# Are You Living In a Computer Simulation?

> An annotated edition of this paper on Fermat's Library, where scientists and readers explain its ideas, equations and history in the margins. Annotated edition: https://fermatslibrary.com/s/are-you-living-in-a-computer-simulation

- **Authors:** Nick Bostrom
- **Published:** 2003
- **Original:** https://s3-us-west-1.amazonaws.com/fermatslibrary/uploads%2F1465147047918-idg4vw2q5g9nerj5-5c6444d79c64e830175bcad534391647%2Fsimulation.pdf
- **Pages:** 14
- **Annotations:** 27 comments in 26 threads
- **Featured:** Fermat's Library Journal Club
- **Annotated edition:** https://fermatslibrary.com/s/are-you-living-in-a-computer-simulation

## Annotations

### Annotation 1 (page 1)

Beside the passage: “AREYOULIVINGINACOMPUTERSIMULATION?”

**Annotation** by Micael Oliveira (2016-06-07):

> In this paper, Bostrom describes his **"Simulation Argument"**. This simple argument does not directly state that we live in a simulation but instead leads to a trilemma of three unlikely-seeming propositions, one of which must be true.
>
> One of these three propositions must be true.
> 1. "The fraction of human-level civilizations that reach a posthuman stage is very close to zero"
> 2. "The fraction of posthuman civilizations that are interested in running ancestor-simulations is very close to zero"
> 3. "The fraction of all people with our kind of experiences that are living in a simulation is very close to one"
>
> Read on to learn more about Bostrom's argument.
>
> To dive deeper into this topic: [The Simulation Argument}](http://www.simulation-argument.com/)

### Annotation 2 (page 1)

Beside the passage: “ BYNICKBOSTROM ”

**Annotation** by Micael Oliveira (2016-06-05):

> Nick Bostrom is a Swedish philosopher at the University of Oxford. He is known for his work on existential risk, the anthropic principle, human enhancement ethics, superintelligence risks, the reversal test, and consequentialism. He authored more than 200 publications including New York Times bestseller *Superintelligence: Paths, Dangers, Strategies*.
>
> You can learn more about Nick on his personal home page: [nickbostrom.com](http://www.nickbostrom.com)
> ![Nick Bostrom](https://upload.wikimedia.org/wikipedia/commons/b/b4/Nick_Bostrom.jpg)

### Annotation 3 (page 1)

Beside the passage: “Thispaperarguesthatatleastoneofthefollowingpropositionsistrue:(1) the human species is very likely to go extinct before reaching a “posthuman” stage; (2) any posthuman civilization is extremely unlikely torunasignificantnumberofsimulationsoftheirevolutionaryhistory(or …”

**Annotation** by Micael Oliveira (2016-06-06):

> Here is an interview with with Nick Bostrom at the Future of Humanity Institute Oxford University:
> [![nick bostrom interview](https://i.ytimg.com/vi_webp/nnl6nY8YKHs/hqdefault.webp)](https://www.youtube.com/watch?v=nnl6nY8YKHs "nick bostrom interview")

### Annotation 4 (page 1)

Beside the passage: “Thispaperarguesthatatleastoneofthefollowingpropositionsistrue:(1) the human species is very likely to go extinct before reaching a “posthuman” stage; (2) any posthuman civilization is extremely unlikely torunasignificantnumberofsimulationsoftheirevolutionaryhistory(or …”

**Annotation** by Micael Oliveira (2016-06-06):

> This subject gained attention recently due to an interview with Elon Musk at the code conference. Someone asked him if he believed we were living in a simulation and he replied that there is a one in one billion chance that we are living in base reality. You can watch that part of the interview here:
>
>
> [![Elon Musk interview](https://i.ytimg.com/vi_webp/2KK_kzrJPS8/sddefault.webp)](https://www.youtube.com/watch?list=PLKof9YSAshgyPqlK-UUYrHfIQaOzFPSL4&v=2KK_kzrJPS8 "Elon Musk interview")

### Annotation 5 (page 1)

Beside the passage: “I.INTRODUCTION  Manyworksofsciencefictionaswellassomeforecastsbyserioustechnologists and futurologists predict that enormous amounts of computing power will be available in the future. Let us suppose for a moment that these predictions are correct. One thing that later …”

**Annotation** by Micael Oliveira (2016-06-05):

> There are mentions to simulated worlds in pop culture that date as far back as the 1950s *[(Time Out of Joint)](https://en.wikipedia.org/wiki/Time_Out_of_Joint)*. It is inherent to our civilization to want to embed itself in virtual, simulated worlds. This argument is gaining weight as we are at the dawn of virtual reality and augmented reality. If we take a look at the evolution of video games we can clearly see that they are becoming more photorealistic and that we feel a need to immerse ourselves in these virtual worlds that depict stories of our past, our present and try to imagine our future.
> With the upcoming virtual reality revolution, we should start seeing more and more experiences similar to Second Life which show a real want of humans to embed themselves in simulated realities.

### Annotation 6 (page 1)

Beside the passage: “I.INTRODUCTION  Manyworksofsciencefictionaswellassomeforecastsbyserioustechnologists and futurologists predict that enormous amounts of computing power will be available in the future. Let us suppose for a moment that these predictions are correct. One thing that later …”

**Annotation** by Micael Oliveira (2016-06-06):

> *Thinking is a function of man's immortal soul [they say.] God has given an immortal soul to every man and woman, but not to any other animal or to machines. Hence no animal or machine can think. I am unable to accept any part of this … It appears to me that the argument quoted above implies a serious restriction of the omnipotence of the Almighty. It is admitted that there are certain things that He cannot do such as making one equal to two, but should we not believe that He has freedom to confer a soul on an elephant if He sees fit? ... An argument of exactly similar form may be made for the case of machines.*
>
> ** - Alan Turing **

**Reply** by Marcelo Magallon (2016-06-07):

> Is this the “quite widely accepted position in the philosophy of mind”? I'm probably missing some subtlety in the argument because to me this sounds like “machines can think because machines can think.” My position is that thinking is a physical process and as such it can be carried out by a sufficiently finegrained simulation. I was expecting an argument for or against why thinking is or is not a physical process.

### Annotation 7 (page 2)

Beside the passage: “II.THEASSUMPTIONOF SUBSTRATE‐INDEPENDENCE  A common assumption in the philosophy of mind is that of substrate‐ independence.Theideaisthatmentalstatescansuperveneonanyofabroadclass of physical substrates. Provided a system implements the right sort of computational …”

**Annotation** by Micael Oliveira (2016-06-05):

> **Substrate-Independence:** Substrate-Independence is a term used to denote how the mind is a dynamic process and is not tied to a specific set of atoms. The functions of the mind need to carry out their processing on a platform, a substrate. If these functions can be implemented on different processing substrates, then we can call it a Substrate-Independent mind. This simple fact would allow the mind to exist in other forms and it could be transferred to a computer network. A good analogy is to think of programs that are written in platform independent code, they are independent of the architecture where they are running and can be easily transferred.

