Why we are not living in a Simulation
From Simulism
Consider what we would actually need to do to program a universe to the detail that we see around us - not the computing power, but the knowledge & understanding of physical, psychological, sociological, biological and cosmological processes, together with an algorithmic implementation of these. It has been hypothesised that "an intelligent operating system could fill in the details in an ad hoc manner", and circumvent the computing power requirements, but there is no circumventing the main issue here, that we would need to know just about everything about everything in order to do it. In addition we would have to have a mechanism for being able to implement this on some sort of computational substrate; which either means programming line by line or evolving some sort of neural net (or equivalent).
Let us say that all this is possible. The beings who have reached this stage have effectively solved all the problems in the universe. They understand biology, they understand themselves, they know all about physics; if they don't they can't create a reality simulation. So the question is, if anyone can do this, why would they wish to run a simulation? What would be its purpose?
Simulations currently have two main purposes: information-gathering and entertainment. Bostrom presupposes that societies would want to run ancestor-simulations for the former reason, the purpose of such simulations being to find out how people behave under situation X, or what would have happened if Y had occurred intead of Z. That is only valid if the simulation is a model of the society, containing less detail, and can run in real-time at a faster rate than the original. However, that means that in that case the simulation is not a simulation of reality - it is a simulation of less-than-reality. The first purpose for creating a simulation is therefore is ruled out, and unless the superintelligent beings are creating a simulation just to watch how we behave when when we get up in the morning and go to the toilet, the second one would be too If it turns out to be true that we are really nothing but contestants in Big Brother 10^17, waiting to be ejected into the arms of some 961st century Davina McCall then I despair about the future of the universe. If that fate awaits us, we should all roll over and wait for the men in white coats instead of getting up in the morning.
There is one, and only one get out and that is Tipler's Omega Point theory. In Tipler's restaurant at the end of the universe, all of creation's resources would be used to fabricate simulations which could be run at exponentially increasing rates to cheat the demise of everything. If the reason that we are here is that this is the only way to survive the Big Crunch (or to witness the second coming of the Great Prophet Zarquon) then I can go along with that. However, when you consider what that entails, if the entire universe is the substrate on which the simulation is running, then how is that different to what we currently regard as reality?
The world as we experience it, by scientific reckoning, has been evolving for the last 15 billion years, starting with the Big Bang, and passing through lots of different manifestations to arrive at this point in time. The notion of 'Last Thursdayism' or 'The five-minute universe' posits that all of this was created some time in the recent past, e.g. 5 minutes ago, with all the fossils, memories of people, the broken beer bottle in my bin, and the scar on my knee all created to make us believe that it has been in existence for 15 billion years. That is an interesting diversionary tactic, but it does not account for the fact that the complexity that we see around us would need to have been developed in order to convince us of such a hoax.
I am not 100% sure of this point, but my intuition tells me that has got to be some sort of conservation of information rule that says that something equivalent to the level of complexity and organisation we observe in the universe around us could not have been created without an equivalent degree of effort to that which was needed to get us to this point. I am sure that this can be related to the second law of thermodynamics (Looyd, 2007). It has taken us (apparently) 15 billion years to get to this point; we think it has taken us that time, not because we have guessed, but because we have calculated a reasoned prediction made on the basis of consistent scientific theory & observations. That means the program has been running for the equivalent of 15 billion years, or that it took 15 billion years to develop. They have either got an amazingly dedicated set of programmers who must like each other a lot, or they are amazingly patient.
Looked at from our point of view, we observe a universe which seemingly has used all of its 'computing power' to get to this point, and it has apparently taken 15 Billion years of our time to get to this point. Suppose that we decide, in our future, to develop a program to do simulate what we see around us now. In order to do this, we would need to either (1) use ALL the universe's resources and take 15 billion years to develop it, which is effectively what has happened, and we call it 'reality', or (2) create the simulation in the same detail, but use a subset of the Universe's resources, thus having a run-time of longer than 15 billion years to get to this point (which for practical reasons, no-one would realistically attempt to do), or (3) create a simulation with far less detail than we think we see, using only a fraction of the Universe's resources.
It is case (3) which is most likely, and therefore most interesting. This is tha case which has been used as an 'exit clause' by proponents of the Simulation Hypothesis, in order to explain how we could simulate the Universe by using far fewer resources than we apparently need. However, if we could actually do this, then we would need to be albe to access a huge store of knowledge and understanding about physical and other processes. Just think about what would have to happen to implement this. Suppose I am working ar CERN, and I decide to investigate Higgs' Bosons. The intelligent operating system sees that I am about to create some new physics, and decides that it needs to do some calculations about what needs to happen, just to make sure that I see what he I am supposed to see.
This means that some being, somewhere has enough knowledge and computing power to calculate everything down to the last qubit, or at least invent it when it becomes necessary. The knowledge, understanding and the computational ability to do this is phenomenal. Remember this is a simulation, so things would have to be there when they were required, which means that they would have to be calculated and generated in real time. This would thererefore require detailed and complete deterministic knowledge of the way highly complex and non-linear systems operate. Currently when we have such complex dynamical or non-linear systems, we solve them by running simulations. This is not an option here, otherwise we would need to run a simulation to find out what to put in the simulation, which would inevitably involve simulations within the sub-simulation and so on into infinite regress.
Therefore I cannot see how a simulation of reality this complex would work. Either we would never achieve the programming capapbility to do it, or if we did, why would we bother? If we could do it, we would be able to calculate answers to every question in nanoseconds; if we really wanted to find out how our ancestors behaved we would calculate it, not simulate it. There would be no point to running ancestor simulations except for their entertainment value.
Effectively this means that we will either never reach the stage at which we are capable of running ancestor-simulations, or if we do, we will not be interested in them, because we have better ways of finding answers to questions. Either way, using Bostrom's argument, cases (1) & (2) from his disjuction are far more likely than his case (3), ergo we are not living in a simulation.
If I'm wrong, cart me off to the Diary room so I can talk to Big Brother.
Tony Fleet
March 2007
[edit] References & Notes
Lloyd, 2007 ( Seth Lloyd,in an article "Black Hole Computers" by Seth Lloyd and Y.Jack Ng, in Scientific American Special Issue on Astrophysics Vol 17, No. 1 (2007), says "At each epoch in its history, the maximum number of bits the universe can contain is approximately equal to the number of operations it could have performed up to that moment", and further in the article "The universe has performed the maximum possible number of operations allowed by the laws of physics". This would seem to support my original intuition that Last Thursdayism is essentially incorrect, as it would have taken the Universe 15 billion years to get to this point, even if it had been programmed.
