Update on civilizational collapse research

I spent a few months in 2019 re­search­ing civ­i­liza­tional col­lapse sce­nar­ios, and came to some ten­ta­tive con­clu­sions. One ques­tion that drove my re­search was: “How difficult would it be to launch a pro­ject that sig­nifi­cantly im­proved the re­silience ca­pac­ity of civ­i­liza­tion?” and “How likely is it that such a pro­ject could im­prove the long term prospects for hu­man­ity?”

For con­text, I also gave a talk about re­duc­ing the un­cer­tainty in col­lapse sce­nar­ios, which you can watch here.

My con­clu­sion af­ter in­ves­ti­gat­ing po­ten­tial col­lapse sce­nar­ios. (I origi­nally wrote this as an EA Fo­rum com­ment.)

  1. There are a num­ber of plau­si­ble (>1% prob­a­bil­ity) sce­nar­ios in the next hun­dred years that would re­sult in a “civ­i­liza­tional col­lapse”, where an un­prece­dented num­ber of peo­ple die and key tech­nolo­gies are (tem­porar­ily) lost.

  2. Most of these col­lapse sce­nar­ios would be tem­po­rary, with com­plete re­cov­ery likely on the scale of decades to a cou­ple hun­dred years.

  3. The high­est lev­er­age point for in­ter­ven­tion in a po­ten­tial post-col­lapse en­vi­ron­ment would be at the state level. In­di­vi­d­u­als, even wealthy in­di­vi­d­u­als, lack the in­fras­truc­ture and hu­man re­sources at the scale nec­es­sary to re­build effec­tively. There are some de­cent miti­ga­tions pos­si­ble in the space of in­for­ma­tion archival, such as seed banks and in­ter­net archives, but these are far less likely to have long term im­pacts com­pared to state efforts.

Based on these con­clu­sions, I de­cided to fo­cus my efforts on other global risk anal­y­sis ar­eas, be­cause I felt I didn’t have the rele­vant skills or re­sources to em­bark on a state-level pro­ject. If I did have those skills & re­sources, I be­lieve (low to medium con­fi­dence) it would be worth­while pro­ject, and if I found a per­son or group who did pos­sess those skills /​ re­sources, I would strongly con­sider offer­ing my as­sis­tance.

Col­lapse Q&A

In re­view­ing a grant pro­posal re­lated to the pro­ject above, Oliver Habryka noted he had a num­ber of cruxes about col­lapse sce­nar­ios & miti­ga­tions. I thought these were good ques­tions, so re­cently wrote out my re­sponses.

  • Is there a high chance that hu­man pop­u­la­tion com­pletely col­lapses as a re­sult of less than 90% of the pop­u­la­tion be­ing wiped out in a global catas­tro­phe?

I think the an­swer in the short term is no, if “com­pletely col­lapses” means some­thing like “is un­able to get back to at least 1950′s level tech­nol­ogy in 500 years”. I think think there are a num­ber of things that could re­duce hu­man­ity’s “tech­nolog­i­cal car­ry­ing ca­pac­ity”. I’m cur­rently work­ing on ex­pli­cat­ing some of these fac­tors, but some ex­am­ples would be dras­tic cli­mate change, long-lived ra­dionu­clides, in­crease in per­sis­tent pathogens.

  • Can we build any rea­son­able mod­els about what our bot­tle­necks will be for re­cov­ery af­ter a sig­nifi­cant global catas­tro­phe? (This is likely de­pen­dent on an anal­y­sis of what spe­cific catas­tro­phes are most likely and what state they leave hu­man­ity in)

I think we can. I’m not sure we can get very con­fi­dent about ex­actly which po­ten­tial bot­tle­necks will prove most sig­nifi­cant, but I think we can nar­row the search space and put forth some good hy­pothe­ses, both by rea­son­ing from the best refer­ence class ex­am­ples we have and by think­ing through the eco­nomics of po­ten­tial sce­nar­ios.

  • Are there ma­jor risks that have a chance to wipe out more than 90% of the pop­u­la­tion, but not all of it? My mod­els of biorisk sug­gests it’s quite hard to get to 90% mor­tal­ity, I think most nu­clear win­ter sce­nar­ios also have less than a 90% food re­duc­tion impact

I’m not sure about this one. I can think of some sce­nar­ios that would wipe out 90%+ of the pop­u­la­tion but none of them seem very likely. Eng­ineered pan­demics seem like one can­di­date (I agree with Denken­berger here), and the worst-case nu­clear win­ter sce­nar­ios might also do it, though I haven’t read the nu­clear win­ter pa­pers in a while, and there has been sev­eral new pa­pers and com­ments in the last year, in­clud­ing real dis­agree­ment in the field (yay, fi­nally!)

  • Are there non-pop­u­la­tion-level de­pen­dent ways in which mod­ern civ­i­liza­tion is frag­ile that might cause wide­spread col­lapse and the end of sci­en­tific progress? If so, are there any ways to pre­pare for them?

Pop­u­la­tion seems like one im­por­tant vari­able in our tech­nolog­i­cal car­ry­ing ca­pac­ity, but I ex­pect some of the oth­ers are as im­por­tant. The one I men­tioned in my other post is ba­si­cally I think a huge one is state plan­ning & co­or­di­na­tion ca­pac­ity. I think post-WWII Ger­many and Ja­pan illus­trate this quite well. How­ever, I don’t have a very good sense of what might cause most states to fail with­out also de­stroy­ing a large part of the pop­u­la­tion at the same time. But what I’m say­ing is that the pop­u­la­tion fac­tor might not be the most im­por­tant one in those sce­nar­ios.

  • Are there strong rea­sons to ex­pect the ex­is­ten­tial risk pro­file of a re­cov­ered civ­i­liza­tion to be sig­nifi­cantly bet­ter than for our cur­rent civ­i­liza­tion? (E.g. maybe a bad ex­pe­rience with nu­clear weapons would make the world much more aware of the dan­gers of tech­nol­ogy)

I’m very un­cer­tain about this. I do think there is a good case for in­ter­ven­tions aimed at im­prov­ing the ex­is­ten­tial risk pro­file of post-dis­aster civ­i­liza­tion be­ing com­pet­i­tive with in­ter­ven­tions aimed at im­prov­ing the ex­is­ten­tial risk pro­file of our cur­rent civ­i­liza­tion. The gist is that there is far less com­pe­ti­tion for the former in­ter­ven­tions. Of course, given the huge un­cer­tain­ties about both the cir­cum­stances of global catas­tro­phes and the po­ten­tial in­ter­ven­tion points, it’s hard to say whether it would pos­si­ble to ac­tu­ally al­ter the post-dis­aster civ­i­liza­tion’s pro­file at all. How­ever, it’s also hard to say whether we can al­ter the cur­rent civ­i­liza­tion’s pro­file at all, and it’s not ob­vi­ous to me that this lat­ter task is eas­ier.