Quantum suicide seems like a good idea to me if we know that the assumptions behind it (both quantum and identity-related) are true, if we’re purely selfish (eg don’t care about the bereaved left behind), and if we don’t assume our actions are sufficiently correlated with those of others to make everyone try quantum suicide and end up all alone in our own personal Everett branch.
However, I might have the same “It’s a good idea, but I am going to refuse to do this for reasons of personal sanity” reaction as I have with Pascal’s Mugging.
Quantum suicide seems like a good idea to me if we know that the assumptions behind it (both quantum and identity-related) are true, if we’re purely selfish (eg don’t care about the bereaved left behind), and if we don’t assume our actions are sufficiently correlated with those of others to make everyone try quantum suicide and end up all alone in our own personal Everett branch.
Fortunately, if you combine the second and third potential problems you end up with a solution that eliminates both of them. Then you just have the engineering problem involved in building a bigger death box.
However, I might have the same “It’s a good idea, but I am going to refuse to do this for reasons of personal sanity” reaction as I have with Pascal’s Mugging.
I hope so. Your position is entirely consistent—I cannot fault it on objective grounds and what you say in your post does directly imply what you confirm in your comment. That said, the preferences you declare here are vastly different to those that I consider ‘normal’ and so there remains the sneaking suspicion that you are wrong about what you want. That is, that you incorrectly extrapolate your volition.
On the other hand the existence of people with the preferences you describe here is a great potential boon to the rest of us. Whenever parties have vastly different values there is the potential for trade between them. And the difference in values between those that care about measure and those that don’t rounds off to absolute. When you act on your preferences we can essentially just inherit all of your stuff in exchange for a (from our perspective) token probabilistic payout. Everybody wins!
(Not sure where to put this:) Yvain’s position doesn’t seem sane to me, and not just for reasons of preference; attempting to commit suicide will just push most of your experienced moments backwards to regions where you’ve never heard of quantum suicide or where for whatever reason you thought it was a stupid idea. Anticipating ending up in a world with basically no measure just doesn’t make sense: you’re literally making yourself counterfactual. If you decided to carve up experience space into bigger chunks of continuity then this problem goes away, but most people agree that (as Katja put it) “anthropics makes sense with shorter people”. Suicide only makes sense if you want to shift your experience backwards in time or into other branches, not in order to have extremely improbable experiences. I mean, that’s why those branches are extremely improbable: there’s no way you can experience them, quantum suicide or no.
Yvain’s position doesn’t seem sane to me, and not just for reasons of preference; attempting to commit suicide will just push most of your experienced moments backwards to regions where you’ve never heard of quantum suicide or where for whatever reason you thought it was a stupid idea.
This doesn’t seem to be a problem that comes from being quantum suicidal but rather from an entirely different kind of anthropic based suicidal insanity. That is, I would not predict that experience as evaluated by Yvain’s model of caring would be perceived this way. It certainly could be but that would be an additional insanity to the one that makes quantum roulette desirable. (No offense to Yvain and his Quantum Suicidal ilk by referring to this as ‘insanity’. I mean only ‘drastically different preferences preferences to my own in an agent similar enough to me that such comparison is meaningful’. In fact, if you’re going to limit your optimisation to tiny amounts of measure then go ahead and exterminate humanity to maximise paperclips for all I care!)
To expand somewhat: Quantum suiciding at (subjective) time t results in you at time t-1 having more measure than you at time t+1 but under default quantum-suicidal preferences these are in no way in competition. Relative measure between past and future selves isn’t any particular issue. There are just various different subjective experiences at t-1, t and t+1, a desire to have each of them as positive-on-average as can be but no particular inclination to trim measure in one part of a timeline to increase it in another. For example, I wouldn’t expect Yvain to (consider it rational to) commit conventional-and-complete suicide whenever it seemed like all his peak experiences are in the past and all that remained in life is to make the most of the remaining dregs.
You’re not saying that if I perform QS I should literally anticipate that my next experience will be from the past, are you? (AFAICT, if QS is not allowed, I should just anticipate whatever I would anticipate if I was going to die everywhere, that is, going to lose all of my measure.)
(Not Will, but I think I mostly agree with him on this point)
There is no such thing as an uniquely specified “next experience”. There are going to be instances of you that remember being you and consider themselves the same person as you, but there is no meaningful sense in which exactly one of them is right. Granted, all instances of you that remember a particular moment will be in the future of that moment, but it seems silly to only care about the experiences of that subset of instances of you and completely neglect the experiences of instances that only share your memories up to an earlier point. If you weight the experiences more sensibly then in the case of a rigorously executed quantum suicide the bulk of the weight will be in instances that diverged before the decision to commit quantum suicide. There will be no chain of memory leading from the QS to those instances, but why should that matter?
So, if Omega was willing to put a copy of you in an Everett branch that didn’t already have one, how much money would you be willing to bid for this service?
If Omega was going to charge $100, and the offer remained open for as many Everett branches as you wanted, how many $100s would you give Omega?
So, if Omega was willing to put a copy of you in an Everett branch that didn’t already have one, how much money would you be willing to bid for this service?
