# Quantum versus logical bombs

Child, I’m sorry to tell you that the world is about to end. Most likely. You see, this mad­woman has de­signed a dooms­day ma­chine that will end all life as we know it—painlessly and im­me­di­ately. It is at­tached to a su­per­com­puter that will calcu­late the 10100th digit of pi—if that digit is zero, we’re safe. If not, we’re doomed and dead.

How­ever, there is one thing you are al­lowed to do—switchout the log­i­cal trig­ger and re­placed it by a quan­tum trig­ger, that in­stead gen­er­ates a quan­tum event that will pre­vent the bomb from trig­ger­ing with 1/​10th mea­sure squared (in the other cases, the bomb goes off). You ok pay­ing €5 to re­place the trig­gers like this?

If you treat quan­tum mea­sure squared ex­actly as prob­a­bil­ity, then you shouldn’t see any rea­son to re­place the trig­ger. But if you be­lieved in many wor­lds quan­tum me­chan­ics (or think that MWI is pos­si­bly cor­rect with non-zero prob­a­bil­ity), you might be tempted to ac­cept the deal—af­ter all, ev­ery­one will sur­vive in one branch. But strict to­tal util­i­tar­i­ans may still re­ject the deal. Un­less they re­fuse to treat quan­tum mea­sure as akin to prob­a­bil­ity in the first place (mean­ing they would ac­cept all quan­tum suicide ar­gu­ments), they tend to see a uni­verse with a tenth of mea­sure-squared as ex­actly equally val­ued to a 10% chance of a uni­verse with full mea­sure. And they’d even do the re­verse, re­place a quan­tum trig­ger with a log­i­cal one, if you paid them €5 to do so.

Still, most peo­ple, in prac­tice, would choose to change the log­i­cal bomb for a quan­tum bomb, if only be­cause they were slightly un­cer­tain about their to­tal util­i­tar­ian val­ues. It would seem self ev­i­dent that risk­ing the to­tal de­struc­tion of hu­man­ity is much worse than re­duc­ing its mea­sure by a fac­tor of 10 - a pro­cess that would be un­de­tectable to ev­ery­one.

Of course, once you agree with that, we can start squeez­ing. What if the quan­tum trig­ger only has 120 mea­sured-squared “chance” of sav­ing us? 1/​000? 1/​10000? If you don’t want to fully ac­cept the quan­tum im­mor­tal­ity ar­gu­ments, you need to stop—but at what point?

• I’m not sure what quan­tum me­chan­ics has to do with this. Say hu­man­ity is spread over 10 planets. Would you rather take a log­i­cal 910 chance of wiping out hu­man­ity, or de­stroy 9 of the planets with cer­tainty (and also de­stroy 90% of un­in­hab­ited planets to re­duce the po­ten­tial for fu­ture growth by the same de­gree)? Is there any eth­i­cally rele­vant differ­ence be­tween these sce­nar­ios?

• There are two differ­ences I can see :

1. The “planets” ex­am­ple ad­mit the MWI is cor­rect. Without MWI, the quan­tum trig­ger is ex­actly a nor­mal ran­dom trig­ger, not kil­ling 9/​10th of the wor­lds, but kil­ling ev­ery­one with a 1/​10th prob­a­bil­ity. The thought ex­per­i­ment is a way to force peo­ple to quan­tify their ac­cep­tance of MWI.

2. Com­mu­ni­ca­tion. There is no com­mu­ni­ca­tion pos­si­ble be­tween the MW, while there is be­tween the planets, and that will have mas­sive long term effects.

• Good point.

The two sce­nar­ios have some­what differ­ent in­tu­ition pumps, but are oth­er­wise similar.

• It seems like some­thing has gone ter­ribly wrong when our eth­i­cal de­ci­sions de­pend on our in­ter­pre­ta­tion of quan­tum me­chan­ics.

My un­der­stand­ing was that many-wor­lds is in­dis­tiguish­able by ob­ser­va­tion from the Copen­hagen in­ter­pre­ta­tion. Has this changed? If not, it fright­ens me that peo­ple would choose a higher chance of the world end­ing to res­cue hy­po­thet­i­cal peo­ple in un­ob­serv­able uni­verses.