### Annotation 8 (page 2)

Beside the passage: “Argumentsfor thisthesishavebeengiven intheliterature,andalthough itis notentirelyuncontroversial,weshallheretakeitasagiven.”

**Annotation** by Micael Oliveira (2016-06-05):

> There is a lot of debate about whether we will be able to simulate consciousness in machines. The central problem lies in the fact that we have no real understanding of how the brain gives rise to the mind, of how neurons and action potentials create consciousness.
>
> Henry Markram is a neuroscientist who heads the [Blue Brain Project](https://en.wikipedia.org/wiki/Blue_Brain_Project), a project that attempts to create a synthetic brain by reverse-engineering mammalian brain circuitry. The project successfully recreated a single neocortical column of a two-week old rat, which contains about 10,000 neurons (the human brain contains 100 billion neurons). **Markram believes that it will be possible to build a model of the human brain within 10 years.**
>
> This is a very interesting topic and you can dig deeper with these interesting reads:
> - [Big Think: Can Computers be Conscious?](http://bigthink.com/going-mental/can-computers-be-conscious)
> - [Popular Science: Is A Simulated Brain Conscious?](http://www.popsci.com/article/science/simulated-brain-conscious)
> - [What It Will Take for Computers to Be Conscious](https://www.technologyreview.com/s/531146/what-it-will-take-for-computers-to-be-conscious/)

### Annotation 9 (page 3)

Beside the passage: “III.THETECHNOLOGICALLIMITSOFCOMPUTATION  At our current stage of technologicaldevelopment, we have neither sufficiently powerful hardware nor the  requisite software to create conscious minds in computers. But persuasive arguments have been given to the effect that if …”

**Annotation** by Micael Oliveira (2016-06-05):

> Ray Kurzweil predicts that by 2029 a personal computer will be 1,000 times more powerful than the human brain.
>
> This chart shows the evolution of computational power over time and Kurzweil's predictions.
> ![computer power is growing exponentially.](https://upload.wikimedia.org/wikipedia/commons/d/df/PPTExponentialGrowthof_Computing.jpg)
>
> Kurzweil argues that time periods between events in human history shrink exponentially. He believes that we are very close to having the computational power required to simulate a human brain.
>
> The following chart by Kurzweil depicts his law of accelerating returns.
> ![Countdown to singularity](https://upload.wikimedia.org/wikipedia/commons/7/71/PPTCountdowntoSingularityLog.jpg)

### Annotation 10 (page 3)

Beside the passage: “III.THETECHNOLOGICALLIMITSOFCOMPUTATION  At our current stage of technologicaldevelopment, we have neither sufficiently powerful hardware nor the  requisite software to create conscious minds in computers. But persuasive arguments have been given to the effect that if …”

**Annotation** by Micael Oliveira (2016-06-06):

> The simulation argument is time independent. It does not matter that it will take 100 or 100,000 years, all that matters is that we reach the posthuman stage before going extinct.

### Annotation 11 (page 4)

Beside the passage: “theoretical limits on the information processing attainable in a given lump of matter. We can with much greater confidence establish lower bounds on posthuman computation, by assuming only mechani sms that are already …”

**Annotation** by Micael Oliveira (2016-06-05):

> The author invokes new physical phenomena yet to be discovered and states that we can only put lower bounds on the posthuman computational power. He believes that the computational power required for the simulation hypothesis is within our reach.

### Annotation 12 (page 4)

Beside the passage: “The amount of computing power needed to emulate a  human mind can likewise be roughly estimated. One estimate, based on how computationally expensive it is to replicate the functionality of a  piece of nervous tissuethatwe have already understood and whose functionality has …”

**Annotation** by Micael Oliveira (2016-06-05):

> There are  $10^{11}$ neurons in a human brain operate by sending electrical pulses to one another via approximately $10^{15}$ synapses. Assuming 10 impulses/second this suggests that an adult human brain carries out about $10^{16}$ logical operations per second.
>
> You can read more about this topic here: [Energy Limits to the Computational Power of the Human Brain.](http://www.merkle.com/brainLimits.html)

### Annotation 13 (page 4)

Beside the passage: “Memory seems to be a no more stringent constraint than processing power. 8 Moreover,sincethemaximumhumansensorybandwidthis~10 8 bitsper second, simulating all sensory events incurs a negligible cost compared to simulating the cortical activity. We can therefore use the …”

**Annotation** by Micael Oliveira (2016-06-05):

> In the late 1980s Thomas K. Landauer performed a study in order to determine the functional capacity of human memory. In the study people were asked to read text, look at pictures, and hear words, short passages of music, sentences, and nonsense syllables. After delays ranging from minutes to several days the subjects were tested to determine how much they had retained. The remarkable result of Landauer's study was that human beings remembered **two bits per second**. Extrapolating this value over a lifetime (say 80 years), this rate of memorization would produce:
> $$3600\,\times\,24\,\times\,365.25\,\times\,80\,\times\,2 = 5,049,216,000$$
> which is somewhat over **$10^9$ bits for the entire lifetime of a human being**.
>
> You can read the study here: [Thomas K. Landauer: How Much Do People Remember? Some Estimates of the Quantity of Learned Information in Long-term Memory](http://www.cs.colorado.edu/~mozer/Teaching/syllabi/7782/readings/Landauer1986.pdf)

### Annotation 14 (page 5)

Beside the passage: “Moreover,aposthumansimulatorwouldhaveenough computing power to keep track of the detailed belief‐states in all human brains at all times. Therefore, when it saw that a human was about  to make an observation of the microscopic world, it could fill in sufficient detail in …”

**Annotation** by Micael Oliveira (2016-06-05):

> We can easily compute this number of oprations.
>
> Let's assume that:
> - H = 107 billion humans have ever lived
> - L = 50 years as the average human lifetime
> - T = 31,557,600 seconds in a year
> - O = a range of [$10^{14}$, $10^{17}$] operations in each human brain per second
>
> $$ H \times L \times T \times O \simeq [10^{33}, 10^{36}] \,\,\mathrm{operations} $$

### Annotation 15 (page 5)

Beside the passage: “5”

**Annotation** by Micael Oliveira (2016-06-05):

> As you go deeper into the simulation and you consider multi-level simulations (simulations within simulations), the cost of simulating the human history becomes marginal to the cost of simulating machines capable of creating simulations.

