# NunoSempere

Karma: 918

I stud­ied Maths and Philos­o­phy, but dropped out. I have helped im­ple­ment the Euro­pean Sum­mer Pro­gram on Ra­tion­al­ity dur­ing 2017, 2018 and 2019, and SPARC dur­ing 2020.

I like to spend most of my time ac­quiring deeper mod­els of things, and I keep a record of in­ter­est­ing pro­jects at nunosem­pere.github.io. I’ve also sold soft­ware and worked as a con­trac­tor for var­i­ous fore­cast­ing pro­jects. I’m Lok­iOdinevich on GoodJudge­men­tOpen, and Loki on CSET-Foretell.

Fu­ture of Hu­man­ity In­sti­tute 2020 Sum­mer Re­search Fel­low.

• Many of my in­ter­ests are re­lated to Gen­eral Se­man­tics, so I’d like to un­der­stand it bet­ter.

Are you look­ing for any par­tic­u­lar poin­t­ers?

• In­cen­tives might be a prob­lem, see here.

# Fore­cast­ing Newslet­ter: Au­gust 2020.

1 Sep 2020 11:35 UTC
22 points
• As a side-note, the max­i­mum en­tropy prin­ci­ple would tell you to choose the max­i­mum en­tropy prior given the in­for­ma­tion you have, and so if you in­tuit the in­for­ma­tion that the balls are likely to be pro­duced by the same pro­cess, you’ll get a differ­ent prior that if you don’t have that in­for­ma­tion.

I.e., your dis­agree­ment might stem from the fact that the max­i­mum en­tropy prin­ci­ple gives differ­ent an­swers con­di­tional on differ­ent in­for­ma­tion.

I.e., you ac­tu­ally have in­for­ma­tion to differ­en­ti­ate be­tween draw­ing n balls and flip­ping a fair coin n times.

• I have some notes about this topic, re­pro­duced be­low, which might be in­ter­est­ing to you. As a thread to pull, you might want to talk with Miranda Dixon-Luinen­burg (if she’s available), who re­ceived a grant from the Long-Term fund to work on “[w]rit­ing EA-themed fic­tion that ad­dresses X-risk top­ics” (see the first link in the next sec­tion).

Other­wise, Un­cle Tom’s Cabin did have the effect of in­spiring The Clans­man and the sub­se­quent film, which were also in­fluen­tial in the op­po­site di­rec­tion.

## Bite sized case studies

### The Works of Jules Verne

Ob­ject class: Safe

Con­tain­ment pro­ce­dure: None.

De­scrip­tion: Jules Verne wrote a se­ries of highly en­ter­tain­ing nov­els which are still widely read, and in­spired countless young peo­ple. I had a hunch that Von Braun (de­scribed as “the ar­chi­tect of the Moon land­ings”) might have been in­spired by Verne, and this hunch was con­firmed:

Wern­her von Braun, the Ger­man-born sci­en­tist who be­came crit­i­cal to the suc­cess of the Amer­i­can space pro­gram, hoped the rock­ets he de­signed would en­able space ex­plo­ra­tion as de­scribed by Verne and H. G. Wells… When von Braun’s Saturn rock­ets pow­ered the US Mer­cury and Apollo pro­grams, he gave credit to Verne’s self-fulfilling proph­e­sies: “The sci­ence in 1865’s From the Earth to the Moon is nearly as ac­cu­rate as the knowl­edge of the time per­mit­ted… He was read with great re­spect by work­ing sci­en­tists, so care­fully did he do his sci­en­tific home­work.” Ac­cord­ing to von Braun’s ac­count­ing, “the debt mod­ern as­tro­nauts owe Verne is ap­par­ent.” Source

Be­sides nudg­ing Von Braun to have an in­ter­est in rock­ets (as op­posed to, say, trains), the works of Jules Verne also in­spired plenty of other pi­o­neers (see Wikipe­dia: Cul­tural in­fluence of Jules Verne for a full list). This is re­mark­able in light of other similarly pop­u­lar writ­ers, which var­i­ously wrote about:

• A dar­ing spy for the Bri­tish crown fails to have a cooler greet­ing than Íñigo Mon­toya.