If any­thing this seems like a (weak) ar­gu­ment in favour of to­tal util­i­tar­i­anism, in that it doesn’t suffer from giv­ing differ­ent an­swers ac­cord­ing to one’s choice among in­dis­tiguish­able the­o­ries.

• Why should you not have prefer­ences about some­thing just be­cause you can’t ob­serve it? Do you also not care whether an in­ter­galac­tic colony-ship sur­vives its jour­ney, if the colony will be be­yond the cos­molog­i­cal hori­zon?

• The de­pa­ture of an in­ter­galac­tic colony-ship is an ob­serv­able event. It’s not that the fu­ture of other wor­lds is un­ob­serv­able, it’s that their ex­is­tance in the first place is not a testable the­ory (though see army1987′s com­ment on that is­sue).

To make an anal­ogy (though ad­mit­tedly an un­fair one for be­ing a more com­plex rather than an ar­guably less com­plex ex­pla­na­tion): I don’t care about the lives of the fairies who carry rain­drops to the ground ei­ther, but it’s not be­cause fairies are in­visi­ble (well, to grown-ups any­way).

• The de­pa­ture of an in­ter­galac­tic colony-ship is an ob­serv­able event.

By the ex­act same to­ken, the world-state prior to the “split­ting” in a Many Wor­lds sce­nario is an ob­serv­able event.

I think the spirit of the ques­tion is ba­si­cally: In what situ­a­tions do we give cre­dence to hy­pothe­ses which posit sys­tems that we can in­fluence, but can­not in­fluence us?

• By the ex­act same to­ken, the world-state prior to the “split­ting” in a Many Wor­lds sce­nario is an ob­serv­able event.

The fal­ling of rain­drops is also ob­serv­able, you ap­pear to have missed the point of my re­ply.

To look at it an­other way, there is strong empyri­cal ev­i­dence that sen­tient be­ings will con­tinue to ex­ist on the colony-ship af­ter it has left, and I do not be­lieve there is analo­gous ev­i­dence for the con­tinued ex­is­tence of split-off par­allel uni­verses.

The spirit of the ques­tion is ba­si­cally this:

Can the most par­si­mo­nious hy­poth­e­sis ever posits sys­tems that you can in­fluence, but can­not causally in­fluence you? And if so, what does that mean for your prefer­ences?

No, the spirit of the ques­tion in con­text was to un­der­mine the ar­gu­ment that the untesta­bil­ity of a the­ory im­plies it should have no prac­ti­cal im­pli­ca­tions, a crit­i­cism I opened my­self up to by talk­ing about ob­serv­abil­ity rather than testa­bil­ity. The an­swer to the ques­tion was re­dun­dant to the ar­gu­ment, which was why I clar­ified my ar­gu­ment rather than an­swer it.

But since you want an an­swer, in prin­ci­ple yes I could care about things I can’t ob­serve, at least on a moral level. On a per­sonal level it’s a strong can­di­date for “some­body else’s prob­lem” if ever I’ve seen one, but that’s a whole other is­sue. Usu­ally the in­abil­ity to ob­serve some­thing makes it hard to know the right way to in­fluence it though.

• you ap­pear to have missed the point of my re­ply.

Let’s check: “I can only have prefer­ences over things that ex­ist. The ship prob­a­bly ex­ists, be­cause my mem­ory of its de­par­ture is ev­i­dence. The par­allel wor­lds have no similar ev­i­dence for their ex­is­tence.” Is that cor­rect para­phras­ing?

Be­fore the ship leaves, you know that some­time in the fu­ture there will be a fu­ture-ship in a lo­ca­tion where it can­not in­ter­act with fu­ture-you.

By the same to­ken, you can ob­serve the laws of physics and the pre­sent-state of the uni­verse. If, for some rea­son, your in­ter­pre­ta­tion of those laws in­volves Many Wor­lds split­ting off from each other, then, be­fore the wor­lds split, you know that some­time in the fu­ture there will be a fu­ture-world un­able to in­ter­act with fu­ture you.

