Summary: The Fermi Paradox is no paradox—there really are no aliens out there that we can see. But that’s only because there weren’t many in the past. Now they’re popping up more and more often, and soon the universe will be teeming with them. The earliest ones to harvest their own star’s energy, maybe including us, start expanding at close to the speed of light in all directions, and soon these “Grabby Civilizations” control everything.
The good news: This means humanity is not necessarily doomed to annihilate itself, and we’ll have the opportunity to build a massive enduring civilization. In fact, it’s evidence against the Zoo Hypothesis or the risk of a Great Filter ahead of us. And even if humanity fails to spread, the universe will have many more tries with other civilizations in the future.
This the picture drawn by considering two fascinating papers together. I highly recommend both of these, and they complement each other well. After reading them, thinking about anything else in life just seems small.
The first demonstrates that it’s likely “pretty easy” to expand across a huge number of galaxies from a single origin star, at close to the speed of light. The second argues that advanced life arises more and more often over time, and builds a model showing that the first civilizations to begin expanding are likely to start at nearly the same time, and not see each other until later in the expansion.
Together, these ideas have profound implications on how to think about the Fermi Paradox and what we should expect for humanity’s long-term future. This post will summarize the two papers and extend some further thoughts.
Who Am I?
My qualifications or lack thereof aren’t that important, but I should emphasize I have no particular expertise in this field, and I’ve read very little of the literature. So I’ll own up to likely mistakes, and I’m sure I’m missing other related work. I’d be really interested to hear expert opinion on what’s new in these papers, what is already well-known, and what might be mistaken.
I just happened to come across both of these papers at the same time:
“Eternity in six hours” was mentioned on this episode of the 80000 Hours podcast. Ajeya Cotra sold it well, and it didn’t disappoint.
“Grabby Aliens” was linked from Tyler Cowen’s blog.
Eternity in six hours
This first paper is the easier of the two to understand and to summarize. In brief, the authors argue that it’s likely “pretty easy” to colonize the universe.
A single solar system contains the resources necessary to launch self-replicating probes, at half-light-speed or greater, to every other galaxy that is reachable at that speed (from 100 million to several billion galaxies, depending on speed). Up to 99% of the speed of light is plausible. The time and energy required are minimal at cosmological scales, no more than ~10,000 years with a single star, and perhaps much less. Upon arriving at a destination galaxy, a probe could self-replicate and quickly build a civilization within that galaxy.
The main assumptions in this result are:
Exploratory engineering estimating the order-of-magnitude possibilities in terms of probe size, energy requirements, launch technique, and deceleration technique. A range of possibilities is considered, plausible based on the laws of physics, replicators found in nature, and existing technology.
The level of intergalactic dust—a greater amount than assumed would raise the technological challenge enormously.
The efficiency of rocket designs for decelerating upon reaching the target galaxy.
I’m not qualified to evaluate these assumptions, but if we take the conclusion of the paper at face value, we should expect that if human civilization survives and colonizes the solar system, it can pretty soon afterward colonize the universe at close to the speed of light. Unless someone else got there first!
This makes the Fermi Paradox even more extreme. Not only do we observe no alien life in our galaxy, but also in none of the millions of galaxies that could have launched probes to the Milky Way within the last few billion years. This suggests that if we do manage to launch a universe-colonization program, we are the only civilization to do so across millions of galaxies and billions of years. (Unless the alien probes are hiding from us.)
One addition I would make… The authors show how probes need only be launched from the origin star and coast to the destination galaxy—no waypoints or leaping from galaxy to galaxy needed. However, given that probes effectively lose speed as they coast due to the universe’s expansion, it would seem useful to stop at an intermediate galaxy, build more probes, and launch them further in the same direction—effectively “boosting” back up to whatever fraction of the speed of light is feasible. This would make even more galaxies reachable. (Unless I’m just not understanding general relativity properly.)
Grabby Aliens
As far as I can tell, the key innovations in this paper, under-appreciated in previous work, are:
We should expect that advanced life arises much more often at later times than earlier times—the “advanced civilization formation rate” should rise with time as a power law, tN where N is probably in the range of 3-12. So whatever number of advanced civilizations are arising now, we should expect far fewer arose in the past, and far more will arise in the future.
