“All possible civilizations will never create simulations”—seems to be very implausible… “All possible civilizations will go extinct before they create simulations”—it also seems implausible
We only need to be concerned about actual civilizations, not all possible civilizations. We don’t know how many actual civilizations there are, so there could be very few (we know of only one). We also don’t know how difficult creating a sufficiently realistic simulation will be—obviously the more difficult it is to achieve such a simulation the more likely it is that civilizations will tend to go extinct before they create simulations. Finally, Bostrom’s propositions use “almost all” rather than “all”, e.g. “almost all civilizations at our current level of development go extinct before reaching technological maturity”.
In light of these considerations, Bostrom’s first two propositions do not seem implausible to me.
It all depends of our understanding of actuality. If modal realism is true, there is no difference between actual and possible. If our universe is really very large because of MWI, inflation and other universes, there should be many civilizations. But it all require some difficult philosophical questions, so it is better to use simple model with cut-of. (I think that model realism is true and all possible is actualy exist somewhere in the universe, if actuality is not depending of human consciousness, but it is long story for another post, so I will not concentrate on proving it here)
Imagine that in our Galaxy (or any other sufficiently large part of the Universe) exists 1000 our-tech level civilizations. If 990 of them go extinct in x-risks, 9 decide not to create simulations and 1 decided to model all civilizations in the galaxy 100 000 000 times in order to solve Fermi paradox numerically.
That is why I didn’t use the word “almost”. Because in this example almost all go extinct, and almost all will not make simulations, but it doesn’t prevent one civilization to create so many simulations that overweight it.
The only condition in which we are not in simulation is that ALL possible civilization will not make them.
In with case we have 100 001 000 total number of the civilizations.
In this example we see that even if most civilization will go extinct, and most of survived civilizations will decide not to run simulation, 1 will do it based on its practical needs, and proportion of real to simulated will be 1 to 100 000.
It means that we are with a chance of 100 000 is in simulated civilization.
This example is also predicts the future of our simulation: it will simulate extinction event with 99 per cent probability, it will simulate simulation-less civilization with 0.9 probability and it will result in two level “matryoshka simulation” with 0.1 per cent simulation.
It also demonstrate that Bostrom’s preposition is not alternative: all 3 conditions are true in this case (And Bostrom said that “at least one of three condition is true”). We will go extinct, we will not run simulations, we are in simulation.
With those assumptions (especially modal realism), I don’t think your original statement that our simulation was not terminated this time doesn’t quite make sense; there could be a bajillion simulations identical to this, and even if most of them we’re shut down, we wouldn’t notice anything.
In fact, I’m not sure what saying “we are in a simulation” or “we are not in a simulation” exactly means.
What you say is like quantum immortality for many simulation world. Lets name it “simulation immortality” as there is nothing quantum about it. I think that it may be true. But it requires two conditions: many simulations and identity problem solution (is a copy of me in remote part of the universe is me.) I wrote about it here: http://lesswrong.com/lw/n7u/the_map_of_quantum_big_world_immortality/
Simulation immortality precisely neutralise risks of the simulation been shut down. But if we agree with quantum immortality logic, it works even broader, preventing any other x-risk, because in any case one person (observer in his timeline) will survive.
In case of simulation shutdown it works nicely if it will be shutdown instantaneous and uniformly. But if servers will be shut one by one, we will see how stars will disappear, and for some period of time we will find ourselves in strange and unpleasant world. Shut down may take only a millisecond in base universe, but it could take long time in simulated, may be days.
Slow shutdown is especially unpleasant scenario for two reasons connected with “simulation immortality”.
Because of simulation immortality, its chances are rising dramatically: If for example 1000 simulations is shutting down and one of them is shutting slowly, I (my copy) will find my self only in the one that is in slow shut down.
If I find my self in a slow shutdown, there is infinitely many the same simulations, which are also experience slow shutdown. In means that my slow shutdown will never end from my observer point of view. Most probably after all this adventure I will find my self in the simulation which shutdown was stopped or reversed, or stuck somewhere in the middle.
TL;DR: So shut down of the simulation may be observed and may be unpleasant and it is especially likely if there are infinitely many simulations. It will look like very strange global catastrophe from the observer point of view. I also wrote a lot of such things in my simulation map, but idea of slow shut down I got only now
In fact, first two propositions in Bostrom’s article are very improbable, especially if we include in the consideration all other civilizations.