For fu­ture-you, the ex­is­tence of the fu­ture-ship is not a testable the­ory, but the fact that you have a mem­ory of the ship leav­ing counts as ev­i­dence.

For fu­ture-you, the ex­is­tence of the Other-Wor­lds is not a testable the­ory, but if Many-Wor­lds is your best model, then your mem­ory of the past-state of the uni­verse, com­bined with your knowl­edge of physics, counts as ev­i­dence for the ex­is­tence of cer­tain spe­cific other wor­lds.

In your Faeries ex­am­ple, the Faeries do not merit con­sid­er­a­tion be­cause it is im­pos­si­ble to get ev­i­dence for their ex­is­tence. That’s not true in the quan­tum bomb sce­nario—if we ex­cept Many Wor­lds, then for the sur­vivors of the quan­tum bomb, the mem­ory of the ex­is­tence of a quan­tum bomb is ev­i­dence that there ex­ist many branches with Other Wor­lds in which ev­ery­one was wiped out by the bomb.

So, the ac­tual ques­tion should be:

1) Does Many-Wor­lds fit in our on­tol­ogy—as in, do uni­verses on other branches con­structed in the Many-World for­mat even fit within the defi­ni­tion of “Real­ity” or not? (For ex­am­ple, if you told me there was a par­allel uni­verse which never in­ter­acted with us in any way, I’d say that your uni­verse wasn’t Real by defi­ni­tion. Many Wor­lds branches are a gray area be­cause they do in­ter­act, but cur­rent Other Wor­lds only in­ter­act with the past and the pre­sent only in­ter­acts with fu­ture Other Wor­lds, not cur­rent ones )

2a) If we de­cide that the Other Wor­lds from Many Wor­lds qual­ify as “Real”, can Many Wor­lds ever be a hy­poth­e­sis which is Par­si­mo­nious enough to not be Pas­cal-Wager-ish? The Faeries qual­ify as “Real” be­cause they do cause the rain­drops to fall, but be­cause of the na­ture of that hy­poth­e­sis it can never be par­si­mo­nious enough to rise above Pas­cal-Wager-thresh­olds. Is Many-Wor­lds the same way? (From your an­swer, I gath­ered that your an­swer is “yes”, but I dis­agreed with your rea­son—see para­graph that be­gins with “In your Faeries ex­am­ple...” which is why I pointed out that if you ac­cept Many Wor­lds then you can have ev­i­dence that points to cer­tain sorts of wor­lds ex­ist­ing in my first re­ply.)

2b) If we de­cide that the other branches do not qual­ify as Real, can we make a defi­ni­tion of re­al­ity that does not ex­clude light-cone-leav­ing-space­ships?

3) And how do we con­struct our prefer­ences, in re­la­tion to what we have defined as “Real”? (For ex­am­ple, we could sim­ply say that de­spite hav­ing an on­tol­ogy that ac­knowl­edges all the branches of Many Wor­lds as Real, our prefer­ences only care about the world that we end up in.)

• The space­ship “ex­ists” (I don’t re­ally like us­ing ex­ists in this con­text be­cause it is con­fus­ing) in the sense that in the fu­tures where some­one figures out how to break the speed of light, I know I can in­ter­act with the space­ship. What is the prob­a­bil­ity that I can break the speed of light in the fu­ture?

Then for Many Wor­lds, what is the prob­a­bil­ity that I will be able to in­ter­act with one of the Other Wor­lds?

I would not care more about things if I gain in­for­ma­tion that I can in­fluence them, un­less I also gain in­for­ma­tion that they can in­fluence me. If I gain cre­dence in Many Wor­lds, then I only care about Other Wor­lds to the ex­tent that it might be more likely for them to in­fluence my world.

• We’re as­sum­ing you can’t break the speed of light or in­ter­act with the other wor­lds.

It’s a one-way in­fluence. You can in­fluence the space­ship be­fore it leaves your light cone (you can give them sup­plies, etc). The MW ar­gu­ment is that you can in­fluence par­allel uni­verses be­fore they split off.

