According to standard models of cosmology, there will be habitable planets in our universe for a very long time, and some of these planets will be much longer-lived than our Earth. If we expect the emergence of intelligent life to require multiple unlikely hard steps to accidentally happen in a row (as is a common model, usually assuming between 3 and 10 hard steps), then it should be very surprising that we find ourselves on a relatively short-lived planet early in the history of the universe. There are planets that are likely habitable for hundreds of billions of years - the chance that 3 to 10 unlikely steps happen in their long lifetime is overwhelmingly more likely than that it happens in the short lifespan of our Earth.[1] Hanson’s Grabby Aliens paper does some modeling, and almost all their models show that we should be among the first 0.1% of intelligent species to emerge.
So what explains being so early? Hanson’s answer is that some grabby alien civilizations start spreading through the galaxies with near-lightspeed, and conquer everyone in their way. Therefore, there will be no unconquered land left hundreds of billions of years from now, so intelligent life can’t develop on the long-lived planets, so Earth is actually a typical example of a cradle of civilization - you need to develop early, before the grabby aliens arrive at your planet.
Hanson also uses this as an argument that aliens likely exist - maybe you previously thought that life is rare enough that we are alone in the universe. But then it would be a huge mystery why we are not living on a long-lived planet in the far-future. The best explanation for our earliness is the existence of grabby aliens, so aliens very likely exist.
I think this is not a bad argument, but I have two objections, which make me think that I should only bet at around 30% that we will ever encounter another powerful alien civilization in the observable universe.
We are not that early if weighted by cosmic endowment
There are two possible theories about the rarity of intelligent life.
Theory A says that it’s extremely hard for intelligent life to emerge, to the point where we should expect most observable universe regions of similar size to ours to be completely empty. However, in the vast quantum multiverse, sometimes strong enough coincidences happen that intelligent life still emerges in a universe. By the anthropic principle, we find ourselves in one such universe. We shouldn’t expect any other intelligent life in the observable universe - it would be a crazy coincidence for two extremely unlikely events to happen within a universe, when in expectation the event happens much less than once.
Theory B says that the emergence of intelligent life is rare, but not that rare. We don’t need to invoke the anthropic principle to explain why our universe has life - we should expect it to evolve intelligent life many times, and we are one such instance.
One can rephrase the grabby aliens argument this way:
Suppose Alice believes in Theory A and Bob believes in Theory B. Suppose they have both read half an astronomy book, and learned everything that scientists now think about cosmology (how frequent various types of stars are, how long planets might remain habitable, how the expansion of the universe works). However, neither of them have ever looked at the sky, and they've never read the other half of the astronomy book, so they have no idea what kind of planet they live in, at which point in the Universe’s history.
Before reading the other half of the book, they take a bet.
Based on calculations similar to the ones in the Grabby Aliens paper, Alice bets with over 99.9% confidence that we live on a long-lived planet later in the history of the universe - after all, those are much more likely to evolve intelligent life according to the model.
Meanwhile, Bob thinks it’s pretty likely we live fairly early - everything gets conquered by grabby aliens later.
They read the textbook and they learn they live on Earth, barely 13 billion years after the Big Bang, on a relatively short-lived planet. Alice loses very badly.
Bayesian updating is analogous to theories Kelly-betting with their influence points, so this should be an argument for Theory A losing almost all its influence in our minds.
However, I believe Alice was betting wrongly. If Alice scope-sensitively cares about making the world better[2], then she should care much more about making money (and using that money to make the world X% better) in cases where the world she can affect is bigger.
(Stepping out of the metaphor: The existence of aliens might be strategically relevant for some decisions we make now. I care more about being right in the worlds where the affectable world is bigger, so I will take that into account.)
This means that when Alice makes a bet about how early we are in the history of the universe, she should weight her probability distribution by the cosmic endowment we should expect to conquer if we are at that particular point in our universe's history.
