Things usually end slowly
Epistemic status: Low confidence. These findings updated me slightly towards worrying more about slow declines vs. sudden collapse/extinction. Most sources are things I quickly googled so are probably off the mark.
TL;DR It’s sometimes claimed or suggested that if an extinction event were to happen, it would happen very quickly (i.e. days, weeks, months). However, when empires fall or species go extinct, they usually do so very slowly (years, decades, millennia). These base rates suggest we should have a reasonably strong prior on slow decline when thinking about civilisational collapse and, at more of a stretch, existential risks. This sketch is very rough, and I’d love to see someone work out these base rates, and how they might be changing, more carefully.
Thinkers in and around the EA community sometimes claim that x-risk scenarios will cause extinction rapidly. For example:
- Eliezer Yudkowsky argues that the likely scenario with AGI takeoff is “growth stays at 5%/year and then everybody falls over dead in 3 seconds and the world gets transformed into paperclips”
- Toby Ord, in The Precipice (p.28), writes that “As technology continues to advance, new threats appear on the horizon. These threats promise to be more like nuclear weapons than like climate change: resulting from sudden breakthroughs, precipitous catastrophes”
- Relatedly, the metaphor of a precipice implies a very steep descent from stability to extinction.
Perhaps it’s right to expect that an existential catastrophe will take 3 seconds. Your inside view might well take into account things happening awfully fast (such as rapidly increasing compute efficiency or pathogens spreading). But holding this view should require a very strong update from the prior that extinction events and civilisational collapses usually take a very long time.
Data from civilisations
- I expect the way I’m most likely to be wrong about all this is that historical extinctions and civilisational collapses are not appropriate reference classes for x-risks. I think they’re very useful because they’re a rough approximation of x-risks from actual data and I’m sceptical of claims along the lines of “the thing I’m worried about is totally unique and has never happened before”, but I could be wrong.
- While we often refer to the “fall” of Rome and the “collapse” of the Mayans, it’s rarely appropriate to consider human empires and civilisations extinct. In this post, I usually use years when the empire or civilisation is considered by scholars to have “ended.” The specific year, or whether they “ended” at all, is usually up for debate.
To form a rough prior on how long things (civilisations, empires, species) take to decline, I scanned some encyclopedic pages to find:
- The start of the end: the last period in time when the civilisation was still ‘flourishing’, usually just before the start of a collapse or invasion.
- The earliest plausible date and the latest plausible date which could constitute the “end” of the civilization, to form a kind of confidence interval.
This sheet lists some empires, their last points of flourishing, their earliest reasonable dissolution points and their latest reasonable dissolution points. From this, I calculated a lower guess on decline length and an upper guess on decline length (in years). You can see this summary graph below. I expect anyone could review these dates and find points of disagreement, even by reviewing the same sources I did. Furthermore, I’d love to see someone expand this list (perhaps using this useful list compiled by Luke Kemp).
You don’t need to read this data too carefully to notice that rapid dissolution is clearly the exception to the rule.
One interesting thing to note, however, is that declines do seem to be happening faster over time. The data is certainly weak but the trend is also relatively clear.
Why is this? Here are some hypotheses (some of which overlap):
- Empires and nation states are interacting with each other more and more over time.
- Economic growth is accelerating, and this allows challengers to rise much more quickly and defeat the incumbent empire more quickly.
- Many empires end when conquered, and wars happen much more quickly now (I’m not sure why — maybe because planes are faster than walking?)
- The “fitness landscape” of empires changes faster due to accelerating global cultural and technological changes.
Data from species
An obvious response to claims based on the above dataset is that these are all civilizations that are part of humanity and that we should think very differently about humanity itself. But we do have evidence for how long species extinction events usually take.
There have been 5 major extinction events in history, where mass extinction is defined as over 70% of species going extinct (from Our World in Data’s excellent resource on this).
You’ll often see the causes of these extinction events described as “rapid” or “sudden” but it’s worth putting these numbers in context (even the term “event” is somewhat confusing). Most of these mass extinctions occurred over a timescale of over a million years, with some claims of more “rapid” extinction events lasting tens of thousands of years.
- E-O: A series of extinctions that took place over 30 million years (445 to 415 million years ago).
