Notes on ‘Atomic Obsession’ (2009)

Lately I’ve been read­ing about the his­tory of nu­clear weapons so that I can ex­tract po­ten­tial les­sons for the gov­er­nance of pow­er­ful AI tech­nolo­gies. One in­ter­est­ing and fairly con­trar­ian book is John Muel­ler’s 2009 book Atomic Ob­ses­sion (h/​t Holden Karnofsky).

In short, I think Muel­ler is right that nu­clear weapons have turned out to be less im­por­tant than most peo­ple in the 1940s thought they would be, and less im­por­tant than many (most?) se­cu­rity-re­lated poli­cy­mak­ers seem to think. I’d guess they’re also less im­por­tant than most GCR-fo­cused EAs seem to think (in­clud­ing my­self 12mo ago), but that’s hard to as­sess. I also agree with Muel­ler that nu­clear weapons have pro­lifer­ated less than origi­nally ex­pected, and that’s prob­a­bly as much or more due to their sur­pris­ingly low mil­i­tary and diplo­matic util­ity and high cost as it is to counter-pro­lifer­a­tion efforts. Their dan­ger is of­ten ex­ag­ger­ated, and nu­clear ter­ror­ism ap­pears quite un­likely. Co­op­er­a­tive se­cu­rity treaties may have con­tributed lit­tle to nu­clear se­cu­rity, and may even have been net harm­ful.

That said, (1) the book doesn’t say much about the worst-case sce­nar­ios I worry most about (large-scale nu­clear war and nu­clear win­ter), (2) I con­tinue to worry that the so-far (ap­par­ently) perfect safety and se­cu­rity record for nu­clear weapons will even­tu­ally end, which could (but prob­a­bly won’t) have global catas­trophic effects, and (3) I sus­pect there are highly cost-effec­tive things worth do­ing in nu­clear se­cu­rity (but they might re­quire lots of hard-won ex­per­tise to iden­tify).

Ob­vi­ously I’m not an ex­pert on nu­clear weapons his­tory, and my im­pres­sions might evolve dra­mat­i­cally on some points as I con­tinue to read more or hear from ex­perts on the topic. If you know of ev­i­dence for or against the claims be­low, es­pe­cially ev­i­dence not cov­ered in Muel­ler’s book, I hope you’ll post it in the com­ments.

Below, I quote the au­thor’s own sum­mary (from the epi­logue), with my own quick re­ac­tions added be­low.

Ob­ses­sion with nu­clear weapons, some­times based on ex­ag­ger­a­tions of the weapons’ de­struc­tive ca­pac­ity, has of­ten led to poli­cies that have been un­wise, waste­ful, and damaging

Yes, many ex­ag­ger­a­tions have definitely oc­curred, even at the high­est lev­els, as Muel­ler doc­u­ments.

Nu­clear weapons have been of lit­tle his­toric con­se­quence and have not been nec­es­sary to pre­vent World War III or a ma­jor con­flict in Europe.

I think this was one of the weaker parts of the book. I think it’s un­clear and hard to know how im­por­tant nu­clear weapons have been to re­duc­ing con­flict in­volv­ing nu­clear-armed states (in­clud­ing the pos­si­bil­ity of WWIII or “a ma­jor con­flict in Europe”) via de­ter­rence, but Muel­ler seems to ar­gue for some­thing like “pretty/​very un­likely that WWIII or ma­jor Euro­pean con­flict would’ve hap­pened with­out nukes,” and I think that’s too con­fi­dent. I think he gives too lit­tle weight to:

  1. WWII hap­pened im­me­di­ately af­ter WWI even though ev­ery­body could see the dev­as­ta­tion of WWI.

  2. In 1945, Stalin had just seized half of Europe, and it seems plau­si­ble a pri­ori he’d have seized more if he’d been ca­pa­ble, which he plau­si­ble was at the end of WWII if the U.S. hadn’t had the bomb; his troop counts in Europe were much larger than those of all of Western Europe at that time. Muel­ler pre­tends all the ev­i­dence points to Stalin not hav­ing fur­ther ex­pan­sion­ist am­bi­tions, but ac­tu­ally the ev­i­dence is mixed, e.g. see here.