• A cyn­i­cal ge­nius solves crim­i­nal mys­ter­ies through de­duc­tions which wouldn’t work in real life.

• A group of kids travel to a mys­te­ri­ous land which was pre­vi­ously ruled by a lion which rep­re­sents God.

• A kid goes to a mag­i­cal school and fights a villain who wants to be im­mor­tal.

• An at­trac­tive yet nor­mal teenager rebels against a pas­tiche dystopia by go­ing on TV.

Honor­able men­tion:

• A fel­low­ship seeks to de­stroy an ar­ti­fact which offers the user im­mense power, but which will cor­rupt and ma­nipu­late even the most in­no­cent wielder.

### A mes­sage to Gar­cia.

Ob­ject Class: Safe

Spe­cial Con­tain­ment Pro­ce­dures: None. In par­tic­u­lar see the Streisand effect. Nonethe­less, much like At­las Shrugged, op­er­a­tives are cau­tioned not to ex­pose the wrong 14 year old. In case of ex­po­sure with acute nega­tive effects, the speech A More Perfect Union may be pre­scribed, as it pro­duces a similarly sized effect in a differ­ent ide­olog­i­cal di­rec­tion. Nonethe­less, the effects of A Mes­sage to Gar­cia are more likely to be mildly pos­i­tive.

De­scrip­tion: A Mes­sage to Gar­cia is a short pub­li­ca­tion which nar­rates a fic­tion­al­ized his­tor­i­cal de­pic­tion of a dar­ing ad­ven­ture by an Amer­i­can sol­dier. The reader will tend to re­gard the val­ues of in­di­vi­d­ual ini­ti­a­tive and con­scien­tious­ness more fa­vor­ably.

It may also cause the com­pul­sion to share the pub­li­ca­tion with other peo­ple, to sym­pa­thize for those par­ties which are called liber­tar­ian in Amer­ica and liberal in the rest of the world, and to view the works of Ayn Rand, Friedrich Hayek, et al. with less skep­ti­cism. The ob­ject has been widely shared, and con­tinues to be read, since 1899. As it has reached the in­ter­net, con­tain­ment is now un­fea­si­ble.

### The Ones Who Walk Away from Omelas

Ob­ject Class: Safe.

Spe­cial Con­tain­ment Pro­ce­dures: None (see above). Oper­a­tives who are not nega­tive util­i­tar­i­ans might want to promptly ex­pose read­ers to Toby Ord’s Why I’m Not a Nega­tive Utili­tar­ian.

De­scrip­tion: The Ones Who Walk Away from Ome­las is a short story by famed writer Ur­sula K. Le Guin. Among util­i­tar­i­ans, it has the effects of mak­ing one won­der whether clas­si­cal util­i­tar­i­anism might not, af­ter all, be wrong, and whether other eth­i­cal sys­tems such as virtue ethics, or nega­tive util­i­tar­i­anism might be cor­rect af­ter all. Among nonu­til­i­tar­ian de­mo­graph­ics, it also sparks heated de­bate.

### The Fable of The Dragon Tyrant.

Ob­ject Class: Safe.

Spe­cial Con­tain­ment Pro­ce­dures: None (see above).

De­scrip­tion: The Fable of The Dragon Tyrant is a short story by Nick Bostrom, in which a dragon sav­ages a land, and the king and the pop­u­la­tion learn to ap­pease it and, de­spite it at first be­ing seem­ingly im­pos­si­ble, kill it. The twist is that at the end of the story the au­thor com­ments on how the Dragon Tyrant is similar to senes­cence, and by re­fraiming death as a Dragon which can be defeated, the reader may be able to over­come cached thoughts, in­stinc­tive turns of phrases

It is un­clear what the im­pact of this spe­cific story has been in the Zeit­geist, but I think it has the po­ten­tial to be shared widely and be as cul­turally per­va­sive as the above works of fic­tion, and that this would be a good thing.

