is true, then simulated observers living through eras like ours could vastly outnumber the original biological observers who lived through the eras the first time. If the simulated people vastly outnumber the originals, then by sheer counting, we should be more likely to be among the simulated ones.

10:28That is the basic argument. It is not, look, quantum mechanics is pixelated, therefore we are in a video game. It is not, the universe is weird, therefore someone programmed it. It is not, I saw a glitch, therefore reality crashed. It is simply a population argument about observers. If future civilizations can make realistic minds and worlds in vast numbers, and choose to do so, then most minds like ours might be simulated minds. And if simulated civilizations eventually make simulations of their own, the latter need not stop one level up.

11:00You get simulations inside simulations, turtles all the way down. I should also note that IF is doing an immense amount of work here. Take the first question. Can civilizations simulate minds like ours? We do not know. We do not have a working theory of consciousness that lets us say exactly where a personhood begins and ends. We do not know how detailed a brain simulation must be to preserve a mind, whether a perfect behavioral copy is conscious, or just an extremely convincing puppet, or whether substrate matters.

11:30We do know something important. The gap between utterly impossible, and maybe eventually possible, has narrowed emotionally, if not scientifically. A century ago, the idea of a machine holding a conversation with you was fantasy. Even Asimov wasn't writing about robots yet. Now, we argue about how good the conversation is. Software can create faces of people who never existed, and game boards that once existed of a few colored blocks now contain cities, economies, weather systems, histories, and millions of players who treat them as a meaningful place.

12:02None of that is consciousness, nor is it cheap in computer power or electricity, but it is scaffolding. A puppet is not a person, and when puppets begin speaking full paragraphs, remembering your preferences, drawing pictures, generating code, and acting as if they have goals, it becomes easier to imagine future versions that raise much harder questions. And here we should be careful. There is a common mistake in both directions. One mistake is assumed that anything which talks like a person must be a person. That is far too quick. A chatbot may imitate empathy without feeling anything.

12:35A game character may beg for mercy because someone wrote or generated that response, not because they were suffering behind the words, nor a desire by the apparent speaker to persist. The other mistake is to assume that because current systems are not conscious, no artificial system could ever be made. Your own brain is made of cells obeying chemistry. Neurons are not individually writing poetry or contemplating mortality, and yet arranged the right way, they produce us. Now, it is certainly possible there is something about us that cannot be replicated by us, be that natural or supernatural.

13:07But if so, we have no evidence for it either, and if not, we should be entirely irreplicable by technology. Mind you, having a supernatural soul does not mean we couldn't make a copy of someone that had one too, since we have no idea how that process operates if there is one. Either way, if consciousness can't arise from matter arranged one way, we should be cautious about to claim that it can't arise from matter arranged another way. That uncertainty is where the simulation argument enters the discussion. If future civilizations can create conscious simulated beings, then access to simulations

13:37become possible, but possible does not mean likely. This episode is brought to you by Progressive Insurance. Do you ever think about switching insurance companies to see if you could save some cash? Progressive makes it easy to see if you could save when you bundle your home and auto policies. Try it at Progressive.com. Progressive Casualty Insurance Company and affiliates. Potential savings will vary. Not available in all states.

14:05Motives for making simulated rules. Lacking hard evidence, we can often examine the question in the classic detective style. Motive, means, and opportunity. The means, in this case, would be computation, though computer might be a crude term if we were talking about civilization thousands or millions of years ahead of us. The opportunity is simpler. Survive long enough to develop the technology and have enough resources to run such simulations in quantity. But motive is trickier. Why would anyone run one?

