I have been extremely confused about why anthropics is treated the way it is, I am asking for clarification. I will first explain my current position:
There is no such thing as an “observer.” The problem with anthropic reasoning is the same as the problem with the Copenhagen interpretation of quantum mechanics, namely, it invokes non-real magical properties of human minds which I thought we had safely dissolved at this point.
Take your pick of the various anthropic assumptions. Each of them treats the “observer” as some kind of ontologically granular spirit, and that there are somehow equal odds that you might have found your observer-spirit situated in the head of any given human throughout history and into the future. (Setting aside the unjustified assertion that my observer-spirit can inhabit human skulls but not animal skulls or distant alien skulls.) This sense that we have of ourselves as embodied discrete entities is an illusion, a holographic accident of physics.
You could not have been anyone but what you are—namely, one specific dynamic pattern of matter and energy embedded within physics.
If you’ve lived on an island your whole life with a hundred fellow tribespeople, you would not be right in saying, “There probably aren’t any people on any other islands, because if there were a lot of other people, the odds would be really small that I was born on this island.” I don’t know the name for this error in thinking, but I’m sure it has one.
Perhaps I’m just profoundly misunderstanding the point of anthropics.
You could not have been anyone but what you are—namely, one specific dynamic pattern of matter and energy embedded within physics.
So imagine that post-singularity scientists find evidence of a Parallel Earth, but aren’t able to measure or observe it directly. After a bunch of work, they come up with a device that can interact with the parallel world, but only in a very specific way: it connects to the mind of a single intelligent life-form there, and transmits its thoughts and sensations back to you in a sort of “virtual reality”. Only problem is they don’t know how to aim it: as far as anyone knows, it samples randomly across all space and time when choosing its subject. Also, they only have enough funding to use the device once.
So they aim it and they end up in the body of some guy. The tech level around him seems to be approximately medieval, and he seems to be speaking a Tocharian language. Unfortunately, just then some barbarians show up and kill him, and the device explodes.
What information can we glean about Parallel Earth from this experiment? Well, we know that at least one person spoke Tocharian there in medieval times. But it’s unlikely that just one person speaks Tocharian there—if one person spoke Tocharian, and everyone else speaks (let’s say) Basque, then it would be hugely improbable that the random device would have chosen the one Tocharian speaker. Because our random sampling device chose a Tocharian speaker, we have some evidence that Tocharian is probably one of the more common languages on Parallel Earth.
We can go even further. Our subject lived in medieval times, we live in post-singularity times and observe a galactic population of five hundred trillion. If Parallel Earth also experiences a singularity with a population of five hundred trillion, and we are genuinely selecting at random from everyone who ever lived, it would be extremely weird for our machine to randomly select one of the (let’s say) 500 million people in ancient times as opposed to the 500 trillion people in post-singularity times: in fact, the chances are only (500 mil/500 tril) = (1/1 million). This provides Bayesian evidence that Parallel Earth humanity managed to destroy itself before reaching a singularity.
Anthropics is just the belief that just as I can draw inferences from some machine placing me a randomly selected Parallel Earthling’s body, so I can draw inferences from blind luck placing me in a randomly selected Real Earthling’s body.
I’m curious to know where you disagree here. Do you think the scientists shouldn’t draw inferences based on their device selecting that ancient Tocharian-speaker, or do you think they are justified in their assumptions but present-day anthropic reasoners are not?
1) The first objection, which I hint at in my original post, is that the selection of a reference class of human beings seems to be selected specifically to make whatever point the anthropic reasoner is trying to make. Why don’t I have an equal likelihood of being
any available 1.4 kg oblate lump of matter, including rocks and large jellyfish, or
any animal possessing more than 100 neurons, or
any human being who is aware of the concept of anthropic reasoning, or
any sentient being capable of self-awareness …
Obviously each of these reference classes generate completely different answers. You would guess that, e.g. for each reference class,
most 1.4 kg lumps of matter must be brains; or
that most animals possessing more than 100 neurons must be humans; or
that the concept of anthropic reasoning or the human race dies out pretty soon; or
that there must not be any other self-conscious life forms in the universe.
In other words, there is no good reason to pretend that human consciousness works anything like a virtual-parallel-Earth device rather than being like any other conceivable reference class.