  3. After Stalin, the USSR did se­ri­ously ex­plore the op­tions for launch­ing a new Euro­pean war (but it’s un­clear how much they were de­terred by nukes vs. other fac­tors).

And of course, a few years af­ter Muel­ler’s book came out, Rus­sia did en­gage in West­ward in­va­sion and ex­pan­sion, and only in a state not un­der the nu­clear um­brella (Ukraine).

Through­out the book Muel­ler does give some credit to the no­tion that nu­clear weapons have de­ter­rent power (in con­trast to their low util­ity for con­quest or com­pel­lence).

Mili­tar­ily, the weapons have proved to be use­less and a very sub­stan­tial waste of money and of sci­en­tific and tech­ni­cal tal­ent: there never seem to have been mil­i­tar­ily com­pel­ling rea­sons to use them, par­tic­u­larly be­cause of an in­abil­ity to iden­tify suit­able tar­gets or ones that could not be at­tacked about as effec­tively by con­ven­tional munitions

Seems true so far. I pretty con­fi­dently agree the nukes were not needed for Ja­panese sur­ren­der. Ja­pan was already ask­ing Rus­sia to me­di­ate sur­ren­der talks ~1mo be­fore Hiroshima. Not sure they would’ve agreed to un­con­di­tional sur­ren­der with­out nukes, but con­di­tional sur­ren­der would’ve been ~fine.

Although nu­clear weapons seem to have at best a quite limited sub­stan­tive im­pact on ac­tual his­tor­i­cal events, they have had a tremen­dous in­fluence on our ag­o­nies and ob­ses­sions, in­spiring des­per­ate rhetoric, ex­trav­a­gant the­o­riz­ing, waste­ful ex­pen­di­ture, and fre­netic diplo­matic posturing

As men­tioned el­se­where, it’s hard to tell what the de­ter­rence effect of nukes has been, and I think the de­ter­rence effects might be sub­stan­tially big­ger than Muel­ler thinks, but it’s also plau­si­ble Muel­ler’s right. Cer­tainly the threat has been ex­ag­ger­ated via “des­per­ate rhetoric” and re­sult­ing in “waste­ful ex­pen­di­ture” and so on — as is also true with e.g. ter­ror­ism.

Wars are not caused by weapons or arms races, and the quest to con­trol nu­clear weapons has mostly been an ex­er­cise in irrelevance

The first clause seems mostly true at least as far back as WWI; I’m less fa­mil­iar with wars be­fore that. One ex­cep­tion is the 2003 Iraq war, which Muel­ler seems to agree (ch. 10) was about pre­vent­ing Iraq from get­ting nu­clear weapons (cer­tainly that was the pub­lic rea­son).

The 2nd clause might be true, though even Muel­ler praises some nu­clear con­trol mea­sures such as Nunn-Lu­gar. My guess is that Muel­ler is un­der­weight­ing the im­pact of nu­clear con­trol/​counter-pro­lifer­a­tion mea­sures, but not by a ton. I’d guess he’s closer to right than peo­ple in­volved in im­ple­ment­ing those mea­sures tend to be.

The atomic bombs were prob­a­bly not nec­es­sary to in­duce the sur­ren­der of the Ja­panese in World War II

True, as men­tioned above.

Those who stole Amer­i­can atomic se­crets and gave them to the Soviet Union did not sig­nifi­cantly speed up the Soviet pro­gram; how­ever, ob­ses­sion about that es­pi­onage did detri­men­tally af­fect Amer­i­can for­eign and do­mes­tic policy, some­thing that led to a very sub­stan­tial in­fla­tion in the es­ti­ma­tion of the dan­gers that ex­ter­nal and in­ter­nal en­e­mies presented

Muel­ler ad­mits that Soviet spy­ing may have sped up the Soviet pro­gram by “a year or two,” and “18mo or more” is the most com­mon schol­arly es­ti­mate I’ve seen (for the speed-up from Klaus Fuchs pass­ing along ba­sic facts about the plu­to­nium path, which is the one the Soviets pur­sued). In the con­text of a 4-year U.S. nu­clear monopoly, a ~2yr speed up for the Soviet pro­gram is pretty sig­nifi­cant! Muel­ler seems right that if the Soviets had in­stead pur­sued the ura­nium path (which AFAIK they would’ve done if not for learn­ing about the plu­to­nium path via spy­ing), they might’ve got­ten there just as quickly, but that’s un­clear.