### Annotation 16 (page 6)

Beside the passage: “IV.THECOREOFTHESIMULATIONARGUMENT  Thebasicideaof thispapercanbeexpressedroughlyasfollows:Iftherewerea  substantialchancethat ourcivilizationwillevergettotheposthumanstageand run many ancestor‐simulations, then how come you are not living in such a simulation?  …”

**Annotation** by Micael Oliveira (2016-06-05):

> The author develops a probabilistic consideration of the argument of this paper in order to determine whether or not we live in a simulation.

### Annotation 17 (page 7)

Beside the passage: “(1) 0 P f  (2) 0 I f  (3) 1  sim f  ”

**Annotation** by Micael Oliveira (2016-06-05):

> **(1)** $f_p \approx 0 \Rightarrow  $ *"The fraction of human-level civilizations that reach a posthuman stage is very close to zero"*
>
> **(2)** $f_I \approx 0 \Rightarrow $*"The fraction of posthuman civilizations that are interested in running ancestor-simulations is very close to zero"*
>
> **(3)** $f_{sim} \approx 1 \Rightarrow $ *"The fraction of all people with our kind of experiences that are living in a simulation is very close to one"*

### Annotation 18 (page 8)

Beside the passage: “The same reasoning holds if S is not the property of having a certain genetic sequence but instead the property of being in a simulation, assuming only that  we have no information that enables us to predict any differences betweentheexperiences ofsimulatedminds …”

**Annotation** by Micael Oliveira (2016-06-05):

> **The Brain in a Vat** thought-experiment is most commonly used to illustrate global or Cartesian skepticism. Imagine that you are a brain hooked up to a sophisticated computer program that can perfectly simulate experiences of the outside world. If you cannot be sure that you are not a brain in a vat, then you cannot rule out the possibility that all that you know about the external world are false. We can construct the following skeptical argument. Let “B” stand for any belief or claim about the external world, for example, the grass is green.
>
> 1. If I know that B, then I know that I am not a brain in a vat.
> 2. I do not know that I am not a brain in a vat.
> 3. Thus, I do not know that B.
>
> You can read more about it here: [The Brain in a Vat Argument
> ](http://www.iep.utm.edu/brainvat/)

### Annotation 19 (page 9)

Beside the passage: “Further, one can consider a sequence of possible situations in which an increasing fraction of all people live in simulations: 98%, 99%, 99.9%, 99.9999%, and so on. As one approaches the limiting case in which everybody is in a simulation (from which one can deductively …”

**Annotation** by Andrew Robinson (2018-04-20):

> Couldn't you consider the other limiting case with decreasing fraction of people living in simulations?  Can someone explain why the upward bound is more likely than the lower bound.  I feel like the hypothesis hinges on this assumption.

### Annotation 20 (page 10)

Beside the passage: “There are many ways in which humanity could become extinct before reachingposthumanity.Perhapsthemostnaturalinterpretationof(1)isthatwe are likely to go extinct as a result of the development of some powerful but dangerous technology. 13  One candidate is molecular …”

**Annotation** by Micael Oliveira (2016-06-05):

> We could go extinct trying to reach posthuman stage. Here is a video where Nick Bostrom goes into more detail about this possibility: [![Nick Bostrom asks, Will we engineer our own extinction?](http://dwgyu36up6iuz.cloudfront.net/heru80fdn/image/upload/c_fill,d_placeholder_thenewyorker.png,fl_progressive,g_face,h_450,q_80,w_800/v1447449753/thenewyorker_nick-bostrom-2.jpg)](https://www.youtube.com/watch?v=5LStef_kV6Q "nbi2")

### Annotation 21 (page 11)

Beside the passage: “individuals who have sufficient resources and interest to run ancestor‐ simulations; or else they have reliably enforced laws that prevent such individualsfromactingontheirdesires.  What force could bring about such convergence? One can speculate that …”

**Annotation** by Micael Oliveira (2016-06-05):

> Running ancestor simulations is either banned by future civilizations or posthumans might simply lose interest in creating ancestor simulations.

### Annotation 22 (page 11)

Beside the passage: “individuals who have sufficient resources and interest to run ancestor‐ simulations; or else they have reliably enforced laws that prevent such individualsfromactingontheirdesires.  What force could bring about such convergence? One can speculate that …”

**Annotation** by Micael Oliveira (2016-06-05):

> Humans experience the world through their senses. What we call "reality" is what we  interact with via our senses and our beliefs. There could be a base reality to which we have no direct access that turns out to be a different "reality" than the one we have access to.

### Annotation 23 (page 12)

Beside the passage: “machine,andsoon,inarbitrarilymanystepsofiteration.Ifwedogoontocreate ourownancestor‐simulations,thiswouldbestrongevidenceagainst(1)and(2), andwewouldthereforehavetoconcludethatweliveinasimulation.Moreover, we would have to suspect that the  posthumans running our …”

**Annotation** by Micael Oliveira (2016-06-05):

> If we believe that we are in a simulation and that our simulators are in a simulation too then we live in a multi-level simulation. We can safely assume that we are on our way to create our own ancestor-simulations.

### Annotation 24 (page 12)

Beside the passage: “machine,andsoon,inarbitrarilymanystepsofiteration.Ifwedogoontocreate ourownancestor‐simulations,thiswouldbestrongevidenceagainst(1)and(2), andwewouldthereforehavetoconcludethatweliveinasimulation.Moreover, we would have to suspect that the  posthumans running our …”

**Annotation** by Micael Oliveira (2016-06-05):

> This is an interesting argument against the multi-level hypothesis. This would mean that if we are in a simulation then we ourselves will never achieve the posthuman level and will never be able to create ancestor-simulations.

### Annotation 25 (page 12)

Beside the passage: “machine,andsoon,inarbitrarilymanystepsofiteration.Ifwedogoontocreate ourownancestor‐simulations,thiswouldbestrongevidenceagainst(1)and(2), andwewouldthereforehavetoconcludethatweliveinasimulation.Moreover, we would have to suspect that the  posthumans running our …”

**Annotation** by Micael Oliveira (2016-06-05):

> The author makes a good analogy with religion. **The creators of ancestor-simulations can be seen as Gods.** They are omnipotent and decide when and how to create, and stop the simulations. Assuming we live in multi-level simulations then only the creators in base reality are really omnipotent and omniscient and have control over all the simulations.

