From Simulism

New Page to store Text currently on Userpage

---

Contents 1 Links 2 Notes For an Article 2.1 What exactly is meant by a simulation? 2.1.1 Understanding Simulations 2.1.2 Some Features of Simulations 2.1.3 How Simulations Work 2.1.4 The Fire in The Equations 2.2 What is the Purpose of Simulation? 2.2.1 Utilitarian Simulations 2.2.2 Training Simulations 2.2.3 Entertainment Simulations 2.2.4 'Exploratory' Simulations 2.2.5 'Escapist' Simulations 2.2.6 For What Purposes might we use simulations in the future? 2.3 Discussion of Bostrom's assertion that ancestor-simulations will be a commonplace 2.4 Discussion of consciousness in relation to the simulation environment 2.5 Evolutionary versus Creationist Simulations 2.6 Discussion about the computing power needed to run a convincing simulation 2.6.1 Estimates of Computing Requrements 2.6.2 The Current Situation 2.6.3 Extrapolations 2.6.4 Summary 2.6.5 A note on Mips, Flops & Mtops 2.7 Discussion about the difficulty of debugging/validating simulations 2.7.1 Program Design & Development 2.7.2 Some Expected Consequences 2.7.3 Possible Solutions to the Problem 2.8 Discussions on the Difficulties inherent in simulating Reality 2.8.1 The Scale of the Problem 2.8.2 Simulating Time 2.8.3 Simulating Relativistic Events 2.8.4 Simulating at the Human Level 2.8.4.1 Practical problems in managing intelligent beings 2.8.4.1.1 Keeping Control 2.8.4.1.2 Unpredictability 2.8.4.1.3 Memory, Points of View and the Historical Record 2.8.4.2 Moral and Legal Issues 2.8.5 Simulating events at the Quantum level 2.8.6 Summary 2.9 Re-examination of Bostrom's argument 2.10 Bostrom's Argument & the Liar's Paradox 2.11 Conclusions

[edit] Links

I am intending to collect together 'interesting' papers here.

How to live in a simulation Explores what kind of person should we be or atempt to be if the hypothesis is true.

How to tell if you live in a simulation Attempts to explore whether or not the hypothesis is true.

When will Computer hardware match the human brain? Discussion on computing power, and when it will achieve the complexity and computing performance required to mimic human brain function.

The Matrix as MetaphysicsLong, philosophical article discussing the many issues raised by and surrounding the central tenet in the Matrix trilogy.

A Review of Chalmer's essay A Review of the above essay, criticising it and coming to different conclusions

Historical Simulations - Motivational, Ethical and Legal Issues Paper discussing the rationale for creating 'ancestor' simulations', and what would be the ethical (and legal) status of intelligences within such a simulation

Wikipedia Entry I personally think that entry this needs transporting here, tidying up by anyone who is willing to lend a hand, then transplanting back again into Wikipedia.

Complexity is subject to diminishing returns by Jason Godesky An interesting article which argues that as society becomes more technologically advanced, the progress it will make will eventually occur in amaller and smaller steps; ultimately technological advances will become so burdensome, that they actually result in a society which is less advanced, because it is top-heavy with beaurocracy and conflict.

-not sure whether these will work-

Simulation in the next Millinnium Abstract only. For those who have access to IEEE sources, the full paper is interesting in that it speculates on what simulations will be capable of, and the uses to which they will be put during the next 1000 years. The paper concludes that simulations will be a pervasive feature of life.

Debugging Simulation Models Abstract of Technical paper discussing the difficulties of ensuring that suimulations are bug-free. (Need rights access to ACM for full paper)

Letters and Articles From New Scientist related to Simulation

Virtual Outbraks: Real World Ramifications Discussion of the plague which swept through World of Warcraft, and how epidemiologists are now seriously considerning infecting other MMORPGs in order to simulate plagues, to investigate human behaviour in otbraks.

'Simulation Software' a letter expressing views that relativity and quantum mechanics are software bugs in the simulation.

'Free Will' Letter suggesting that free will and consciousness arises form the fact that human minds are capable of carrying out parallel 'what if' simulations.

'Wet Simulation' letter making the point that simulated consciousnesses may not in fact need to be conscious.

Simulation Falls Over Letter making the point that Bostrom's point (1) may in fact be correct, quioting Von Neumann's argument about self-replicating probes.

Simulation Falls Over (2) letter discussing the resources needed to run a realistic simulation, and concluding that it would take the resources of the Universe to run such a simulation.

'Platonic Simulation' a letter discussing the relationship between a simulation, and its execution. It is very reminiscent of Bart's posting on Paper Minds.

Simulating Quantum MechanicsA Simulation run on supercomputer Blue Gene/L performing very complex quantum calculations to simulate the behaviour of thousands of atoms in three dimensions. The computer Gene/L consists of 131,072 individual processors wired together using specialised high-speed networking gear. It is capable of a peak performance of 360 trillion calculations per second (teraflops).

Simulation Book

[edit] Notes For an Article

Ivo, please do not remove until I have finished

[edit] What exactly is meant by a simulation?

This section discusses the nature of simulations, how they are constructed and validated, and whether a computer is necessary. The final sections discusses whether the act of simulation is a necessary part of the process or are the rules and structures themselves sufficient to define a simulation. This final part relates to Stephen Hawking’s musing: “What is it that breathes fire into the equations and makes a universe for them to describe?”.

[edit] Understanding Simulations

There is no doubt that simulation is now a hugely important and widely-used research methodology, covering fields as diverse as cell biology, cosmology and manufacturing processes. Some idea of the range and scope of simulation can be obtained by following up some of the links to [Web-Based Simulation Forums and Societies.

It is useful to try to understand what exactly we mean by a simulation. The term itself implies that we are simulating, or mimicking a target object, probably a real or imagined world. Sometimes the term is used synonymously with 'computer model', although simulations have an important history which pre-date the computer era. Role-Playing, Wargames, and even the Big Brother House can all be considered to be types of simulation. However, it was with the advent of computers, especially when computing power increased dramatically in the 70s and 80s that large scale simulations of global events (e.g. climate change, macroeconomic forecasting etc.), and gaming scenarios (e.g. Sim City, Populus) that the term 'simulation' began to take on its current meaning as a computer 'model' of some aspect of real life.

We should however, not automatically assume that the term 'simulation' necessarily signifies an attempt to mimic reality, let alone incorporate all the features we see around us. Taken on its own terms, a simulation is simply an attempt to create some sort of viable artificial world, with its own rules, its own logic, its own events, artifacts and possibly its own characters which interact in a vaguely analogous fashion to the experience we label as 'reality'. We should also not assume that these artificial worlds would, even in the far future, would necessarily contain conscious entities. Extrapolating from some of the current offerings, sufficiently complex artificial world might admit conscious beings, but again these need not necessarily be human - such beings might be aliens, anthropomorphisations, or even superintelligent shades of the colour blue.

In addition, having opened up the possibility of 'artificial worlds', many other types of computer-generated environments now fall under this umbrella. There have been many attempts to create Artifical Life scenarios, from Conway's Game of Life to Gene Pool. Clearly these have a long way to go before they achieve any realistic mimickry of living creatures, let alone achieving any form of intelligence.

[edit] Some Features of Simulations

Computer simulations are normally used where deterministic solutions are impossible or intractable. Such uses can range from simple Monte-Carlo methods of integrating mathematical functions, to modelling bird migration behaviour. These simulations typically have stochastic features built into their models; this is normally to mimic the 'random' events which occur from time to time, but can be simply a device for obtaining a more reliable or statistically valid answer.

It is actually not strictly necessary for a computer to be involved, even in a mathematical simulation. In the 60's and 70's many simulations were paper-based, and involved lots of calculations done by hand, with tabulations. A computer merely makes the calculations easier, and facilitates the whole. In essence, what is required in a simulation, is a set of clearly-defined events (normally randomised), a set of rules governing the interations beteeen events, and a systematic method of recording.

Traditionally, those simulations which involve the occurrences of sequences of events are said to be discrete- or continuous- time based. Discrete simulation 'time-slice', that is they move time on in small amounts, and events occur at the point at which the clock 'ticks'. Continuous simulations are event-based, and events are executed in order; in principle the time between events is as long as it takes. This could be years or nanoseconds. However, because a computer is a finite state machine, all computer simulations are in essence based on discrete units of time, and therefore simulated time by default must, in some sense, be quantized. To that extent, if time is continuous, a computer simulation of an external reality, by its very nature, can only ever be an approximation.

