Really interesting article, which should be on arxiv.
A few notes:
I expected to find here a link on the Grace SIA Doomsday argument. She uses the same logic as you, but then turns to the estimation of the probability that Great filter is ahead. It looks like you ignore possible high extinction rate implied by SIA (?). Also, Universal DA by Vilenkin could be mentioned.
Another question, which is interesting for me, is how all this affects the possibility of SETI-attack—sending malicious messages with the speed of light on the intergalactic distance.
The third idea I had related to this is the possibility that “bad fine tuning” of the universe will overweight the expected gain of the civilisation density from SIA. For example, if a universe will be perfectly fine-tuned, every star will have a planet with life; however, it requires almost unbelievable fidelity of its parameters tuning. The more probable is the set of the universes there fine tuning is not so good, and the habitable planets are very rare.
Now the question arises: where should I find myself: in super-fine-tuned universe or in the set of all bad-fined-tunes universes? The answer depends on the number of parameters in the fine-tuning, or the number of dimension of functional space. I will illustrate this by the following example: the biggest density is in the center of the Sun (150 g/cm3), however, an average atom inside the Sun is located rather far from the center on much smaller densities (1.4 g/cm3.), that is, hundred times less.
This is the key possible objection to your argument: the total number of the observers and the total density of observers may not correlate, if we take into account existence of very larger set of universes with very small densities. I think it could be addressed by analysing actual distribution of parameters important for fine-tuning.
I expected to find here a link on the Grace SIA Doomsday argument. She uses the same logic as you, but then turns to the estimation of the probability that Great filter is ahead. It looks like you ignore possible high extinction rate implied by SIA (?). Also, Universal DA by Vilenkin could be mentioned.
Yup, I talk about this in the section Prior probability distribution. SIA definitely predicts doomsday (or anything that prevents space colonisation), so this post only applies to the fraction of possible Earths where the probability of doom isn’t that high. Despite being small, that fraction is interesting to a total consequentialist, since it’s the one where we have the best chance at affecting a large part of the Universe (assuming that our ability to reduce x-risk gets proportionally smaller as the probability of spreading to space goes below 0.01 % or so).
Another question, which is interesting for me, is how all this affects the possibility of SETI-attack—sending malicious messages with the speed of light on the intergalactic distance.
There was a bunch of discussion in the comments of this post about whether SETI would even be necessary to find any ETI that wanted to be seen, given that the ETI would have a lot of resources available to look obvious. At least Paul concluded that it was pretty unlikely that we could have missed any civilisation that wanted to be seen. I think that analysis still stands.
Including the possibility of SETI-attacks in my analysis would mean that no early civiliation could ever develop in an advanced civilisation’s light cone, but the borders between advanced civilisations would still be calculated with the civilisations’ actual expansion speed (with the additional complication that advanced civilisations could ‘jump’ to any early civilisation that appears in their light cone). If we assume that the time left until we become invulnerable to SETI-attacks is negligible (a dangerous assumption?), I think that’s roughly equivalent to the scenario under Visibility of civilisations in Appendix C, from Earth’s perspective.
The third idea I had related to this is the possibility that “bad fine tuning” of the universe will overweight the expected gain of the civilisation density from SIA. For example, if a universe will be perfectly fine-tuned, every star will have a planet with life; however, it requires almost unbelievable fidelity of its parameters tuning. The more probable is the set of the universes there fine tuning is not so good, and the habitable planets are very rare.
If I understand you correctly, this is an argument that our prior probability of fl should be heavily weighted towards life being very unlikely? That could change the conclusion if the prior probability of fl was inversely proportional to fl, or even more extremely tilted towards lower numbers. I don’t see any particular reason why we would be that confident that life is unlikely, though, especially since the relevant probability mass in my analysis already puts fl beneath 10−10. Having a prior that puts 1010 times more probability mass on fl=10−20 than fl=10−10 is very extreme, given the uncertainty about this area.
My objection was that highest population and highest population density don’t necessary correlate even on Earth. For example, in India, the highest population density maybe in Bombey, there around 20 million people live, but most people (1 billion) live in rural areas with lower population density. It means that anthropic reasoning can’t be used to estimate density without some prior consideration of the density distribution and the size of low populated ares.
Hm, I still can’t find a way to interpret this that doesn’t reduce it to prior probability.
