Against the argument that we are living in a simulation
The original article can be found here:
- Against the argument that we are living in a simulation - Henry S. Sturman
This wiki is a good place to discuss the arguments in Henry's article. Below is a copy of his arguments, accompanied by views by the wiki users.
1. The assumption is that if people start simulating people, they will generally not tell the simulated people they are simulated people. But is that a valid assumption? If most people believe it is unethical to simulate people without telling them they are simulated, the argument fails. Hence, the fact that nobody told us that we are simulated is a case for thinking we might not be simulated.
- If the universe is a simulation, there's no way of telling who or what is running the simulation, and what the purpose of the simulation is. If we take ourselves as a reference, then we would run simulations for a number of reasons. The most common reason is to study. If we run a simulation to learn from it, we would want it to mimic reality as close as possible; if we would tell the simulated people that they are simulated, the simulation would immediately become virtually useless.
- Another purpose of a simulation could be recreational, i.e., games. On the Simulism page, where Simulism is explained using computer game history, we show an evolution of simulations; in none of them, the simulated characters are aware that they are simulated.
- The other part of this statement is about ethics. If the simulation is "not-real", there is no ethical requirement for informing the simulated people they are simulated; these people are simply not real so it would be 'ok' to treat them this way. If, on the other hand, the individuals are "real"--that is, real beings in the outer reality connected to their avatars in our (inner) reality--then it is always possible that they have already given their prior consent to the simulation. Furthermore, it is possible that they are giving their consent repeatedly and continuously, but simply do not have a memory of this from within the simulation, just as they apparently do not recall anything else about the outer reality from within the simulation.
- If the consciousness is real then it doesn't matter how the simulated people came into being. One has no more right to kill them than one has the right to kill one's own children.
2. Similarly, there is a lot of suffering in the world. If people simulating people were ethical they would not allow their simulated worlds to contain suffering. So again, the argument assumes that people simulating people are unethical and the validity of that assumption can be questioned.
- If you take computer games as an example here, it is sometimes hard to believe that it wouldn't be the other way around; many games revolve around suffering. The statement is closely related to the first one and a matter of ethics as well: is it ok if simulated people suffer? If it would not be ok, then it would also not be ok to have video games where simulated game characters are killed.
- It is interesting to see that The Matrix addresses this issue as well; Agent Smith explains to Morpheus why the simulated world is not a perfect world. His argument is that people reject a perfect world; the human mind is built to deal with suffering.
- Life is a rich tapestry of experience that must include suffering. Our suffering and the way we respond to the suffering of others are significant axes of the human experience. How many times have we heard the stories of a catastrophic event or disease leading people down new, better paths with renewed appreciation for life? How much of our art is created as the sublimination of suffering? How many people find their life's value defined by their work to reduce the suffering of others?
- I also submit that you can't eliminate suffering even if you remove the big causes (disease and so forth) without rewiring our minds, because most suffering happens in our minds, because of our attitudes towards events and change.
3. It is striking that there is such a great deal of mathematical logic and consistency in the laws of nature. Such features are not necessary in a simulation. In a simulation there is no need for our heads to contain brains rather than green cheese, for example. And in a simulation we might as well have been given the option to magically move objects from A to B with the flash of a thought. Thus, the fact that no such magic or inconsistency exist is an argument for our world not being a simulation.
- This arguments makes assumptions on the purpose of the simulation; if the purpose of the simulation is to be as close to reality as possible, it might be important to have heads that contain brains instead of green cheese. An interesting article in this area is 'The Expanding Simulation'.
- Even so, the absence of these 'features' does not form a strong argument; neither is gravity necessary in Quake, but that doesn't prove that it's not a simulation, nor that there's no good reason for it.
4. As the author admits, the argument requires acceptance of the assumption that the hardware doesn't matter for the arrisal of consciousness. I'm skeptical about that. If the biological structure of our brain is in fact a requirement for consciousness, then a computer which simulates the brain has no consciousness, and then the simulation argument fails. Except one could still argue that the space of the universe can contain more biological brains in vats (connected via wires to a computer simulating a world) than it has space for complete biological bodies. That would leave room for the argument that if people start putting brains in vats and simulating worlds for them, the probability is larger that you are a brain in a vat than a complete biological body in the real world.
- "I'm skeptical" is not much a counter-argument. The best that can be said for this is that the "brain in a vat" alternative does have more intuitive appeal than the "total simulation" alternative.
- Who is to say that our experience of consciousness is the same as the folks who designed the simulation? That is to say, once a sufficiently intelligent artificial life has been backwards-engineered from the original (assuming that's even the case) it would have to do a fair job of simulating consciousness. The sim would have to have sensory input, memory, free will, biological predispositions, interior monologue, the capacity to learn and so forth in order to perform in a realistic manner. Maybe what we experience as consciousness is merely that facsimile.