If it’s possible for a civilization to expand at close to the speed of light (50% or higher), then we’d be unlikely to see that civilization before it reaches us. Either it was too far away and not long enough ago—so its light hasn’t reached us—or it already reached us. The closer to the speed of light, the smaller the time window between arrival of the civilization’s light and arrival of its spaceships.
The paper argues it’s plausible that our civilization could begin such an expansion (i.e. become a “Grabby Civilization”) with not many more steps or much time required, and so the present moment is representative of a time when Grabby Civilizations could form. This doesn’t mean that we definitely will become Grabby. It’s entirely consistent with this model that 999,999,999 out of every billion civilizations die at our stage. The universe could be filled with civilizations like ours that are born, then stagnate or die. This death rate would delay the explosion of Grabby Civilizations into the universe, but eventually the tN formation rate would catch up and produce the explosion.
I found the basic arguments very compelling. There aren’t many assumptions needed to derive the Grabby Civilizations expansion model, and they seem likely true.
I’ll begin with my interpretation of the implications of the paper, extending a bit beyond the paper. I think the key implications follow from the key points above, rather than the specific assumptions of the model implemented in the paper. Then I’ll go back to describe and review the paper.
Implications of #1: Advanced Life Formation Rate Increasing Over Time
The implications of point #1 (increasing rate of advanced life formation) are profound. We might naively expect that advanced life is equally likely to evolve at any point in earth’s history. But assuming that many different extremely unlikely steps must occur to form advanced life, we’re much more likely to arise late in the planet’s history (and indeed we did). The same is true for other planets, so we expect many more civilizations to arise late than early.
This means that at first, new advanced life basically never appears, then suddenly it explodes in a very short time, then it’s popping up everywhere (unless the earlier advanced life expands and snuffs out the later emergences). It only stops when the universe becomes inhospitable to new life (stars dying out, universe expanded into black nothingness).
The paper argues, in Section 4, that the universe seems to have at least as much “habitable” time after the present moment as existed before, and perhaps much more. So given the much greater rate later, we seem to be in the first sub-percent of civilizations of our type that will ever exist. That is, we’re remarkably early!
The fact that we’re early makes me optimistic for two reasons:
If human civilization fails to reach its potential (say we annihilate ourselves), the universe will have a lot more tries with later civilizations. Even if we’re currently alone, our failure doesn’t mean that the universe will be permanently sterile.
It argues against two distasteful solutions to the Fermi Paradox:
Great Filter Hypothesis: If civilizations always annihilate themselves shortly after our present stage, a random civilization like ours should be somewhere in the middle percentiles of formation dates. But we’re super-early!
Zoo Hypothesis: If our region of the universe is controlled by a civilization that is letting “primitive” civilizations like ours form naturally, the same argument applies—we should be in the middle of the formation date distribution, but we’re super-early. (Though another version of the Zoo Hypothesis, where “primitive” civilizations are seeded rather than forming naturally, isn’t necessarily ruled out by this argument.)
To relate this back to “Eternity in six hours”, the power-law rate of advanced life formation reduces the sting of the Fermi Paradox implications in that paper. Most of the galaxies that could have sent probes to our galaxy by now would have had to send them billions of years ago, when the rate of advanced life emergence was much lower. So the Fermi Paradox is basically “correct” to think that we don’t see other alien life, but in the Grabby Aliens model that’s because that life is only beginning to emerge at our present moment. We don’t see it in our past, but we will definitely see it in our future!
Implications of #2: Unlikely to See a Civilization That Expands Close to the Speed of Light
The paper simulates the formation and expansion of “Grabby Civilizations” (the ones that succeed in expanding before being swallowed by another Grabby Civilization), and determines it’s very likely that a Grabby Civilization won’t see any others before it begins expanding. That is, if humanity will eventually become a GC, our experience of not seeing any evidence of other life is typical of GCs. The apparent absence of other life is no reason to expect that human civilization doomed to fail.
The paper assumes essentially 4 pieces of “data”:
Acceleration in advanced life formation rate (power law)
We exist now
There are no hard steps left for us to become a Grabby Civilization. We might totally fail, but if we succeed we will succeed very soon.