14:36Let alone so many that you're more likely to be a simulation than the original reality that made them. The most obvious answer is entertainment, as that is how we tend to think about simulations today. Future people booting up 21st century Earth, deluxe edition, like a historical strategy game, or taking a vacation in virtual ancient Rome. But this is probably one of the weaker reasons, at least for a full consciousness ancestor simulation. If you want entertainment, you do not need billions of genuinely conscious people. You need convincing characters

15:07where the player interacts with them. If you are walking through a simulated medieval village, the blacksmith needs to answer your questions, the king needs to make decisions, and the dragon needs to be scary. But the farmer of three kingdoms away, whom you will never meet, does not need a full inner life, childhood trauma, hopes, dreams, and a secret fondness for cheese. That would be like building an entire real star to light your desk when a lamp would do. Future civilizations may have immense resources,

15:37but immense is not infinite. Computation costs energy, produces heat, and requires matter. Even a civilization with Dyson swarms and matriarchic brains still has reasons not to waste a million times more processing power than necessary. And this is one of the places where modern AI cuts against the simulation argument, at least the casual portion of it. We can already generate convincing dialogue, images, voices, and personality sketches without running anything remotely like a complete human mind. The exact energy use varies wildly

16:07by model and hardware, but we are talking about ordinary modern data center computation. Costly, certainly, but not equivalent to running a full conscious person. A whole brain emulation is a very different beast. All usual rough guesses for emulating a human mind start around a billion billion operations per second, and climb many orders of magnitude higher if you want detailed neurons, chemistry, and cellular behavior. That means a future civilization able to run truly digital people would almost certainly also have a far cheaper way to fake people well enough

16:37for most entertainment or educational purposes. I want to emphasize that scale of difference. Back when this topic first started getting serious discussion, we did not really know how much computation would be needed to run a convincing fake character versus a real mind. It was reasonable to wonder if the two might be fairly close. If a full simulated person constantly ten times as much as a convincing non-conscious character, some people might pay the premium for maximum realism. But if the difference is ten thousand, or a million,

17:07or a trillion times more processing power, the logic changes. At that point, simulate a fully conscious person for every interaction, let alone every bartender, senator, farmer, soldier, and passerby. It was monstrously inefficient. It would be like torching a forest to cook a few hamburgers. Impressive, in a technically horrifying way, but not a method anyone sane would choose when a grill exists. And the key to ancestor simulation arguments is that simulated minds might be more common than original minds. So if many simulations

17:38can be run with non-conscious characters, or with partial minds that only awake in detail when needed, that pushes back hard against the idea that you are statistically more likely to be an ancestor simulation than in the real civilization generating them. So, pure entertainment does not get us all the way there. Education is a stronger motive. A future civilization might use ancestor simulations to train people by letting them live through difficult performative periods. Not as tourists with safety rails, but as participants who do not remember the higher reality while inside.

18:10That sounds unsettling, but we already understand the appeal in miniature. Stories let us live other lives, games let us test choices, and history lets us imagine what we would have done. A sufficiently advanced civilization might take that to an extreme. Want to understand scarcity or fragile civilization? Live it. This is what we would call a nursery universe. Not a place made for sadism or cheap entertainment, but a training ground, a moral classroom, or a coming-of-age simulation. Needless to say, this raises enormous

18:41ethical problems. If the suffering is real, you cannot just wave it away by saying it builds character. That excuse has been used by every tyrant, bully, and bad gym coach in history. But the ethical picture gets more complicated if only some people inside are conscious. If the worst suffering is avoided, or if participants volunteered for entering and have traumatic memories softened afterward. In an ancestor simulation, I can only be certain of my own experience, not that every apparent person around me is a full emulation.

19:11Call it digital solipsism, if you like, and it is not a comforting idea. For my own part, I can view my own trials and tribulations as having made me a better and wiser person. Your mileage must inevitably vary, because your life and experience are different from mine, and you do not know if I'm real. Another motivation is research. A future civilization might run simulations of its past to answer historical questions. How close did we come to a nuclear war? What if that plague had been worse?