### Annotation 26 (page 13)

Beside the passage: “There is also the possibility of simulators abridging certain parts of the mental lives  of simulated beings and giving them  false memories of the sort of experiencesthattheywouldtypicallyhavehadduringtheomittedinterval.Ifso, one can consider the following (farfetched) …”

**Annotation** by Micael Oliveira (2016-06-05):

> **The simulation hypothesis cannot be proven or disproven**. It also cannot matter, it cannot make any difference to us. If it's true, it's not that trees and clouds and other people don't exist, it's just that trees and clouds and other people have a different ultimate nature than we thought.

## Full text

Extracted from the PDF. [Annotation N] marks where each annotation above sits.

#### Page 1

1

AREYOULIVINGINACOMPUTERSIMULATION? [Annotation 1]

 [Annotation 2] BYNICKBOSTROM 

[PublishedinPhilosophicalQuarterly(2003)Vol.53,No.21 1,pp.243‐255.(Firstversion:2001)]  

Thispaperarguesthatatleastoneofthefollowingpropositionsistrue:(1) [Annotation 3] the human species is very likely to go extinct before reaching a “posthuman” stage; (2) any posthuman civilization is extremely unlikely torunasignificantnumberofsimulationsoftheirevolutionaryhistory(or variations thereof); (3) [Annotation 4]  we are almost certainly living in a computer simulation.It follows thatthebeliefthatthereisa significant chance that we will one day become posthumans who run ancestor‐simulations is false, unless we are currently living in a simulation. A number of other consequencesofthisresultare alsodiscussed.  

I.INTRODUCTION  Manyworksofsciencefictionaswellassomeforecastsbyserioustechnologists and futurologists predict that enormous amounts of computing power will be available in the future. Let us suppose for a moment that these predictions are correct. One thing that later generations might [Annotation 5] do with their super‐powerful computers is run detailed simulations of their forebears or of people like their forebears.Becausetheircomputerswouldbesopowerful,theycouldrunagreat many such simulations. Suppose that these simulated people are conscious (as they would be if the simulations were sufficiently fine [Annotation 6] ‐grained and if a certain quite widely accepted position in the philosophy of mind is correct). Then it could be the casethatthevast majority of minds like ours do not belong tothe original race but rather to people simulated bythe advanced descendants of an originalrace. Itisthenpossibletoarguethat,ifthiswerethecase,wewouldbe rational to think that we are likely among the simulated minds rather than among the original biological ones. Therefore, if we don’t think that we are currentlylivingina computersimulation,wearenotentitled tobelievethatwe will have descendants who will run lots of such simulations of their forebears. Thatisthebasicidea.Therestofthispaperwillspellitoutmorecarefully.

#### Page 2

2

 Apartformtheinterestthisthesismayholdforthosewhoareengagedin futuristic speculation, there are also more purely theoretical rewards. The argument provides a stimulus for formulating some methodological and metaphysical questions, and it suggests naturalistic analogies to certain traditional religious conceptions, which some may find amusing or thought‐ provoking.  Thestructureofthepaperisasfollow s.First,weformulateanassumption thatweneedtoimportfromthephilosophyofmindinordertogettheargument started. Second, we consider some empirical reasons for thinking that running vastly many simulations of human minds would be within the capability of a future civilization that has developed many of those technologies that can already be  shown to be compatible with known physical laws and engineering constraints.Thispartisnotphilosophicallynecessarybutitprovidesanincentive for paying attention to the rest. Then follows the  core of the argument, which makes use of some simple probability theory, and a section providing support foraweak indifference principle thattheargument employs.Lastly, wediscuss someinterpretationsofthedisjunction,mentionedintheabstract,thatformsthe conclusionofthesimulationargument. 

II.THEASSUMPTIONOF SUBSTRATE‐INDEPENDENCE  A common assumption in the philosophy of mind is that of substrate‐ [Annotation 7] independence.Theideaisthatmentalstatescansuperveneonanyofabroadclass of physical substrates. Provided a system implements the right sort of computational structures and processes, it can be associated with conscious experiences. It is not an essential property of consciousness that it is implemented on carbon‐based biological neural networks inside a cranium: silicon‐basedprocessorsinsideacomputercouldinprincipledothetrickaswell.

Argumentsfor thisthesishavebeengiven intheliterature,andalthough itis [Annotation 8] notentirelyuncontroversial,weshallheretakeitasagiven.

The argument we shall present does not, however, depend on any very strong version of functionalism or computationalism. For example,weneednot assume that the thesis of substrate‐independence is necessarily true (either analyticallyormetaphysically)–just that,infact,acomputerrunningasuitable program would be conscious. Moreover, we need not assume that in order to createamindon a computer it  would be sufficient toprogram it in such away that it behaves like a human in all situations, including passing the Turing  test etc.Weneedonlytheweakerassumptionthatitwouldsufficeforthegeneration ofsubjectiveexperiences that the computational processesof ahumanbrainare structurally replicated in suitably fine‐grained detail, such as on the level of

#### Page 3

3

individual synapses. This attenuated version of substrate‐independenceis quite widelyaccepted.

Neurotransmitters, nerve growth factors, and other chemicals that are smallerthanasynapseclearly playa roleinhumancognitionandlearning.The substrate‐independencethesis isnotthattheeffectsofthese chemicalsaresmall orirrelevant, butratherthattheyaffectsubjectiveexperienceonlyviatheirdirect orindirectinfluenceoncomputationalactivities.Forexample,iftherecanbeno difference in subjective experience without there also being a difference in synapticdischarges,thentherequisitedetailofsimulationisatthesynapticlevel (orhigher). 

III.THETECHNOLOGICALLIMITSOFCOMPUTATION  At our current stage of technologicaldevelopment, we have neither sufficiently powerful hardware nor the  requisite software to create conscious minds in computers. But persuasive arguments have been given to the effect that if technological progress continues unabated then these shortcomings will eventually be [Annotation 9]  overcome. Some authors argue that this stage may be only a few decades away. 1  Yet present purposes require no assumptions about the time‐ scale. The simulation argument works equally well for those who think that it will take hundreds of thousands of years to reach a “posthuman” stage of civilization, where humankind has acquired most of the technological [Annotation 10] capabilities that one cancurrently show tobeconsistentwithphysicallawsand withmaterialandenergyconstraints.