## Path­ways to im­pact, some speculation

Most so­cial move­ments don’t sur­vive many gen­er­a­tions. OpenPhilantropy looks like they want to donate all of Good Ven­ture’s cap­i­tal, and it’s un­clear whether they will up­date in light of Tram­mel’s pa­per on pa­tient philantropists. If Effec­tive Altru­ism ceases to ex­ist in a gen­er­a­tion or two, or be­comes pro­gres­sively more mediocre, liter­a­ture can re­main.

Given that gen­er­a­tion from gen­er­a­tion the poli­ti­cal and tech­nolog­i­cal land­scape changes dras­ti­cally, it might be difficult to leave spe­cific and con­crete ad­vice for our in­tel­lec­tual de­scen­dants. They may dis­cover new cru­cial con­sid­er­a­tions, or bet­ter tech­nolo­gies.

Thus, it is per­haps eas­ier to leave liter­ary bread­crumbs. For ex­am­ple, a sense of sci­en­tific won­der, the drive to make the world bet­ter and the dis­po­si­tion to do so effi­ciently, the ex­pan­sion of one’s moral cir­cle. Th­ese ideas can be made formidable and long-last­ing through liter­a­ture. More am­bi­tiously, one might try to pull off a Jules Verne and write sto­ries about (benefi­cial) in­ven­tions which might be pos­si­ble with the tech­nol­ogy level of 50 years from now.

And liter­a­ture is also a pos­i­tive if your move­ment doesn’t van­ish into the ashes. The search for bet­ter turns of phrase, more wor­thy cog­ni­tive soft­ware, and metaphors which one would want to live by might con­tribute to our col­lec­tive flour­ish­ing. And liter­a­ture may also serve to stump, con­fuse and dis­tract any ad­ver­saries, en­er­gize your al­lies, and en­chanter the in­de­ci­sive. Other mod­els of im­pact ex­ist.

The caveat is Stur­geon’s law: that 90% of ev­ery­thing is crap. And of the frac­tion which is not crap, most is not ac­tu­ally good.

• Yes, I think that this cor­re­sponds to the Ger­man tank prob­lem af­ter you see the first tank.

• there are a few out­liers that only get widely rec­og­nized af­ter decades, much longer than for typ­i­cal insights

Th­ese sleep­ing beau­ties might hap­pen more of­ten the younger a field is. In par­tic­u­lar, I don’t par­tic­u­larly care that (per­haps lesser) in­sights are spread quickly once a field is pro­duc­ing a lot of pa­pers.

Any­ways, some other ex­am­ples are Tay­lor polyno­mi­als (!) and var­i­ous dis­cov­er­ies by Tsiolkovsky on the me­chan­ics of space travel.

• Yep, ex­actly right.

• With 50% prob­a­bil­ity, things will last twice as long as they already have.

In 1969, just af­ter grad­u­at­ing from Har­vard, Gott was trav­el­ing in Europe. While tour­ing Ber­lin, he won­dered how long the Ber­lin Wall would re­main there. He re­al­ized that there was noth­ing spe­cial about his be­ing at the Wall at that time. Thus if the time from the con­struc­tion of the Wall un­til its re­moval were di­vided into four equal parts, there was a 50% chance that he was in one of the mid­dle two parts. If his visit was at the be­gin­ning of this mid­dle 50%, then the Wall would be there three times as long as it had so far; if his visit was at the end of the mid­dle 50%, then the Wall would last 13 as long as it had so far. Since the Wall was 8 years old when he vis­ited, Gott es­ti­mated that there was a 50% chance that it would last be­tween 2.67 and 24 years. As it turned out, it was 20 more years un­til the Wall came down in 1989. This suc­cess of this pre­dic­tion spurred Gott to write up his method for pub­li­ca­tion. (It ap­peared in the jour­nal Na­ture in 1993.)

I have used this method with great suc­cess to es­ti­mate, among other things, the prob­a­bil­ity that friends will break up with their ro­man­tic part­ners.

I also car­ried out some ex­per­i­ments a while ago to find out what the prior prob­a­bil­ity was for me “be­ing re­ally sure about some­thing”, or the prob­a­bil­ity as­so­ci­ated to “I would be highly sur­prised to learn if this were false.” That is, for the feel­ing of be­ing highly sure, how does that pan out?