(For a further intuition pump on this, see my recent shortform on whether you should feel surprised to be unusually influential.)
According to our current cosmological models, the accelerating expansion of the universe is making more and more of the universe inaccessible at an exponential rate. If we start spreading with near-lightspeed today, 90 billion galaxies might be reachable. If we only start spreading 50 billion years from now, only 14 million galaxies can be reached - more than a thousand times fewer. In 100-150 billion years, we will get locked into a pocket forever with just the Milky Way and Andromeda and a few smaller galaxies.[3] Everything else will be forever unreachable.
The modeling in the Grabby Aliens paper suggests that, in the absence of alien conquests, we should expect the vast majority of instances of intelligent life to emerge in the far future. But most of that is after everyone gets locked into tiny pockets. When Alice makes a bet before reading the astronomy textbook, she is not going to bet on living in a tiny pocket - she cares about money much more in the worlds where humanity’s destiny is to conquer 90 billion galaxies than when it is to conquer two.
This is (according to Claude) a representation of the mainline model that appears in the Grabby Aliens paper, with the assumption that the emergence of intelligent life takes 3 hard steps, which is within the range that the paper considers.[5]
You can see that the vast majority of intelligent life is supposed to arise in the far future, but once we weight civilizations by their importance (how much of the universe they can conquer), our position is not that atypical - we are 10th percentile early. And if you use a bit different assumptions about the habitability of red dwarfs, it becomes even less surprising that we live this early.
So I think a scope-sensitive altruist shouldn’t update heavily from the fact that we live early in the history of the universe.
Aliens require a somewhat narrow range of parameters
How easy is it for intelligent life to emerge? If we only looked at Earth, and didn’t know anything about the rest of the cosmos, I think it would have been a plausible guess that life is quite simple, evolution is pretty strong, and most planets evolve intelligent life within a few billion years.
It would also be (and still is) a plausible guess that intelligent life is in fact crazy hard to emerge, and the emergence of self-replicating life that is able to build up complex organisms requires a monkeys-typing-Shakespeare level coincidence of a lot of chemicals accidentally falling in the right place at once and forming a sufficiently capable self-reproducing system.
Theory A says that intelligent life requires a crazy coincidence, and most universes don’t have intelligent life. According to Claude, this means that if the probability of life emerging within a billion years on a given planet is below , then Theory A is true.
Meanwhile, we know that no intelligent life emerged in the past few billion years in our few billion light-years neighborhood, otherwise someone would have already conquered us.[6][7] Claude tells me that this means that the probability of life emerging in a billion years on any given planet is less than .
So there is this important parameter, the probability of intelligent life emerging on a random planet in a billion years. A priori, I had huge uncertainty over its value - it can be close to 1 or super-astronomically small. It also feels intuitive that I should mostly think of having a natural-looking distribution over the logarithm of its value - how many bits of coincidence need to happen for intelligent life to emerge.
Theory A is equivalent to saying that the required bits of coincidence are at least 90[8], while Theory B pins it down to the range between 80 and 90.
It’s not crazy for Theory B to be true, but playing with some models that fit some intuitive numbers I guessed, Claude tells me that Theory B only gets about 10% probability in the distribution.
Counter-point: What would the Presumptuous philosopher say?
Note: This section will be somewhat esoteric and might be hard to understand without having read my sequence on Probabilities and Infinities.
Even if we take all the arguments above into account, shouldn’t we assume that we live in a universe that has a lot of civilizations similar to us in it? If the wide Tegmark IV multiverse has both universes where the emergence of life is easy and where it is very hard, most civilizations will emerge in the universes where it is easy. Therefore, it’s more likely that we are in a universe teeming with life rather than being alone.
(This is the classic Presumptuous philosopher argument under the self-indicating assumption theory of anthropics.)
However, I once again think that this is not the correct way to reason for a scope-sensitive altruist. When making bets about meeting aliens, you should weight your probabilities by how important each scenario is, which is proportionate to how much territory you can conquer.