Suchamaturestageoftechnologicaldevelopmentwillmakeitpossibleto convert planets and other astronomical resources into enormously powerful computers. It is currently hard to be confident in any upper bound on the computing power that may  be available to posthuman civilizations. As we are still lacking a “theory of everything”, we cannot  rule out the possibility that novel physical phenomena,notallowedforincurrentphysicaltheories,maybe utilizedtotranscendthoseconstraints 2 thatinourcurrentunderstandingimpose

1  See e.g. K. E. Drexler, Engines of Creation: The Coming Era of Nanotechnology, London, Forth Estate, 1985; N. Bostrom, “How Long Before Superintelligence?” International Journal of Futures Studies, vol. 2, (1998); R. Kurzweil, The Age of Spiritual Machines: When computers exceed human intelligence,NewYork,VikingPress,1999;H. Moravec,Robot:MereMachinetoTranscendentMind, OxfordUniversityPress,1999. 2  Such as the Bremermann‐Bekenstein bound and the black hole limit (H. J. Bremermann, “Minimumenergyrequirements of information transfer and computing.” InternationalJournalof TheoreticalPhysics 21: 203‐217(1982);J. D. Bekenstein, “Entropy contentand informationflowin systemswithlimitedenergy.”PhysicalReviewD30:1669‐1679 (1984);A.Sandberg,“ThePhysics

#### Page 4

4

theoretical limits on the information processing attainable in a given lump of matter. We can with much greater confidence establish lower bounds on [Annotation 11] posthuman computation, by assuming only mechani sms that are already understood.Forexample,EricDrexlerhasoutlinedadesignforasystemthesize of a sugar cube  (excluding cooling and power supply) that would perform  10 21  instructionspersecond. 3 Anotherauthorgivesaroughestimateof10 42 operations per second for a computerwitha mass on order of alarge planet. 4  (If wecould createquantumcompu ters,orlearntobuildcomputersoutofnuclearmatteror plasma, we could push closer to the theoretical limits. Seth Lloyd calculates an upperboundfora1kg computerof5*10 50 logi cal operations persecondcarried out on ~10 31  bits. 5  However, it suffices for our purposes to use the more conservativeestimatethatpresupposesonlycurrentlyknowndesign‐principles.)

The amount of computing power needed to emulate a  human mind can likewise be roughly estimated. One estimate, based on how computationally expensive it is to replicate the functionality of a  piece of nervous tissuethatwe have already understood and whose functionality has been replicated in silico, contrastenhancementintheretina,yieldsafigureof~10 14 operationspersecond for the entire human brain. [Annotation 12] 6  An alternative estimate, based the number of synapses in the brain and their firing frequency, gives a figure of ~10 16 ‐10 17  operationspersecond. 7 Conceivably,evenmorecouldberequiredifwewantto simulate in detail the internal workings of synapses and dendritic trees. However,itislikelythatthehumancentralnervoussystemhasahighdegreeof redundancy on the mircoscale to compensate forthe unreliability and noisiness ofits neuronal components.Onewouldtherefore expect a substantialefficiency gainwhenusingmorereliableandversatilenon‐biologicalprocessors.

Memory seems to be a no more stringent constraint than processing power. 8 Moreover,sincethemaximumhumansensorybandwidthis~10 8 bitsper second, simulating all sensory events incurs a negligible cost compared to [Annotation 13] simulating the cortical activity. We can therefore use the processing power

of Information Processing Superobjects: The Daily Life among the Jupiter Brains.” Journal of EvolutionandTechnology,vol.5(1999)). 3  K. E. Drexler, Nanosystems: Molecular Machinery, Manufacturing, and Computation, New York, JohnWiley&Sons,Inc.,1992. 4  R. J. Bradbury, “Matrioshka Brains.” Working manuscript (2002), http://www.aeiveos.com/~bradbury/MatrioshkaBrains/MatrioshkaBrains.html. 5 S.Lloyd,“Ultimatephysicallimitstocomputation.”Nature406(31August):1047‐1054(2000). 6 H.Moravec,MindChildren,HarvardUniversityPress(1989). 7 Bostrom(1998),op.cit. 8 Seereferencesinforegoingfootnotes.

#### Page 5

5 [Annotation 15]

required to simulate the central nervous system as an estimate of the total computationalcostofsimulatingahumanmind.

If the environment is included in the simulation, this will require additionalcomputingpower–howmuchdependsonthescopeandgranularity of the simulation. Simulating the entire universe  down to the quantum level is obviouslyinfeasible, unless radically new physics is discovered. But in order to get a realistic simulation of human experience, much less is needed – only whateverisrequiredtoensurethatthesimulatedhumans,interactinginnormal human ways with their simulated environment, don’t notice any irregularities. ThemicroscopicstructureoftheinsideoftheEarthcanbesafelyomitted.Distant astronomicalobjectscanhavehighlycompressedrepresentations:verisimilitude need extend to the narrow band of properties that we can observe from our planetorsolarsystemspacecraft.OnthesurfaceofEarth,macroscopicobjects in inhabited areas may need to be continuously simulated, but microscopic phenomena could likely be filled in ad hoc. What you see through an electron microscope needs to look unsuspicious, but you usually have no way of confirming its coherence with unobserved parts of the microscopic world. Exceptions arise when  we deliberately design systems to harness unobserved microscopicphenomenathatoperateinaccordancewithknownprinciplestoget resultsthatweareabletoindependentlyverify.Theparadigmaticcaseofthisis a computer. The simulation may therefore need to include a continuous representationofcomputersdowntothelevel ofindividuallogicelements.This presents no problem, since our current computing power is negligible by posthumanstandards.

Moreover,aposthumansimulatorwouldhaveenough computing power to keep track of the detailed belief‐states in all human brains at all times. Therefore, when it saw that a human was about  to make an observation of the microscopic world, it could fill in sufficient detail in the simulation in the appropriatedomainon an as‐needed basis.Should anyerror occur, thedirector couldeasilyeditthestates ofanybrainsthathavebecomeawareofananomaly before it spoils the simulation. Alternatively,the director could skip back a few secondsandrerunthesimulationinawaythatavoidstheproblem.  It thus seems plausible that the main computational cost in creating simulations that are indistinguishable from physicalreality for human minds in thesimulationresidesinsimulatingorganic [Annotation 14] brainsdowntotheneuronalorsub‐ neuronallevel. 9 Whileitisnotpossibletogetaveryexactestimateofthecostofa realistic simulation of human history, we can use ~10 33 ‐10 36  operations as a

9 As webuild moreand faster computers,thecostofsimulating ourmachinesmighteventually cometodominatethecostofsimulatingnervoussystems.