I’m not used to evaluating the worth of Everett branches by count. But for the purpose of this question may I assume you mean “another Everett branch of equal measure to this one, as of the time I click ‘comment’”?
As for an answer… um… I’m not sure, a fair bit? Working out my preferences—quantitatively - in situations so far outside the usual realm of operation is tricky.
If Omega was going to charge $100, and the offer remained open for as many Everett branches as you wanted, how many $100s would you give Omega?
After I gave him everything I had I would get a new job that more closer matched my potential for financial gain.
Two extra considerations:
Even aside from a terminal preference for measure maximization I would consider buying more measure purely for the purpose of giving me acausal bargaining power. (I’m even less sure about qualitatively evaluating the usefulness of acausal bargaining power.)
Buying more equal-measure branches is different to trying to preserve measure in the one we are in. While I think I have preferences such that I would buy a new one I’m not sure if the default behavior of humans would be to do so.
But would you take a partial amnesia pill before you go to sleep (which acts while you’re not conscious and does or does not dissolve based on a quantum event)? As a partial quantum suicide. What about total amnesia? How much of your state has to be lost until its a suicide where QM will save you? Why won’t QM save you from quantum-hangover where you do or don’t get a terrible headache and IQ impairment of 20 points?
Why won’t QM save you from quantum-hangover where you do or don’t get a terrible headache and IQ impairment of 20 points?
Just because. Really, that’s all there is to it. It so happens that Yvain’s (alleged) subjectively objective preferences are such that given the existence of a non-hangover branch the existence of a hangover branch is less desirable than if the hangover was substituted for death in that branch.
This is not entirely unrelated to the natural outcome of ‘average utilitarianism’. Under that value system the most simple solution is to go around killing the unhappiest person in the world until such time as the remaining people would be made on average less happy by the removal of the lowest happiness people. Sure, this is totally crazy. But if you have crazy preferences you should do crazy things.
Well, that was kind of rhetorical question; I want to make him be less into quantum-suicide by appeal to how (my guess) he’s not expecting MWI to save him from experiencing quantum-hangover.
edit: to explain, I myself believe death to be a real value from epsilon to 1 rather than a binary value, the epsilon being the shared memories remaining in other people, other people nearby who think like you, etc. and the 1 being the state where you are never forgetting anything at all, with living being somewhere close to 1, and amnesias somewhere between 0 and that. Having binary valued anything seem to result in really screwed up behaviours. (by other people nearby i mean within distances that are much smaller than your information content; the other people at distances that have as many bits in their relative coordinate as you have in your brain, it seems to me those really should be counted as substantially different even though I can’t quite pin point why)
It was a question that got a straight answer that is in accord with Yvain’s position. To the extent that such rhetorical questions get answers that do not require the rhetoric victim to change their mind, the implied argument can be considered refuted.
In practical social usage rhetorical questions can, indeed, often be used as a way to make arguments that an opponent is not allowed to respond to. Here on lesswrong we are free to reject that convention and so not only are literal answers to rhetorical questions acceptable, arguments that are hidden behind rhetorical questions should be considered at least as subject to criticism as arguments made openly.
Well, it is the case that humans often strive to have consistent goal systems (perhaps minimizing their descriptive complexity?), so while he can just say ‘because I defined my goal system to be this’ he is also likely to think and try to come up with some kind of general principle that does not have weird discontinuities at some point when the amnesia is too much like death for his taste.
edit: i think we are talking about different issues; i’m not making a point about his utility function, i’m making a point that he is expecting to become ‘him 20 points dumber and with a terrible headache’, who’s a rather different person, rather than to become someone in a galaxy far far away who doesn’t have the hangover and is thus more similar to him before the hangover.
Well, it is the case that humans often strive to have consistent goal systems (perhaps minimizing their descriptive complexity?), so while he can just say ‘because I defined my goal system to be this’
Yvain does present a consistent goal system. It is one that may appear either crazy or morally abhorrent to us but all indications are that it is entirely consistent. If you were attempting to demonstrate to Yvain an inconsistency in his value system that requires arbitrary complexity to circumvent then you failed.
I think you’re misunderstanding me, see edit. The point I am making is not so much about his values, but about his expectations of subjective experience.
Well, my argument is that you can propose a battery of possible partial quantum suicide set ups involving a machine that partially destroys you (e.g. you are anaesthetised and undergo lobotomy with varying extent of cutting, or something of this sort such as administration of a sublethal dose of neurotoxin). At some point, there’s so little of you left that you’re as good as dead; at some other point, there’s so much of you left that you don’t really expect to be quantum-saved. Either he has some strange continuous function inbetween, that I am very curious about, or he has a discontinuity, which is weird. (and I am guessing a discontinuity but i’d be interested to hear about function)
Quantum suicide seems like a good idea to me if we know that the assumptions behind it (both quantum and identity-related) are true, if we’re purely selfish (eg don’t care about the bereaved left behind), and if we don’t assume our actions are sufficiently correlated with those of others to make everyone try quantum suicide and end up all alone in our own personal Everett branch.
However, I might have the same “It’s a good idea, but I am going to refuse to do this for reasons of personal sanity” reaction as I have with Pascal’s Mugging.