1) “All possible civilizations will never create simulations”—seems to be very implausible, we are already good in creating movies, dreams and games.
2) “All possible civilizations will go extinct before they create simulations—it also seems implausible.
We only need to be concerned about actual civilizations, not all possible civilizations. We don’t know how many actual civilizations there are, so there could be very few (we know of only one). We also don’t know how difficult creating a sufficiently realistic simulation will be—obviously the more difficult it is to achieve such a simulation the more likely it is that civilizations will tend to go extinct before they create simulations. Finally, Bostrom’s propositions use “almost all” rather than “all”, e.g. “almost all civilizations at our current level of development go extinct before reaching technological maturity”.
In light of these considerations, Bostrom’s first two propositions do not seem implausible to me.
It all depends of our understanding of actuality. If modal realism is true, there is no difference between actual and possible. If our universe is really very large because of MWI, inflation and other universes, there should be many civilizations. But it all require some difficult philosophical questions, so it is better to use simple model with cut-of. (I think that model realism is true and all possible is actualy exist somewhere in the universe, if actuality is not depending of human consciousness, but it is long story for another post, so I will not concentrate on proving it here)
Imagine that in our Galaxy (or any other sufficiently large part of the Universe) exists 1000 our-tech level civilizations. If 990 of them go extinct in x-risks, 9 decide not to create simulations and 1 decided to model all civilizations in the galaxy 100 000 000 times in order to solve Fermi paradox numerically.
That is why I didn’t use the word “almost”. Because in this example almost all go extinct, and almost all will not make simulations, but it doesn’t prevent one civilization to create so many simulations that overweight it.
The only condition in which we are not in simulation is that ALL possible civilization will not make them.
In with case we have 100 001 000 total number of the civilizations.
In this example we see that even if most civilization will go extinct, and most of survived civilizations will decide not to run simulation, 1 will do it based on its practical needs, and proportion of real to simulated will be 1 to 100 000.
It means that we are with a chance of 100 000 is in simulated civilization.
This example is also predicts the future of our simulation: it will simulate extinction event with 99 per cent probability, it will simulate simulation-less civilization with 0.9 probability and it will result in two level “matryoshka simulation” with 0.1 per cent simulation.
It also demonstrate that Bostrom’s preposition is not alternative: all 3 conditions are true in this case (And Bostrom said that “at least one of three condition is true”). We will go extinct, we will not run simulations, we are in simulation.
With those assumptions (especially modal realism), I don’t think your original statement that our simulation was not terminated this time doesn’t quite make sense; there could be a bajillion simulations identical to this, and even if most of them we’re shut down, we wouldn’t notice anything.
In fact, I’m not sure what saying “we are in a simulation” or “we are not in a simulation” exactly means.
What you say is like quantum immortality for many simulation world. Lets name it “simulation immortality” as there is nothing quantum about it. I think that it may be true. But it requires two conditions: many simulations and identity problem solution (is a copy of me in remote part of the universe is me.) I wrote about it here: http://lesswrong.com/lw/n7u/the_map_of_quantum_big_world_immortality/
Simulation immortality precisely neutralise risks of the simulation been shut down. But if we agree with quantum immortality logic, it works even broader, preventing any other x-risk, because in any case one person (observer in his timeline) will survive.
In case of simulation shutdown it works nicely if it will be shutdown instantaneous and uniformly. But if servers will be shut one by one, we will see how stars will disappear, and for some period of time we will find ourselves in strange and unpleasant world. Shut down may take only a millisecond in base universe, but it could take long time in simulated, may be days.
Slow shutdown is especially unpleasant scenario for two reasons connected with “simulation immortality”.
Because of simulation immortality, its chances are rising dramatically: If for example 1000 simulations is shutting down and one of them is shutting slowly, I (my copy) will find my self only in the one that is in slow shut down.
If I find my self in a slow shutdown, there is infinitely many the same simulations, which are also experience slow shutdown. In means that my slow shutdown will never end from my observer point of view. Most probably after all this adventure I will find my self in the simulation which shutdown was stopped or reversed, or stuck somewhere in the middle.
TL;DR: So shut down of the simulation may be observed and may be unpleasant and it is especially likely if there are infinitely many simulations. It will look like very strange global catastrophe from the observer point of view. I also wrote a lot of such things in my simulation map, but idea of slow shut down I got only now
http://lesswrong.com/lw/mv0/simulations_map_what_is_the_most_probable_type_of/