The expected amount of territory conquered by civilizations with similar values to mine is equal under the scenario where there are a thousand civilizations each conquering 1/1000 of the Universe, or one civilization conquering all of it. So if the Presumptuous Philosopher is scope-sensitive, he shouldn’t actually prefer one hypothesis over the other.
There is a harder question, however. Theory A doesn’t posit that every universe will have one civilization which will conquer it. It posits that most universes in the quantum multiverse are devoid of intelligent life, and only a tiny fraction of them evolves a civilization. Therefore, one might argue that from a scope-sensitive perspective, Theory B has much higher importance than Theory A which assumes that most of the multiverse is empty anyway, so we should act like Theory B was true.
Note that the same argument can also be used to argue that theories in physics that require the parameters of the universe to be finetuned for life must be disregarded too. According to the finetuning hypothesis, most of the possible parameter space is covered by dead matter, so we can only affect a tiny fraction, so the hypothesis is negligible from a scope-sensitive perspective. But I intuitively think that the finetuning argument is pretty reasonable, so this argument is pretty suspicious.
Ultimately, I think that reasonable theories of infinite ethics would mostly reject this argument, and weight theories by something like “what fraction of living beings we can affect if they are true” and not “what fraction of all matter we can affect”. So one theory predicting that most of the universe is dead matter we can't possibly affect shouldn't count as a big strike against the theory.
I believe UDASSA would probably come to a similar conclusion,[9] and my attempted synthesis on probabilities and infinities also only gives a mild preference to Theory B over Theory A based on this argument.
I currently think that humanity’s earliness doesn’t provide a strong reason to bet on the existence of intelligent aliens in the observable universe, though according to some of the mainline models, we are still a bit surprisingly early. Meanwhile, for aliens to exist but not to have conquered us yet, some parameters need to be in a somewhat narrow range.[10] And the Presumptuous philosopher argument only provides a mild update in favor of having aliens in the observable universe.
I think that the argument from aliens requiring a narrow range of parameters is more robust than the calculation showing that we are somewhat early even in the importance-weighted distribution, and the presumptuous philosopher can’t outweigh this difference.
So altogether I would only bet at 30% implied probability of ever meeting other advanced alien civilizations. Obviously, this number is pretty unstable, but I at least feel fairly confident that importance-weighting weakens the earliness argument enough that I won’t update back to thinking that I should bet at over 90%[11] that we will meet aliens.
If the emergence of intelligent life takes n hard steps, then the probability that intelligent life emerges on a planet by time scales with . So if T is at least 3, then longer-lived planets have a vastly better chance for evolving life.
General note: The ideas in this post come from me, but the cosmology facts and the calculations mostly come from Claude, and I only did weak sanity-checks on them.
More precisely, we know that at most a few emerged, otherwise probably there would have been one that emerged early enough and close enough to reach us and which decided to go grabby.
One might object: how do we know they haven’t conquered us yet? How do we know that we don’t just live in a nature preserve, and the night sky with the stars is not just a dome built around the Earth? My answer is the same as for the versions of the simulation hypothesis that don’t assume we will take part in some important acausal trade affecting the outside world. These theories might be likely by some definitions of probability, but from a scope-sensitive perspective, they don’t really matter compared to the worlds where our destiny is to conquer a huge chunk of the universe.
The universes with intelligent life will probably have a lot of interesting stuff in them which will make them have shorter description-length than the dead universes. So most of the reality fluid is distributed among living universes, so the percentage of living matter you control is a better approximation of what UDASSA cares about than the percentage of all matter. But UDASSA is pretty cursed in many ways and I don’t really endorse it.