#### Page 6

6

roughestimate 10 .Aswe gain moreexperience with virtual reality,we will get a better grasp of the computational requirements for making such worlds appear realistic to their visitors. But in any case, even if our estimate is off by several ordersofmagnitude,thisdoesnotmattermuchforourargument. Wenotedthat a rough approximation of the computational power of a planetary‐mass computer is 10 42 operations per second, and that assumes only already known nanotechnologicaldesigns,whichareprobablyfarfromoptimal.Asinglesucha computer could simulate the entire mental history of humankind (call this an ancestor‐simulation) by using less than one millionth of its processing power for one second. A posthuman civilization may eventually build an astronomical numberofsuchcomputers.Wecanconcludethatthecomputingpoweravailable to a posthuman civilization is sufficient to run a huge number of ancestor‐ simulations even it allocates only a minute fraction of its resources to that purpose.Wecandrawthisconc lusioneven whileleavingasubstantialmarginof errorinallourestimates. 

 Posthuman civilizations would have enough computing power to run

hugelymanyancestor‐simulationsevenwhileusingonlyatinyfractionof theirresourcesforthatpurpose. 

IV.THECOREOFTHESIMULATIONARGUMENT  Thebasicideaof [Annotation 16] thispapercanbeexpressedroughlyasfollows:Iftherewerea  substantialchancethat ourcivilizationwillevergettotheposthumanstageand run many ancestor‐simulations, then how come you are not living in such a simulation?  Weshalldevelopthisideaintoarigorousargument.Let usintroducethe followingnotation: 

P f : Fraction of all human‐leveltechnological civilizations that survive to reachaposthumanstage  N : Average number of ancestor‐simulations run by a posthuman civilization  H :Averagenumberofindividualsthathavelivedinacivilizationbefore itreachesaposthumanstage

10  100 billion humans  50 years/human  30 million secs/year  [10 14 , 10 17 ] operations in each humanbrainpersecond  [10 33 ,10 36 ]operations.

#### Page 7

7

 The actual fraction of all observers with human‐type experiences that live in simulationsisthen 

HHNf

HNf f

P

P sim  

)( 

 Writing I f  for the fraction of posthuman civilizations that are interested in running ancestor‐simulations (or that contain at leastsome individuals  whoare interested in that and have sufficient resources to run a significant number of such simulations), and I N  for the average number of ancestor‐simulations run bysuchinterestedcivilizations,wehave 

II NfN    andthus: 

1)(  

IIP

IIP sim Nff

Nff f    (*)

 Because of the immense computing power of posthuman civilizations, I N  is extremelylarge,aswesawintheprevioussection.Byinspecting(*)wecanthen seethatatleastoneofthefollowingthreepropositionsmustbetrue: 

(1) 0 P f  (2) 0 I [Annotation 17] f  (3) 1  sim f  

V.ABLANDINDIFFERENCEPRINCIPLE  Wecantakeafurtherstepandconcludethatconditionalonthetruthof(3),one’s credence in the hypothesis that one is in a simulation should be close to unity. More generally, if we knew that a fraction x of all observers with human ‐type experiencesliveinsimulations,andwedon’thaveanyinformationthatindicate that our own particular expe riences are any more or less likely than other human‐type experiences to have been implemented in vivo rather than in machina,thenourcredencethatweareinasimulationshouldequal x: 

#### Page 8

8

xxfSIMCr sim  )|( (#)  This stepis sanctioned bya veryweakindifference principle.Letusdistinguish twocases.Thefirstcase,whichistheeasiest,iswhereallthemindsinquestion are like your own in the sense that they are exactly qualitatively identical to yours:theyhaveexactlythesame informationandthesameexperiencesthatyou have.Thesecondcaseiswherethemindsare“like”eachotheronlyintheloose senseofbeingthesortofmindsthataretypicalofhumancreatures,but theyare qualitativelydistinctfromoneanotherandeachhasadistinctset ofexperiences. I maintain that even in the latter case, where the minds are qualitatively different, the simulation argument still works, provided that you have no information that bears on the question of which of the various minds are simulatedandwhichareimplementedbiologically.  Adetaileddefenseofa strongerprinciple,whichimpliestheabovestance forbothcasesastrivialspecialinstances,hasbeengivenintheliterature. 11 Space doesnotpermita recapitulationofthatdefensehere,but wecanbringoutoneof the underlying intuitions by bringing to our attentionto an analogous situation ofamorefamiliarkind.Supposethatx%ofthepopulationhasacertaingenetic sequence Swithinthe partof theirDNAcommonlydesignatedas“junkDNA”. Suppose,further,thattherearenomanifestationsofS(shortofwhatwouldturn up in a gene assay) and that there areno knowncorrelations between having S andanyobservablecharacteristic.Then,quiteclearly, unlessyouhavehadyour DNA sequenced, it is rational to assign a credence of x% to the hypothesis that youhaveS.Andthisissoquiteirrespectiveofthefactthatthepeoplewhohave Shavequalitativelydifferentmindsandexperiencesfromthepeoplewhodon’t haveS.(Theyaredifferentsimplybecause allhumanshavedifferentexperiences from one another, not because of any known link between S and what kind of experiencesonehas.)

The same reasoning holds if S is not the property of having a certain genetic sequence but instead the property of being in a simulation, assuming only that [Annotation 18]  we have no information that enables us to predict any differences betweentheexperiences ofsimulatedminds andthoseof the originalbiological minds.

Itshouldbestressedthattheblandindifferenceprincipleexpressedby(#) prescribes indifference only between hypotheses aboutwhich observer you are, when you have no informationabout which of these observers you are. It does

11 Ine.g.N.Bostrom,“TheDoomsdayargument,Adam&Eve,UN ++ ,andQuantumJoe.”Synthese 127(3): 359‐387 (2001); and most fully in my book Anthropic Bias: Observation Selection Effects in ScienceandPhilosophy,Routledge,NewYork,2002.

#### Page 9

9

notingeneralprescribeindifferencebetweenhypotheseswhenyoulackspecific informationabout whichofthehypothesesistrue.Incontrast to Laplacean and other more ambitious principles of indifference, it is therefore immune to Bertrand’s paradox and similar predicaments that tend to plague indifference principlesofunrestrictedscope.

Readers familiar  with the Doomsday argument 12  may worry that the bland principle of indifference invoked here is the same assumption that is responsible for getting the  Doomsday argument off the ground, and that the counterintuitivenessofsomeoftheimplicationsofthelatterincriminatesorcasts doubtonthevalidityoftheformer.Thisisnot so.TheDoomsdayargumentrests on a much stronger and more controversial premiss, namely that one should reasonas ifone werea randomsample from the set of allpeoplewhowill ever havelived(past,present,andfuture)eventhoughweknowthatwearelivinginthe early  twenty‐first century rather than at some point in the distant past or the future.Theblandindifferenceprinciple,bycontrast,a ppliesonlytocaseswhere wehavenoinformati onaboutwhichgroupofpeoplewebelongto.

