In this paper, Anders Sandberg, Stuart Armstrong and Milan M. Cirkovic argue that
If a civilization wants to maximize computation it appears rational to aestivate until the far future in order to exploit the low temperature environment: this can produce a 1030 multiplier of achievable computation.
Later Charles H. Bennett, Robin Hanson, C. Jess Riedel disagree, claiming
In fact, while this assumption may apply in the distant future, our universe today contains vast reservoirs and other physical systems in non-maximal entropy states, and computer-generated entropy can be transferred to them at the adiabatic conversion rate of one bit of negentropy to erase one bit of error. This can be done at any time, and is not improved by waiting for a low cosmic background temperature. Thus aliens need not wait to be active. As Sandberg et al. do not provide a concrete model of the effect they assert, we construct one and show where their informal argument goes wrong.
Who was right?
With all due respect to the first set of authors, I wouldn’t argue with Charles Bennett on the subject of thermodynamics. https://link.springer.com/article/10.1007%2FBF02084158
Aestivation is not sound strategy, as both its assumptions are shaky: Heat death of the universe is not an inevitable outcome: Big Rip now seems more plausible, when growing acceleration will tear apart the universe in like 20 billion years of now. Moreover, the fate of the universe can’t be known for sure until much larger scale physical experiments will be performed, like building galactic size accelerators.
This turns us to the second assumption: the need to maximise computations. Future AI may not just maximise computations, but it has to do it as soon as possible, as it may need the results of computations earlier in order to control the fate of the universe.
The computing resources in one star system are already huge and it’s not clear to me that you need more than that to be certain for all practical purposes about both the fate of the universe and how best to control it.
Besides computational resources one need experimental data and observations. Sabine Hossenfelder wrote somewhere that to discover new physics we probably need accelerators which have energy like 10E15 of LHC. Also to measure changes in the speed of universe acceleration probably billions of years are needed.
Also, the problem of other ETI arises. If one civilisation is sleeping in aestivation, another could come to power and eradicate its sleeping seeds. Thus sleeping seeds should be in fact berserkers, which come to a activity from time to time, check the level of other civilizations (if any) and destroy or upgrade them. It is a rather sinister perspective.
Unimaginably large amounts of theory can often compensate for small amounts of missing empirical data. I can imagine the possibility that all of our current observations truly underdetermine facts about the universe’s future large-scale evolution, but it wouldn’t be my default guess.
For what it’s worth, my intuition agrees that any superintelligence, even if using an aestivation strategy, would leave behind some sort of easily visible side effects, and that there aren’t actually any aestivating aliens out there.
I think an argument could be made that they have left subtle visible effects, and we just haven’t been able to reach consensus that that’s what it is, and one of these days we’re going to correlate the universe’s contents, and when we do, we’re going to be a bit upset.
We don’t seem to be sure what the deal was with oumuamua, and we’re constantly getting reports of what look like alien probes on earth, but we (at least, whatever epistemic network I’m in) can only shrug and say “These things usually aren’t aliens.”
How so? I looked on the web for a defense of Big Rip being more plausible than heat death but couldn’t immediately find it.
One interesting argument against the heat death is that in this case the most probable observers will be Boltzmann brains, but in the universes with “cutoff” (that is the end like 20 billion years from now), real observer should dominate.
That doesn’t sound like it would work in UDT or similar decision theories. Maybe in Heat Death world there’s one me and a thousand Boltzmann brains with other observations (as per the linked post), and in Big Rip world there’s only the one me. If I’m standing outside the universe trying to decide what response to the observation that I’m me would have the best consequences, why shouldn’t I just ignore the Boltzmann brains? (This is just re-arguing the controversy of how anthropics works, I guess, but considered by itself this argument seems strong to me.)
Yes, it is a good point that based on UDT one could ignore BBs.
But it looks like that cosmologists try to use the type of experiences that an observer has to deduce either he is BB or not. Cosmologists assume that BB has very random experience, and that their current experience is not very random. (I find both claims weak.) They conclude that as their current observations are not random, they are not BBs, and thus BBs are not dominating type of the observers. There are several flaws in this logic, one of of them is that a BB can’t make a coherent conclusions if its experiences are random or not.
Also, even if there are worlds, with Big Rip and another with Heat Death with many BBs, SIA favors the heat death world.
However, some BBs could last longer than just one observer-moment and they could have time to think about the type of experience they have. Such BBs more likely to find themselves in the empty universe than in the one full of stars. Also, interesting thing is that some BBs could appear even in Big Rip scenarios by the process called “nucleation”, where they jump out of cosmological horizon of the accelerating universe.
So, my point is that the idea of the heat death of the universe is getting more alternatives after the discovery of the dark energy, and there are some new arguments against it like the ones based on BBs. All this is not enough to conclude now which form of the end on the universe is more probable.