We don’t see evidence of any Grabby Civilizations, either present here and now, or as light in part of the sky.
Taken together, the paper derives that this means most Grabby Civilizations are being born around the present moment (via #1-3), and they expand at greater than half the speed of light (via #4). This latter point is supported by “Eternity in six hours”. I’d add that the other possible reason for #4 is that Grabby Civilizations aren’t clearly visible—in that case they might have filled the universe long ago.
I’ve seen other posts suggesting this expansion close to the speed of light is scary, since by the time we see another Grabby Civilization in the sky, it is almost upon us, about to snuff us out. But I think that’s extremely unlikely. It’s likely very little time, at cosmological scales, to go from our current stage to beginning a Grabby expansion. It would be an extraordinary coincidence for another GC to hit us right now. In fact, the paper’s model produces probability distributions indicating we’re unlikely to meet another GC for hundreds of millions of years. By then, we’ll have the tools to fight back against invaders.
More Detail on the Grabby Aliens Paper
“A Simple Model of Grabby Aliens” is brilliantly innovative in its core ideas, and its core model is brilliantly simple, with just 3 parameters. (One implementation of the core model is just 165 lines of code!) And it’s incredibly awe-inspiring to picture civilizations popping up all at once and bubbling out to fill all space, each absorbing millions or billions of galaxies—here’s a video.
I’d recommend starting by reading the Conclusion (Section 13) - which states the key points in a very clear and accessible way.
The remainder of this post goes into detail on specific sections of the paper. It might be useful if you decide to read the paper (which I recommend!), but feel free to stop reading this post here.
Section 2 (The Hard-Steps Model) has a good description and justification of the hard-steps model that leads to the power-law probability of advanced life formation. It’s about as easy to follow as this topic can be (with the exception of Equation (1), which is very difficult). Another paper, The Timing of Evolutionary Transitions Suggests Intelligent Life Is Rare, seems to be a more detailed treatment.
Section 4 (How We Seem Early) shows that across a range of plausible parameter estimates on star formation rates and habitable duration of stars, life on earth seems among the earliest to arise in the universe. Almost all parameter values put us in the first 10%, and many put us in the first 1%. Interestingly, both “Grabby Aliens” and “Timing of Evolutionary Transitions...” cite the long lifetime of M dwarf stars and their potentially long period of habitability. But while “Timing of Evolutionary Transitions...” assumes life on earth is average (not early) and thereby argues M dwarf stars are uninhabitable, “Grabby Aliens” goes the opposite direction. It takes the long potentially habitable lifetime of M dwarf stars as evidence that life on earth is very early.
Section 5 (Model Rationale) provides good qualitative justification of the assumption that some civilizations will choose to initiate an expansion across the universe. It also argues well that such civilizations are likely to be quite visible and noticeable, though I think this second claim is not quite as strong, as there are potential arguments for a civilization remaining invisible even after expansion.
Section 6 (The Model) is a clear explanation of the model, which as mentioned before is quite simple.
The approach to setting the time and distance scales in the model is pretty hard to follow. To be honest, I don’t think I fully understand Section 7 (Heuristic 1D Model) or the rescaling described in Section 9 (Simulating the Model).
The treatment of the expanding universe in Section 8 (Cosmology) was hard to follow, and confusing that they followed a power-law expansion rather than a more accurate expansion model. The authors acknowledge this approximation and compare to the real expansion of the universe in Figure 16. I think that the results aren’t especially sensitive to this—it’s only important to get it right in the time period when Grabby Civilizations are emerging and expanding, before that and after that don’t matter. I also found it odd that the expansion model seems to suggest a civilization could spread to arbitrary long distances given enough time, whereas my understanding of the universe’s expansion is that more distant galaxies would be unreachable. Perhaps this is allowed in the power-law approximation for the expansion, but not in the actual expected expansion? I’d need to know more about general relativity and cosmology to evaluate this.
