> Simulations Are Things In The World ... they do not step out of Musk’s base reality. They are still base reality. They are made of the same stuff everything else is made of.
The basic objection here -- that Musk misunderstands the nature of simulation on a philosophical-technical level, and that even if our existence was simulated, it is still as meaningful as it would be if it weren't -- is nicely written and argued, but it feels woefully irrelevant to the debate. The author has done nothing to address the fundamental probabilistic argument that Musk uses in favor of simulation. This article is mostly an attack on definitions, not one on substance.
The author doesn't consider the probabilities because he refutes the whole notion altogether. The point is that being able to create a simulation (imitation) of reality (a game for example) is completely different from creating a different, separate reality. The simulations that we create are always, and by necessity, inside the real world, the only one that really exists.
Inside a game, there is no virtual reality. The only thing that exists inside a machine are electric voltages. The simulation provided by a game only woks for us, who are already in the real world, observing the simulation. And these simulations only work because they replicates real things that we already know. For any kind of "being" "living" in the simulation itself, there is nothing of importance other than a bunch of electrical impulses in silicon.
> For any kind of "being" "living" in the simulation itself, there is nothing of importance other than a bunch of electrical impulses in silicon.
Considering the fact that "a bunch of electrical impulses in silicon" is enough to encompass all computable functions, including all of the laws of physics, I don't know of anything "other" that a being might care about...
> Whatever computers are capable of doing, they're still just part of the real universe, not of a separate, simulated universe.
I don't recall anyone ever claiming that computers or simulations are separate from the real universe. Dualists make similar claims, but they usually claim their non-physical quackery doesn't apply to computers. So, I don't know what point you're trying to make.
Just consider that when a computer simulation looks incredibly good, it just looks great for us, human observers. The simulation itself doesn't recognize anything that is generated by computer screens (even when there is some level of AI involved), it is our brain that perceives everything as real. So, my point is that the simulations we create don't live in a separate universe where they "experience" another reality. Just the opposite, they only simulate our own reality. So, it really doesn't matter how good they look, it doesn't make any difference if there is no subtract for this simulated reality to exist independently.
Of course, if a simulation happens to have some output, e.g. pixels on a screen, infrared radiation from a processor, etc. that doesn't effect the simulation itself.
While it's true that a simulation will be part of the real universe, it's not particularly useful to think at that level; in the same way that it's not useful to handle POSTs to a login form by considering the voltage difference across logic gates. It's much more useful to use the "laws of HTTP", "laws of TCP", etc. to reason about "requests", "processes", "sockets", etc. Such entities do exist, but they exist in a "virtual" world; that world is not separate from the physical world, but it can be completely described without reference to the physical world. Physics and reality are merely implementation details, and don't offer any particular insights into the system's behaviour.
In fact, a computer (e.g. one running a simulation) is an incredibly contrived situation; about as pathological as you can get, since (digital) computers are specifically engineered to abstract away the laws of physics, are robust against perturbations, and their overall behaviour is irreducibly complex (as Wolfram would call it). The only results we can derive about the signals in the computer are trivial, i.e. they would apply regardless of what the system's doing; e.g. the signals aren't creating/destroying energy, they're not propagating information faster than light, etc.
As a concrete example, the statistical mechanics underpinning classical thermodynamics requires that we assume ergodicity (all states are reachable from all other states, with some non-zero probability). We can't make such an assumption for a computer, since it would violates Rice's theorem (i.e. if the "halt state" is always reachable, we've solved the halting problem), and hence we can't apply many of those techniques. We could argue that there will be eventually be errors, like cosmic ray interactions, and we can use their (exponentially small) probabilities to gain ergodicity, but that gives us predictions despite the simulation, not about the simulation.