If betting odds provide some guidance to rational belief, it may also be worthto ponder that if everybody wereto placea bet on whether they are in a simulation or not, then if people use the bland principle of indifference, and consequentlyplacetheirmoneyonbeinginasimulationiftheyknowthatthat’s wherealmostallpeopleare,then almosteveryonewillwintheirbets.Iftheybet onnotbeinginasimulation,thenalmosteveryonewilllose.Itseemsbetterthat theblandindifferenceprinciplebeheeded.

Further, one can consider a sequence of possible situations in which an increasing fraction of all people live in simulations: 98%, 99%, 99.9%, 99.9999%, and so on. As one approaches the limiting case in which everybody is in a simulation (from which one can deductively infer that one is in a simulation [Annotation 19] oneself), it is plausible to require that the credence one assigns to being in a simulation gradually approach the  limiting case of complete certainty in a matchingmanner. 

VI.INTERPRETATION  The possibility represented by proposition (1) is fairly straightforward. If (1) is true, then humankind will almost certainly fail to reach a posthuman level; for virtually no species at our level of development become posthuman, and it is hardtoseeany justification for thinkingthat our ownspecies will be especially privileged or protected from future disasters. Conditional on (1), therefore, we

12 Seee.g.J.Leslie,“IstheEndoftheWorldNigh?”PhilosophicalQuarterly40,158:65‐72(1990).

#### Page 10

10

must give  a high credence to DOOM, the hypothesis that humankind will go extinctbeforereachingaposthumanlevel: 

1)0|(  P fDOOMCr  

One can imagine hypothetical situations were we have such evidence as would trump knowledge of P f . For example, if we discovered that we were about to be hit by a giant meteor, this might suggest that we had been exceptionally unlucky. We could then assign a  credence to DOOM larger than our expectation of the fraction of human‐level civilizations that fail to reach posthumanity.In theactualcase,however,weseemtolackevidenceforthinking thatwearespecialinthisregard,forbetterorworse.  Proposition(1)doesn’tbyitselfimplythatwearelikelytogoextinctsoon, only that we are unlikely to reach a posthuman stage. This possibility is compatible  with us remaining at, or somewhat above,  our current level of technologicaldevelopmentfor alongtimebeforegoingextinct.Anotherwayfor (1)tobetrueisifitislikelythattechnologicalcivilizationwillcollapse.Primitive humansocietiesmightthenremainonEarthindefinitely.

There are many ways in which humanity could become extinct before reachingposthumanity.Perhapsthemostnaturalinterpretationof(1)isthatwe are likely to go extinct as a result of the development of some powerful but dangerous technology. 13  One candidate is molecular nanotechnology, which in its mature stage would enable the construction of self‐replicating nanobots [Annotation 20] capable of feeding on dirt and organic matter – a kind of mechanical bacteria. Suchnanobots,designedformaliciousends,couldcausetheextinctionofalllife onourplanet. 14   Thesecondalternativeinthesimulationargument’sconclusionisthat the fraction of posthuman civilizations that are interested in running ancestor‐ simulationisnegligiblysmall.Inorder for (2) tobetrue,theremustbe a  strong convergence among the courses of advanced civilizations. If the number of ancestor‐simulations createdbytheinterestedcivilizationsisextremelylarge,the rarity of such civilizations must be correspondingly extreme. Virtually no posthuman civilizations decide to use their resources to run large numbers of ancestor‐simulations. Furthermore, virtually all posthuman civilizations lack

13 Seemypaper“ExistentialRisks:AnalyzingHumanExtinctionScenariosandRelatedHazards.” Journal of Evolution and Technology, vol. 9 (2001) for a survey and analysis of the present and anticipatedfuturethreatstohumansurvival. 14  See e.g. Drexler (1985) op cit., and R. A. Freitas Jr., “Some Limits to Global Ecophagy by Biovorous Nanoreplicators, with Public Policy Recommendations.” Zyvex preprint April (2000), http://www.foresight.org/NanoRev/Ecophagy.html.

#### Page 11

11

individuals who have sufficient resources and interest to run ancestor‐ simulations; or else they have reliably enforced laws that prevent such individualsfromactingontheirdesires.  What force could bring about such convergence? One can speculate that advancedcivilizationsalldevelopalongatrajectorythatleadstotherecognition of an ethical prohibition against running ancestor‐simulations because of the sufferingthatisinflictedontheinhabitantsofthesimulation.However,fromour present point of view, it is not clear that creating a human race is immoral. On the contrary, we tend to view the existence of our race  as constituting a great ethical value. Moreover, convergence on an ethical view of the immorality of running ancestor‐simulations is not enough: it must be combined with convergence on a civilization‐wide social structure that enables activities consideredimmoraltobeeffectivelybanned.  Another possible convergence point is that almost  all individual posthumansinvirtuallyallposthumancivilizationsdevelopinadirectionwhere theylosetheirdesirestorunancestor‐simulations.Thiswouldrequiresignificant changes to the motivations driving their human predecessors, for there are certainlymanyhumanswhowouldliketorunancestor‐simulationsiftheycould afford todoso.Butperhapsmanyofourhumandesireswillberegardedassilly by anyone who becomes a  posthuman. Maybe the scientific value of ancestor‐ simulations to a posthuman civilization is negligible (which is not too implausible given its unfathomable intellectual superiority), and maybe posthumans regard recreational activities as  merely a very inefficient way of getting pleasure – which can be obtained much more cheaply by direct [Annotation 21] stimulationofthebrain’srewardcenters.Oneconclusionthatfollowsfrom(2)is that posthuman societies will be very different from human societies: they will not contain relatively wealthy independent agents who have the full gamut of human‐likedesiresandarefreetoactonthem.  The possibility expressed  by alternative (3) is the conceptually most intriguing one. If we are living in a simulation, then the cosmos that we are observing isjust atinypiece ofthetotalityof physicalexistence. The physics in theuniversewherethecomputerissituatedthat is running thesimulationmay ormay not resemble thephysics of theworldthatweobserve. While the world weseeisinsomesense“real”,itisnotlocatedatthefundamentallevelofreality.  Itmaybepossi bleforsimulatedcivilizationstobecomeposthuman.They [Annotation 22] maythenruntheirownancestor‐simulationsonpowerfulcomputerstheybuild in their simulated universe. Such computers would be “virtual machines”, a familiar concept in computer science.(Javascript web‐applets,for instance, run on a  virtual machine –  a simulated computer – inside your desktop.) Virtual machinescan bestacked: it’s possibleto simulatea machine simulating another

