author: niplav, created: 2023-07-06, modified: 2023-07-26, language: english, status: in progress, importance: 5, confidence: probable
Hi Anders
The aim is to develop solid arguments about these questions, such as bounds on how much computation could be done given up-to-date physical theory or game-theoretic arguments that civilizations meeting certain assumptions would behave in certain ways when they met.
Perhaps change to:
The aim is to develop solid arguments about these questions, such as bounds on how much computation could be done given up-to-date physical theory, or game-theoretic arguments that civilizations meeting certain assumptions would behave in certain ways if they met.
Performing more than a google (10100) computations
I think it is supposed to be "googol" instead of "google".
Ord and Cotton-Barrett defined “An existential eucatastrophe is an event which causes there to be much more expected value after the event than before.” [684]²
The citation 684 is for "Gravitationally redshifted absorption lines in the x-ray burst spectra of a neutron star", but I think it should be 685, "Existential risk and existential hope: definitions". This off-by-one pattern repeats itself throughout my copy of the book.
The usage of the term civilization in this book needs to be disambiguated. Civilization can be used to denote complex societies with features such as irrigation, urban areas, social stratification, symbolic systems of communication, division of labour etc.
Perhaps use quotation marks around "civilization" here—this feels like a mention more than a use.
I will be using the term civilization not in the narrow high culture sense but to denote as a cohesive, long-range (social) structure with a high degree of coordination
Same with quotation here, also the "as" feels stumbly.
Near-term is more predictable and hence conclusions are more likely to be true
I don't know of a good resource that firmly establishes this. Dillon 2020 and my own investigation don't come to firm conclusions, mostly because of a dearth of good long-term forecasting data. Muehlhauser 2019 also laments the lack of long-term forecasts that are precise enough to be checked. I think later in the book there will be talk about why less accurate forecasts should be expected in dynamical systems, and I look forward to that discussion.
I haven't yet done a full review of that field yet, and my guess is that there are much better references out there (maybe even on my hard drive).
While this book certainty is not intended as science fiction, it may well find use as a source of raw material for science fiction authors.
Typo, should be "certainly".
[Stanislaw Lem — Summa Technologiae, comments on prognostication]
There is a work by Lem which I have only been able to find references to in German. It's called "Das Katastrophenprinzip: Die Welt as Holocaust" (Stanisław Lem, 1983). It gives an interesting perspective on predictability, stating that the universe is less predictable in the short term than in the long term. This feels related to notions of sophistication from complexity theory, which try to capture the property of being complex but not random.
But for having a liveable society with an open future dynamism looks better.
This sentence tripped me up, twice.
In terms of macrohistory, one (self-described) amateur attempt is Muehlhauser 2017a and Muehlhauser 2017b, which find that along 5 different metrics, human well-being has been improving increasingly rapidly since the industrial revolution.
Also relevant to 1.2.2.
A pessimistic possibility is that the set of people skilled in core competences is so small compared to the overall population today that many key technologies would be lost in a fall of civilization. An optimistic counterargument may be that in this case many flock to the frarmer or smith to learn their skills.
For an investigation into such global collapse at current civilizational levels causing human extinction, see Rodriguez 2020. It broadly finds that recovery appears likely in scenarios where less than 99.9% of all humans die, and finds extinction unlikely unless more than 99.999% of all humans die.
although underground storage aims at storage for
$1 \times 10^5 yr$
Word storage is repeated, perhaps use another term? Maybe use "safety for" instead of "storage for".
La tombe du soldat inconnu in Paris has been guarded and with a lit eternal flame since 1920.
There is a large amount of "eternal" flames around the world. I don't know of any dataset that estimates the age of such eternal flames, and their failure rates.
Maybe add the 60 year knuckle-cracking experiment.
This is not merely large constructions such a Bering Strait bridge or a trans-oceanic subway in the world
Shouldn't it be "such as a Bering Strait bridge"?
I.J. Rabi
Should be 'I.I. Rabi'.
It dealt with not just eugenics ([…]) but that human evolution would become a result of cultural or political decisions.
The part after the parentheses might benefit from two or three more words, "but also with the hypothesis that […]".
and be compact enough for a rapid exchanged of in- formation among them.
Two typos (I think), I guess it should be "a rapid exchange of information among them."
More branches?
Perhaps elaborate on distinction between transhumanism/posthumanism/extropianism, somewhere.
I like $r$
more for radial distances.
Loss of biodiversity might be irreversible and regrettable, yet that permanent loss might not mean a permanent loss of potential.