Bibliography

Fishman, G.S., 2001, Discrete-Event Simulation: Modeling, Programming, and Analysis, Berlin: Springer-Verlag, 0-387-95160-1 Douverstein, W, 2006, Continuous Simulation, http://www.cs.uu.nl/docs/vakken/sim/continuous.pdf, accessed February 2007.

[edit] How Simulations Work

The way that a computer simulation is normally developed is by extracting essential features from the world, and finding logical and mathematical relationships between them. These relationships can normally be expressed in the form of algorithms, and work best when the variable changes can be clearly quantified. The model is then created and tested, to see to what extent the model produces results that conform to expectations, normally attempting to match historical data. Initially the modelling will be quite crude, and results only bear a passing resemblance to reality; however, as more features are included, and further logical and mathematical relationships are incorporated, a further round of testing can ensue, and the process begins again. This is called the modelling cycle, and is described in detail on Peter Ball's Website.

This is by no means an exact science. The level of detail involved in doing this modelling can vary enormously. For example, on early Global Weather forecasting systems, the blocks used were of the order of 100km squares and above. The descriptions of the various climate models from The Canadian Centre for Climate Modelling and Analysis illustrate this well; the most complex model comes in two versions, one of which has grid sizes of 3.75 degrees lat/long and the other of grid size 1.85 degrees lat/long. Current levels of forecasting can sometimes use 1km cubes and smaller. Clearly more detail can potentially mean greater accuracy, but there is a trade off here between accuracy, and speed. The more detail a simulation incorporates, the slower the simulation will run.

The trade-off can be countered by increasing the computing power. There is a natural limit to this. If we were to simulate our current 'reality', it would necessarily either run at a slower rate, or it would involve a loss of detail, or , more probably both of these. In any mathematical modelling, the Law of Diminishing Returns eventually begins to kick in. It takes ever more effort to incorporate greater level of detail, with little appreciable difference in the output. Mathematical modellers instinctively know when to stop pedalling round the modelling cycle, and simulation modellers too, tend to stop when good results only get slightly better with huge amounts of effort.

Typically simulations which model real life are 'trained' to approximate the aspect of life that they are modelling as part of the validation process. Many runs of the simulation are carried out, and variables are changed, relationships tweaked, in order to match and already-existing set of data gleaned from the real world. The simulation then will be tested on some new data, to make predictions, to see whether this matches what we would expect. There is a whole raft of formal validation techniques to ensure that the simulation actually does what is is supposed to do. See for example, the collection of papers from Robert Sargent, a regular contributor to the Winter Simulation Conference. None of these is methods is precise, and always there will be anomalies and 'freak' events which occur, exactly as in real life. A major difficulty is to distinguish between natural freak events, and an error in the algorithm or the state of the variables.

At some point the simulation is accepted as 'valid', and it is put to work, making predictions. As simulations normally contain stochastic features, different runs will result in different predictions. A simple way of countering this is to carry out multiple runs, collect data and perform statistical analyses, creating confidence intervals and undertaking statistical tests. In most cases, this is sufficient to get a fairly reliable answer; however there are cases, for example in weather simulations, where the outcomes of different runs vary so wildly that no reliable forecast can be given. These are the situations where 'the butterfly effect' is so pronounced that prediction is deemed impossible.

Bibliography

Sargent R. G., 2005, Verification and validation of Simulation Models, Proceedings of the 37th conference on Winter simulation, Orlando, Florida

[edit] The Fire in The Equations

An important question to ask is whether the simulation is actually the set of entities, algorithms and relationships or whether it is the act of simulating itself. In other words, if the world around us is a simulation, what exactly drives it, and what is the essential element that makes it work? Is it the algorithm, is it the state of the variables and objects or is it the processing of the algorithm or is it a combination of both? This is the parallel question to that asked by Hawking (1996), and further explored by Ferguson in her book " The Fire in the Equations"; that is, what essence makes the mathematical equations that describe the universe change from being symbols on a piece of paper to become particles and energy?

There is a clear difference between an algorithm written on paper, or stored electronically as a program, and what happens when the program is run. Processing makes the simulation 'come alive'. The algorithm and the processing can be regarded as two separate features of the simulation, but with very different outcomes. Stepping through the algorithm is sufficient to cause it to function, to change states of objects and mimic real-world processes. Sequentially following the logic and changing the state of objects in a step-by-step manner appears to breathe life into the system. This not only requires action, it also requires interpretation. Just having a sequence of memory-states alone is not meaningful. IT is the interpretation that we put on those memory states that makes the simulation real.

One interesting conclusion to all this, is that if it were possible to simulate conscious minds by computer program, then that program could be written on paper, and stepped through. In stepping through the program, consciousness would somehow be created - but how, and where? This starkly highlights the dilemma: how can it be possible for a set of algorithms to be anything more than just a collection of symbols or a sequence of memory-states? Do we not need a conscious entity to make sense of these, and if so, how then could it be possible that a conscious entity itself could ever arises from the same process?

Bibliography

Ferguson, K., 2004, The Fire in the Equations: Science, Religion & the Search for God, Templeton Books

Hawking, S., 1996: 'A Brief History of Time', Bantam

[edit] What is the Purpose of Simulation?

Simulations can be grouped together in terms of their purposes. This categorization 'cuts across' the three degrees of simulation described in this Wiki. What follows is not meant to be an exhaustive or mutually exclusive set of categories, as it is quite clear that some simulations would fit into one or more sections.

[edit] Utilitarian Simulations

Simulations of this type would include workflow management models, traffic models, and utility capacity and usage models, all designed to run different type of scenario, on a 'what would happen if' basis. The main purpose of these is to determine the effects of policy changes or how external events might affect a current or future situation. Many of these simulations are highly sophisticated, and the 'butterfly effect' noted in the literature on chaos theory can be observed in many of these simulations. Normally the user in these contexts is not part of the simulation, and is simply an external observer, or even a controller. Because the simulation is normally designed to model the real world closely, belief in these simulations tends to be quite high.

[edit] Training Simulations

Simulations of this type offer a user an environment where situations which are difficult, dangerous or unlikely, can be experienced prior to them being exposed to the real-life version. Flight simulators and surgical procedure simulators allow mistakes to be made without endangering life. The main purpose of these is to allow acquisition of experience. Normally the user is immersed in the simulation to the extent that for a while, they may actually believe that the simulation is real.

[edit] Entertainment Simulations

These simulations are very familiar: they range from simple games such as Pac-Man, through sporting simulations to fantasy and role-playing games. In each case they offer the user a model of an alternative reality, either which does not actually exist, or one in which the user's skills are enhanced. The main purpose of these is for fun & enjoyment. User involvement in the simulation is usually quite high, but the 'belief' factor in the viability of the simulation as an 'alternative reality' can vary considerably.

[edit] 'Exploratory' Simulations

Exploratory simulations are those which have no pre-defined outcome, or answer which is being sought. Included in this category would be historical or evolutionary simulations, or model worlds where users create and let loose creatures to evolve, change live and die. Cyberpets might be regarded as a simulation of this type. The purpose of this is simply knowledge-gathering, to see how a simulated world might change over time. User involvement in these simulations is normally low; belief in them however can be quite high. Cyberpet owners can become very attached to their charges, and become emotional when they die.

[edit] 'Escapist' Simulations

These simulations would be designed to provide a user with such immersion in the simulation that the user actually believes (even if for a short time) that the simulation is a viable alternative reality. The main purpose would be to release the user from their daily confines and provide an almost total level of immersion. Many on-line gamers report the addictive nature of games which provide such alternative realities. This in part may be due to the belief while playing them, that a player's life within the game is preferable to their life outside it. Currently there are few purely 'escapist' simulations; many might have started out as games or training simulations, and have naturally grown. Undoubtedly in the future purpose-built escapist simulations will become more prevalent. Holodecks on Star Trek, or 'Better than Life' in Red Dwarf offer a vision of these.

[edit] For What Purposes might we use simulations in the future?