The only sensible level to understand a simulation is at the level of software; e.g. as instructions, procedures, datastructures, modules, classes, combinators, whatever. If those software components interact such that an "Inhabitant" object's "tastes" attribute references an "Apple" object, then it's completely appropriate to say that 'the simulation's inhabitant tastes an apple'; we can compare/contrast how/if an inhabitant is related to a human, and how its taste attribute is related to humans' sense of taste. In this simplistic example, the simulation would be incredibly crude in comparison to humans and their taste, in which sense we could say the simulation is "low resolution". However, we cannot deny the existence of such simulated individuals, or claim that they cannot taste simulated apples, in the same way we cannot deny the existence of login forms and POST requests. They exist in virtual worlds governed by laws encoded in software and hardware. That those virtual worlds just-so-happen to be implemented by a chunk of the physical world (called "a computer") is undeniable, but it's also almost completely irrelevant if we want to describe, study, predict, etc. such worlds at a meaningful level of abstraction.
Downloading pirated torrent is unethical and probably illegal, but tons of people still use it. So I think "less likely" is not really the case. People will do anything as long as there's value.
Why would it be unethical or illegal? You are also basing the fact that the simulation is run by humans or any other entities with a social structure and a grasp on what we call "morality".
If we are living in a simulation morally it is not much different than the SIMS or minecraft...
This guy basically is saying "the world we live in is physical, because well.. it is physical!". How does someone like this become a professor and publish books?
I have no problem someone saying "it is very likely we live in a simulation" because it's an opinion and he's not trying to hide the fact that it's just an opinion.
But saying stuff like "THAT IS WRONG" and "It's conceptually and empirically incorrect" is really amateur.
By the way does this guy even know what "empirically" means? Our experience is nothing more than what we perceive as human beings, and there's plenty of evidence that supports that what we perceive is not necessarily the reality. And this guy is saying "it's empirically wrong". It is impossible to prove a philosophical theory "empirically wrong".
"But Einstein said that compound interest is the most powerful force in the universe, and I take all my investment advice from flippant remarks by theoretical physicists making small talk at parties."
Musk's comments get held to ludicrously high standards because it's sensational to write headlines like "Elon is wrong."
Someone missed the point.
If we're in a simulation, there is no spoon, man. Nor a worm that needs that apple. The water isn't simulated because there is no water.
I'm getting a bit tired of reading this 'Elon Musk says we live in a simulation'. No he didn't! At least he wasn't the first one to say it. The idea isn't his. The philosophical idea has been around since when people started questioning existence and actual theoretical basis of living in a computer simulation argument has been around since well 2001.
Yes it did and that's between Plato[1] or earlier and Bostrom[2] and later, which was essentially my point. You really should have visited the link in my comment before making that observation. I was speaking about the theoretical basis of the idea.
Elon Musk's caution about AI comes from the book superintelligence[3]
I think Elon is wrong, but for technical reasons, not philosophical ones. For one, we do not have photorealistic simulations for real-time games. We're not even close, frankly. Still-renders can pass the photorealism barrier, as in indistinguishable from a photograph, but even state-of-the-art triple A games are still facilely obvious simulations.
Most importantly, there hasn't been much progress in the past few years. Compare Battlefield 3 (2011) with Battlefield 4 (2013) with Battlefield 1 (2016, strangely enough), and you really won't see that much progress in realism. While you can get pretty screenshots, as soon as it's in motion, the distance from reality is obvious. What's missing is physics. Good physics simulations (ie fluid flow, soft-body, divisible materials, etc) is so computationally expensive that I doubt anything even approaching it will ever be realtime possible.
Why? Because the rate of improvement is slowing. Those paying attention know Moore's law is running out of steam (Intel recently had to insert an extra tick into their tick-tock model), and power draw / heat dissipation are ever more difficult to push. Arguments for the singularity / simulation all rely on an exponential curve being projected as far into the future as the speaker feels necessary, but in reality, there is no such thing as an exponential curve. All physical processes that experience exponential looking growth eventually slow down, forming a logistic curve, and it's looking like the growth curve for FLOPS is going to top out well before "photorealistic realtime simulation of the entire world available to every consumer".
From inside a simulation, it's impossible to tell how fast the simulation is running because you have no external time reference. The visible universe as we know it could be running at a time dilation of of 10^100:1 and it would look exactly the same to us.