On an­other di­rec­tion, su­perfore­cast­ers have some meta-pri­ors, such as “things will take longer than ex­pected, and longer for larger or­ga­ni­za­tions”, or “things will stay mostly as they have.”

• I’m also in­ter­ested in ques­tions around ap­proval vot­ing in gen­eral, and the Cen­ter for Elec­tion Science in par­tic­u­lar.

Some stuff:

• Con­di­tional on less than 5 cities with >=50,000 peo­ple hav­ing im­ple­mented ap­proval vot­ing by Dec 31, 2022, what will the fund­ing for the Cen­ter for Elec­tion Science be dur­ing 2023? Con­text: Ac­cord­ing to the CES’s strate­gic plan con­vert­ing 5 cities with >= 50,000 in­hab­itants is one of their main tar­gets by 2022 (see p. 7). Con­di­tional on them not achiev­ing it, how will their fund­ing look like? This can prob­a­bly be op­er­a­tional­ized with refer­ence to IRS tax re­ports.

• How many US cities with more than 50,000 peo­ple will have im­ple­mented ap­proval vot­ing by [date]?

• What will CES fund­ing look like in 2021, 2022, etc.

• A while ago, Leah Edger­ton, of An­i­mal Char­ity Eval­u­a­tors, gave an AMA, and one of the ques­tions I asked was What are some ques­tions re­gard­ing EAA (effec­tive an­i­mal ad­vo­cacy) which are amenable to be­ing fore­casted?.

Her an­swer is in this video here. In short:

• Will cor­po­ra­tions stick to their an­i­mal welfare com­mit­ments?

• When will spe­cific an­i­mal free food tech­nolo­gies be­come cost-com­pet­i­tive with their tra­di­tional an­i­mal coun­ter­parts?

• Timelines for cul­tured meat com­ing to mar­ket?

• When will tech­nol­ogy ex­ist which al­lows the iden­ti­fi­ca­tion of the sex of a chicken be­fore it hatches? When, if ever, will such a tech­nol­ogy be adopted

• When, if ever, will the global pro­duc­tion and con­sump­tion of farmed an­i­mals stop grow­ing? When will stop com­pletely?

• When will spe­cific coun­tries or states adopt le­gal pro­tec­tion for an­i­mals /​ farmed an­i­mals?

• When will EAA or­ga­ni­za­tions have a bud­get of more than \$500 mil­lion? \$1 billion?

• Ques­tions re­lated to the pan­demic.

• Ques­tions re­lated to the bud­get of EAA or­ga­ni­za­tions in the im­me­di­ate fu­ture.

Oper­a­tional­iz­ing these ques­tions, and find­ing out what the most use­ful things to fore­cast are may in­volve con­tact­ing ACE di­rectly. For ex­am­ple, “cor­po­ra­tions” is pretty gen­eral, so I imag­ine ACE has some par­tic­u­lar ones in mind.

# Fore­cast­ing Newslet­ter: July 2020.

1 Aug 2020 16:56 UTC
31 points
• Good Judge­ment Open might have you cov­ered here; see: Will China’s Three Gorges Dam fail be­fore 1 Oc­to­ber 2020?.

Cur­rent crowd prob­a­bil­ity: 3%. (note the timeline).

Some com­ments I’ve cu­rated from that ques­tion:

https://​​asi­a­times.com/​​2020/​​07/​​three-gorges-dam-de­formed-but-safe-say-op­er­a­tors/​​ Re­lease ap­pears to be con­trol­led. https://​​www.upi.com/​​Top_News/​​World-News/​​2020/​​07/​​23/​​China-braces-for-im­pact-af­ter-mass-flood­ing-at-Three-Gorges-Dam/​​2221595525864/​​ Re­vis­ing to­day af­ter con­sid­er­ing this fur­ther. The chance that the coun­try doesn’t di­vert wa­ter to pre­vent the dam from failing, even if up­stream dams burst, seems very slim.