• Let’s check: “I can only have prefer­ences over things that ex­ist. The ship prob­a­bly ex­ists, be­cause my mem­ory of its de­par­ture is ev­i­dence. The par­allel wor­lds have no similar ev­i­dence for their ex­is­tence.” Is that cor­rect para­phras­ing?

No, not re­ally. I mean, it’s not that far from some­thing I said, but it’s de­part­ing from what I meant and it’s not in any case the point of my re­ply. The mis­take I’m mak­ing is per­sist­ing in try­ing to clar­ify a par­tic­u­lar way of view­ing the prob­lem which is not the best way and which is lead­ing us both down the gar­den path. In­stead, please for­get ev­ery­thing else I said and con­sider the fol­low­ing ar­gu­ment.

The­o­ries have two as­pects. Testable pre­dic­tions, and de­scrip­tive el­e­ments. I would (and I think the se­quences sup­port me) ar­gue that two the­o­ries which make the same pre­dic­tions are not differ­ent the­o­ries, they are the same the­ory with differ­ent flavour. In par­tic­u­lar, you should never make a differ­ent de­ci­sion un­der one the­ory than un­der the other. Many Wor­lds is a flavour of quan­tum me­chan­ics, and if that choice of flavour effects eth­i­cal de­ci­sions then you are mak­ing differ­ent de­ci­sions ac­cord­ing to the flavour rather than con­tent of the the­ory, and some­thing has gone wrong.

Every­thing else I said was in­tended solely to sup­port that point, but some­where along the way we got lost ar­gu­ing about what’s ob­serv­able, what con­si­tutes ev­i­dence and meta-ethics. If you ac­cept that ar­gu­ment then I have no fur­ther point to make. If you do not ac­cept it, then please di­rect com­ments at that ar­gu­ment di­rectly rather than any­thing else I’ve said.

I’ll try to ad­dress the rest of your re­ply with this in mind in the hopes that it’s helpful.

If … your in­ter­pre­ta­tion of those laws in­volves Many Worlds

You could equally have said “If your in­ter­pre­ta­tion of the physics of rain­drops in­volves fairies”. My point is that no-one has any jus­tifi­ca­tion for mak­ing that as­sump­tion. Quan­tum physics is a whole bunch of maths that mod­els the be­havi­our of par­ti­cles on a small scale. Many Wor­lds is one of many pos­si­ble de­scrip­tions of that maths that help us un­der­stand it. If you ar­bi­trar­ily as­sume your de­scrip­tion is a mean­ingful prop­erty of re­al­ity then sure, ev­ery­thing else you say fol­lows log­i­cally, but only be­cause the mis­take was made already.

You com­pare Many Wor­lds to fairies in the wrong place, in par­tic­u­lar post-ar­bi­trary-as­sump­tion for Many Wor­lds and pre-ar­bi­trary-as­sump­tion for fairies. I’ll give you the analo­gous state­ments for a cor­rect com­par­i­son:

the Faeries do not merit con­sid­er­a­tion be­cause it is im­pos­si­ble to get ev­i­dence for their existence

The peo­ple of other wor­lds do not merit con­sid­er­a­tion be­cause it is im­pos­si­ble to get ev­i­dence of their ex­is­tance.

if we ex­cept Many Wor­lds...

If we ac­cept fairies...

… the mem­ory of the ex­is­tence of a quan­tum bomb is ev­i­dence that there ex­ist many branches with Other Wor­lds in which ev­ery­one was wiped out by the bomb

… the sight of a rain­drop fal­ling is ev­i­dence that there ex­ists a fairy a short dis­tance away.

• Ta­boo “jus­tifi­ca­tion”. Jus­tifi­ca­tion is es­sen­tially a poin­ter to ev­i­dence or in­fer­ence. After all the in­fer­ence is said and done, the per­son who needs to provide more ev­i­dence is the per­son who has the more un-par­si­mo­nious hy­poth­e­sis. You re­ject fairies based on a lack of jus­tifi­ca­tion be­cause it’s not par­si­mo­nious. You can’t re­ject Many-Wor­lds on those same grounds, at least not with­out ex­plain­ing more.