I personally do feel a sense of loss at languages, species, cultures & ecosystems going extinct. (I think it's sad trilobites & opabinia regalis with its five eyes are gone, and would like to revive them, even though this conflicts with wild-animal suffering concerns.) Similarly, projections that by 2050 >90% of all languages will have gone extinct seem like a tragedy to me, and I like the fact that there are (some) language revitalization and language documentation projects underway.
many, many worlds worlds
Typo
Finding the right priorization strategy is non-trivial. One heuristic is nick Bostrom's MaxiPOK principle “ “maximize the chance of an OK outcome”.
Typo: after "principle" there are two quotation marks.
[Other strategies] [Priority. Targeted vs broad interventions. Diff tech development. Maxipok. Global public good. Safety beats haste.]
MaxiPOK was already mentioned & explained. For differential development there's also differential intellectual progress, as explained in “Differential Intellectual Progress as a Positive-Sum Project” (Brian Tomasik, 2017).
[?, 232] [Trajectory changes. Planning?]
First reference is broken.
If the total amount of work done is proportional to the number of members and the fraction of time not spent on meetings, then it is
$T_{work}=c_1N(1-c_2 N)$
.
What do $c_1$
and $c_2$
stand for?
but hard for e.g. space when not easy enforcement
Now I'm wondering how spacetime ownership might look like, especially around black holes (and how that spacetime might be taxed in a Georgist fashion).
Even one implemented as an autonomous AGI can have a corrigible value function that it does not resist changing [2700, 2218].
Corrigibility remains (afaik) fundamentally unsolved in the toy cases (at least one that is demanded in Soares et al. 2015, so perhaps a "probably" or "may" would be good here.
However, if these differences can be cached out in normatively relevant unfairness, political power, or social scarce resourcse such as status (see box ), then this may be an issue that may motivate distributive actions. However, redistributing goods may be less important than (if possible) influencing the fairness, power or status.
(Emphases mine)
Some quibbles: "However" is written twice, as is "may" in the first sentence. And there is a stray space after "box".
badly made poetry
I like the image of someone haphazardly slapping together some poetry with their clumsy hands.
Details 15 ("Space solar EROI"), footnote a: what does $S$
stand for?
Footnote 30 links to footnote 31, which links…nowhere?
A civilization that wants to use more energy on Earth without heating it up need to increase the emissivity or albedo throguh surface changes³² or geoengineering, or add an external cooling system for more extreme uses
"needs" would read more smoothly, methinks.
If
$M(r) \propto r^{-α} α > 1$
I think there should be a comma between $r^{-α}$
and $α > 1$
.
and its logarithm the the learning coefficient
This is especially true since once the system has learned the task it might be possible to copy its performance (having it act as an ANI or generate an ANI with the skill
$^{51}$
).
Some of the literature on distilling larger neural networks into smaller ones might be relevant here, but I haven't read any of it.
People recognizes that “life will find a way”
Should be "recognize"
Perhaps somewhat surprising there are no quantum self-replicators. The no cloning theorem means that copying arbitrary quantum states is impossible. A corollary is that there is no universal quantum constructor since it would be able to clone itself, a “no self replication theorem”.
Probably should start with "Perhaps somewhat surprisingly".
If the no cloning theorem says that arbitrary quantum states can't be copied, can't there be a subspace of quantum states that is capable of replication, similar to how arbitrary hard SAT problems are computationally intractable but in practice we can solve real-world instances just fine?
Or is the issues with the universal quantum constructor, since such a replicator wouldn't be able to create arbitrary quantum states?
and New Zealand was likely settled a founder population with between 50 and 100 women
Missing "by": "settled by a founder population with […]"
It can be argued that the heterogeneity of services requires human-level artificial intelligence, but a vast number of services can be performed by more narrow systems adequately (making the human-provided version a valuable luxury in many cases: small talk with the hairdresser may soon be the actual reason to pay the premium).
I somewhat balk at the specific example, as the job of a hair dresser (from a robotic perspective, though I'm no roboticist) seems to me to be one of the hardest professions to automate (hair is non-rigid & highly non-linear & malfunctioning machinery would be dangerously close to a human neck, and the wages aren't high enough for there to be significant pressure for automation). I'd maybe pick legal advice & a charismatic lawyer instead, or medical advice and empathy (though both of these examples suffer from the fact that these two professions have far more legal protection).
Happiness has a strong genetic component, ranging from 30-80% of the variation depending on study
Perhaps this should be "the study".