True, but orthogonal to the argument. Elon's pro-simulation argument is fundamentally grounded in our reality ("video games have progressed a lot and there are lots of video game consoles, so eventually we could run the types of simulations postulated"), and my argument is refuting that ("video games will never be good enough to run those types of simulations")
Your argument is no different than people from the old times saying "the earth is flat". Of course that was the reality back then, but we've found out that's not the case.
When talking about these things you should keep in mind whatever you say is just an opinion and a theory. Just like Newton's law was the reality until Einstein came along, the "reality" is nothing more than our interpretation of the world.
That's why I think it's not a good idea to conclude someone is "wrong" for technical reasons. Our "technical" reality has been evolving and will keep evolving, so our reality today won't be the same as tomorrow's reality.
For all we know, the graphics and physics engine running our universe could be a laughably mediocre imitation of what the external reality is like.
That doesn't help the "ancestor simulation" idea, since that relies on advanced being wanting to simulate conditions encountered in their past (i.e. within the same universe).
It does apply to the idea that we might be some kind of low-res game, running on a crummy hand(/tentacle)-held computer.
Urgh, one nonsensical statement after another. I think I need a shower after reading that.
Most of the authors' misguided ramblings seem to stem from a very naive misunderstanding of the word "computer", i.e. that in a technical context it refers to any system capable of universal computation (AKA "Turing complete"). "Universal computation" has precise Mathematical definitions, but can roughly be understood as the "complexity limit" of the Universe; just as nothing can go faster than light, nothing can exhibit behaviour that's more complex than that of a Turing complete system. The proof boils down to showing that a Turing complete system can, given a specially constructed input stimulus, produce response behaviour which is equivalent to any other system. These days, we call the system the "hardware" and the stimulus the "software".
As an analogy, the author's arguments would be like claiming the laws of thermodynamics don't apply to the brain, since we have no theory of conciousness and it certainly doesn't look like a steam engine. Just like "computer" in computer science, the word "engine" has a precise meaning in thermodynamics (any system which transforms energy from one form to another), and we don't need to know the precise workings of the brain to know that, for example, it's not a perpertual motion machine.
Likewise, we don't need to know the precise hardware or software of the brain to know it obeys the physical Church Turing thesis, and can hence be simulated by a universal Turing machine.
If you are in a simulation you are not the one looking at the picture of the apple, and trying to eat it from the outside: you are the one for whom the only real thing is the apple that you have in front of you. And the apple indeed feeds you, since the rules in the simulated world are such that "consuming apple object provide nutrients to being eating it".
In a simulation, the reality is the simulation, the rules are the rules programmed in the simulation, in fact you will have a hard time reaching out to base level, since the rules in the simulated world are hiding it from you (which gives for interesting science fiction ideas).
> Simulations are things that we use to talk or to think about other things. In this respect, they do not step out of Musk’s base reality. They are still base reality. They are made of the same stuff everything else is made of.
Accepting the simulation premise does contradict the fact that everything is made of stuff. But that everything is made of stuff does not mean that the objects in the simulation are made of the stuff that the simulation suggests.
And he is missing the main point of the simulation theory, namely that all entities in the simulation are not aware of the underlying levels: in our simulations, the apple and the avatar eating the apple are objects in a program (that is, basically 1s and 0s, not matter) running in a computer processor made of silicon matter. But the apple, or the avatar do not know anything about computer programs, 0s or 1s, silicon, or anything else in the underlying level. And, more interestingly, the apple and the avatar can be upgraded, modified, cloned, deleted or shut down by the computer programmer, all while they are running on "stuff".
In the simulations being run by our masters, where we live, we are objects in a "program" running in a "computer" made of "silicon", all that in quotes because we do not know how exactly the underlying level is structured.
31 comments
[ 3.4 ms ] story [ 80.0 ms ] thread> Simulations Are Things In The World ... they do not step out of Musk’s base reality. They are still base reality. They are made of the same stuff everything else is made of.
I stopped reading here.