The differ­ence is that the fairies in­ter­pre­ta­tion of rain­drops has differ­ent maths than the non-fairy in­ter­pre­ta­tion of rain­drops. When the math­e­mat­i­cally-rigor­ous de­scrip­tions for two differ­ent hy­pothe­ses are differ­ent, there is a clear cor­rect an­swer as to which is more par­si­mo­nious.

Many-wor­lds has ex­actly the same math­e­mat­i­cal de­scrip­tion as the al­ter­na­tive, so it’s hard to say which is more par­si­mo­nious. You can’t say that Sin­gle-World is de­fault and Many Wor­lds re­quires jus­tifi­ca­tion. This is why I claim that it is first a ques­tion of on­tol­ogy (a ques­tion of what we choose to define as re­al­ity), and then maybe we can talk about the episte­mol­ogy and whether or not the state­ment is “True” within our defi­ni­tions...af­ter we clar­ify our on­tol­ogy and define the re­la­tion­ship be­tween on­tol­ogy and par­si­mony, not be­fore.

• ‘Copen­hagen’ isn’t so much an in­ter­pre­ta­tion as a rel­a­tively tra­di­tional, rel­a­tively au­thor­i­ta­tive body of physics slo­gans. Depend­ing on which Copen­hagenist you speak to, the in­ter­pre­ta­tion might amount to Ob­jec­tive Col­lapse, or Oper­a­tional­ism, or Me­ta­phys­i­cal Ideal­ism, or Quietism. The lat­ter three aren’t so much al­ter­na­tives to MWI as al­ter­na­tives to the very prac­tice of main­stream sci­en­tific re­al­ism; and Ob­jec­tive Col­lapse is gen­er­ally em­piri­cally dis­tinct from MWI (and, to the ex­tent that it has made testable pre­dic­tions, these have always been falsified.)

Bohmian Me­chan­ics is an al­ter­na­tive to the MWI fam­ily of in­ter­pre­ta­tions that re­ally does look em­piri­cally in­dis­t­in­guish­able. But it’s about as differ­ent from Copen­hagenism as you can get, and is al­most uni­ver­sally dis­missed by physi­cists. Also, it may not solve this prob­lem; I haven’t seen dis­cus­sion of whether the com­plex­ity of the BM pi­lot wave is likely to it­self en­code an over­whelming pre­pon­der­ance of men­tal ‘rip­ples’ that crowd out the moral weight of our own world. Are par­ti­cles needed for com­plex biol­ogy-like struc­ture in BM?

• It seems like some­thing has gone ter­ribly wrong when our eth­i­cal de­ci­sions de­pend on our in­ter­pre­ta­tion of quan­tum me­chan­ics.

Yes. Some­one has hooked up a uni­verse-de­stroy­ing bomb and is offer­ing to make the out­come quan­tum. I think that cov­ers it.

• My un­der­stand­ing was that many-wor­lds is in­dis­tiguish­able by ob­ser­va­tion from the Copen­hagen in­ter­pre­ta­tion. Has this changed?

Ac­cord­ing to MWI you can put ar­bi­trar­ily large sys­tems into quan­tum su­per­po­si­tion, whereas ac­cord­ing to CI when the sys­tem is suffi­ciently large the wave­func­tion will col­lapse.

• Ac­cord­ing to MWI you can put ar­bi­trar­ily large sys­tems into quan­tum superposition

Yes and no. Ac­cord­ing to MWI, there is no the­o­ret­i­cal limit to how a large a sys­tem in quan­tum su­per­po­si­tion can be, yes. But to keep the sys­tem in quan­tum su­per­po­si­tion with­out mak­ing two (or more) words that will never in­ter­act with each other again, you’ve to keep them from in­ter­act­ing with the rest of the world (in a way that is linked to the kind of su­per­po­si­tion). And prac­ti­cally that is very hard to do for large-scale sys­tems. That’s an is­sue with quan­tum com­put­ing for ex­am­ple, the more qbits you try to add, the harder it is to keep them iso­lated.