Jain (1999) discusses the purposes to which we are currently putting simulation, and speculates how simulations might be used in the near future. Two main uses are identified: firstly, simulation will be used to evaluate decisions in all aspects of business operations, extending the use of simulations from manufacturing and production processes to sales and training; secondly, simulation models will likely grow in scope and size, and will be come more fully integrated, so as to more closely model reality, and will be used to validate processes and activities, prior to the construction of plant and the installation of machinery.

Jenkins (2006) specifies four distinct reasons why we might actually wish to run 'historical simulations'. The first is for nostalgic reasons, following the logic that as currently society is interested in the past, often the recent past there is little to assume that any futuire society would be any different in this regard. The second reason offered by Jenkins is that of testing Artifical Intelligence. Simulations would offer the opportunity of a relatively safe environment where machine intelligences could be examined, and whether their programming conformed to legal, ethical and moral expectations. The third category offered by Jenkins is that of Social and Economic Experiments, citing work by Castronova (2001) which explores how future generations of social scientists may investigate policy changes by running experiments on computer-simulated societies. Finally Jenkins forsees a possible use for simulations in the case of an Apocalyptic scenario, where populations may wish to migrate into simulated worlds in the event of a global catastrophe.

It is clear that even though Jenkins' paper is speculative, the notion of Social and Economic experiments are not that far away from Jain's notion of process validation. Such uses (for example Global Warming simulations) are in current use, and it seems highly likely that these will continue to develop and become ever more sophisticated. The notion of testing AI is an important one, and should not be underestimated; the legal and ethical problems in creating AI should not be underestimated, and there may well be legal requirements in the future to 'prove' safety (as in the case of GM Products and drug-testing) before release of AI intelligences were allowed. Jenkins rightly notes that future societies may well be different from ours, and may not have the same 'nostalgia' for past times. What we need to remember here is the time scale. If we are imagine societies in the far future, thay may be so advanced that their interest in running a simulation of the 20th Century may well be equivalent to our interest in running a simulation of the Australopithecus era; it might be interesting to visit for a while, but you certainly wouldn't want to live there. Jenkins final category is totally speculative. The migration of minds into machines might be the stuff of science fiction, but there is no current mechanism that could be suggested whereby this feat could be achieved. The notion presupposes that the brain's software could in principle be detached from its hardware. Many authors question whether or not such a feat is possible, even in theory.

Bibliography

Jain, S., 1999, Simulation in the next Millennium, Proceedings of the 1999 Winter Simulation Conference

Jenkins, Peter S., Historical Simulations - Motivational, Ethical and Legal Issues . Journal of Futures Studies, Vol. 11, No. 1, pp. 23-42, August 2006

[edit] Discussion of Bostrom's assertion that ancestor-simulations will be a commonplace

In Bostrom's argument, the simulations referred to are 'ancestor-simulations', and one of the reasons that Bostrom gives that these will be prevalent in the future is an extrapolation of current trends. These are based on our interest in the past, and our interest in examining the way that previous civilisations have lived. However it is by no means certain that any future society would actually wish systematically to run simulations of its ancestors, and that these would form the core of simulation use. Even if we were to do this, we would need a clear rationale for incorporating intelligent conscious beings as part of the programminmg.

If we do look at the simulations in current use, there are many simulations which simply model inanimate systems, such as the weather, traffic, workflow etc. These are sufficiently sophisticated without the incorporation of intelligent features as to provide us with fairly accurate and quite detailed forecasts. Those simulations which do involve human, animal and other entities are not predominantly historical. Many of these are fantasy-based, futuristic or whimsical and in the main are produced for entertainment rather than for information-gathering. Admittedly, there are various historical simulations on the market. (see for example The Koei Historical Simulation Games. However, it is clear that such offerings are a minority interest, and furthermore although they claim to be set in the historical past, their contextualisation is by no means accurate; neither is their portrayal of historical figures.

Even if a huge number of simulations were to be run by a huge number of civilisations, the fact that there are so many different types of simulation, and the fact that historical simulations are currently not a minstream offering, may well mean that ancestor-simulations would be a very small proportion of the total number of simulations actually run.

[edit] Discussion of consciousness in relation to the simulation environment

Ivo Jansch in the article About simulations argues that there are three distict degrees of simulation. A degree 3 simulation is characterised by an external entity with a viable existence outside the simulation' immersing their physical presence inside the simulation. This might be typified by role-playing, or the Holodeck-type scenario portrayed in the Startrek movies. A Degree 2 simulation is where the entity has some sort of viable physical presence outside the simulation, but the immersion in the simulation is merely 'virtual'. This is typified by the Matrix movie scenario, but equally almost all other on-line role-playing games. A degree 1 simulation occurs where the entity has no physical or other presence whatsoever outside the simulation, and in that regard has total immersion in the simulation.

The Simulated RealityWikipedia Entry defines four separate types:

(a) Brain-Computer Interface, where each observer directly cinnects his/her brain to the computer, possibly (for the time of the simulation) obliterating any notion of an external reality.

(b) Virtual people, where every inhabitant is a native of the simulation, and has no external presence outside it. This is divided into two sub-types - Virtual people - virtual world, in which a simulated world is created, in which artificial consciousnesses exist, and Solipsistic Simulation, in which consciousnesses are simulated, and the 'reality' is purely simulated within the minds of the simulees.

(c) Emigration in which a participant uses 'mind transfer' to relocate themselves within the simulated entity.

(d) Intermingled, which supports both players from an external reality or natives of the simulation who are simulated consciousnesses of one type or another.

The Wikipedia article would appear to be deficient, on the grounds that it is purely speculative, and it is by no means certain that any of the four types defined, could, in fact be created in the way that they are defined. The notion of emigration, for example, is not based on any currently-understood scientific principle. In addition, category (d) adds nothing to the debate, as it merely merges elements from the other categories. Finally, the notion of 'solipsistic simulation' could equally apply to a 'brain-computer interface' or a 'virtual people' type simulation. When we take these factors into account, we are left with two distinct catgories, and these approximate roughly to Janch's degrees 1 and 2, with Jansch further distinguishing between 'brain-computer interface' and 'physical presence', to achieve his degree 3 simulation. Although Jansch's definitions too are speculative, the examples provided demonstrate an identifiable currently-existing basis for the categorisations.

However, whether there is a real and distinguishable difference between Jansch's degree 3 and degree 2, and whether this is useful, is debatable. Under the definition above, the Big Brother house is clearly a degree 3 simulation. If we were to have a 'Big Brother' computer game, with all of the features of the original, then this would be classified as degree 2. The distinction occurs only as a result of the user's physical presence. However, imagine a future world where some of the Big Brother housemates were real, some were holograms; in a likewise manner, imagine that the environment might be a mixture of the two. The boundaries begin to blur and it becomes unclear as to whether this is degree 3 or 2; this is the 'intermingled' category from the Wikipedia definition.

I believe that it is more useful to distinguish between those cases where a conscious entity within the simulation has no viable physical presence whatsoever outside the simulation (i.e if the simulation stopped working they would blink out of existence), and those in which a conscious entity has some external viable physical presence. This incorporates the 'brain-in-a-vat' scenario, because even if the simulation were to cease, the brain would not necessarily die. In this regard, 'brain-in-a-vat' is in fact no different to 'plug-and-play'. This focuses on the real issues, those of attempting to determine how an entity within a simulation can distinguish whether or not it is real, and the issue of what happens to the simulee when the simulation is ended, or they die within the simulation.

An entity which has some sort of viable external reality has a reference framework outside the simulation, whether thay are conscious of it or not. This might involve experiencing physical sensations, having mental sensations unconnected with the simulation, or just experiencing the passage of time in a different, possibly subjective manner. An entity which exists purely within a simulation potentially has none of these things. All of their sensations arise as a result of sensory inputs controlled by the simulation, their mental processes are purely confined within a simulated brain, and their experience of the passage of time is governed entirely by the simulation environment.