Optimistic Resolution of the Fermi Paradox: Eternity in Six Hours & Grabby Aliens
Summary: The Fermi Paradox is no paradox—there really are no aliens out there that we can see. But that’s only because there weren’t many in the past. Now they’re popping up more and more often, and soon the universe will be teeming with them. The earliest ones to harvest their own star’s energy, maybe including us, start expanding at close to the speed of light in all directions, and soon these “Grabby Civilizations” control everything.
The good news: This means humanity is not necessarily doomed to annihilate itself, and we’ll have the opportunity to build a massive enduring civilization. In fact, it’s evidence against the Zoo Hypothesis or the risk of a Great Filter ahead of us. And even if humanity fails to spread, the universe will have many more tries with other civilizations in the future.
This the picture drawn by considering two fascinating papers together. I highly recommend both of these, and they complement each other well. After reading them, thinking about anything else in life just seems small.
Eternity in six hours: Intergalactic spreading of intelligent life and sharpening the Fermi paradox (2013, Stuart Armstrong, Anders Sandberg)
A Simple Model of Grabby Aliens (2021, Robin Hanson, Daniel Martin, Calvin McCarter, Jonathan Paulson) [website]
The first demonstrates that it’s likely “pretty easy” to expand across a huge number of galaxies from a single origin star, at close to the speed of light. The second argues that advanced life arises more and more often over time, and builds a model showing that the first civilizations to begin expanding are likely to start at nearly the same time, and not see each other until later in the expansion.
Together, these ideas have profound implications on how to think about the Fermi Paradox and what we should expect for humanity’s long-term future. This post will summarize the two papers and extend some further thoughts.
Who Am I?
My qualifications or lack thereof aren’t that important, but I should emphasize I have no particular expertise in this field, and I’ve read very little of the literature. So I’ll own up to likely mistakes, and I’m sure I’m missing other related work. I’d be really interested to hear expert opinion on what’s new in these papers, what is already well-known, and what might be mistaken.
I just happened to come across both of these papers at the same time:
“Eternity in six hours” was mentioned on this episode of the 80000 Hours podcast. Ajeya Cotra sold it well, and it didn’t disappoint.
“Grabby Aliens” was linked from Tyler Cowen’s blog.
Eternity in six hours
This first paper is the easier of the two to understand and to summarize. In brief, the authors argue that it’s likely “pretty easy” to colonize the universe.
A single solar system contains the resources necessary to launch self-replicating probes, at half-light-speed or greater, to every other galaxy that is reachable at that speed (from 100 million to several billion galaxies, depending on speed). Up to 99% of the speed of light is plausible. The time and energy required are minimal at cosmological scales, no more than ~10,000 years with a single star, and perhaps much less. Upon arriving at a destination galaxy, a probe could self-replicate and quickly build a civilization within that galaxy.
The main assumptions in this result are:
Exploratory engineering estimating the order-of-magnitude possibilities in terms of probe size, energy requirements, launch technique, and deceleration technique. A range of possibilities is considered, plausible based on the laws of physics, replicators found in nature, and existing technology.
The level of intergalactic dust—a greater amount than assumed would raise the technological challenge enormously.
The efficiency of rocket designs for decelerating upon reaching the target galaxy.
I’m not qualified to evaluate these assumptions, but if we take the conclusion of the paper at face value, we should expect that if human civilization survives and colonizes the solar system, it can pretty soon afterward colonize the universe at close to the speed of light. Unless someone else got there first!
This makes the Fermi Paradox even more extreme. Not only do we observe no alien life in our galaxy, but also in none of the millions of galaxies that could have launched probes to the Milky Way within the last few billion years. This suggests that if we do manage to launch a universe-colonization program, we are the only civilization to do so across millions of galaxies and billions of years. (Unless the alien probes are hiding from us.)
One addition I would make… The authors show how probes need only be launched from the origin star and coast to the destination galaxy—no waypoints or leaping from galaxy to galaxy needed. However, given that probes effectively lose speed as they coast due to the universe’s expansion, it would seem useful to stop at an intermediate galaxy, build more probes, and launch them further in the same direction—effectively “boosting” back up to whatever fraction of the speed of light is feasible. This would make even more galaxies reachable. (Unless I’m just not understanding general relativity properly.)