The grabby aliens argument
According to standard models of cosmology, there will be habitable planets in our universe for a very long time, and some of these planets will be much longer-lived than our Earth. If we expect the emergence of intelligent life to require multiple unlikely hard steps to accidentally happen in a row (as is a common model, usually assuming between 3 and 10 hard steps), then it should be very surprising that we find ourselves on a relatively short-lived planet early in the history of the universe. There are planets that are likely habitable for hundreds of billions of years - the chance that 3 to 10 unlikely steps happen in their long lifetime is overwhelmingly more likely than that it happens in the short lifespan of our Earth.[1] Hanson’s Grabby Aliens paper does some modeling, and almost all their models show that we should be among the first 0.1% of intelligent species to emerge.
So what explains being so early? Hanson’s answer is that some grabby alien civilizations start spreading through the galaxies with near-lightspeed, and conquer everyone in their way. Therefore, there will be no unconquered land left hundreds of billions of years from now, so intelligent life can’t develop on the long-lived planets, so Earth is actually a typical example of a cradle of civilization - you need to develop early, before the grabby aliens arrive at your planet.
Hanson also uses this as an argument that aliens likely exist - maybe you previously thought that life is rare enough that we are alone in the universe. But then it would be a huge mystery why we are not living on a long-lived planet in the far-future. The best explanation for our earliness is the existence of grabby aliens, so aliens very likely exist.
I think this is not a bad argument, but I have two objections, which make me think that I should only bet at around 30% that we will ever encounter another powerful alien civilization in the observable universe.
We are not that early if weighted by cosmic endowment
There are two possible theories about the rarity of intelligent life.
Theory A says that it’s extremely hard for intelligent life to emerge, to the point where we should expect most observable universe regions of similar size to ours to be completely empty. However, in the vast quantum multiverse, sometimes strong enough coincidences happen that intelligent life still emerges in a universe. By the anthropic principle, we find ourselves in one such universe. We shouldn’t expect any other intelligent life in the observable universe - it would be a crazy coincidence for two extremely unlikely events to happen within a universe, when in expectation the event happens much less than once.
Theory B says that the emergence of intelligent life is rare, but not that rare. We don’t need to invoke the anthropic principle to explain why our universe has life - we should expect it to evolve intelligent life many times, and we are one such instance.
One can rephrase the grabby aliens argument this way:
Suppose Alice believes in Theory A and Bob believes in Theory B. Suppose they have both read half an astronomy book, and learned everything that scientists now think about cosmology (how frequent various types of stars are, how long planets might remain habitable, how the expansion of the universe works). However, neither of them have ever looked at the sky, and they've never read the other half of the astronomy book, so they have no idea what kind of planet they live in, at which point in the Universe’s history.
Before reading the other half of the book, they take a bet.
Based on calculations similar to the ones in the Grabby Aliens paper, Alice bets with over 99.9% confidence that we live on a long-lived planet later in the history of the universe - after all, those are much more likely to evolve intelligent life according to the model.
Meanwhile, Bob thinks it’s pretty likely we live fairly early - everything gets conquered by grabby aliens later.
They read the textbook and they learn they live on Earth, barely 13 billion years after the Big Bang, on a relatively short-lived planet. Alice loses very badly.
Bayesian updating is analogous to theories Kelly-betting with their influence points, so this should be an argument for Theory A losing almost all its influence in our minds.
However, I believe Alice was betting wrongly. If Alice scope-sensitively cares about making the world better[2], then she should care much more about making money (and using that money to make the world X% better) in cases where the world she can affect is bigger.
(Stepping out of the metaphor: The existence of aliens might be strategically relevant for some decisions we make now. I care more about being right in the worlds where the affectable world is bigger, so I will take that into account.)
This means that when Alice makes a bet about how early we are in the history of the universe, she should weight her probability distribution by the cosmic endowment we should expect to conquer if we are at that particular point in our universe's history.
(For a further intuition pump on this, see my recent shortform on whether you should feel surprised to be unusually influential.)
