Epistemic: generally agree with S.Armstrong here, but adding some links and ideas
SSA and SIA are designed in the way that they work only in a finite universe. FNС and anthropic decision theory also assume that the number of observers can be countable. Therefore, they are toy examples which are known not to generalize to infinite multiverse.
Why infinite universe is almost certain?
A) Because any time we count a finite number of observers or universes, we can ask – why not one more causally independent universe just pop up into existence the same way as our universe?
B) If we assume SIA (the fact that I exist at all means that many observers exist), it updates with infinite Bayesian power for infinite universe.
С) Different Tegmark’s levels.
The fact that we are rather typical in the checkable ways says that adding infinities doesn't distort our typicality – for example, the world around doesn't look like chaotic Boltzmann brain (BB) experience (with some caveats) – and the typicality of my age and name is preserved.
There are several ways how to add infinities into account of observer selection – and it is an active field of discussion between astrophysicist with practical consequences like Youngness paradox and BBs, as well as panspermia, the size of the observable universe and some constants fine-tuning.
There are several ways to count observers in the infinite universe
Regions counting. We assume the whole chaotic inflation universe as a very large continuous space and we count densities of different regions. The infinite distribution appears as limit of final distribution when we add more and more space may be with different physical constants.
Time counting. We count how many observers of different types appear in each moment through the whole infinite universe. This creates Youngness paradox.
Concentration counting. We don’t count observers, but instead take the limit of known shares of observers. It is similar to the type of claim as the one that the share of all numbers which can be divided on 3 is one third. Here we need to take into account the sizes of the universes (classical Hubble volumes); the consequence is that the universe with the highest concentration of observers per parsec may be not the most probable as it can turn out to be small. This is important to predict fine-tuning parameters as not-perfect fine-tuning may produce more universes than perfect fine-tuning and we would find ourselves in less stable and less populated universe than theoretical maximum. A good example here is that most of sun mass is not located in its center where density is highest.
Infinite timelines connected with trajectory of some elementary particle as described in the article "Watchers of the Infinity"
No infinities. The number of possible Hubble volumes is countable, as well as the number of possible observers and observations. This means that a finite map is possible which connects observations and worlds and allow to calculate shares of the worlds corresponding to observations. Infinities can reappear here if the whole multiverse would appear many times or infinite times, which potentially allows creative counting of branches, but the basic rate will be set for finite case. This also put the end to the idea of linear immortality as the finite number of observer states means that any life is finite until it starts to repeat or ends.
Empirical research. This approach assumes that a real correct multiverse distribution exists but can’t be learned from the first principles and instead suggests existence of unknown А distribution and tries to guess it from empirical observations as is described in the article “The Xerographic Distribution: Scientific Reasoning in a Large Universe”. The core idea is that the real theory of weighting observers can’t be derived theoretically, but can be measured empirically, and empirically we know that we should discount BBs – and thus don’t weigh all minds equally.
Mutual cancelation of different exotic counting methods. We can imagine an counting method of observers on the multiverse in which the name Aaron Aalexander becomes typical – one such counting methods may be like: every time we find a different name, we search for 10 Aarons Aalexanders (btw, at least one person with such (nick?)- name does exist on Earth). And if this counting method would be valid, I should expect to have such name. However, there are also other counting methods which favors any other exotic or non-exotic name. Such methods mutually cancel-out each other creating some form of noise. For example, remote paths in Feynman’s model of quantum electrodynamic also cancels out each other.
I located in the biggest (observer-wise) universe and thus the effect of other universes is small because they are sparsely populated. First, I can suggest that I live in the type of universe which produces most of the observers. In other words, Earth-like planets around Sun-like stars created by chaotic inflation are the best observers’ producers ever. This also means that the share of other types of observers – like the ones in 4-dimensional space or living on neutron stars or whatever – is minuscule. This means that counting observer’s density in the observable universe is good approximation of total observer density in the multiverse as other regions are much smaller. This would not work if there are several equal size (observer-wise) regions, but existence of the same size regions seems unlikely. This model suffers from circularity as we can’t define the biggest universe without postulating some counting method.
Other approaches that try to kill infinity sampling on its core:
1. No sampling. Often people try to dissolve self-sampling by claiming that there is no sampling at all – I just exist. The problem of this approach is that I can obviously observe my typicality in some aspects (name, date of birth etc) – and I have to add some artificial cut-off and say that this typicality is not applicable to civilization type or its life-duration.