Inside a game, there is no virtual reality. The only thing that exists inside a machine are electric voltages. The simulation provided by a game only woks for us, who are already in the real world, observing the simulation. And these simulations only work because they replicates real things that we already know. For any kind of "being" "living" in the simulation itself, there is nothing of importance other than a bunch of electrical impulses in silicon.
Considering the fact that "a bunch of electrical impulses in silicon" is enough to encompass all computable functions, including all of the laws of physics, I don't know of anything "other" that a being might care about...
I don't recall anyone ever claiming that computers or simulations are separate from the real universe. Dualists make similar claims, but they usually claim their non-physical quackery doesn't apply to computers. So, I don't know what point you're trying to make.
While it's true that a simulation will be part of the real universe, it's not particularly useful to think at that level; in the same way that it's not useful to handle POSTs to a login form by considering the voltage difference across logic gates. It's much more useful to use the "laws of HTTP", "laws of TCP", etc. to reason about "requests", "processes", "sockets", etc. Such entities do exist, but they exist in a "virtual" world; that world is not separate from the physical world, but it can be completely described without reference to the physical world. Physics and reality are merely implementation details, and don't offer any particular insights into the system's behaviour.
In fact, a computer (e.g. one running a simulation) is an incredibly contrived situation; about as pathological as you can get, since (digital) computers are specifically engineered to abstract away the laws of physics, are robust against perturbations, and their overall behaviour is irreducibly complex (as Wolfram would call it). The only results we can derive about the signals in the computer are trivial, i.e. they would apply regardless of what the system's doing; e.g. the signals aren't creating/destroying energy, they're not propagating information faster than light, etc.
As a concrete example, the statistical mechanics underpinning classical thermodynamics requires that we assume ergodicity (all states are reachable from all other states, with some non-zero probability). We can't make such an assumption for a computer, since it would violates Rice's theorem (i.e. if the "halt state" is always reachable, we've solved the halting problem), and hence we can't apply many of those techniques. We could argue that there will be eventually be errors, like cosmic ray interactions, and we can use their (exponentially small) probabilities to gain ergodicity, but that gives us predictions despite the simulation, not about the simulation.
The only sensible level to understand a simulation is at the level of software; e.g. as instructions, procedures, datastructures, modules, classes, combinators, whatever. If those software components interact such that an "Inhabitant" object's "tastes" attribute references an "Apple" object, then it's completely appropriate to say that 'the simulation's inhabitant tastes an apple'; we can compare/contrast how/if an inhabitant is related to a human, and how its taste attribute is related to humans' sense of taste. In this simplistic example, the simulation would be incredibly crude in comparison to humans and their taste, in which sense we could say the simulation is "low resolution". However, we cannot deny the existence of such simulated individuals, or claim that they cannot taste simulated apples, in the same way we cannot deny the existence of login forms and POST requests. They exist in virtual worlds governed by laws encoded in software and hardware. That those virtual worlds just-so-happen to be implemented by a chunk of the physical world (called "a computer") is undeniable, but it's also almost completely irrelevant if we want to describe, study, predict, etc. such worlds at a meaningful level of abstraction.
If we are living in a simulation morally it is not much different than the SIMS or minecraft...
I have no problem someone saying "it is very likely we live in a simulation" because it's an opinion and he's not trying to hide the fact that it's just an opinion.
But saying stuff like "THAT IS WRONG" and "It's conceptually and empirically incorrect" is really amateur.
By the way does this guy even know what "empirically" means? Our experience is nothing more than what we perceive as human beings, and there's plenty of evidence that supports that what we perceive is not necessarily the reality. And this guy is saying "it's empirically wrong". It is impossible to prove a philosophical theory "empirically wrong".
"But Einstein said that compound interest is the most powerful force in the universe, and I take all my investment advice from flippant remarks by theoretical physicists making small talk at parties."
Musk's comments get held to ludicrously high standards because it's sensational to write headlines like "Elon is wrong."
http://www.simulation-argument.com
Elon Musk's caution about AI comes from the book superintelligence[3]
[1] https://en.wikipedia.org/wiki/Allegory_of_the_Cave
[2] https://en.wikipedia.org/wiki/Nick_Bostrom
[3] https://en.wikipedia.org/wiki/Superintelligence:_Paths,_Dang...