We might call the first type an 'extrinsic' consciousness simulation, in recognition that there is an external mind involved, and the second type an 'intrinsic' consciousness simulation, as in this case the mind emerges from programming. We might also note that in the case of extrinsic simulation, there may be differing degrees to which the participant is immersed, from simply using a viewscreen, to having their brain plugged directly into the computer, with all memory of an external reality erased. Similarly, with the intrinsic simulation, there may be different ways in which this can be achieved, such as creating an entire world, which contains conscious intelligences alongside inanimate objects and (possibly) other non-player characters which, while appearing to have some sort of intelligence do not actually achieve consciousness. It is possible to blur the boundaries between this situation and a 'solipsistic' simulation in which each mind is a 'simulation in itself', and the simulated reality is 'fed' to the minds, either simultaneously, or in some hugely complex asynchronous but interrelated manner. It would even be possible, for an 'intrinsic simulee' to be a simulated 'brain-in-a-vat', experiencing a second-level simulated world, which might then be classed as 'extrinsic' with relation to the simulated brain.

In the case of an 'intrinsic' simulation, one interesting feature to note is that the operating system and the run-time environment would both have access to the same type of sophisticated programming used to create and support the 'minds' in the simulation. This is not necessarily therefore a reactive inanimate environment, but it could be proactive and intelligent. It might anticipate actions, deal with flaws in the programming, circumvent problems, create 'simulations on the fly', or invent new classes and objects when required. In other words, it too might be a conscious entity.

[edit] Evolutionary versus Creationist Simulations

A simulation which approximates the reality that we see around us may have arrived in that state via one of two mechanisms. An Evolutionary Simulation will have started out at some earlier state (possibly the Big Bang), and changed gradually to the current situation. A Creationist Simulation will have been created with elements such as historical records, fossils and rusting cars, then started at some arbitrary point in time.

There would seem to be huge computational challenges in creating a realistic Evolutionary Simulation. It has taken the Universe 15 Billion years to evolve to this point, and has taken the entire resources of the Universe to do it. Simulating this precisely on a computer which is strictly contained within this Universe would therefore take much longer than 15 Billion years, unless (a) detail is lost, or (b) the external reality which is doing the simulating has far more resources than our apparent reality. In the case of (b), such a Universe would necessarily be more complex than ours, in which case, our 'reality' is merely a simplified version to the true reality, and therefore not a valid simulation of it.

Given these facts, it is difficult to see how an Evolutionary simulation would be compatible with extrinsic consciousnesses, as the evolutionary timescale would seem to prohibit their involvement. It would in theory be possible for the simulation to have 'evolved' rapidly, with the OIT flowing large orders of magnitude greater than OET; then slowing down OIT at the point that the extrinsic consciousnesses entered, so that from that point OET and OIT match precisely. However, the feat of simulating 15 billion years of evolution within a reasonable period of OET, say fifteen years, would require computational resources one billion times greater than those of our entire universe. For these reasons, we can probably rule out the entire class of Evolutionary extrinsic simulations as impracticable.

A Creationist Simulation, on the other hand, would be entirely suitable for extrinsic consciousnesses; the simulated world could be created, and the minds could enter or migrate, just as we play games or log on today. Creating intrinsic consciousnesses however, is more of an issue, as a whole history of memories would have to be simulated, along with false memories, memories of memories, partially-remembered and forgotten events. This appears to be hugely problematic; it would clearly be much simpler to 'evolve' the consciouness at an accelerated rate of OIT.

From this discussion we can conclude that if we are living in a simulation, then it is likely to be either an extrinsic creationist simulation or an intrinsic evolutionary simulation.

[edit] Discussion about the computing power needed to run a convincing simulation

UNDER CONSTRUCTION - CALCULATIONS NEED TO BE CHECKED & VERIFIED

In discussions with a colleague about the kind of computer needed to run a similation convincing enough so that concious entities would believe it to be real, his first reaction was 'You'd need a bloody good graphics card!"

[edit] Estimates of Computing Requrements

In fact an extrinsic simulation need not be very powerul, as effectively the consciousness exists outside it. Current computer graphics are almost powerful enough to be able to fool us at times into thinking the images we are watching are real; it is not unreasonable to assume that a few years of development away, virtual reality booths will be totally convincing, and human beings will spend much of their leisure time immersed in highly-believable fantasy scenarios. We could probably do this with 1 Million MIPS, and 1 Million Gb.

Conversely, intrinsic simulations will require enormous power. They not only have to be convincing to the users, they need to create and manage them as well. This is not just a doubling, or even a tripling of computing power required; each simulated mind will need at least the the complexity of an extrinsic simulation (if only to be able to imagine, to fantasise and to dream), and then some. Moravec (1997) estimates that an intrinsic simulation of a single mind would require around 100 Million MIPS of computer processing power occupying about 500 Billion Gb. If we were to include the environment in this, a reasonable estimate would be 1 Billion MIPS and around 1 Trillion Gb. Introducing other minds into the simulation not only involves issues of parallel processing, but also the interactions between individuals also cause scalability issues. If we were to imagine simulating 5 Billion people on this planet, this would not just be 5 Billion times the problem, but inevitably involve other factors. As is argued in another section, any simulation which is capable of supporting conscious intelligent entities as part of its run-time environment will almost certainly use conscious intelligences in its operation. This means that intelligent monitoring of the environment, its reuirements, use of power, memory etc. can be carefully and consciously controlled. This affects arguments about how much memory usage would be needed to model the world, and would almost certainly increase the computing power and memory requirements. A not unreasonable estimate for simulating a planet full of people might be 10 Quadrillion MIPS, and 10 Quintillion Gb.

[edit] The Current Situation

Currently the worlds most advanced Computer is Blue Gene/L. This operates at a speed of around 300 teraflops, using a connected array of 65536 dual processors, each with 0.5Gb of local memory. on the same scales as above, this converts to 3000 MIPS with 30,000 Gb. If we take this as a benchmark, as a conservative estimate, we would need a processing capability increase of a factor of about 300 to create a realistic extrinsic real-time simulation. It is probably less than that for a specific-purpose simulation, with a vary narrow remit, such as a small, but detailed cityscape, or a single user domain. By comparison, we would need an increare in processing power by a factor of around 3 trillion to create a convincing and realistic instrinsic simulation.

[edit] Extrapolations

Moore's Law in its simplest form offers the promise of computer power doubling every two years. Provided that this law holds, this means that within 16 years (eight doublings), the first convincing extrinsic simulation could potentially be created, provided that we have the programming capability to do it. The factor of 3 trillion to create an intrinsic simulation sounds daunting by comparison; however, this factor can be exceeded in just over 40 doublings, or 80 years. Even if there is some tailing off in Moore's Law, by the end of this century we should have the computer capability to create an intrinsic simulation; whether we will have the undersanding and the know-how is another matter. It is interesting to compare these calculations with those on the Moore's Law page concerning the time from first simulation of ahuman being to the first simulation of a world population (66 years).

In this discussion, I have not invoked the idea of Quantum Computing. As suggested by Penrose and others, it may be that quantum processes lie at the heart of consiousness, and it would be impossible to create an intrinsic simulation without incorporating quantum processes. It is certainly true that if quantum computers are used, then the processing power requirements plummet, and much more can be achieved with far less hardware than that outlined above. In those terms, we would expect to achieve first viable simulation well before the dawn of 2100. However, this would be contingent on having achieved fully-functional quantum computation within the next 20 years or so.

[edit] Summary

As a conservative estimate, we should expect to have convincing virtual-reality booths which provide the user with a convincing extrinsic simulationm in operation within about 20 years, and probably in wide used by 2030. By the end of the century, we should be withessing our first full-scale simulations capable of supporting intrinsic intelligences, which are fully-working, convincing and believable to the intelligences within them.

[edit] A note on Mips, Flops & Mtops

MIPS is defined as 'Millions of instructions per second", or by some as "meaningless indication of processor speed"; FLOPS are 'floating point instructions per second". MTOPS is defined as 'Millions of theoretical operations per second". The distinction between them is fine-grained. An 'instruction' is an instruction at the processor level, and could be an instruction to add, store data, compare etc. A Floating Point operation is done by a calculation with decimal numbers, sometimes requiring a few, sometimes requiring hundres of instructions. The measuring system is further confused because of the different mathematical scales used when describing larger numbers. we talk of millions of MIPS, but we have Giga-Flops, Tera-Flops etc. An already confused situation is then take advantage of by games console manufacturers who make outrageous claims for their devices. The XBox 360, for example claims to have operating speeds of around 1 TeraFLOPS; however this is because the power is focussed on a single activity, that of pixel-generation. If the processer were used in general purpose calculation it would have nothing like this speed.