5. As the author, in effect, admits there is doubt whether his calculations are correct. Maybe more rather than less calculating power is required to simulate a consciousness + environment than nature requires for a direct implementation of a conscious being. For example, suppose we build a very powerful computer, based on nanotechnology of very small detail. If we build such a computer within a simulated world, then for it to work properly all those small nanodetails would have to be simulated rather than that they are directly implemented physically. I would think that this only adds another layer of complexity, such that a computer simulating such a computer would actually have to be larger and more complex than the computer it's simulating. Similarly, if one emulates an Apple on a PC or vice versa, you also generally lose computer speed. This provides a case for turning the whole simulation argument on its head: because a simulation of consciousnessss + environment costs more calculation resources than a direct implementation, we can simulate fewer people than the universe can contain directly. Therefore, the probability we are a real person is larger than the probability we are a simulated person.
- This is easily covered by Moore's Law. The real world, or the next level up, would have technology thousands of years beyond ours. One can emulate a 20-year-old Intel PC on last year's G5 Apple with zero loss of speed. Modern mobile phones, if the software existed, could also emulate that 20-year-old PC with no loss of speed. In a thousand years? 10,000 years? 100,000 years? 4,000,000,000 years?
- (It's easy to underestimate the power of Moore's law. By one estimate, if it continues for another 600 years or so, we will have enough processing power to simulate the known universe at the smallest level of detail we can currently notice. The result of 10,000 years of growth under Moore's law, were it possible, would be simply inconceivable to us.). More on the implications of this on the dedicated 'Moore's Law' page.
6. The validity of counter argument that if the simulation is no longer correct, the computer could simply change our memory content, is questionable. I believe that we can't reason anyway, unless we assume that our brains operate relatively correctly and there is no God in the background playing with our consciousness. However, if you do not make that assumption, then there is still no basis for accepting the simulation argument, for accepting any argument is only reasonable if we have faith in our ability to think straight. It's inconsistent to draw a conclusion based on our reasoning, while that very reasoning is based on the assumption that we can't trust our own reasoning.
- Our minds play tricks on us all the time. We exist in the context of this reality, whatever it is. If it is a simulation, it is a rich and elegantly-detailed simulation that (apparently) allows us reason, allows us free will. We just have to keep pursuing what we believe to be the best course of inquiry. Did Heisenberg give up when he discovered uncertainty?
7. But the fundamental error is that the whole argument is based on the application of a conclusion about our own world to another world about which we don't know anything. That is invalid. Even if all assumptions about computer power and motives about the people running simulations are correct, this only proves something about beings that we simulate ourselves. For the class of beings consisting of us plus the simulated beings it is correct to say that the probability one of them is simulated is greater than the probability that it is not. But, paradoxically, that's a conclusion only knowable to us, and not to the simulated beings. For it is only valid to apply this reasoning to our world and not to an imagined world above us. It's impossible for us to know or assume that the facts in our world, on which the whole simulation argument is based, are also true in a world above us. For example, any world above us might be a world in which all beings are already made with maximum efficiency, and so has more physical space for directly implemented beings than for simulated beings. In other worlds, the fatal flaw in the whole argument is that it hypothesizes a world above our world, while the very assumption that we are living in a simulation implies it is impossible for us to know anything about what such a world would be like and what would be the motives or actions of its inhabitants. The conclusion that we're probably living in a computer simulation immediately leads to the next conclusion that we can't trust the assumptions which led us to the first conclusion. Thus, the argument destroys itself.
- This is arguably Sturman's strongest argument. This argument does not lead us to dismiss the possibility that we are living in a simulation, however. Rather, it leads us to believe that we simply cannot know one way or the other. Some might then argue that we should assume that we are not, a la Russell's teapot. Bostrom's 'The Simulation Argument' acknowledges this, but provides a statistical model that illustrates that, even though we cannot tell for sure, the probability is in favour of the simulation.
- We have a hypothesis. At this time we know of no way to prove it or disprove it. That doesn't mean it can't happen. Even if we are completely computer-generated constructs, once we have the technology to build our own simulations perhaps we'll even be invited to visit a virtual version of the outside world and interact with its inhabitants.
8. The conclusion that we are living in a simulated world inside another world is a violation of Occam's razor.
- For those unfamiliar with this concept, here is an abstract from the wikipedia entry:
- This is often paraphrased as "All things being equal, the simplest solution tends to be the best one." In other words, when multiple competing theories are equal in other respects, the principle recommends selecting the theory that introduces the fewest assumptions and postulates the fewest hypothetical entities.
- In this light, the idea that the world is not a simulation would be simpler than the idea that it is a simulation. However, there are things that are actually easier to explain when the world is a simulation, such as what the origin of the universe is, the purpose of life, etc. For some theories, it would be against Occam's razor if we would assume that the world is not a simulation.
- In addition, simulism can only explain this world in terms of the external simulating world, which is mysterious to us. Therefore, one mystery is removed only to be replaced by another. Since the total amount of mystery under simulism is at least the same, and it is plainly more complex, it should be rejected.
- In any case, Occam's razor is a heuristic--essentially a piece of advice--and not capable of introducing a contradiction. In other words, this argument against simulism is exceedingly weak.