- L-D: Estimates range from 500,000 to 25 million years, extending from the mid-Givetian to the end-Famennian. Some consider the extinction to be as many as seven distinct events, spread over about 25 million years.
- E-P: Quoting from Brittanica: “Many geologists and palaeontologists contend that the Permian extinction occurred over the course of 15 million years during the latter part of the Permian Period (299 million to 252 million years ago). However, others claim that the extinction interval was much more rapid, lasting only about 200,000 years, with the bulk of the species loss occurring over a 20,000-year span near the end of the period.”
- E-T: Less than 10,000 years.
- E-C: Some evidence suggests extinction over a period of less than 10,000 years. Other evidence points to a duration of 1,000–71,000 years. Of course, many species were probably killed within a few years of the impact (perhaps instantly) - I didn’t look into this.
- At the current rate of species extinction, which is extremely fast in historical terms, it would take us 37,500 years to lose 75%.
Some of the shortest estimates for the length of these extinction events are still longer than the entire history of humanity (~200,000 years).
Of course, this is for 75% of species on Earth, and we’re just one. I struggled to find an authoritative source for the time it takes for individual species to decline. Some data points:
- The Neanderthals peaked in population around 52,000 years ago, before disappearing around 28,000 to 35,000 years ago with the arrival in Europe of Homo sapiens — a decline of 17,000–24,000 years.
- More recently, The Pinta Island tortoise was first encountered and hunted in 1877 and was classified as extinct in 2012 (spare a thought for the last surviving male, Lonesome George) — a decline of >100 years.
- Schomburgk’s deer was first encountered in 1863 and the last captive individual was killed in 1938 — a decline of 75 years.
- The Desert rat-kangaroo was first encountered in the early 1840s. It was last sighted in 1935 and first pronounced extinct in 1994 — a decline of 100–160 years.
I pulled the last three examples from the International Union for Conservation of Nature’s Red List of Threatened Species’ Extinction list — it would be cool if someone built a better dataset of how long extinctions usually take from this list.
Why we might not trust the data from species
It’s hard to draw conclusions from the data on species’ extinction, and the uncertainties do not pull in favour of a prior on very slow decline:
- We won’t notice rapid declines in fossil records. The base rates on the length of decline we can get from the fossil record must have a high lower bound since we just don’t have that many fossils and the time granularity is poor. If species were rising and falling at a rate of years or months, this wouldn’t show up in the fossil record.
- For mass extinctions, it may take a long time for the total number of extinctions to creep up because not all species become maladapted at once. Much like the asteroid impact, we should probably expect a large number of individual species to die out quite rapidly over the course of a mass extinction event.
- This data can help us form a prior on the length of climate-induced catastrophes and natural extinction events (e.g. asteroids) but not the anthropological threats that we face today.
This doesn’t cause me to totally ignore the data here — species do seem to die out slowly — but we should take all the data here with a very big pinch of salt.
Before we look at potential x-risk scenarios, base rates suggest we should have a reasonably strong prior on extinction events occurring very slowly — over years, decades and millennia rather than days, weeks and months.
More specifically, it seems that mass extinctions from natural causes tend to occur over the course of millennia while civilisational collapses brought on by anthropogenic causes tend to occur over the course of years, decades or centuries. We should want to see very strong evidence that extinction events might happen faster than this to update towards something like a days, weeks and months expectation.
There might be some more interesting takeaways here. Perhaps this should give us some hope that we can notice the beginning of an existential decline and prevent it, or that we should look harder for signs of slow decline to protect ourselves against. However, I’m not confident enough in the data or the argument to make such claims.
And to reiterate, this sketch is very rough. I’d love to see someone work out this base rate, and how it might be changing, more carefully.
My thanks to Luisa Rodriguez, Lizka Vaintrob, Sophie Rose, Stephen Clare, Matthew van der Merwe and Justis Mills for comments and encouragement.
- Careful readers will note it’s actually very funny I’m discussing base rates, because it is the name of this blog which is named after me.
- Look out for my forthcoming book on this topic, “The Concerningly Steep Slope”
- This is probably the most ill-defined part of the analysis, but there is lots of competition for this title.
- My thanks to Stephen Clare, who suggested most of these.
- Note that, for many individual species, the first observation of a species is, sadly, also the point at which the discoverers start hunting it to extinction.
- My thanks to Matthew van der Merwe and Stephen Clare for pointing these out.