There have been heavy rains in the re­gion which con­tinue as of the time of writ­ing. The dam is 181m high and the de­sign max­i­mum wa­ter level is 176m. Dams are de­signed to last hun­dreds of years, though cli­mate change could mean that the origi­nal de­sign as­sump­tions have be­come out­dated. There is a very slight chance of the dam “failing” within the next few months and re­leas­ing a sud­den rapid un­con­trol­led flow down­stream—just above zero. Some re­ports say the dam was built to hold 145 me­ters of wa­ter but ac­tu­ally that figure refers to the level at which wa­ter is re­leased down­stream in or­der to smooth out flood flows and main­tain ca­pac­ity in the reser­voir. Discharge in re­cent days and weeks has been be­tween 20 and 30 thou­sand cumecs, but this has gone up to 40k in the past, so there is still some cush­ion. Prob­a­bly the great­est risk is of failure of one or more ma­jor dams up­stream, un­leash­ing a flood surge that could over­top the dam.
(Says a civil en­g­ineer)

This is a grav­ity dam, and it re­lies upon the con­struc­tion it­self to stand. It was 50 ft Above flood level. I see this as a con­cern with­out be­ing a high prob­a­bil­ity event.

This ques­tion talks about the failure of a \$ 32 Billion pro­ject com­pleted in 2012. We have roughly 9 weeks from to­day till when the ques­tion is re­solved. 5% for Yes is a good baseline to start. While the dam is cur­rently hold­ing more wa­ter than it is de­signed for, wa­ter can always be re­leased if things get bad. The rea­son why they would be hold­ing more wa­ter is in or­der to pre­vent the catch­ment ar­eas from get­ting flooded.

This looks un­likely, but how un­likely seems difficult to es­ti­mate: on the one hand, qual­ity of con­struc­tion in China is poor and cut­ting cor­ners is a way of life. On the other this is a flag­ship pro­ject, which means that there must have been stringent qual­ity con­trols (in con­trast with the stan­dard situ­a­tion in China). Un­for­tu­nately this is in­side view. I did not try to make a his­toric re­view of dams failing around the world or in China. How­ever the Ban­qiao dam failure in 1975 read­ily comes to mind: https://​​en.wikipe­dia.org/​​wiki/​​Ban­qiao_Dam https://​​en.wikipe­dia.org/​​wiki/​​1975_Ban­qiao_Dam_failure and catas­trophic floods have been a com­mon re­cur­rence along Chi­nese his­tory.

There were 3,523 in­ci­dents of dam failure from 1954 to 2013 (He et al. 2008; Zhao 2014) that caused sig­nifi­cant loss of life and eco­nomic losses in china. This av­er­ages out to 67 dam failures, of var­i­ous sizes per year. There are ap­prox 87,000 dams in to­tal. Given this, the three gorges dam broadly speak­ing has .08 chance of failing this year. There is sig­nifi­cant flood­ing atm, which could in­crease prob­a­bil­ity of failure, but the con­crete dam wall is 181 m (594 ft) high above the rock ba­sis and has a max ca­pac­ity wa­ter level of 175 m (wiki) Water level cur­rently seems to be at 145-7 m from the ar­ti­cles I can find, which is well within ca­pac­ity. The dam has passed qual­ity checks, is a rel­a­tively new pro­ject (old dams fail more of­ten) and there’s a large amount of re­search done on seis­mic ac­tivity in the area. https://​​www.sci­encedi­rect.com/​​sci­ence/​​ar­ti­cle/​​pii/​​S1674984715300756; https://​​jour­nal.probein­ter­na­tional.org/​​2014/​​04/​​07/​​three-gorges-dam-trig­gers-fre­quent-seis­mic-ac­tivi­ties/​​ https://​​www.slu.edu/​​news/​​2018/​​septem­ber/​​earth­quake-re­search.php There are ru­mors of buck­ling and de­for­ma­tion and a google earth image go­ing around (https://​​www.foxnews.com/​​world/​​in­tegrity-of-chi­nas-three-gorges-dam-ques­tioned-de­spite-china-offi­cials-dis­miss­ing-it-as-safe) but I looked my­self, and google earth cur­rently shows shows no buck­ling, nor could I find any in­con­sis­ten­cies, so Ima say probs not, image seems fake. Long story short, is the dam gonna pre­vent flood­ing down­stream? Maybe not, its effec­tive­ness at do­ing so seems ques­tion­able based off the ar­ti­cles. This, how­ever, isn’t the is­sue at hand. We’re ask­ing is the wor­lds largest dam gonna fail in the next three months af­ter pass­ing safety checks, hav­ing re­search available about seis­mic ac­tivity in the area and cur­rently within ca­pac­ity? Highly un­likely. If up­stream dams start to fail and/​​or if wa­ter lev­els breach ca­pac­ity, then it gets more likely. But til then, low low chance.