2) The second objection is, again, that the existence of “observers” in the first place is an illusion. You have a 100% likelihood of being you because you are identically you (minus some small increment allowing for insanity, etc.). We aren’t souls injected into bodies from heaven. We are matter that thinks it has identity.
Let me start with the second question, since I think I have a little more of a clue how to answer it.
Anthropics doesn’t really rely on you being you. You being you is just...I guess I could call it a convenient Schelling point. We’ve got to choose someone to do anthropics on. Suppose we chose Genghis Khan. We could say that Genghis was a conquering warlord, so therefore most people throughout history must be conquering warlords. But this would fail, because the only reason we selected Genghis Khan to begin with was that he was a conquering warlord.
But this selection bias is inherent in anything we try. If we were to deliberately select some random peasant from Khan’s era to do anthropics on, so that we avoided that first bias, we would be biasing ourselves towards peasants, biasing ourselves toward people who weren’t important, biasing ourselves towards people from the past, and biasing ourselves to Earthlings.
The fact that you are necessarily you is part of why anthropics works. If we were souls who chose bodies at the moment of birth, I couldn’t condition on my own existence in 2012, because my soul might have been really excited at the prospect to go into one of those super-rare presingularity bodies. As it is, I know I have no selection bias in selecting myself with my specific characteristics, because I did not select myself or my personal characteristics. So I can look at those characteristics—white human male born in 1984 - and consider them a random sample of the characteristics of all people everywhere and everywhen, and do anthropics on them. The fact that the person who is a suitable random sample for anthropics also happens to be me is probably overemphasized, but I don’t think it’s that important.
And I think this also goes part of the way to solving your first objection. We can’t do anthropics on my brain as a representative sample of all 1.4 kg lumps of matter, because it’s getting selected specifically as a 1.4 kg lump of matter that is especially interesting to me—most 1.4 kg lumps of matter were disqualified before they even had a chance to be the one we’re doing anthropics on.
I admit I am still quite confused on how this works in more complicated scenarios—see this post for the same argument in the opposite direction.
I have been extremely confused about why anthropics is treated the way it is, I am asking for clarification. I will first explain my current position:
There is no such thing as an “observer.” The problem with anthropic reasoning is the same as the problem with the Copenhagen interpretation of quantum mechanics, namely, it invokes non-real magical properties of human minds which I thought we had safely dissolved at this point.
Take your pick of the various anthropic assumptions. Each of them treats the “observer” as some kind of ontologically granular spirit, and that there are somehow equal odds that you might have found your observer-spirit situated in the head of any given human throughout history and into the future. (Setting aside the unjustified assertion that my observer-spirit can inhabit human skulls but not animal skulls or distant alien skulls.) This sense that we have of ourselves as embodied discrete entities is an illusion, a holographic accident of physics.
You could not have been anyone but what you are—namely, one specific dynamic pattern of matter and energy embedded within physics.
If you’ve lived on an island your whole life with a hundred fellow tribespeople, you would not be right in saying, “There probably aren’t any people on any other islands, because if there were a lot of other people, the odds would be really small that I was born on this island.” I don’t know the name for this error in thinking, but I’m sure it has one.
Perhaps I’m just profoundly misunderstanding the point of anthropics.
So imagine that post-singularity scientists find evidence of a Parallel Earth, but aren’t able to measure or observe it directly. After a bunch of work, they come up with a device that can interact with the parallel world, but only in a very specific way: it connects to the mind of a single intelligent life-form there, and transmits its thoughts and sensations back to you in a sort of “virtual reality”. Only problem is they don’t know how to aim it: as far as anyone knows, it samples randomly across all space and time when choosing its subject. Also, they only have enough funding to use the device once.
So they aim it and they end up in the body of some guy. The tech level around him seems to be approximately medieval, and he seems to be speaking a Tocharian language. Unfortunately, just then some barbarians show up and kill him, and the device explodes.
What information can we glean about Parallel Earth from this experiment? Well, we know that at least one person spoke Tocharian there in medieval times. But it’s unlikely that just one person speaks Tocharian there—if one person spoke Tocharian, and everyone else speaks (let’s say) Basque, then it would be hugely improbable that the random device would have chosen the one Tocharian speaker. Because our random sampling device chose a Tocharian speaker, we have some evidence that Tocharian is probably one of the more common languages on Parallel Earth.