Grabby Aliens
As far as I can tell, the key innovations in this paper, under-appreciated in previous work, are:
We should expect that advanced life arises much more often at later times than earlier times—the “advanced civilization formation rate” should rise with time as a power law, tN where N is probably in the range of 3-12. So whatever number of advanced civilizations are arising now, we should expect far fewer arose in the past, and far more will arise in the future.
If it’s possible for a civilization to expand at close to the speed of light (50% or higher), then we’d be unlikely to see that civilization before it reaches us. Either it was too far away and not long enough ago—so its light hasn’t reached us—or it already reached us. The closer to the speed of light, the smaller the time window between arrival of the civilization’s light and arrival of its spaceships.
The paper argues it’s plausible that our civilization could begin such an expansion (i.e. become a “Grabby Civilization”) with not many more steps or much time required, and so the present moment is representative of a time when Grabby Civilizations could form. This doesn’t mean that we definitely will become Grabby. It’s entirely consistent with this model that 999,999,999 out of every billion civilizations die at our stage. The universe could be filled with civilizations like ours that are born, then stagnate or die. This death rate would delay the explosion of Grabby Civilizations into the universe, but eventually the tN formation rate would catch up and produce the explosion.
I found the basic arguments very compelling. There aren’t many assumptions needed to derive the Grabby Civilizations expansion model, and they seem likely true.
I’ll begin with my interpretation of the implications of the paper, extending a bit beyond the paper. I think the key implications follow from the key points above, rather than the specific assumptions of the model implemented in the paper. Then I’ll go back to describe and review the paper.
Implications of #1: Advanced Life Formation Rate Increasing Over Time
The implications of point #1 (increasing rate of advanced life formation) are profound. We might naively expect that advanced life is equally likely to evolve at any point in earth’s history. But assuming that many different extremely unlikely steps must occur to form advanced life, we’re much more likely to arise late in the planet’s history (and indeed we did). The same is true for other planets, so we expect many more civilizations to arise late than early.
This means that at first, new advanced life basically never appears, then suddenly it explodes in a very short time, then it’s popping up everywhere (unless the earlier advanced life expands and snuffs out the later emergences). It only stops when the universe becomes inhospitable to new life (stars dying out, universe expanded into black nothingness).
The paper argues, in Section 4, that the universe seems to have at least as much “habitable” time after the present moment as existed before, and perhaps much more. So given the much greater rate later, we seem to be in the first sub-percent of civilizations of our type that will ever exist. That is, we’re remarkably early!
The fact that we’re early makes me optimistic for two reasons:
If human civilization fails to reach its potential (say we annihilate ourselves), the universe will have a lot more tries with later civilizations. Even if we’re currently alone, our failure doesn’t mean that the universe will be permanently sterile.
It argues against two distasteful solutions to the Fermi Paradox:
Great Filter Hypothesis: If civilizations always annihilate themselves shortly after our present stage, a random civilization like ours should be somewhere in the middle percentiles of formation dates. But we’re super-early!
Zoo Hypothesis: If our region of the universe is controlled by a civilization that is letting “primitive” civilizations like ours form naturally, the same argument applies—we should be in the middle of the formation date distribution, but we’re super-early. (Though another version of the Zoo Hypothesis, where “primitive” civilizations are seeded rather than forming naturally, isn’t necessarily ruled out by this argument.)
To relate this back to “Eternity in six hours”, the power-law rate of advanced life formation reduces the sting of the Fermi Paradox implications in that paper. Most of the galaxies that could have sent probes to our galaxy by now would have had to send them billions of years ago, when the rate of advanced life emergence was much lower. So the Fermi Paradox is basically “correct” to think that we don’t see other alien life, but in the Grabby Aliens model that’s because that life is only beginning to emerge at our present moment. We don’t see it in our past, but we will definitely see it in our future!
Implications of #2: Unlikely to See a Civilization That Expands Close to the Speed of Light
The paper simulates the formation and expansion of “Grabby Civilizations” (the ones that succeed in expanding before being swallowed by another Grabby Civilization), and determines it’s very likely that a Grabby Civilization won’t see any others before it begins expanding. That is, if humanity will eventually become a GC, our experience of not seeing any evidence of other life is typical of GCs. The apparent absence of other life is no reason to expect that human civilization doomed to fail.