As for the effects of ob­ses­sion about es­pi­onage, that’s definitely true w.r.t. the 1946 rev­e­la­tion of Soviet spies in the Man­hat­tan Pro­ject, which dra­mat­i­cally af­fected do­mes­tic leg­is­la­tion w.r.t. con­trol of atomic en­ergy, and maybe af­fected at­tempts at in­ter­na­tional con­trol also. Muel­ler talks more about the 1950 ar­rests (in par­tic­u­lar the Rosen­bergs), which may also have af­fected policy but I haven’t read as much about that yet so I’m not sure.

Changes in anx­ieties about nu­clear de­struc­tion have not cor­re­lated at all well with changes in the sizes or the de­struc­tive ca­pac­i­ties of nu­clear arsenals

True AFAICT, though Muel­ler ad­mits that “Nu­clear fears have been de­ter­mined far more by the lev­els of poli­ti­cal ten­sion,” which seems rea­son­able if peo­ple aren’t scope sen­si­tive af­ter “mil­lions kil­led quickly,” which was fea­si­ble ev­ery year af­ter ~1949 (but fluc­tu­ated in prob­a­bil­ity based largely though not en­tirely on fluc­tu­at­ing de­grees of poli­ti­cal ten­sion). On the other hand, I’m not that sym­pa­thetic to scope-in­sen­si­tive views. :)

Arms re­duc­tion will pro­ceed most ex­pe­di­tiously if each side feels free to re­verse any re­duc­tion it later comes to re­gret; for­mal disar­ma­ment agree­ments are likely sim­ply to slow and con­fuse the process

This claim is in­ter­est­ing and I don’t know whether it’s true. This and sev­eral other con­trar­ian takes in the book about the ac­tual effects of well-in­tended policy are worth think­ing about in the con­text of what might be done w.r.t. in­ter­na­tional con­trol of AI (e.g. ML ASICs).

The eco­nomic and or­ga­ni­za­tional costs of fabri­cat­ing a nu­clear ar­se­nal can be mon­u­men­tal, and a failure to ap­pre­ci­ate this has led to con­sid­er­able over­es­ti­ma­tions of a coun­try’s abil­ity to do so

Prob­a­bly true. I was sur­prised by how many quotes Muel­ler had of peo­ple say­ing it’s triv­ial to build a nuke, given that we already know it some­times takes en­tire na­tion-states decades to suc­ceed even when de­vot­ing sub­stan­tial re­sources to the challenge (e.g. Pak­istan).

The pro­lifer­a­tion of nu­clear weapons has been far slower than rou­tinely pre­dicted be­cause, in­so­far as most lead­ers of most coun­tries (even rogue ones) have con­sid­ered ac­quiring the weapons, they have come to ap­pre­ci­ate sev­eral defects: the weapons are dan­ger­ous, dis­taste­ful, costly, and likely to rile the neighbors

The pro­lifer­a­tion of nu­clear weapons has definitely been slower than rou­tinely pre­dicted be­fore 1945 and there­after. Of course the causes are harder to as­sess.

The nu­clear diffu­sion that has tran­spired has proved to have had re­mark­ably limited, per­haps even im­per­cep­ti­ble, consequences

Mostly de­pends on your view about their effec­tive­ness in de­ter­rence.

Nu­clear pro­lifer­a­tion is not par­tic­u­larly de­sir­able, but it is also un­likely to ac­cel­er­ate or prove to be a ma­jor danger

Cer­tainly that is the trend so far.

Stren­u­ous efforts to keep “rogue states” from ob­tain­ing nu­clear weapons have been sub­stan­tially coun­ter­pro­duc­tive and have been a cause of far more deaths than have been in­flicted by all nu­clear deto­na­tions in history

It’s not his paradig­matic ex­am­ple, but the ex­am­ple of Taiwan is in­ter­est­ing. In the late 70s, the U.S. pres­sured Taiwan to aban­don any ac­tivi­ties rele­vant to de­vel­op­ing nukes, but I do won­der if the world would be bet­ter off if Taiwan had nukes, as a de­ter­rent against main­land China. And in the late 70s, China’s strug­gles were such that it might not have been able to stop Taiwan it­self.

The part about “far more deaths…” is of course true.