• But the point would re­main in that case that there is in prin­ci­ple an ex­per­i­ment to dis­t­in­guish the the­o­ries, even if such an ex­per­i­ment has yet to be performed?

Although (and I ad­mit my un­der­stand­ing of the topic is be­ing stretched here) it still doesn’t sound like the cen­tral is­sue of the ex­is­tance of par­allel uni­verses with which we may no longer in­ter­act would be re­solved by such an ex­per­i­ment. It seems more like Copen­hagen’s lat­est at­tempt to define the con­di­tions for col­lapse would be dis­proven with­out par­tic­u­larly ne­ces­si­tat­ing a fun­da­men­tal change of in­ter­pre­ta­tion.

• For Copen­hagen, yes, but MWI and Copen­hagen aren’t the only two in­ter­pre­ta­tions of quan­tum me­chan­ics worth think­ing about.

In truth, you’ll find few physi­cists who treat the Copen­hagen In­ter­pre­ta­tion as any­thing but con­ve­nient short­hand, not usu­ally for MWI.

• If we man­aged to put hu­man-sized sys­tems into su­per­po­si­tion, that’d rule out CI AFAICT. And be­fore that, the larger the sys­tems we man­age to put into su­per­po­si­tion the less likely CI will seem.

• Good point. De­co­her­ence makes MWI de facto in­dis­t­in­guish­able from CI ex­cept that the max­i­mum size of sys­tems you can put into su­per­po­si­tions de­pends on the tem­per­a­ture rather than grav­ity/​con­scious­ness/​what­ever.

• If any­thing this seems like a (weak) ar­gu­ment in favour of to­tal util­i­tar­i­anism, in that it doesn’t suffer from giv­ing differ­ent an­swers ac­cord­ing to one’s choice among in­dis­tiguish­able the­o­ries.

Oh, to­tal util­i­tar­i­anism has its own prob­lem with in­dis­t­in­guish­able the­o­ries :-)

• Fair point, it sounds like it’s a co-in­ci­den­tal vic­tory for to­tal-util­i­tar­i­anism in this par­tic­u­lar case.

• In my un­der­stand­ing of Many Wor­lds In­ter­pre­ta­tion, I’m un­der the im­pres­sion that there should ex­ist some mea­sure in which the su­per­com­puter ac­ci­den­tally com­putes 0 on a non zero log­i­cal trig­ger be­cause too many quan­tum events hap­pened to hit a bit and hence doesn’t blow up, and for that mat­ter, the re­verse: Too many too quan­tum events hit a bit, and turned a log­i­cally com­puted 0 into some other num­ber. Pre­sum­ably these chances are there, and they rep­re­sent some small, non zero mea­sure.

But as far as I can tell, the fact that this oc­curs seems to ren­der a sub­stan­tial premise of the ques­tion moot, be­cause it means nei­ther trig­ger hits 0 quan­tum mea­sure, which is what we were pos­tu­lat­ing pay­ing money to avoid.

So to ver­ify this:

1: Am I cor­rect that MWI does im­ply that there will be some quan­tum mea­sure left even when us­ing the log­i­cal trig­ger?

2: Does this ac­tu­ally ren­der a sub­stan­tial part of the origi­nal sce­nario moot, or does it still ap­ply for rea­sons I don’t un­der­stand?

• That de­pends if you have some kind of “man­gled wor­lds” hy­poth­e­sis (in short, an hy­poth­e­sis that wor­lds with a too low prob­a­bil­ity will be un­sta­ble and col­lapse due to con­tam­i­na­tion from “nearby” wor­lds).

As long as we don’t know where the Born rule comes from in MWI, it’s hard to say if all wor­lds are “real” and how much “real” they are, or if there is some kind of bound­ary be­low which the world isn’t “real” for prac­ti­cal pur­pose (like, not sta­ble enough to al­low a con­scious­ness to ex­ist in it).

1. Yes.

2. It de­pends—those will­ing to go all the way up to quan­tum suicide are fine with very very low mea­sure, but oth­ers might not go there.