The paper assumes essentially 4 pieces of “data”:
Acceleration in advanced life formation rate (power law)
We exist now
There are no hard steps left for us to become a Grabby Civilization. We might totally fail, but if we succeed we will succeed very soon.
We don’t see evidence of any Grabby Civilizations, either present here and now, or as light in part of the sky.
Taken together, the paper derives that this means most Grabby Civilizations are being born around the present moment (via #1-3), and they expand at greater than half the speed of light (via #4). This latter point is supported by “Eternity in six hours”. I’d add that the other possible reason for #4 is that Grabby Civilizations aren’t clearly visible—in that case they might have filled the universe long ago.
I’ve seen other posts suggesting this expansion close to the speed of light is scary, since by the time we see another Grabby Civilization in the sky, it is almost upon us, about to snuff us out. But I think that’s extremely unlikely. It’s likely very little time, at cosmological scales, to go from our current stage to beginning a Grabby expansion. It would be an extraordinary coincidence for another GC to hit us right now. In fact, the paper’s model produces probability distributions indicating we’re unlikely to meet another GC for hundreds of millions of years. By then, we’ll have the tools to fight back against invaders.
More Detail on the Grabby Aliens Paper
“A Simple Model of Grabby Aliens” is brilliantly innovative in its core ideas, and its core model is brilliantly simple, with just 3 parameters. (One implementation of the core model is just 165 lines of code!) And it’s incredibly awe-inspiring to picture civilizations popping up all at once and bubbling out to fill all space, each absorbing millions or billions of galaxies—here’s a video.
I’d recommend starting by reading the Conclusion (Section 13) - which states the key points in a very clear and accessible way.
The remainder of this post goes into detail on specific sections of the paper. It might be useful if you decide to read the paper (which I recommend!), but feel free to stop reading this post here.
Section 2 (The Hard-Steps Model) has a good description and justification of the hard-steps model that leads to the power-law probability of advanced life formation. It’s about as easy to follow as this topic can be (with the exception of Equation (1), which is very difficult). Another paper, The Timing of Evolutionary Transitions Suggests Intelligent Life Is Rare, seems to be a more detailed treatment.
Section 4 (How We Seem Early) shows that across a range of plausible parameter estimates on star formation rates and habitable duration of stars, life on earth seems among the earliest to arise in the universe. Almost all parameter values put us in the first 10%, and many put us in the first 1%. Interestingly, both “Grabby Aliens” and “Timing of Evolutionary Transitions...” cite the long lifetime of M dwarf stars and their potentially long period of habitability. But while “Timing of Evolutionary Transitions...” assumes life on earth is average (not early) and thereby argues M dwarf stars are uninhabitable, “Grabby Aliens” goes the opposite direction. It takes the long potentially habitable lifetime of M dwarf stars as evidence that life on earth is very early.
Section 5 (Model Rationale) provides good qualitative justification of the assumption that some civilizations will choose to initiate an expansion across the universe. It also argues well that such civilizations are likely to be quite visible and noticeable, though I think this second claim is not quite as strong, as there are potential arguments for a civilization remaining invisible even after expansion.
Section 6 (The Model) is a clear explanation of the model, which as mentioned before is quite simple.
The approach to setting the time and distance scales in the model is pretty hard to follow. To be honest, I don’t think I fully understand Section 7 (Heuristic 1D Model) or the rescaling described in Section 9 (Simulating the Model).
The treatment of the expanding universe in Section 8 (Cosmology) was hard to follow, and confusing that they followed a power-law expansion rather than a more accurate expansion model. The authors acknowledge this approximation and compare to the real expansion of the universe in Figure 16. I think that the results aren’t especially sensitive to this—it’s only important to get it right in the time period when Grabby Civilizations are emerging and expanding, before that and after that don’t matter. I also found it odd that the expansion model seems to suggest a civilization could spread to arbitrary long distances given enough time, whereas my understanding of the universe’s expansion is that more distant galaxies would be unreachable. Perhaps this is allowed in the power-law approximation for the expansion, but not in the actual expected expansion? I’d need to know more about general relativity and cosmology to evaluate this.