Meta-analysis show that positive psychology interventions can have medium (standardized mean difference 0.3) effects for subjective well-being
If this is a plural, I believe the plural of "analysis" is "analyses".
$μ$
-opoid
Missing 'i'.
The introspective training required may preclude widespread use at present
$^{94}$
, but future neurotechnologies may well simplify achieving it.
Might be worth referencing some of the people working on this, e.g. the semalab. More information on them in “Spirit Tech” (Wesley J. Wildman/Kate J. Stockly, 2021) ch. 2 — unfortunately they don't publish that much.
There is also “Transcranial Focused Ultrasound for Modulation of Attention Networks in Humans” (Maria Elizabeth Fini, 2020), but I have read very little of it.
Also potentially worth mentioning is Jhourney? Although I'm not sure about the policy you have about mentioning startups and speculative business ventures in an academic book.
Maybe add a remark on how energy efficient pleasure vs. suffering are? See Shulman 2012 for some pontification. Or maybe put in the abolitionism section?
In many psychological experiments subjective magnitude of a stimuli typically grows as the logarithm (Fechner's law) or a power-law (Steven's law) with the objective intensity of the stimuli
If I'm not mistaken the singular of "stimuli" is "stimulus".
One can also argue that the process is more about understanding and explanations, the opposite of compression: the formulation of forward models that allow predictions.
I like this point. Pithier: "everything is compression" is wrong, it ought to be "everything is compression and execution".
of real world-data
Shouldn't this be "of real-world data" instead?
The cubic (or quadratic?) limit to growth also means that if there are limits to growth or scientific knowledge, their end might arrive comparatively suddenly this growth starts asymptoting.
While buying and selling may be reversible operations¹²⁷ there is a need to perform error correction and there might even be desirable to ensure that the transaction is not too readily undone.
"and there might even be desirable" -> "and it might even be desirable" or some other modification.
a civilization (or society, or institution) collapses or transforms if it no longer can or want to pursue any multigenerational project that it used to regard as highly valuable.
"want" should be "wants".
Footnote 3 should have a period at the end.
Garrett then equates the absorption ability with accumulated global economic value, arriving at an energy consumption equal to λC where C is amount of accumulated value and λ=(9.7±0.3)mW per 1990 US dollar
"is amount of" should be "is the amount of".
Recognizing that the lifespan is set by the overall energy release (L) and releasable energy (0.007 Mc²) gives the information.
What information exactly?
as they accumulate, the number of rules needed seem to diminish
Emphasis mine.
There also appears to be a bias in people towards adding rather removing things when improving a situation.
Missing "than" between "rather" and "removing".
but it might also be that civilizations on average tend to make artefacts and processes that in make retrodiction far easier than in the natural state
"that in make retrodiction far easier" is either confusingly written or grammatically incorrect.
The Anthropocene is (maybe) Hthe current geological age
Typo.
human-generated patters
Typo, should be "patterns"
Figure 6.2: Where 10 000 should be there is only 10 on the chart.
Footnote 42: "but as far as I know none of the children have returned" reads like they're still in Antarctica, and haven't returned to the countries they're citizens of, while the opposite is the case.
Footnote 48: "Hf B_2" should be "Hf B₂", I reckon
In particular, methane poisoning 3.6 gya, the oxygen catastrophe 2.8 gya, snowball earth 2.3 gya and 790-630 mya […]
Shouldn't "mya" and "gya" be "Mya" and "Gya", for consistency's sake?
This contrasts to the fear/hope that change may go in *any* direction
I assume the asterisks are supposed to make "any" italic.
[The fit with demandite, ]
Maybe explain what demandite is, Wikipedia doesn't contain that word.
far more habitable than any planetary body could be expected tot be.
Typo, emphasis mine.
Hence settling space for land is not a radical improvement over Earth.
In terms of raw mass
Practical habitat design is somewhat outside the topic of this book.
Stray sentence fragment. Also I can't believe anything is outside the topic of this book.
while at 90 atm pressure on Venus is challenging³⁹
Missing period at the end of the sentence.
To approach relativistic velocities the barrel need to be longer than 100 AU
"need" should be "needs".
The most fundamental differences are that the environment between galaxies appear far more dust-free than inside galaxies
Should be "appears" instead of "appear".
As the settlement wave progresses initial conditions matters less as it approaches an asymptotic form.
although most of the effect is is mild heating
Repeated word.
This happens for
$n_0=10^{-6} cm^{-3}$
in 1060 Gyr.
Should this be $10^{60}$
or $1060$
Gyr?