Most importantly, there hasn't been much progress in the past few years. Compare Battlefield 3 (2011) with Battlefield 4 (2013) with Battlefield 1 (2016, strangely enough), and you really won't see that much progress in realism. While you can get pretty screenshots, as soon as it's in motion, the distance from reality is obvious. What's missing is physics. Good physics simulations (ie fluid flow, soft-body, divisible materials, etc) is so computationally expensive that I doubt anything even approaching it will ever be realtime possible.
Why? Because the rate of improvement is slowing. Those paying attention know Moore's law is running out of steam (Intel recently had to insert an extra tick into their tick-tock model), and power draw / heat dissipation are ever more difficult to push. Arguments for the singularity / simulation all rely on an exponential curve being projected as far into the future as the speaker feels necessary, but in reality, there is no such thing as an exponential curve. All physical processes that experience exponential looking growth eventually slow down, forming a logistic curve, and it's looking like the growth curve for FLOPS is going to top out well before "photorealistic realtime simulation of the entire world available to every consumer".
When talking about these things you should keep in mind whatever you say is just an opinion and a theory. Just like Newton's law was the reality until Einstein came along, the "reality" is nothing more than our interpretation of the world.
That's why I think it's not a good idea to conclude someone is "wrong" for technical reasons. Our "technical" reality has been evolving and will keep evolving, so our reality today won't be the same as tomorrow's reality.
That doesn't help the "ancestor simulation" idea, since that relies on advanced being wanting to simulate conditions encountered in their past (i.e. within the same universe).
It does apply to the idea that we might be some kind of low-res game, running on a crummy hand(/tentacle)-held computer.
Most of the authors' misguided ramblings seem to stem from a very naive misunderstanding of the word "computer", i.e. that in a technical context it refers to any system capable of universal computation (AKA "Turing complete"). "Universal computation" has precise Mathematical definitions, but can roughly be understood as the "complexity limit" of the Universe; just as nothing can go faster than light, nothing can exhibit behaviour that's more complex than that of a Turing complete system. The proof boils down to showing that a Turing complete system can, given a specially constructed input stimulus, produce response behaviour which is equivalent to any other system. These days, we call the system the "hardware" and the stimulus the "software".
As an analogy, the author's arguments would be like claiming the laws of thermodynamics don't apply to the brain, since we have no theory of conciousness and it certainly doesn't look like a steam engine. Just like "computer" in computer science, the word "engine" has a precise meaning in thermodynamics (any system which transforms energy from one form to another), and we don't need to know the precise workings of the brain to know that, for example, it's not a perpertual motion machine.
Likewise, we don't need to know the precise hardware or software of the brain to know it obeys the physical Church Turing thesis, and can hence be simulated by a universal Turing machine.
In a simulation, the reality is the simulation, the rules are the rules programmed in the simulation, in fact you will have a hard time reaching out to base level, since the rules in the simulated world are hiding it from you (which gives for interesting science fiction ideas).
> Simulations are things that we use to talk or to think about other things. In this respect, they do not step out of Musk’s base reality. They are still base reality. They are made of the same stuff everything else is made of.
Accepting the simulation premise does contradict the fact that everything is made of stuff. But that everything is made of stuff does not mean that the objects in the simulation are made of the stuff that the simulation suggests.
And he is missing the main point of the simulation theory, namely that all entities in the simulation are not aware of the underlying levels: in our simulations, the apple and the avatar eating the apple are objects in a program (that is, basically 1s and 0s, not matter) running in a computer processor made of silicon matter. But the apple, or the avatar do not know anything about computer programs, 0s or 1s, silicon, or anything else in the underlying level. And, more interestingly, the apple and the avatar can be upgraded, modified, cloned, deleted or shut down by the computer programmer, all while they are running on "stuff".
In the simulations being run by our masters, where we live, we are objects in a "program" running in a "computer" made of "silicon", all that in quotes because we do not know how exactly the underlying level is structured.