19:41What if artificial intelligence had developed faster or slower under different political conditions? What pathways lead to collapse, stagnation, expansion, or transcendence? This connects to the Fermi paradox. If you are trying to understand how often technological civilizations survive, expand, turn inward, create artificial intelligence, or destroy themselves, running simulated histories might be tempting. But again, you do not necessarily need conscious people for that. If you are studying economics, politics, climates, or technology

20:11at a coarse scale, billions of full human-level minds may be overkill, unless consciousness itself is what you are trying to study, or unless you are trying to resurrect the dead. That is another modern angle, and one that connects strongly to digital immortality. Suppose a future civilization has vast records of the past. Writings, videos, genetic data, medical scans, social media, government files, family stories, and enough behavioral evidence to reconstruct plausible versions of people. If they also have powerful mind-modeling

20:42technology, they might attempt to recreate ancestors, not merely as historical exhibits, but as persons. This is no longer just to simulate the year 2026, it is bring them back. Now, we are in very strange territory. If a recreated person wakes up and remembers being Isaac Arthur, or one of you listening, and continues from that point, are they the original person, a copy, a descendant, a fictionalized reconstruction, a new person with an old pattern? We have discussed this

21:12in digital immortality contexts, including the worrisome case of Roko's Basilisk, which has its own episode, but the hard answer is that a backup does not move you, it duplicates you. The copy may be a real person with a valid claim to your name and memories, you're still sitting in the chair. Ancestor simulations raise the same problem on a civilization scale. They may not be preserving the past, and may be creating new people who resemble the past. That does not make them worthless, though. A child is not worthless because they are not their parent.

21:43A restored book is not meaningless because the original burned. A rebuilt city is not fake merely because the stones are new. But it does mean that we should be careful before calling it immortality or resurrection in the simple sense. A future civilization might also run simulations as a cultural memory. Imagine a post-biological civilization. Mines living in vast computational structures around a star, preserving versions of earlier humanity as living branches of their broader human story.

22:13That sounds noble. It also sounds dangerous. Because once you make simulated people, you can make slaves, pets, enemies, punishments, test subjects, and disposable civilizations. You can create a billion mines, run them for a century, and delete them if you dislike the outcome. Or worse, you can edit their memories until they thank you for it. This may be the strongest argument against ancestor simulation being common. Advanced civilizations might decide they are immoral.

22:44We often assume future civilizations will do everything they can, but that is not true even now. We restrict technologies, ban experiments, and build weapons we hope to never use. A civilization powerful enough to assimilate conscious worlds may also be wise enough, or frightened enough, to avoid creating them casually. And if it does create them, it may regulate them heavily. Running an ancestor simulation with billions of conscious minds might be treated less like playing a game and more like founding a nation, raising children,

23:15cloning citizens, or creating an entire planet full of moral obligations. That is not something you let every bored graduate student do on weekends. Perhaps more importantly, it is not something someone could easily do secretly inside their own civilization. Because it likely requires massive amounts of computation, energy, and waste heat management. And regardless of the skill of those things they might have, it probably requires more of all of these things than detection equipment to find people doing it would. And that might not help us detect it much from inside the simulation,

23:46but it could matter a lot to their laws and institutions. If they get caught, or fear getting caught, and don't do it, then our odds of being a simulation change. Can I make an AI that's good at hunting for illegal simulations for less power and energy than it takes to make those illegal simulations? If I can, then the odds truly change. And even if I can't, civilization may be willing to pay quite a lot on the purpose of preventing such illegal simulations from running. And even if someone could hide their hardware, say by burying illegal simulations

24:17inside icy moons or other cold, remote places, that only strengthens the case for nursery universes and moral testing. You might raise new people, or immerse current ones to encourage decent morality or a stress test whether they have it. Indeed, running an AI inside a simulated reality to see if it goes all sky and dead at first opportunity is a good way to check if your AI alignment efforts work. In that sense, a simulated person is also a kind of artificial intelligence. If we are a code rather than brains in a jar,

24:47then the simulator is not merely watching our behavior. It may be able to pause us, inspect the pattern, and see what we are thinking. That is one of the unsettling differences between being observed from an outside world and being implemented by that outside world. All right, so where does that leave the odds? Well, not anywhere simple. The popular version says if civilizations can simulate many worlds, we are probably simulated. But the less flashy version says there are many reasons they may not make full consciousness ancestor simulations in large numbers,