According to our current cosmological models, the accelerating expansion of the universe is making more and more of the universe inaccessible at an exponential rate. If we start spreading with near-lightspeed today, 90 billion galaxies might be reachable. If we only start spreading 50 billion years from now, only 14 million galaxies can be reached - more than a thousand times fewer. In 100-150 billion years, we will get locked into a pocket forever with just the Milky Way and Andromeda and a few smaller galaxies.[3] Everything else will be forever unreachable.
The modeling in the Grabby Aliens paper suggests that, in the absence of alien conquests, we should expect the vast majority of instances of intelligent life to emerge in the far future. But most of that is after everyone gets locked into tiny pockets. When Alice makes a bet before reading the astronomy textbook, she is not going to bet on living in a tiny pocket - she cares about money much more in the worlds where humanity’s destiny is to conquer 90 billion galaxies than when it is to conquer two.
Claude made this nice illustration:[4]
This is (according to Claude) a representation of the mainline model that appears in the Grabby Aliens paper, with the assumption that the emergence of intelligent life takes 3 hard steps, which is within the range that the paper considers.[5]
You can see that the vast majority of intelligent life is supposed to arise in the far future, but once we weight civilizations by their importance (how much of the universe they can conquer), our position is not that atypical - we are 10th percentile early. And if you use a bit different assumptions about the habitability of red dwarfs, it becomes even less surprising that we live this early.
So I think a scope-sensitive altruist shouldn’t update heavily from the fact that we live early in the history of the universe.
Aliens require a somewhat narrow range of parameters
How easy is it for intelligent life to emerge? If we only looked at Earth, and didn’t know anything about the rest of the cosmos, I think it would have been a plausible guess that life is quite simple, evolution is pretty strong, and most planets evolve intelligent life within a few billion years.
It would also be (and still is) a plausible guess that intelligent life is in fact crazy hard to emerge, and the emergence of self-replicating life that is able to build up complex organisms requires a monkeys-typing-Shakespeare level coincidence of a lot of chemicals accidentally falling in the right place at once and forming a sufficiently capable self-reproducing system.
Theory A says that intelligent life requires a crazy coincidence, and most universes don’t have intelligent life. According to Claude, this means that if the probability of life emerging within a billion years on a given planet is below
Meanwhile, we know that no intelligent life emerged in the past few billion years in our few billion light-years neighborhood, otherwise someone would have already conquered us.[6] [7] Claude tells me that this means that the probability of life emerging in a billion years on any given planet is less than
So there is this important parameter, the probability of intelligent life emerging on a random planet in a billion years. A priori, I had huge uncertainty over its value - it can be close to 1 or super-astronomically small. It also feels intuitive that I should mostly think of having a natural-looking distribution over the logarithm of its value - how many bits of coincidence need to happen for intelligent life to emerge.
Theory A is equivalent to saying that the required bits of coincidence are at least 90[8], while Theory B pins it down to the range between 80 and 90.
It’s not crazy for Theory B to be true, but playing with some models that fit some intuitive numbers I guessed, Claude tells me that Theory B only gets about 10% probability in the distribution.
Counter-point: What would the Presumptuous philosopher say?
Note: This section will be somewhat esoteric and might be hard to understand without having read my sequence on Probabilities and Infinities.
Even if we take all the arguments above into account, shouldn’t we assume that we live in a universe that has a lot of civilizations similar to us in it? If the wide Tegmark IV multiverse has both universes where the emergence of life is easy and where it is very hard, most civilizations will emerge in the universes where it is easy. Therefore, it’s more likely that we are in a universe teeming with life rather than being alone.
(This is the classic Presumptuous philosopher argument under the self-indicating assumption theory of anthropics.)
However, I once again think that this is not the correct way to reason for a scope-sensitive altruist. When making bets about meeting aliens, you should weight your probabilities by how important each scenario is, which is proportionate to how much territory you can conquer.