Imagine anti-Boltzmann-brain-situation: all atoms which form Earth and me just collide and form our planets and me. This is physically possible but extremely unlikely as physical laws are time-reversible. And in infinite universe such world should exist. The fact that this is not how we appeared tells us that sampling works all over possible worlds and chose the most probable way of our appearance.
Also, often people say that there is no difference if I am in real world or in simulation, if I am the same, but simulation have different futures distribution: miracles are more likely.
Also, we can assign different measure metrics to different types of minds – let’s say 0.01 to digital minds and 0.00000001 to BBs. This may depend on the energy used for computations. Alternatively, it depends the complexity of the realization of the underlying substrate where biological minds win (???). Anyway, an assumption is needed about the way how to count the weight of minds in different universes – what if they live 4 dimensions or based on non-carbon life.
2. No randomness. If everything exists, there is no true randomness. Illustrative example: numbers after 1 in square root from 2 are not random. But there are things which may be not random but we describe them as random and they behave for us as if they are random.
Something like conclusion
In some sense, the idea of typicality works against the idea of the infinite number of observers, because it says that despite infinite variety of actually existing observers, only some of them should be counted as the real ones: Only physically real observers should be counted and Boltzmannian absurd observer should be severely discounted. Absurd observers exist but should be ignored in our calculations – and it is almost the same as to say that they don’t exist at all.
However, there are situations when such absurd observers become important in our calculations, for example, when we discuss the strange ways of survival in quantum immortality. For example, if I am falling on the Sun, the almost only way how I can survive is a random appearance nearby of an intergalactic spaceship.
Epistemic: generally agree with S.Armstrong here, but adding some links and ideas
SSA and SIA are designed in the way that they work only in a finite universe. FNС and anthropic decision theory also assume that the number of observers can be countable. Therefore, they are toy examples which are known not to generalize to infinite multiverse.
Why infinite universe is almost certain?
A) Because any time we count a finite number of observers or universes, we can ask – why not one more causally independent universe just pop up into existence the same way as our universe?
B) If we assume SIA (the fact that I exist at all means that many observers exist), it updates with infinite Bayesian power for infinite universe.
С) Different Tegmark’s levels.
The fact that we are rather typical in the checkable ways says that adding infinities doesn't distort our typicality – for example, the world around doesn't look like chaotic Boltzmann brain (BB) experience (with some caveats) – and the typicality of my age and name is preserved.
There are several ways how to add infinities into account of observer selection – and it is an active field of discussion between astrophysicist with practical consequences like Youngness paradox and BBs, as well as panspermia, the size of the observable universe and some constants fine-tuning.
There are several ways to count observers in the infinite universe
Other approaches that try to kill infinity sampling on its core:
1. No sampling. Often people try to dissolve self-sampling by claiming that there is no sampling at all – I just exist. The problem of this approach is that I can obviously observe my typicality in some aspects (name, date of birth etc) – and I have to add some artificial cut-off and say that this typicality is not applicable to civilization type or its life-duration.
Imagine anti-Boltzmann-brain-situation: all atoms which form Earth and me just collide and form our planets and me. This is physically possible but extremely unlikely as physical laws are time-reversible. And in infinite universe such world should exist. The fact that this is not how we appeared tells us that sampling works all over possible worlds and chose the most probable way of our appearance.
Also, often people say that there is no difference if I am in real world or in simulation, if I am the same, but simulation have different futures distribution: miracles are more likely.
Also, we can assign different measure metrics to different types of minds – let’s say 0.01 to digital minds and 0.00000001 to BBs. This may depend on the energy used for computations. Alternatively, it depends the complexity of the realization of the underlying substrate where biological minds win (???). Anyway, an assumption is needed about the way how to count the weight of minds in different universes – what if they live 4 dimensions or based on non-carbon life.
2. No randomness. If everything exists, there is no true randomness. Illustrative example: numbers after 1 in square root from 2 are not random. But there are things which may be not random but we describe them as random and they behave for us as if they are random.
Something like conclusion
In some sense, the idea of typicality works against the idea of the infinite number of observers, because it says that despite infinite variety of actually existing observers, only some of them should be counted as the real ones: Only physically real observers should be counted and Boltzmannian absurd observer should be severely discounted. Absurd observers exist but should be ignored in our calculations – and it is almost the same as to say that they don’t exist at all.
However, there are situations when such absurd observers become important in our calculations, for example, when we discuss the strange ways of survival in quantum immortality. For example, if I am falling on the Sun, the almost only way how I can survive is a random appearance nearby of an intergalactic spaceship.