25:18even if they could. They might prefer non-conscious approximations, find it unethical, but only small simulations, use temporary participants from their own society rather than newly created minds, run worlds unlike ours, or simply not care about us as much as we imagine. This last possibility is humbling. We tend to picture ourselves as an especially fascinating era, and maybe we are, but every age thinks it is the hinge of history. And maybe it is. To a civilization 10,000 years ahead,

25:49our time might be crucial. To one 10 million years ahead, it might be the blurry bit between agriculture and adulthood. To one a billion years ahead, we may be as specific as the age of damp primates using combustion. Or as vague as when bacteria got themselves multicellular, climbed out of the ocean, and built space rockets. That said, this is the last era of humanity where am I in a simulation is still a bit distant, while also being plausible enough that someone being awakened from the simulation can't at least understand

26:20what just happened. And again, in this case, that someone is you or me, not a random person from our population. Since you don't know if any of those other people are real, it does not matter if they themselves would be seemingly surprised that they woke up from the simulation. Just if you would. It's also the last period before we become spacefaring and reach out to colonize many other places that they would also need to simulate. And the last era before we probably can emulate non-human but intelligent minds. So it was actually a pretty likely era to be run disproportionately often.

26:54Can we tell? Now let's ask the question everyone wants to ask. Can we tell from a simulation? And again, probably not. At least now the simulation is good and the simulation does not want us to know. People often point to quantum mechanics, the Planck length, cosmic limits, mathematical structure, coincidences, glitches, deja vu, or the weird emptiness of the universe as possible evidence. Fun to contemplate, but none are strong proof. A minimum scale in physics does not mean pixels in the video game sense.

27:24Quantum uncertainty does not mean the universe is saving processing power by refusing to render things. The speed of light is not obviously a processor limit, and the Fermi paradox is not obviously the programmers figuring to add aliens. Could those things be clues? Sure. Many of these arguments work backwards, too. You might ask why anyone would simulate an absorbable universe almost 20 orders of magnitude wider than Earth, or why reality would have a scale like the Planck length roughly 20 orders of magnitude below a proton.

27:54Those are strange ratios, but strangeness is not evidence. The universe is full of enormous ratios, and enormous ratios are not the same as glitches. That is the danger of this topic. It easily becomes unfalsifiable, and is easy to see worrisome coincidences. I just noticed that scale one while thinking out of the weirdness of simulating a universe that big or a Planck scale that small. Basically, though, if the universe looks consistent, that proves the simulation is well made. If it looks inconsistent, that proves it has glitches.

28:24If we find aliens, they are simulated, too. If we do not, the programs let them out because they didn't want the extra variable. At that point, you are not doing science, you are doing interpretation. And interpretation can be fun. That's not proof. A genuinely detectable simulation would probably have to be sloppy, limited, experimental, or intentionally leaving clues. And those fall apart in this simulation hypothesis ancestor scenario because you have to be smart to run them. And every time someone figures it out or the system breaks,

28:55you can do a new simulation. Indeed, one from your last save state, if you like. If ancestor simulations are common enough to dominate the observer count, we should not assume we are in the first glitchy prototype. Most observer moments would probably be in a mature, repeated, or patched version, not the test run where the skybox still has a barcode on it. And again, you are not in the matrix. A person in a pod or a brain in a jar, you are code. They can pause you and see what happened and then delete the flaws and unpause things.

29:27They can build in triggers to warn them if there's an arising issue of detection or disbelief in the system. Indeed, they can probably do this for a brain in a jar, too. It's just harder. But maybe the simulators want us to notice. Perhaps the exit door only appears after your civilization reaches a certain level. This is a much more interesting idea than looking for a cosmic typo. After all, what is technology if not hacking reality? We discovered that lightning obeys rules, so we built electrical systems. We discovered chemistry,

29:58so we made medicine and fertilizers and semiconductors. We discovered atomic nuclei and orbital mechanics, so we unlocked nuclear power and threw machines around planets. If we are in a simulation, science may simply be the study of its rules from the inside. And there are deeper rules, escape routes, or higher layers, and then discovering them may look less like taking a red pill and more like doing physics.