One of the things I’ve been think­ing of....dam failure means any amount of wa­ter that they didn’t in­tend to let through pass­ing the dam. So Im won­der­ing, is there a higher per­cent chance of some­thing small hap­pen­ing (whups, a cou­ple gal­lons seeped through, or we lost a cou­ple thou­sand gal­lons over the edge, our bad) or is it an all or noth­ing deal where when she goes, she goes, rip wuhan. Good point on the time it takes for flood level to get there. Also, think­ing of ways it could po­ten­tially fail, I could con­ceiv­ably imag­ine a sce­nario where the dam is at or slightly over ca­pac­ity due to flood­ing, seis­mic ac­tivity hap­pens thats un­prece­dented and the con­crete slips off the bot­tom rock. There’s his­tory of other large con­crete dams do­ing so (https://​​www.ncbi.nlm.nih.gov/​​pmc/​​ar­ti­cles/​​PMC4997708/​​) but seems to be only at filling? There’s also this ar­ti­cle: https://​​asi­a­times.com/​​2020/​​07/​​three-gorges-dam-de­formed-but-safe-say-op­er­a­tors/​​ which is re­cent and does men­tion some de­for­ma­tion to ‘non struc­tural parts of the dam’ I don’t ex­actly know what that means lol. And then fi­nally, there was one sci­en­tist dude who has been talk­ing about failure for a while, but stat­ing cracks in the con­crete dur­ing early stages of the build­ing pro­cess and in­stances of sub­stan­dard con­crete, not buck­ling as the in­ter­net seems in­tent on por­tray­ing lol. https://​​www.rfa.org/​​en­glish/​​news/​​china/​​three­gorges-safety-07082019085631.html Ah, one more thing: https://​​damsafety.org/​​dam-failures#:~:text=Dam failures are most likely,the top of a dam.&text=Na­tional statis­tics show that over­top­ping,of all U.S. dam failures. The way dams are most likely to fail (this is US based, but i read a sci­en­tific study about chi­nese dams that was say­ing the same things) over­top­ping num­ber one rea­son of failure, within the very low per­centage chance of a dam failing. Slip­ping sec­ond most likely. So if the dam does fail, it’s most likely gonna ei­ther overfill or slip. Will be good to pay at­ten­tion to wa­ter lev­els up­stream, pre­cip­i­ta­tion, and how much out­put their let­ting through in an­ti­ci­pa­tion.

This is a topic I have some sub­ject-knowl­edge on, and I think the ques­tion re­quires clar­ifi­ca­tion: The Forestry refer­ence to “sud­den, rapid, and un­con­trol­led re­lease of im­pounded wa­ter” is in­cluded in their defi­ni­tion of “Dam failure” and the key el­e­ment of the failure is the re­lease of the im­pounded wa­ter. The dam is de­signed to man­age a 1:100 year flood, de­rived statis­ti­cally, by hav­ing the im­pound­ment re­serve ca­pac­ity and con­trol­led discharge of this amount of wa­ter. Floods greater than the 1:100 year value are man­aged by the sluice gates, tur­bine chan­nel flow and, ul­ti­mately, by the dam’s spillway. The spillway is the low­est part of the dam crest and is de­signed to per­mit much larger flows (be­yond the “Prob­a­ble Max­i­mum Flood”) As defined, I think the ques­tion asks whether there will be a failure of the dam which re­leases the wa­ter im­pounded be­low the spillway level, e.g., struc­tural/​geotech­ni­cal failure, un­der­min­ing or un­con­trol­led by­pass, which is highly un­likely. How­ever, the ques­tion may be in­ter­preted to ask whether the dam will be ‘over­topped’, with un­con­trol­led, rapid re­lease over the spillway – which is quite prob­a­ble this year. BTW the dam was built to re­duce the fre­quency of flood­ing down­stream, where mil­lions have died from flood­ing of the Yangtze River. From a flood risk man­age­ment per­spec­tive the dam is small at 1:100 year ca­pac­ity. Nega­tive press in com­pet­ing or de­vel­oped coun­tries fo­cussed on dis­plac­ing 1.3M peo­ple in the in­ter­est of power pro­duc­tion, not on flood risk man­age­ment. Also, “The In­ter­preter” ar­ti­cle is ac­cu­rate in de­scribing older dams in China (and around the world) as be­ing po­ten­tial ‘black swans’: these dams were of­ten not de­signed to spill “prob­a­ble max­i­mum floods” and their failures may well jeop­ar­dize life down­stream.

