The main inspiration for my ideas here has been Stuart Armstrong and Anders Sandberg’s paper “Eternity in six hours: Intergalactic spreading of intelligent life and sharpening the Fermi paradox”. It’s main point is to argue that intergalactic colonization of an intelligent civilization is highly feasible in a cosmic timescale, and to discuss the implications of this on the Fermi paradox. In doing so, it also discusses a particular method intergalactic colonization can occur: it argues a single starting solar system has enough materials and energy to directly send a probe to every reachable galaxy, and in turn each probe can self-replicates and sends probes to every star in that galaxy. While thinking through this scenario, I decided that there’s a more efficient and more plausible method that intergalactic colonization can occur. This does not substantially affect Armstrong and Sandbergs’ main points about the Fermi paradox. While re-examining this paper I found it responded to Robin Hanson’s paper “Burning the Cosmic Commons: Evolutionary Strategies for Interstellar Colonization”, which in many respects is closer to my picture of intergalactic colonization, and by all rights should have been an inspiration for me.
Armstrong and Sandberg were careful justifying their assumptions on the technological capability of an intelligent species, and trying to make their results robust to conservative technological assumptions. My approach is more optimistic — roughly speaking, anything that appears to be physically possible I assume to be technologically possible with enough research time. A good follow-up question to my proposal to figure out the exact technological requirements and their feasibility.
In Armstrong and Sandberg’s strategy, a single probe is created at the starting solar system and sent directly to a target galaxy. It spends most of its journey — a vast amount of time — completely inert. This is wasteful. Instead, the probe should spend that time gathering resources from the surrounding space while remaining in motion. It should use those resources to self-replicate while in transit rather than when it reaches a target star and galaxy. That way, it will be able to settle an entire galaxy at one rather than make a time-consuming second hop to colonize the galaxy. Though now there is no reason for a probe’s target to be exactly one galaxy. Instead, single probe now targets a cone-shaped region with the starting solar system as the apex.
Even if this method is more efficient, does that matter? If both strategies are more than capable of colonizing the future lightcone, isn’t it just a matter of which method the civilization chooses, rather than which one is “better”? No it is not, because the second stage for the inert probe really adds a serious delay. Imagine people implemented Armstrong and Sandberg’s proposal first, and launched a probe at 99% lightspeed to every reachable galaxy. Then, it takes ten thousand years until someone successfully launches a self-replicating-in-transit probe at the same speed to a particular galaxy. For comparison, Armstrong and Sandberg’s most pessimistic estimate is that will take eleven thousand years to launch every probe, and a more representative estimate is in the paper’s title, six hours. Then the inert probe arrives to the galaxy twenty thousand years earlier, and has time to create and send secondary probes to a twenty thousand light year radius at best. Meanwhile, the self-replicating-in-transit arrives at the entire galaxy at once. If the galaxy is as large as the Milky Way, one hundred thousand light years across, then the active probe gets to most of the galaxy first. The fifty thousand years it takes the inert probe to colonize the rest of the galaxy is tiny from a cosmological perspective, which is why it was ignored in Armstrong and Sandberg, but huge from a human/historical perspective. Since we are comparing two approaches that can both be initiated by the same technological civilization as part of its historical development, it is the historical timescales that are relevant for comparing which approach wins over.
An active probe in transit won’t just gather resources and self-replicate, it can host an entire society. It can host intelligent entities, whether AIs or real or simulated humans, that are thinking and planning for their future. In particular, they can make scientific and technological advantages that will make the probe work better, either decrease its mass or energy requirements or increase its speed. This gives another reason to expect active-in-transit probes to be more successful than inert probes: If in situ research can lead to an early-launched probe accelerating from 99% lightspeed to 99.9% lightspeed then that speed difference will really add up over millions or billions of light years, beating inert probes launched earlier. It will also beat probes launched later at 99.9% lightspeed from the starting solar system due to its head start.
To make reasonable guesses on the behavior of these intelligent entities and their society, we should think about their incentives. Of all these entities, all in motion, whichever one moves fast would have first access to galaxies, stars, interstellar gas, and other natural resources. These resources have energy in a low entropy form, and both the energy and the low entropy are useful for performing computations and self-replicating. However, these resources are (relatively) stationary, and usable energy from the perspective of our fast-moving society must also have a lot of momentum. So the matter in this space must be accelerated to the speed of our society to be usable, with the remaining matter accelerated to an equal and opposite momentum. This opposing momentum will make it more difficult for any later entity from making use of these resources, especially if it’s also trying to move very fast. Moreover, due to transformation law for energy-momentum, the faster you go the more energy in stationary form is necessary to obtain the same amount of usable energy from your mobile perspective. So the faster the first movers are going, the more they’ll need to propel naturally-occurring matter in the opposite direction and make it difficult to use. So there’s a huge first-mover advantage.
This is absolutely terrible.
Really. Very. Bad.
It’s bad because it’s saying a large proportion of the resources in the future lightcone, perhaps most of them, will be burnt for nothing. Specifically, burnt as rocket fuel to make a huge number of rockets which, due to time dilation, only exist for long enough for their inhabitants to figure out how to make the rocket go so fast. I’m not sure what will be left behind after such a process, whether there will be an absolute vacuum or a maximum-entropy heat bath or whether there will still be some sort of usable energy after this entire process. Either way, it will be a terrible loss. This is what I believe will happen if intergalactic colonization is guided by the incentives of individual colonizers.
 Both estimates from Table 5 of the paper.