#### Page 12

12

machine,andsoon,inarbitrarilymanystepsofiteration.Ifwedogoontocreate ourownancestor‐simulations,thiswouldbestrongevidenceagainst(1)and(2), andwewouldthereforehavetoconcludethatweliveinasimulation.Moreover, we would have to suspect that the  posthumans running our simulation are themselvessimulated beings; and their creators, in turn, may also be simulated [Annotation 23] beings.  Reality may thus contain many levels. Even if it is necessary for the hierarchy to bottom out at some stage – the metaphysical status of this claim is somewhat obscure – there  may be room for a large number of levels of reality, and the number could be increasing over time. (One consideration that counts against the multi‐level hypothesis is that the computational cost for the basement‐level simulators would be very great. Simulating eve n a  single posthuman civilization might be [Annotation 24] prohibitively expensive. If so, then we should expect our simulation to be terminated when we are about to become posthuman.)  Although all the elements of such a system can be naturalistic, even physical,itispossibletodrawsomelooseanalogieswithreligiousconceptionsof theworld.Insomeways, theposthumansrunningasimulationare likegodsin relation to the people inhabiting the simulation: the posthumans created the [Annotation 25] world we see; they are of superior intelligence; they are “omnipotent” in the sense that they can interfere in the workings of our world even in ways that violate its physical  laws; and they are “omniscient” in the sense that they can monitoreverythingthathappens.However,allthedemigodsexceptthoseatthe fundamental level of reality are subject to sanctions by the morepowerful  gods livingatlowerlevels.

Furtherruminationonthesethemescouldclimaxinanaturalistic theogony thatwouldstudythestructureofthishierarchy,andtheconstraintsimposedon its inhabitants by the possibilitythat their actions on their own level may affect thetreatmenttheyreceivefromdwellersofdeeperlevels.Forexample,ifnobody can be sure that they are at the basement‐level,  then everybody would have to consider the possibility that their actions will be rewarded or punished, based perhaps on moral criteria, by their simulators. An afterlife would be a real possibility. Because of this fundamental uncertainty, even the basement civilization may have a reason to behave ethically. The fact that it has such a reason for moral behavior would of course add to everybody else’s reason for behaving morally,  and so on, in truly virtuous circle. One might get a kind of universal ethical imperative, which it would be in everybody’s self‐interest to obey,asitwere“fromnowhere”.

In additionto a ncestor‐simulations,one may alsoconsiderthepossibility of more selective simulations that include only a small group of humans or a

#### Page 13

13

single individual. The rest of humanity would then be zombies or “shadow‐ people” – humans simulated only at a level sufficient for the fully simulated people not to notice anything suspicious. It is not clear how much cheaper shadow‐peoplewouldbetosimulatethanrealpeople.Itisnoteven obviousthat itispossibleforanentitytobehaveindistinguishablyfromarealhumanandyet lack conscious experience. Even if there are such selective simulations, you should not think that you are in one of them unless you think they are much more numerous than complete simulations. There would have to be about 100 billion times as many “me‐simulations” (simulations of the life of only a  single mind)asthereareancestor‐simulationsinorderformostsimulatedpersonstobe inme‐simulations.

There is also the possibility of simulators abridging certain parts of the mental lives  of simulated beings and giving them  false memories of the sort of experiencesthattheywouldtypicallyhavehadduringtheomittedinterval.Ifso, one can consider the following (farfetched) solution to theproblem of evil: that thereisnosufferingintheworldandallmemoriesofsuffering areillusions.Of course,thishypothesiscanbeseriouslyentertainedonlyatthosetimeswhenyou arenotcurrentlysuffering.  Supposing we live in a simulation, what are the implications for us humans? The foregoing remarks notwithstanding, the implications are not all that radical. Our best guide to how our posthuman creators have chosen to set up our world is the standard empirical  study of the universe we see. The revisionstomostpartsofourbeliefnetworkswouldberatherslightandsubtle– inproportionto our lack of confidence inourability to understand theways of posthumans.  Properly understood, therefore, the truth of (3) should have no tendencytomakeus“gocrazy”ortopreventusfromgoingaboutourbusiness [Annotation 26] andmakingplansandpredictionsfortomorrow.Thechiefempiricalimportance of (3) at the current time seems to lie in its role in the tripartite conclusion established above. 15  We may hope that (3) is true since that woulddecrease the probability of (1), although if computational constraints make it likely that simulators would terminate a simulation before it reaches a posthuman level, thenoutbesthopewouldbethat(2)istrue.

If we learnmoreabout posthuman motivations and resource constraints, maybeasa result of developing towards becoming posthumans ourselves, then the hypothesis that we are simulated will come to have a much richer set of empiricalimplications.

15  For some reflections by another author on the consequences of (3), which were sparked by a privatelycirc ulated earlierversionofthis paper, see R. Hanson, “HowtoLiveina Simulation.” JournalofEvolutionandTechnology,vol.7(2001).

#### Page 14

14



VII.CONCLUSION  A technologically mature “posthuman” civilization would have enormous computing power. Based on this empirical fact,thesimulation argument shows that at least one of the following propositions is true: (1) The fractionof human‐ level civilizations that reach a posthuman stage is very close to zero; (2) The  fraction of posthuman civilizations that are interested in running ancestor‐ simulations is very close to zero; (3) The fraction of all people with our kind of experiencesthatarelivinginasimulationisveryclosetoone.

If (1) is true, then we will almost certainly go extinct before reaching posthumanity.If(2)istrue,thentheremustbea strongconvergenceamongthe courses of advanced civilizations so that virtually none contains any relatively wealthyindividualswhodesiretorunancestor‐simulationsandarefreetodoso. If (3) is true, then we almost  certainly live in a  simulation. In the dark forest of our current ignorance, it seems sensible to apportion one’s credence roughly evenlybetween(1),(2),and(3).

Unless we are now living in a simulation, our descendants will almost certainlyneverrunanancestor‐simulation.   Acknowledgements I’m grateful to many people for comments, and  especially to Amara Angelica, Robert Bradbury, Milan Cirkovic, Robin Hanson, Hal Finney, Robert A. Freitas Jr., John Leslie, Mitch Porter, Keith DeRose, Mike Treder, Mark Walker, Eliezer Yudkowsky,andseveralanonymousreferees.            

www.nickbostrom.com www.simulation‐argument.com 

---

Source: [Fermat's Library](https://fermatslibrary.com/) — https://fermatslibrary.com/s/are-you-living-in-a-computer-simulation. More annotated papers: https://fermatslibrary.com/llms.txt