In the discussion above, I have taken the decision to convert all measurements to MIPS, on the assumption that 1 flop takes around 100 processor instructions on average. This is a deliberate overestimate, and is doen to ensure that we obtain realistic estimates for the time needed to obtain 'convincing' simulations.

[edit] Discussion about the difficulty of debugging/validating simulations

[UNDER CONSTRUCTION]

In any given piece of software the number of bugs is countably infinite. This is because no matter how many bugs you find, there is always at least one more that you failed to spot before you released the software to the client.

All programmers will vouch for this; it is a major issue, because with any piece of software which has millions of lines of code, it is not humanly possible to check for all classes of errors, even if there are automatic code-checking routines which perform the task. (Who is to say that these do not have logical flaws which systematically fail to spot particular types of error?) Syntactical errors are in theory easy to spot; however, there are cases where syntactical errors in programs have caused spaceships to malfunction simply due to the misplacement of a single character. As Holzmann (2002) puts it: "After some fifty years of practice, few people today would say that we understand the problem of software quality well enough that we could outline a development process that could lead to zero-defect code."

We should not be surprised by this: software writing is a human activity, and it suffers from the same level of error as any equivalent activity. Amster & McLain (2002) report that on average there is one mistake per 1000 sentences in an average weekday issue of the New Yor Times. Holzman (2002) asserts that the recognised industry estimate is that a typical piece of software contains one to ten residual software defects per 1000 lines of code.

[edit] Program Design & Development

Programming a large system is hugely complex, expensive and time-consuming. Cronqvist (2004) documents a system in which some 2.1 million lines of code were written by around 300 people, some of whom were novice programmers in the Erlang language used at the start of the project.

Attempts at rigorous software development processes have been made over the years, and typically these processes might include: A requirements capture (incorporating systems and user needs analyses), High-Level, then Low-level Design in which the program blocks are structured, gradually being broken down into finer components until the code is written, Testing, in various forms, and finally Customer Acceptance and Maintenance. As Holzmann (2002) points out, bugs can be introduced in each of these stages, and that even in principle such techniques cannot be expected, despite the most diligent application, reduce the residual error rate to zero.

The reason for this, can be seen in the following typical software development scenario, which attempts to encapsulate the process in seven stages from idea to release:

In Stage 0 the programmer, or team of programmers work on conceptual designs, and create specifications, making all sorts of charts and diagrams, systematising variable names, and attempting to manage the software development. Quite a lot of errors are removed at this stage, but equally, errors can easily be manufactured, and often are.

In Stage 1 the software actually gets written. This is often done by writing (or appropriating) individual sections of code to do various tasks. There will be some testing of code at this stage in the vain hope that they can shortcut problems later, but the ploy rarely works.

At Stage 2, the code is glued together. The first rule of software writing is that nothing ever runs the first time; variables will not talk to other variables, the logic will be wrong; it just does not function. This causes head scratching, then light begins to dawn. Changes are made, and eventually the program begins to run. Unfortunately it gives completely the wrong answers.

In Stage 3, attempts are made to tweak variables, to revisit the logic, and gradually answers are obtained, and a pattern of behaviour is obtained which is a good approximation to what is required. At this stage, the program might be said to be working. Unfortunately, it is completely unreliable. When different types of data are thrown at it, it can sometimes go completely out of control, crashing, giving bizarre answers, often seemingly without reason.

In Stage 4, we need to do some rigorous testing, almost like a scientific experiment, varying inputs, to see what happens with extreme data, and pursuing why crashes occur. Eventually the programme will work 99% or 99.9% of the time, possibly even 99.99% of the time. The question is, when to end this process? This is a difficulty, because a piece of software may be released containing an error which is never spotted in the whole of its lifetime. A developer's time is expensive, and at some point a halt needs to be called. No software development house can possibly rule out all bugs before it is released.

In Stage 5 the software is released for testing in the field, by specialist users are effectively the guinea pigs. Many errors and bugs are spotted at this stage; crashes and anomalies are found, but again, the question of how long to hold off general release is problematic. Do we wait until the error reports reduce to 5 a day, 1 a day or 10 days between reports? They will never reduce entirely.

Stage 6, the software is released. As any software house will attest, the first few weeks are a nightmare; bugs are found that you would not believe; how on earth they were not spotted earlier is amazing. By the time we get to v1.4.17 the software is beginning to settle down, and we have faith that it is not going to crash planes, ruin economies or blow up power stations.

With that background, how can we write a believable simulation?

[edit] Some Expected Consequences

If we are going to code for even a fairly modest extrinsic 'believable' simulation, it is going to require the equivalent of billions of lines of code at the very least. No one person could possibly have a complete overview of this, and the simulation could only be produced over a lengthy period of time by a huge number of individuals.

The consequences of this would appear to be that we would expect a simulation to contain some flaws, inconsistencies, bugs, glitches, viruses or other nasties which might case anything from irritating, almost unnoticeable little annoyances to full blown system failures. Cronqvist describes four separate classes of coding error:

• API Mismatches (calling the wrong function, or using the wrong arguments);
• Race Conditions (two parallel processes wroking simultaneously with incompatible aims)
• Wrong Context (performing an action on an object which does not exist in the form required in the location specified)
• Typos (simple spelling and other syntactical errors)

A simulee within a simulated environment might experience this in different ways: there would be elements which simply refused to function properly for no apparent reason, e.g. all books would suddenly become meaningless mumbo jumbo, then suddently revert back to making perfect sense, architectural elements might mysteriously appear suspended in mid-air; birds flying through the air would freeze-frame, then restart. All of these could be manifestations of program glitches.

System crashes would, in an extrinsic simulation, mean that the simulee experiences absolutely nothing; in an intrinsic simulation, the situation is slightly better, but undoubtedly if the system stops and restarts, there would be discontinuties which would easily be detected by the inhabitants.

The fact that we do not appear to experience anything like this, on this sort of scale, has led some writers to argue that we cannot possibly be living in a simulation. This argument is based on the premise that the methodology described above is the one which would be adopted in creating a sophisticated simulation. However, what is clear, is that the software designers of the future responsible for creating a believable simulation would be wise enough to realise that leaving bugs in the system would cause simulees to lose belief in their simulated environment. They would therefore program into the simulation a variety of procedures to circumvent these inevitable bugs.

Clearly there would be sophisticated code-checking, logic verification and program architecture validation methodologies, to ensure that simple, avoidable errors do not occur. Much of the code will have been generated automatically, and individual elements will have been tested to destruction. The computers on which the software runs might be very different from those at present; they may be organic, use parallel processing with DNA, quantum computing techniques, or other as yet undreamed of procedures. However, despite this, it is inconceivable that the errors and viruses being created now, and lurking unnoticed in systems and development software will lie dormant forever. As we procress computationally, these errors will replicate themselves, or generate further errors. At some point, at least some of these will manifest themselves in an unexpected way despite the checking and the validation.

It would therefore appear that unadvoidable errors might serve to prohibit the creation of a believable simulation. However, before we submit to this too soon, we should look at ways which future programmers might use to circumvent the problem.

[edit] Possible Solutions to the Problem

Simulation programmers of the future would have to find ways around this. One obvious way is to use an intelligent operating system. Such a system would anticipate bugs, glitches and even system crashes, and program 'on the fly' to circumvent these in simulated time, so that the simulees do not notice that for a nanosecond all purple objects suddenly vanished & then reappeared. This approach would clearly work if the simulation were of the intrinsic variety. As has been noted earlier, if the simulation contains conscious minds as part of its programming, it has the potential to use a conscious and intelligent operating system. In such a simulation, there would be opportunities to slow down, or even stop simulated time completely, without the simulees noticing. In this temporal void all kinds of repairs could be carried out.

In an extrinsic simulation, we might have neither the computing power nor the sophistication to employ such methods, and the requirements of ensuring that time flows continuously for the simulees could cause great strains on the operating system if it attempted to carry out repairs on the fly. An alternative approach might be to use a network of programming structures, so that if a logical route through the code is blocked, or is bugged, an alternative logical route exists as a pathway elsewhere in the code. This is basically what happens in a neural network; however here the suggestion is not that the network is 'grown' and 'trained', but that it is 'constructed' and 'educated'.