• Are we as­sum­ing that ev­ery in­hab­ited Everett branch has such a dooms­day de­vice, and the same de­ci­sion about the trig­ger will be made in each branch? If you’re only one of a huge num­ber of uni­verses, then 90% of your branch dy­ing vs. a 90% chance of your branch dy­ing isn’t go­ing to make much of a differ­ence.

• How do you see that? Most con­se­quen­tial­ist the­o­ries would as­sume that “par­allel uni­verses” that you can’t af­fect have limited im­pact on your choices in this one.

• You could use a quan­tum ran­dom num­ber gen­er­a­tor to make your de­ci­sion. Then you en­sure there is an Everett branch in which hu­man­ity con­tinues to ex­ist, but you only pay the \$5 1% (say) of the time.

• That’s only true if your util­ity func­tion is lin­ear. If your util­ity func­tion is non­lin­ear, and you care about hu­man­ity ex­ist­ing, but you don’t care as much about how much hu­man­ity ex­ists, then a dooms­day de­vice isn’t nearly as bad if you know hu­man­ity will con­tinue to ex­ist in a par­allel uni­verse. I as­sumed that this is why some­one would pre­fer 90% of the mea­sure of the uni­verse be­ing de­stroyed to a 90% chance of the whole thing be­ing de­stroyed. Is there an­other rea­son you’d pre­fer the former?

• Is there an­other rea­son you’d pre­fer the former?

There are some to­tal util­i­tar­i­ans who are (or would like to be) in­differ­ent be­tween the two op­tions—I’ve chat­ted with them.

• If the dilemma is only tak­ing place in a small por­tion of the branches, the other branches will sur­vive re­gard­less of the choice, which breaks the ar­gu­ment about many-wor­lds to­tal ex­tinc­tion risk.

• I as­sumed that, even if most branches don’t have the ma­chine, the ma­chine’s in­fluence reaches to all branches, so that it can de­stroy all of them along with ours.

• The thought ex­per­i­ment is about elic­it­ing some of the nor­ma­tive con­tent of truth vs. falsity of MWI, in terms that don’t as­sume MWI. The mean­ing of “de­stroy all MWI branches” is given in terms of MWI, so this clause wouldn’t re­spect the mo­ti­va­tion of the thought ex­per­i­ment.

• The thought ex­per­i­ment is about elic­it­ing some of the nor­ma­tive con­tent of truth vs. falsity of MWI, in terms that don’t as­sume MWI.

That is not my read­ing. Con­sider this part:

If you treat quan­tum mea­sure squared ex­actly as prob­a­bil­ity, then you shouldn’t see any rea­son to re­place the trig­ger. But if you be­lieved in many wor­lds quan­tum me­chan­ics (or think that MWI is pos­si­bly cor­rect with non-zero prob­a­bil­ity), you might be tempted to ac­cept the deal—af­ter all, ev­ery­one will sur­vive in one branch.

The post then goes on to ar­gue that there is a dilemma here, that an ap­par­ently plau­si­ble case can be made for ei­ther choice, as­sum­ing that MWI is true.

I take the post to be say­ing, “Here’s an in­ter­est­ing dilemma. Well, it’s only in­ter­est­ing if there’s a pos­si­bil­ity that MWI is true. That is, if you know that MWI is false, then the an­swer is ob­vi­ous. But, grant­ing the pos­si­bil­ity of MWI for the sake of ar­gu­ment, what would you do?”

• If many qual­i­ta­tively differ­ent branches are con­se­quen­tial­is­ti­cally op­ti­miz­ing a com­mon goal, all these branches be­come bet­ter ac­cord­ing to that goal, even if the spe­cific situ­a­tions and ac­tions taken in them are sig­nifi­cantly differ­ent. On the other hand, if these branches re­spond to your ar­gu­ment and aban­don their op­ti­miza­tion due to rar­ity of their par­tic­u­lar situ­a­tions and pos­si­ble ac­tions, all the branches would re­main poorly op­ti­mized.

(More gen­er­ally, an op­ti­mizer doesn’t care about the scope of their in­fluence, only about com­par­i­son of available choices, how­ever in­signifi­cant.)