We can go even further. Our subject lived in medieval times, we live in post-singularity times and observe a galactic population of five hundred trillion. If Parallel Earth also experiences a singularity with a population of five hundred trillion, and we are genuinely selecting at random from everyone who ever lived, it would be extremely weird for our machine to randomly select one of the (let’s say) 500 million people in ancient times as opposed to the 500 trillion people in post-singularity times: in fact, the chances are only (500 mil/500 tril) = (1/1 million). This provides Bayesian evidence that Parallel Earth humanity managed to destroy itself before reaching a singularity.
Anthropics is just the belief that just as I can draw inferences from some machine placing me a randomly selected Parallel Earthling’s body, so I can draw inferences from blind luck placing me in a randomly selected Real Earthling’s body.
I’m curious to know where you disagree here. Do you think the scientists shouldn’t draw inferences based on their device selecting that ancient Tocharian-speaker, or do you think they are justified in their assumptions but present-day anthropic reasoners are not?
I have two separate objections.
1) The first objection, which I hint at in my original post, is that the selection of a reference class of human beings seems to be selected specifically to make whatever point the anthropic reasoner is trying to make. Why don’t I have an equal likelihood of being
any available 1.4 kg oblate lump of matter, including rocks and large jellyfish, or
any animal possessing more than 100 neurons, or
any human being who is aware of the concept of anthropic reasoning, or
any sentient being capable of self-awareness …
Obviously each of these reference classes generate completely different answers. You would guess that, e.g. for each reference class,
most 1.4 kg lumps of matter must be brains; or
that most animals possessing more than 100 neurons must be humans; or
that the concept of anthropic reasoning or the human race dies out pretty soon; or
that there must not be any other self-conscious life forms in the universe.
In other words, there is no good reason to pretend that human consciousness works anything like a virtual-parallel-Earth device rather than being like any other conceivable reference class.
2) The second objection is, again, that the existence of “observers” in the first place is an illusion. You have a 100% likelihood of being you because you are identically you (minus some small increment allowing for insanity, etc.). We aren’t souls injected into bodies from heaven. We are matter that thinks it has identity.
Let me start with the second question, since I think I have a little more of a clue how to answer it.
Anthropics doesn’t really rely on you being you. You being you is just...I guess I could call it a convenient Schelling point. We’ve got to choose someone to do anthropics on. Suppose we chose Genghis Khan. We could say that Genghis was a conquering warlord, so therefore most people throughout history must be conquering warlords. But this would fail, because the only reason we selected Genghis Khan to begin with was that he was a conquering warlord.
But this selection bias is inherent in anything we try. If we were to deliberately select some random peasant from Khan’s era to do anthropics on, so that we avoided that first bias, we would be biasing ourselves towards peasants, biasing ourselves toward people who weren’t important, biasing ourselves towards people from the past, and biasing ourselves to Earthlings.
The fact that you are necessarily you is part of why anthropics works. If we were souls who chose bodies at the moment of birth, I couldn’t condition on my own existence in 2012, because my soul might have been really excited at the prospect to go into one of those super-rare presingularity bodies. As it is, I know I have no selection bias in selecting myself with my specific characteristics, because I did not select myself or my personal characteristics. So I can look at those characteristics—white human male born in 1984 - and consider them a random sample of the characteristics of all people everywhere and everywhen, and do anthropics on them. The fact that the person who is a suitable random sample for anthropics also happens to be me is probably overemphasized, but I don’t think it’s that important.
And I think this also goes part of the way to solving your first objection. We can’t do anthropics on my brain as a representative sample of all 1.4 kg lumps of matter, because it’s getting selected specifically as a 1.4 kg lump of matter that is especially interesting to me—most 1.4 kg lumps of matter were disqualified before they even had a chance to be the one we’re doing anthropics on.
I admit I am still quite confused on how this works in more complicated scenarios—see this post for the same argument in the opposite direction.
I’m skeptical of many of the conclusions of “anthropic” reasoning, but I don’t think it can be rejected out of hand.
“Cogito ergo sum” seems like a valid argument, after all.