All this would probably mean that bugs and anomalies still exist, and if simulees were to look hard enough for them, they would find them; however thay might not recognise them as such. Objects might appear slighly different, as if the simulee were seeing them for the first time; odd coincidences might suddenly happen for no apparent reason; strange feelings of situations repeating themselves might occur. These are the sorts of manifestations of bugs that the simulee might experience - not too 'off the wall' too be clearly anything wrong, but just enough to be disconcerting, and not strong enough to shake the belief that the simulee is living in anything but a reality.

Bibliography

Bayazit, A. A., & Malik, S., 2005, Complementary Use of Runtime Validation and Model Checking, [Source Needed]

Cronqvist,M., 2004, Troubleshooting a Large Erlang System, [Source Needed]

Devadas, S., Ghosh, A. & Keutzer, K., 1996, An Observability-Based Code Metric for Functional Simulation, [Source Needed]

Holtzman G.J., 2002, The Logic of Bugs, [Source Needed]

Ohba, N. & Takano, K., 2006, Hardware Debugging Method Based on Signal Transitions and Transactions, [Source Needed]

Penix,P., Martin, D., Frey,P., Radhakrishnan, R., Alexander,P. & Wilsey, P.A., 1998, Wxperiences in Verifying Parallel Simulation Algorithms [Source Needed]

Poutakidis, D., Padgham, L. & Winikoff, M., 2003 , An Exploration of Bugs and Debugging in Multi-Agent Systems, [Source Needed]

Sargent, R.G., 1988 , Simulation Moduel Verification and Validation, [Source Needed]

[edit] Discussions on the Difficulties inherent in simulating Reality

<UNDER CONSTRUCTION>

Starting from the premise that we are living in a simulation, the following essay will attempt to explore the practical problems in setting up a believable simulation.

As far as we can make out, the world we inhabit contains around six billion individuals, almost all of whom appear to be conscious, intelligent agents with free will. In addition there is an unimaginable number of animals of bewildering variety, and of varying levels of intelligence. Each of these living creatures is part of a complex planet-wide ecological system which has an intricate web of inter-dependences. The universe itself appears to be constructed on a huge scale, in terms of billions of light-years, and at this distance and at the huge speeds required to cross these distances, relativistic physics appears to operate effectively. On the medium scale, such as that experienced by humans, classical physics and Newtonain Mechanics offers a good approximation to everyday experience. At very small distances, and on very small timescales, quantum mechanics offers a coherent picture of the world.

The Universe appears to have been in existence for some 15 billion years, and there is evidence of processes which take place over huge timescales, as well as some processes at the quantum level which take place during fractions of a second.

What is also clear, is that the world we inhabit demonstrates regular and repeatable patterns of behaviour, which allow for prediction and verification. These patterns of behaviour can often be encapsulated and codified in mathematical form, and these formulations often lead to highly accurate predictions.

[edit] The Scale of the Problem

Consider for one moment, the task of attempting to simulate what you see around you. The the size of the problem is daunting. Although we may reach the technological capacity to do this soon, our current level of understanding of physics, biology, psychology, sociology, meteorology and ecology (to name but six sciences) would rule out any realistic attempt to do this within the next hundred, and probably within the next few thousands of years. In addition, the scale, and the level of detail on which this would need to be programmed is almost unimaginable, and it is inconceivable that this could be done without introducing a multitudinous array of bugs, glitches and programming inconsistencies. Furthermore, either the level of detail is pervasive, and the entire universe has been fully simulated from the quantum level right up to the level of galactic clusters, or much of it does not exist, and it is interpolated by a machine intelligence when a human being probes, for example using a telescope or microscope.

The problem with this 'ad hoc' approach, is that the ability to do this leads to circular reasoning. It has been assumed by many writers that the reason not to produce detail at the micro-level would be to save on computing power; however, as we have seen in the exploration od Moore's Law, computing power is not the issue, it is our ability to program, and our understanding of the world that are the barriers. In order to produce micro-level detail 'on demand' when a simulee suddenly decides to look through an electron microscope, or observes subatomic particles in a cloud chamber, means that all of the history of those particles would need to have been known (or at least able to be calculated) prior to the observation, otherwise the sudden 'invention' could lead to all sorts of possible disjunctions between 'near-simultaneous' observations. In essense, everything would need to be predictable and computable. (The alternative to this is that all decisions observers might make about possible future observations and experimentation are either pre-determined, or at best totally predictable.)

This predictability is paradoxical. If everything in the simulation is is predictable, then why is the simulation in existence? If the simulators knew enough about a world to simulate it down to the minutest detail, inclusing all the subtle interrelationships which would constitute ecology and meteorology, then they would effectively be able to solve all the world's major problems deterministically. If they were not able to solve these problems, then it is extremely difficult to see why the simulation techniques they would be using to explore possible solutions would require this level of detail and the inclusion of conscious minds as part of the programming, and, more particularly, how they could achieve it. In short, the notion of 'simulation as exploration' presupposes that we cannot predict in advance where simulations would lead, otherwise, why bother?

At the moment in our 'reality', we are clearly not advanced enough to create a simulation which even begins to approach the level of detail we see in our 'reality', and so, if we are a simulation, we must therefore conclude that we are either in a 'what would happen if' simulation, the purpose of which would be to evolve to a point at sometime in the future where a possible future policy-shift could be trialled , or we are in one of Bostrom's 'Ancestor-Simulations'. Both of these cases are deeply disturbing, as the first assumes that simulators would have enough knowledge to 'evolve' a system to a particular point in its history; the ramifications of this for our abilities to affect our future by actions that we might consider 'free will', would be profound. Again, the very nature of an 'Ancestor-Simulation', means that our freedom to act would be very limited; if we really are a society which is the 39th century equivalent of re-runs of 'I Love Lucy', then our ability for self-determination and free will are negligible.

Summary

The sheer size and complexity of running a simulation of the magnitude which comes close to what we are currently experiencing poses immense threats to our notion of free will. We are left therefore with the unedifying proposition that if we are in a simulation, it is like a cosmic version of The Truman Show, and this entire universe has been created for our, or more likely, someone else's entertainment.

[edit] Simulating Time

In currently-existng simulations, time is dealt with either as discrete time-slices, or as a continuous sequence of events. In principle, in the first case, time flows in the simulation at a constant rate, but in the second, time will flow at a different rate depending upon the number of, and intervals between, events. However, because of the computational complexity involved in processing, the simulated time may, to an external observer, appear to flow at different rates in either case. If we are to posit an observer internal to this simulation, this apparently would not matter: objectively time within the simulation would clearly flow at the rate at which the clock ticks.

All of this poses some conceptual and linguistic difficulties when discussing time. If we assume that there is an external reality, which has an objective time frame, we can call this 'Objective External Time' (OET). The rate at which the clock ticks in the simulation we can call 'Objective Internal Time' (OIT). Added to this, for the simulees experiencing the passage of time, they have their own 'Subjective Internal Time' (SIT), and in the case where the consciounesses are extrinsic to the simulation, there is also the possibility of 'Subjective External Time' (SET)

The first question to be asked is whether these are all distinct. From current simulations, and the discussion in paragraph 1 above, it is clear that OET & OIT do not necessarily flow at the same rate; it is possible, for example in global climate simulations for millennia to pass in a few minutes. Subjective time of whatever nature, will depend on the rate of thought processes; these might be made slower or faster depending upon what else is going on in the simulation at the same time. Again, it seems obvious that SIT and OIT could be different, especially in the case where the simulation is solipsistic. It is unclear as to whether SIT and SET could be different for an extrinsic consciousness. It would seem likely that they would be the same, as there is an already-existing external time-frame in which the consciousness had an external reality. Such a time-frame would override any internal impressions of time to establish its own 'subjective' clock. In this case SIT and SET would be the same. For an intrinsic consciousness, SET has no meaning. Finally, from experience as human beings, OET and SET can appear to flow at very different rates: "Doesn't time drag when you are bored?"

In the debate about the passage of time, we should not make the mistake of thinking that our conventional Newtonian view of time is the one which applies. This assumes that there is an OET which is universal and everyone in the universe experiences the passage of time at the same rate, and that Greenwich Mean Time, could in principle be used throughout the galaxy. This is a fiction. Relativity destroys this, and effectively any simulation which includes relativity as part of its programming would not have an OIT which is consistent with SIT for all observers. This would mean that each consciousness would have thir own Individual Objective Internal Time (IOIT). Again, this might be perceived differently to SIT, as there could potentially be a distinction between the objective and the subjective passages of time for an observer within the simulation.