The dam is already con­tro­ver­sial, so any story on it will be far reach­ing and po­ten­tially ex­ag­ger­ated. The Chi­nese have ad­mit­ted to some move­ment to the dam but say it’s within nor­mal pa­ram­e­ters. While un­der scrutiny for COVID and strug­gling with it’s in­ter­na­tional image I’d like to think that evac­u­a­tions would be in place if the risk was high. Be­cause this may be a naive thought and catas­trophic ac­ci­dents have oc­curred in the past due to bu­reau­cratic failings in similar regimes (think Ch­er­nobyl), I have in­put 2%.

The 3GD is a grav­ity dam, but the blocks are rest­ing on the riverbed, not dug in. This is caus­ing de­for­ma­tion through­out the struc­ture. Con­struc­tion is likely shoddy and the qual­ity team that in­spected the dam were from the same com­pany that built it—not in­de­pen­dently done. The CCP came out yes­ter­day(?) to say that there is some de­for­ma­tion in the dam, but noth­ing to worry about. This alone from the CCP is un­usual. In my es­ti­ma­tion, the CCP would rather flood Wuhan down­stream than see their flag­ship fall. That said, they may not have a choice. As the heavy rains in the re­gion con­tinue, wa­ter is mak­ing its way back into the Yangtze and thus the reser­voir. Not all of the rain we’ve seen in the past has yet en­tered the reser­voir. The wa­ter level is already at 164m, 175m is the warn­ing of col­lapse, 185m and it’s gone. The fi­nal and most likely catas­trophic failure is the spillways be­ing dam­aged—we’re see­ing this on at least two of the spillways on the livestream. A large chunk ap­pears to be bro­ken off spillway #6 and there is ev­i­dence of cav­i­ta­tion in spillway #2. If the dam­age to the spillways con­tinues, the dam will fail, and badly.

(This last com­ment is from a new and un­ex­pe­rienced fore­caster, with a Brier score worse than the ag­gre­gate. He still only gives 15%)

My own im­pres­sion is that the ag­gre­gate seems cor­rect, i.e., a 1:30 bet seems roughly fair.

I also some­what dis­agree with “It is worth ex­am­in­ing even if the risk is small;” it seems to me that de­ci­sions will be taken by the CCP, and that there is prob­a­bly no lev­er­age to be found here.

# Fore­cast­ing Newslet­ter: June 2020

1 Jul 2020 9:32 UTC
45 points
• If you look at your fore­cast­ing mis­takes, do they have a com­mon thread?

• How is your ex­pe­rience ac­quiring ex­per­tise at fore­cast­ing similar/​differ­ent to ac­quiring ex­per­tise in other do­mains, e.g. ob­scure board-games? How so?

• Any fore­cast­ing ressources you recom­mend?

• Who do you look up to?

• How does the dis­tri­bu­tion skill /​ hours of effort look for fore­cast­ing for you?

• Do you want to wax po­et­i­cally or ram­ble di­s­or­ga­nizedly about any as­pects of fore­cast­ing?

• Any se­crets of re­al­ity you’ve dis­cov­ered & which you’d like to share?

• Hi Linch! So what’s up with the Utili­tar­ian Memes page? Can you tell more about it? Any deep les­sons from util­i­tar­ian memes?

• In­dia v. China con­flict is per­haps more im­me­di­ately wor­ry­ing than US v. China.