• Still, most peo­ple, in prac­tice, would choose to change the log­i­cal bomb for a quan­tum bomb,

“Most peo­ple” in which refer­ence class?

• The thing miss­ing here is that there are plenty of branches where the dooms­day de­vice was never built, or a differ­ent digit was cho­sen, and hu­man­ity will con­tinue in those.

• Wedrifid’s com­ment here is similar to this post. If you re­ally value mea­sure and prob­a­bil­ity differ­ently, then you should be open to quite strange trade pro­pos­als. For me, this serves as a re­duc­tio ad ab­sur­dum, my con­clu­sion is that one re­ally should value them the same.

• Would you similarly con­sider 90% chance of to­tal ex­tinc­tion and de­struc­tion of the uni­verse, equiv­a­lent with kil­ling 9/​10th of the hu­man race (and 9/​10th of the uni­verse)?

• Hm, in that case I guess I would pre­fer de­ter­minis­ti­cally de­stroy­ing 90% of the uni­verse. A uni­verse with 10% per­sons is a big­ger im­prove­ment over the empty uni­verse than a uni­verse with 100% per­sons is over the 10% uni­verse, so we have a con­vex util­ity func­tion and risk aver­sion. As you write, to be in­differ­ent I guess I would have to sub­scribe to a “strong to­tal util­i­tar­ian” prin­ci­ple, that I value a uni­verse ex­actly ac­cord­ing to the num­ber of peo­ple in it.

I take it that your ar­gu­ment is that the same rea­son­ing should ap­ply to mul­ti­verses, and we should pick the al­ter­na­tive that leaves a guaran­teed 10% re­main­ing Everett branches? That’s a neat per­spec­tive which I had not ap­pre­ci­ated be­fore: the quan­tum pro­cess is more de­ter­minis­tic than the log­i­cal pro­cess, since it de­stroys (part of) the mul­ti­verse de­ter­minis­ti­cally in­stead of sub­jec­tively-ran­domly.

I note that for this ar­gu­ment to go through, you need that our util­ity func­tion over differ­ent mul­ti­verses re­ally is con­vex, which you didn’t ar­gue for. We choose be­tween differ­ent uni­verses ev­ery day, so we have rel­a­tively well-de­vel­oped in­tu­itions about which uni­verse we pre­fer. I rarely think about which mul­ti­verse I pre­fer to live in, so I’m less con­fi­dent that the util­i­tar­ian prin­ci­ple is wrong there. And re­ject­ing it leads to the strange trades.

I feel this dis­cus­sion doesn’t cap­ture my first in­tu­ition about the prob­lem, though, which is about sub­jec­tive ver­sus ob­jec­tive prob­a­bil­ity. If I knew what the 10^100th digit was, then of course I would have an opinion about whether I wanted the digit or the qubit to de­cide. But when I don’t know ei­ther way, it seems weird to care about the ex­act mechanism used to set off the bomb. So it seems that not only do I value the col­lec­tion of quan­tum fu­ture wor­lds as a sum weighted by their mea­sure, I would also ex­tend this to the col­lec­tion of pos­si­ble fu­ture wor­lds. (Where a world is pos­si­ble if as far as I know it could hap­pen).

• I note that for this ar­gu­ment to go through, you need that our util­ity func­tion over differ­ent mul­ti­verses re­ally is con­vex, which you didn’t ar­gue for.

I would agree with that (on grounds of ex­ces­sive du­pli­ca­tion of al­most-iden­ti­cal agents, for one), but the point for this post is more to get peo­ple re­ac­tions than to push a par­tic­u­lar the­ory.

• This is a dis­cus­sion about sa­cred val­ues with no prac­ti­cal con­se­quences, so it’s hard to get hon­est an­swers. I mean, I’d pay what­ever it took to use the quan­tum trig­ger, how­ever small the mea­sure on it was, be­cause I care about hu­man­ity. If you wouldn’t, what kind of mon­ster are you?

• It’s prob­a­bly ab­stract enough that the peo­ple who fre­quent less wrong are will­ing to be hon­est.