[edit] Simulating Relativistic Events

Argue that it is easy to underestimate the size of the programming problem and the amount of sheer computing power needed to attempt to simulate events at relativistic scale. See for example the 3-D animations produced by Pretorius (2001) Simulation of Gravitational Collapse in General Relativity Pretorius suggests that in order to simulate the collision of two black holes in 3-D Graphics, it would require 1 CPU week on a 1 TFlop system.

The Twins Paradox Argue using the twins paradox that simultaneity is prohibited by relativity, and that each observer must therefore have his/her unique point of view. This means that any simulation is effectively solipsistic. Argue that this has profound implications for extrinsic simulations, as it would place observers in two subjectively different internal time-frames in the same simulation at the same 'external' time. This could potentially allow one simulee knowledge of the others' future actions, effectively denying free will.

This section explores the constraints imposed on simulated reality by relativity, in particular, time dilation effects. This has some serious implications for a would-be simulation programmer. For example, if we are to take the normal reading of the twins papradox, where one twin stays at home, and the other twin nips off to Proxima Centauri, and returns, depending upon the speed that is travelled, one twin can end up with a journey time of a year, and the other, who has stayed at home has passed ten years. This means that we now have two minds in the same simulation at the same point who have experienced different time rates. This is fine, if the consciousness is intrinsic to the simulation, but not if the consciousness is extrinsic (e.g. brain in a vat). An extrinsic consciousness has an external reference frame, which could (and probably would) include an objective way of measuring time. This means that when the twins meet together after the journey, one is effectively 10 years further on in the simulation than the other.

A programmer would be faced with two choices in this situation: either to make time slow down for the travelling twin, but then this defeats the whole notion of time-dilation; the twin would notice, and declare that relativity is a sham. On the other hand, and this is the more alarming prospect, we now have a situation where two brains in vats are experiencing different times in the same simulation. One of these is in the future of the other one, and therefore poterntially knows his actions before the other has acted. This effectively removes any notion of free will. The only way to counter this situation is to suppose that the external consciousness is locked into an individual solipsistic simulation.

There are other issues, too. The whole idea of a simulated reality is based on a misunderstanding of our current 'reality'. It seems to assume that our experience of time is of a common, universal framework, where time passes at the same rate for all, and it is possible to think about what someone else is doing now in another part of the world. Relativity destroys this notion.

The only way a relativistic simulation could function is if each consciouness was intrinsic to the simulation, i.e. there were no brains in vats. That way, the programming could, in effect have the minds experiencing time at different rates, and no one would be any the wiser. Unfortunately, this now presents a slightly different problem. Looking at this simulation from outside, we now do not have a universal simulation, in which minds happen to inhabit, we have lots of separate solipsistic simulations, one for each mind, and 'reality' is constructed for each one of them separately, so that they can react with one another in a consistent manner. This seems to me a pretty good description of what most people regard as 'reality'. The world that I see is a construction of my brain based on sensory perceptions. It may or may not (probably does not) have very many features in common with anyone else.

Once we accept that is the way that the simulation must function, if it contains conscious minds, there are some very profound philosophical, ethical and moral problems to be countered, if we ever start to simulate in this manner.

Bibliography

Pretorius, F., 2001, Analysis, Computation and Collaboration, Lecture given at Simon Fraser University, July 2001

[edit] Simulating at the Human Level

An obvious, but completely overlooked question needs to be asked about the inclusion of intelligent, conscious minds in a simulation. That is the simple question of "Why bother?" In all simulations that we have created so far, many of them very successful, and able to provide predictions which have proved quite accurate, none of them has contained anthing approximating to intelligent, let alone conscious beings.

The inclusion of such entities adds so much complexity, and so many layers of difficulties to the programming that the purpose would not just have to be worthwhile, it would be at the core of the simulation itself. If the simulation were simply to explore policy shifts, or to allow an extrinsic observer to enter an historical world, then it is clear that these could be accomplished without the need for the beings that inhabit the world to be conscious, even though they may well need to have some form of intelligence to allow them to function.

[edit] Practical problems in managing intelligent beings

The main issue with simulating conscious entities is the problem of free will. In our reality, we either do have free will, or we have done a pretty good job of convincing ourselves that we have free will. Conscious beings use free will to explore the boundaries of their world, probing new territory, attempting to take it to pieces to find out how it is made. They can also be perverse, sometimes taking the opposite approach to that which might otherwise have been predicted. In other words, conscious beings represent the ultimate test of a simulation. If there are bugs, anomalies inconsistencies in the programming, conscious entities within the system will expore and exploit these, pushing the programming to its limits.

[edit] Keeping Control

Depending upon the purpose of the simulation, control over the conscious beings might vary, at one extreme, from allowing the simulees total free will to evolve and develop in whatever way they see fit, to a total control scenario, where free will is an illusion, and everything is pre-determined. It is difficult to see why it would be necessary to resort to the subterfuge in the latter case. Clearly if this were the situation, it would be a lot simpler to programme the beings to accept the fact that they did not have free will, and that their lives were pre-ordained, much as the characters do in Greek legend.

An interesting note here is that at least one author believes that this state is in fact a precursor to conciousness. Julian Jaynes in "The Origin of Consciousness in the Breakdown of the Bicameral Mind", discussed the fact that what we currently regard as schizophrenia, may in fact have been a universal state of mind in hellenic times; people believed that they heard voices from gods, controlling them, and advising them ,and telling them how to run their lives. Evidence for this viewpoint arises from the way that Greek Literature makes little use of personal pronouns, and the notion of 'self' is underdeveloped.

Whether this is true or not, what is apparent is that mental health issues are clearly a feature of the reality that we inhabit, and if this world has been programmed, then the programmers have clearly provided us with the ability to become insane. This insanity can take many forms, but in very few cases would this seem to be conducive to assisting programmers in steering a simulation towards its goals, whatever they might be. Sane, rational people have a tendency to behave in a rational, logical and fairly predictable manner. People whose mental states have become distorted, very often follow logic which is unusal, idiosyncratic and therefore inherently less predictable.

[edit] Unpredictability

Of course, an unpredictability of the simulees could well be a feature which a simulator would wish to include in a simulation, specifically to determine what happens when simulees are imbued with free will and the opportunity to make decisions which affect their entire existence, or even to have the freedom to go insane. It is difficult to actually see any other reason for including this feature other than to explore the consequences of it. This is a parallel version of the age-old dilemma: given that there is evil in the world, would a creator have created such a world unless they themselves were evil? The standard response to this dilemma is that the creator allows human beings to choose whether or not to follow particular routes. In that sense both scenarios are 'experiments' in free will.

Simulations can be viewed as 'random walks' through a particular space. Variables of interest can be observed, collected and collated, and the state of the simulation can be summarised as a point in the space of all possible observations. The more unpredictability there is in the simulation, the more random this walk will become. This will have two effects: the first is that the walk itself will be come more frantic, with sudden changes of direction and larger leaps between successive observations, and the second is that the walk may wander off into the more remote areas of the space where more predictable simulations would not venture.

This inclusion of free will as a part of the simulation therefore increases the level of unpredictability, and lessens the likelihood of achieving a particular goal, or desired end-point. It has been noted earlier, that if everything in the simulation were inherently predictable, there would be little point to it as an 'investigation'. However, increasing unpredictability reduces reliability; to counter this one would have to assume that literally thousands of such simulations were being carried out in order to counter these effects, in order to obtain some sort of statistical estimate of likely outcomes.

[edit] Memory, Points of View and the Historical Record

One interesting question arises in all this: where, in the simulation would individual memory lie? Individual memory is unreliable and fragmentory. There is evidence that false memories can be constructed, and selective remembering can cause bias and distortion. This means that each individual will have their own unique point of view of events. This strongly correlates with the features arising from applications of relativity. However, here the situation is probably more complex; not only is there a need to store the memory of an event unique to each individual who witnesses it, but also there should be an 'objective' view of each event, separate from all the viewpoints. On successive 'rememberings' and in discussions, the individual viewpoints may subtly change and alter over time, and so 'individual memory' is not fixed, it is dynamic. However, the objective 'historical' memory should remain fixed.