The expected amount of territory conquered by civilizations with similar values to mine is equal under the scenario where there are a thousand civilizations each conquering 1/1000 of the Universe, or one civilization conquering all of it. So if the Presumptuous Philosopher is scope-sensitive, he shouldn’t actually prefer one hypothesis over the other.
There is a harder question, however. Theory A doesn’t posit that every universe will have one civilization which will conquer it. It posits that most universes in the quantum multiverse are devoid of intelligent life, and only a tiny fraction of them evolves a civilization. Therefore, one might argue that from a scope-sensitive perspective, Theory B has much higher importance than Theory A which assumes that most of the multiverse is empty anyway, so we should act like Theory B was true.
Note that the same argument can also be used to argue that theories in physics that require the parameters of the universe to be finetuned for life must be disregarded too. According to the finetuning hypothesis, most of the possible parameter space is covered by dead matter, so we can only affect a tiny fraction, so the hypothesis is negligible from a scope-sensitive perspective. But I intuitively think that the finetuning argument is pretty reasonable, so this argument is pretty suspicious.
Ultimately, I think that reasonable theories of infinite ethics would mostly reject this argument, and weight theories by something like “what fraction of living beings we can affect if they are true” and not “what fraction of all matter we can affect”. So one theory predicting that most of the universe is dead matter we can't possibly affect shouldn't count as a big strike against the theory.
I believe UDASSA would probably come to a similar conclusion,[9] and my attempted synthesis on probabilities and infinities also only gives a mild preference to Theory B over Theory A based on this argument.
For a more detailed explanation, see the Finetuning and the Presumptuous philosopher section of my last post.
Conclusion
I currently think that humanity’s earliness doesn’t provide a strong reason to bet on the existence of intelligent aliens in the observable universe, though according to some of the mainline models, we are still a bit surprisingly early. Meanwhile, for aliens to exist but not to have conquered us yet, some parameters need to be in a somewhat narrow range.[10] And the Presumptuous philosopher argument only provides a mild update in favor of having aliens in the observable universe.
I think that the argument from aliens requiring a narrow range of parameters is more robust than the calculation showing that we are somewhat early even in the importance-weighted distribution, and the presumptuous philosopher can’t outweigh this difference.
So altogether I would only bet at 30% implied probability of ever meeting other advanced alien civilizations. Obviously, this number is pretty unstable, but I at least feel fairly confident that importance-weighting weakens the earliness argument enough that I won’t update back to thinking that I should bet at over 90%[11] that we will meet aliens.
If the emergence of intelligent life takes n hard steps, then the probability that intelligent life emerges on a planet by time
Which I do, so imaginary personifications of theories I have should do too.
Apparently, they are gravitationally bound together, so they won’t be broken into even smaller pieces by further inflation.
General note: The ideas in this post come from me, but the cosmology facts and the calculations mostly come from Claude, and I only did weak sanity-checks on them.
If it was 4 hard steps, the mass before the present would be 3.3%.
More precisely, we know that at most a few emerged, otherwise probably there would have been one that emerged early enough and close enough to reach us and which decided to go grabby.
One might object: how do we know they haven’t conquered us yet? How do we know that we don’t just live in a nature preserve, and the night sky with the stars is not just a dome built around the Earth? My answer is the same as for the versions of the simulation hypothesis that don’t assume we will take part in some important acausal trade affecting the outside world. These theories might be likely by some definitions of probability, but from a scope-sensitive perspective, they don’t really matter compared to the worlds where our destiny is to conquer a huge chunk of the universe.
The universes with intelligent life will probably have a lot of interesting stuff in them which will make them have shorter description-length than the dead universes. So most of the reality fluid is distributed among living universes, so the percentage of living matter you control is a better approximation of what UDASSA cares about than the percentage of all matter. But UDASSA is pretty cursed in many ways and I don’t really endorse it.
Alternatively, one can phrase this point as saying that according to Theory B, we are a bit surprisingly late.
Which is what I believed based on the grabby aliens argument before I thought through the question in more detail.