The weapons have not proved to be cru­cial sta­tus symbols

Well ex­cept for Rus­sia (a pretty im­por­tant case!) and North Korea (both cases Muel­ler ad­mits), and Libya’s pro­gram (which didn’t get far but Muel­ler ad­mits was largely about ego/​sta­tus).

Not only have nu­clear weapons failed to be of much value in mil­i­tary con­flicts, they also do not seem to have helped a nu­clear coun­try to swing its weight or “dom­i­nate” an area

Hard to say for sure, but AFAICT this is prob­a­bly true.

Given the low value of the weapons and their high costs, any suc­cesses in the an­tipro­lifer­tion effort have been mod­est and might well have hap­pened anyway

Stren­u­ous efforts to pre­vent nu­clear pro­lifer­a­tion can act as a spur to the pro­cess, en­hanc­ing the ap­peal of — or des­per­ate de­sire for — nu­clear weapons for at least a few regimes, an effect that is of­ten ignored

The pa­thetic North Korean regime mostly seems to be en­gaged in a pro­cess of ex­tract­ing aid and recog­ni­tion from out­side, and a vi­able policy to­ward it might be to re­duce the threat level and to wait while con­tin­u­ing to be ex­torted rather than to en­hance the already in­tense mis­ery of the North Korean people

If Iran ac­tu­ally does de­velop some­thing of an atomic ar­se­nal, it will likely find, fol­low­ing the ex­pe­rience of all other states so armed, that the bombs are es­sen­tially use­less and a very con­sid­er­able waste of money and effort

Although there is noth­ing wrong with mak­ing non­pro­lifer­a­tion a high pri­or­ity, it should be topped with a some­what higher one: avoid­ing poli­cies that can lead to the deaths of tens or hun­dreds of thou­sands of peo­ple un­der the ob­ses­sive sway of worst-case sce­nario fantasies

It is likely that no “loose nukes” — nu­clear weapons miss­ing from their proper stor­age lo­ca­tions and available for pur­chase in some way — exist

I haven’t found ev­i­dence against this claim, but didn’t look that hard.

It is likely there is no such thing as a true black mar­ket in nu­clear materials

Less sure about this one. From 1993-2018 there have been 12 state-re­ported cases of theft or loss of highly en­riched ura­nium, but the com­bined amount is much too small even to build a bomb with 1/​10th the yield of Hiroshima. CFR claims that “Rus­sian au­thor­i­ties say that in the past three years alone they have bro­ken up hun­dreds of nu­clear-ma­te­rial smug­gling deals,” but they don’t list a source and I haven’t looked harder.

The ev­i­dence of any de­sire on al-Qaeda’s part to go atomic and of any progress in ac­com­plish­ing this ex­ceed­ingly difficult task is re­mark­ably skimpy, if not com­pletely neg­ligible, while the scariest stuff—a decade’s worth of loose-nuke ru­mor and chat­ter and hype—seems to have no sub­stance whatever

Be­cause of a host of or­ga­ni­za­tional and tech­ni­cal hur­dles, the like­li­hood that ter­ror­ists will be able to build or ac­quire an atomic bomb or de­vice is van­ish­ingly small

Seems true AFAICT, at least for “atomic bomb.”

De­spite the sub­stan­tial ar­ray of threats reg­u­larly is­sued by al-Qaeda (the only ter­ror­ist group that may see at­tacks on the United States as de­sir­able), and de­spite the even more sub­stan­tial an­guish these threats have in­spired in their en­e­mies, the ter­ror­ist group’s ca­pac­ity seems to be quite limited

One rea­son for al-Qaeda’s re­mark­ably low ac­tivity in the last years is that 9/​11 proved to be sub­stan­tially coun­ter­pro­duc­tive from alQaeda’s stand­point; in­deed, with 9/​11 and sub­se­quent ac­tivity, the ter­ror­ist group seems mainly to have suc­ceeded in unit­ing the world, in­clud­ing its huge Mus­lim por­tion, against its vi­o­lent global jihad

Any threat pre­sented by al-Qaeda is likely to fade away in time, un­less, of course, the United States over­re­acts and does some­thing to en­hance their num­bers, pres­tige, and de­ter­mi­na­tion—some­thing that is, need­less to say, en­tirely possible