The question then, is how each of these individual viewpoints relate to the actual 'historical' record. There can be few people who are not familiar with the experience of reading a diary entry or other piece of writing, and failing to remember having written it. In addition, we have collective memories of occurrences which have never taken place: 'Play it Again, Sam', and 'Beam me up Scotty', have fallen into the vernacular, but were never uttered on screen. We therefore have layer after layer of complexity: the original event, memories of the original event, modifications to the memory of the event, mythologies surrounding the event and so on. Each event in effect can spawn its own sub-world of imagery, distortions and denials.

The memory burden and the programming detail needed to create and support the plethora of subtly different memories would be immense, and in order to justify the workload, would need to add huge value to be cost-effective. It is very difficult to imagine what this added value would be. Subtly different memories serve to confuse, since over time, artifacts decay, and the historical record becomes more fragmentary. The 'truth' needs to be reconstructed, and in time all documentary evidence relating to the original event will diappear, leaving an array of different memories. This gradual 'forgetting' will need to be programmed in.

Clearly, it would be simpler, and more cost-effective to forget by erasing. Forgetting by gradual loss is a highly resource-intensive way of achieving the same ends.

[edit] Moral and Legal Issues

What is the legal status of simulees? If they are brins-in-vats do they have rights as human beings? Could they remove themselves? What if the Matrix scenario of death in the simulation meant death in the real world?

What are the moral implications of doing this? Control is a major issue; how can we interpret 'free will' in this context? How can we determine whether or not a person has consented to do this? How would we measure whether a person was a minor, and able to give informed consent if a simulation runs at a different rate?

Bibliography: Jaynes, J., 1977, The Origin of Consciousness in the Breakdown of the Bicameral Mind, Boston, MA: Houghton Mifflin. Jenkins, Peter S., Historical Simulations - Motivational, Ethical and Legal Issues . Journal of Futures Studies, Vol. 11, No. 1, pp. 23-42, August 2006

[edit] Simulating events at the Quantum level

Argue that Quantum mechanics is not an artifact of the simulation, as suggested in :"A Cybernetic Interpretation of Quantum Mechanics", becuse of the tightly-knit mathematical theory with impressive predictive capability, which has yielded all kinds of applications: lasers, superconductive devices etc. In addition, attempts to disprove, eg. the EPR Experiment have shown it to stand up to the most rigorous scrutiny.

At its heart is the 'collapse of the wave-function', requiring an observer. The nature of observations required by this process, would create procedural problems for extrinsic simulations, epecially in the case of repeated observations of quantum-entanglements by different observers.

[edit] Summary

From three distinct arguments - that of producing an apparently bug-free simulation, the constraints imposed by the incorporation of relativity, and the 'collapse of the wave-function' observations of quantum mechanics, there appears to be a convergence on the point of view that if we are living in a simulation, then it must be a solipsistic intrinsic simulation. Effectively this is the most complex kind of simulation, and in that regard, might appear to deny Occam's Razor.

[edit] Re-examination of Bostrom's argument

Will we go extinct before we can perform such a simulation? Will we interested in doing it? Criticism of likelihood estimates.

[edit] Bostrom's Argument & the Liar's Paradox

Bostrom is basing his premises on our collective experience of this 'reality'. From this he deduces that there are three possible events, one of which must occur. One of these is that this 'reality' in fact is all probability is a simulation. If this is true, then the premise of the argument is false, as we cannot extrapolate to events outside this simulation, and events 1 and 2 have no validity.

In fact Bostrom disputes this, claiming that if theis were a simulation, it merely establishes that at least one simulation exists, confirming hiis point (3), and if we are not in a simulation, then the argument follows, and we are highly likely to be living in a simulation.

However, Bostrom's argument has at its heart two distinct flaws: Flaw (A) refers to the way that Bostrom calculates the total number of human-type experiences. He assumes that the average number of individuals that have lived in a civilisation before it reaches a posthuman stage is the same no matter whether that civilsation is real, or whether it is simulated. But why should this be? It could take a very long time for civilisations to reach posthuman-type situations, or it might be that simulations are only run for very short periods. Whatever is the case, it is unlikely that they would be equal. This means that Bostrom's probability calculation needs to be amended. Interstingly, when one follows this argument through, it tends to strengthen, rather than weaken, Bostrom's case, unless the civilisations run ancestor-type simulations for extended periods of time, much longer than their evolution time to posthumanity, which seems unlikely.

Flaw (B) is that within the probability argument, Bostrom claims that "The average number of ancestor-simulation run by interested civilisations is extremely large", and attributes this to their computing power. If we are in a simulation, then we cannot know anything about then number of technologically-capable or interested civilisation, other than there exists at least one of these. Following the Anthropic reasoning principle, this 'reality' will have been manufactured specifically to contain us, and therefore we are predisposed to thinking that simulations will be a commonplace. It may be that there are a myriad other civilisations out there who have not created simulations, the fact that we are living in a simulation cannot be used as evidence for the argument one way or the other. This negates Bostrom's counter-argument

A clearer way to understand the argument is to simplify it. Bostrom's argument can be reformulated as follows:

Only one of the following is true:

(1) Almost all Civilisations reach a point where they are either incapable of, or lack interest in creating artificial worlds.

(2) This 'reality' which we inhabit is an artificial world created by a civilisation about which we have no knowledge.

The probasbility argument which Bostrom uses is identical, and we conclude the probability of (1) is approximately zero or the probability of (2) is approximately 1.

However, If (2) is true, then we cannot discuss whether or not it is either (1) or (2) which is true, as this precludes us from knowing anything about other civilisations. Therefore we cannot conclude anything.

If however (2) is false, then we are living in a 'reality' rather than a simulation of one. This allows us to draw conclusions about the way that civilisations behave. In other words, we can pursue Bostrom's argument, and reduce it to the two-stage argument given above. From the discussions earlier, we can draw conclusions for our experience of this 'reality': it would appear that civilisations will become both technologically mature enough and interested enough in creating artifical worlds. In other words, (1) is false. However, Bostrom's reduced argument would then have us conclude that (2) is true. This clearly leads to a contradiction.

So the conclusions are: Either the argument is self-contradictory, or we cannot draw conclusions from it. In either case the argument does not hold water.

NB Following the Discussion with Peter Jenkins, I now believe that the argument can be interpreted (or possibly reformulated) more clearly as a Proof by Contradiction, in which case it might be valid. See the discussion pages.

[edit] Conclusions

It is clear from this series of articles that the entire thesis that our 'reality' is simulated is beset by a huge number of theoretical and practical difficulties. It is unlikely that an extrinsic consciouness could be a major feature of the system, but at the same time, it is difficult to see why intrinsic consciousnesses would be a necessary feature either; what purpose would it serve to include them? There are tremendous programming and other problems involved in allowing entities free will - effectively it increases unpredictability, and therefore reduces the liklihood of a simulation achieving its aims. In order to cope with the issues raised by relativiy and quantum mechanics, a solipsistic simulation seems most likely; however, this creates a world in which discrete minds experience different versions of the same reality, and again there is no clear purpose to doing this.

In essence, whichever way we turn, means and ends are in conflict. Logically in order to make the simulation workable we would be driven down particular pathways, and to employ particular means; however at the ends of those pathways, we may have a simulation, but there is no clear rationale as to why we would want to produce a simulation in that manner, and what it would be for. The conclusion must be therefore that it is highly unlikley that we are in fact, living in a simulation.

Where then, does that leave Bostrom's argument? Bostrom argues that if we are not living in a simulation, then either we will become extinct before we can produce one, or we will not be interested in doing so. The above analysis may in fact reveal why these two scenarios are, in fact more likely than the simulation hypothesis. The programming difficulties inherent in attempting to produce a realistic simulation may be so great that we effectively will never achieve the level of understanding of the universe required to achieve such a simulation. If we do, then there may be another, more subtle barrier. Having achieved such an understanding, there is little purpose to be served in creating a simualtion of something of which we have full understanding anyway, and therefore as we progress we will have no interest in creating such simulations.

Bostrom's argument may well be correct, but it may lead us towards the less interesting, but ultimately more informative conclusions after all.