Density corresponds to how common life is (?), which is proportional to fl. Then the “size” of an area with a certain density corresonds to the prior probability of a certain fl? Thus, “the total number of people in low density areas is greater than the total number of people in high density areas, because the size of the low density area is so much greater” corresponds to ”p(fl=low)∗low>p(fl=high)∗high, because the prior probability (denoted by p()) of fl=low is so much greater”.
Really interesting article, which should be on arxiv.
A few notes:
I expected to find here a link on the Grace SIA Doomsday argument. She uses the same logic as you, but then turns to the estimation of the probability that Great filter is ahead. It looks like you ignore possible high extinction rate implied by SIA (?). Also, Universal DA by Vilenkin could be mentioned.
Another question, which is interesting for me, is how all this affects the possibility of SETI-attack—sending malicious messages with the speed of light on the intergalactic distance.
The third idea I had related to this is the possibility that “bad fine tuning” of the universe will overweight the expected gain of the civilisation density from SIA. For example, if a universe will be perfectly fine-tuned, every star will have a planet with life; however, it requires almost unbelievable fidelity of its parameters tuning. The more probable is the set of the universes there fine tuning is not so good, and the habitable planets are very rare.
Now the question arises: where should I find myself: in super-fine-tuned universe or in the set of all bad-fined-tunes universes? The answer depends on the number of parameters in the fine-tuning, or the number of dimension of functional space. I will illustrate this by the following example: the biggest density is in the center of the Sun (150 g/cm3), however, an average atom inside the Sun is located rather far from the center on much smaller densities (1.4 g/cm3.), that is, hundred times less.
This is the key possible objection to your argument: the total number of the observers and the total density of observers may not correlate, if we take into account existence of very larger set of universes with very small densities. I think it could be addressed by analysing actual distribution of parameters important for fine-tuning.
Yup, I talk about this in the section Prior probability distribution. SIA definitely predicts doomsday (or anything that prevents space colonisation), so this post only applies to the fraction of possible Earths where the probability of doom isn’t that high. Despite being small, that fraction is interesting to a total consequentialist, since it’s the one where we have the best chance at affecting a large part of the Universe (assuming that our ability to reduce x-risk gets proportionally smaller as the probability of spreading to space goes below 0.01 % or so).
There was a bunch of discussion in the comments of this post about whether SETI would even be necessary to find any ETI that wanted to be seen, given that the ETI would have a lot of resources available to look obvious. At least Paul concluded that it was pretty unlikely that we could have missed any civilisation that wanted to be seen. I think that analysis still stands.
Including the possibility of SETI-attacks in my analysis would mean that no early civiliation could ever develop in an advanced civilisation’s light cone, but the borders between advanced civilisations would still be calculated with the civilisations’ actual expansion speed (with the additional complication that advanced civilisations could ‘jump’ to any early civilisation that appears in their light cone). If we assume that the time left until we become invulnerable to SETI-attacks is negligible (a dangerous assumption?), I think that’s roughly equivalent to the scenario under Visibility of civilisations in Appendix C, from Earth’s perspective.
If I understand you correctly, this is an argument that our prior probability of fl should be heavily weighted towards life being very unlikely? That could change the conclusion if the prior probability of fl was inversely proportional to fl, or even more extremely tilted towards lower numbers. I don’t see any particular reason why we would be that confident that life is unlikely, though, especially since the relevant probability mass in my analysis already puts fl beneath 10−10. Having a prior that puts 1010 times more probability mass on fl=10−20 than fl=10−10 is very extreme, given the uncertainty about this area.
Thanks for the link on the Paul comment.
My objection was that highest population and highest population density don’t necessary correlate even on Earth. For example, in India, the highest population density maybe in Bombey, there around 20 million people live, but most people (1 billion) live in rural areas with lower population density. It means that anthropic reasoning can’t be used to estimate density without some prior consideration of the density distribution and the size of low populated ares.
Hm, I still can’t find a way to interpret this that doesn’t reduce it to prior probability.
Density corresponds to how common life is (?), which is proportional to fl. Then the “size” of an area with a certain density corresonds to the prior probability of a certain fl? Thus, “the total number of people in low density areas is greater than the total number of people in high density areas, because the size of the low density area is so much greater” corresponds to ”p(fl=low)∗low>p(fl=high)∗high, because the prior probability (denoted by p()) of fl=low is so much greater”.