Fortunately, if you combine the second and third potential problems you end up with a solution that eliminates both of them. Then you just have the engineering problem involved in building a bigger death box.
I hope so. Your position is entirely consistent—I cannot fault it on objective grounds and what you say in your post does directly imply what you confirm in your comment. That said, the preferences you declare here are vastly different to those that I consider ‘normal’ and so there remains the sneaking suspicion that you are wrong about what you want. That is, that you incorrectly extrapolate your volition.
On the other hand the existence of people with the preferences you describe here is a great potential boon to the rest of us. Whenever parties have vastly different values there is the potential for trade between them. And the difference in values between those that care about measure and those that don’t rounds off to absolute. When you act on your preferences we can essentially just inherit all of your stuff in exchange for a (from our perspective) token probabilistic payout. Everybody wins!
(Not sure where to put this:) Yvain’s position doesn’t seem sane to me, and not just for reasons of preference; attempting to commit suicide will just push most of your experienced moments backwards to regions where you’ve never heard of quantum suicide or where for whatever reason you thought it was a stupid idea. Anticipating ending up in a world with basically no measure just doesn’t make sense: you’re literally making yourself counterfactual. If you decided to carve up experience space into bigger chunks of continuity then this problem goes away, but most people agree that (as Katja put it) “anthropics makes sense with shorter people”. Suicide only makes sense if you want to shift your experience backwards in time or into other branches, not in order to have extremely improbable experiences. I mean, that’s why those branches are extremely improbable: there’s no way you can experience them, quantum suicide or no.
Here is fine.
This doesn’t seem to be a problem that comes from being quantum suicidal but rather from an entirely different kind of anthropic based suicidal insanity. That is, I would not predict that experience as evaluated by Yvain’s model of caring would be perceived this way. It certainly could be but that would be an additional insanity to the one that makes quantum roulette desirable. (No offense to Yvain and his Quantum Suicidal ilk by referring to this as ‘insanity’. I mean only ‘drastically different preferences preferences to my own in an agent similar enough to me that such comparison is meaningful’. In fact, if you’re going to limit your optimisation to tiny amounts of measure then go ahead and exterminate humanity to maximise paperclips for all I care!)
To expand somewhat: Quantum suiciding at (subjective) time t results in you at time t-1 having more measure than you at time t+1 but under default quantum-suicidal preferences these are in no way in competition. Relative measure between past and future selves isn’t any particular issue. There are just various different subjective experiences at t-1, t and t+1, a desire to have each of them as positive-on-average as can be but no particular inclination to trim measure in one part of a timeline to increase it in another. For example, I wouldn’t expect Yvain to (consider it rational to) commit conventional-and-complete suicide whenever it seemed like all his peak experiences are in the past and all that remained in life is to make the most of the remaining dregs.
You’re not saying that if I perform QS I should literally anticipate that my next experience will be from the past, are you? (AFAICT, if QS is not allowed, I should just anticipate whatever I would anticipate if I was going to die everywhere, that is, going to lose all of my measure.)
(Not Will, but I think I mostly agree with him on this point)
There is no such thing as an uniquely specified “next experience”. There are going to be instances of you that remember being you and consider themselves the same person as you, but there is no meaningful sense in which exactly one of them is right. Granted, all instances of you that remember a particular moment will be in the future of that moment, but it seems silly to only care about the experiences of that subset of instances of you and completely neglect the experiences of instances that only share your memories up to an earlier point. If you weight the experiences more sensibly then in the case of a rigorously executed quantum suicide the bulk of the weight will be in instances that diverged before the decision to commit quantum suicide. There will be no chain of memory leading from the QS to those instances, but why should that matter?
So, if Omega was willing to put a copy of you in an Everett branch that didn’t already have one, how much money would you be willing to bid for this service?
If Omega was going to charge $100, and the offer remained open for as many Everett branches as you wanted, how many $100s would you give Omega?
I’m not used to evaluating the worth of Everett branches by count. But for the purpose of this question may I assume you mean “another Everett branch of equal measure to this one, as of the time I click ‘comment’”?
As for an answer… um… I’m not sure, a fair bit? Working out my preferences—quantitatively - in situations so far outside the usual realm of operation is tricky.
After I gave him everything I had I would get a new job that more closer matched my potential for financial gain.
Two extra considerations:
Even aside from a terminal preference for measure maximization I would consider buying more measure purely for the purpose of giving me acausal bargaining power. (I’m even less sure about qualitatively evaluating the usefulness of acausal bargaining power.)
Buying more equal-measure branches is different to trying to preserve measure in the one we are in. While I think I have preferences such that I would buy a new one I’m not sure if the default behavior of humans would be to do so.
But would you take a partial amnesia pill before you go to sleep (which acts while you’re not conscious and does or does not dissolve based on a quantum event)? As a partial quantum suicide. What about total amnesia? How much of your state has to be lost until its a suicide where QM will save you? Why won’t QM save you from quantum-hangover where you do or don’t get a terrible headache and IQ impairment of 20 points?
Just because. Really, that’s all there is to it. It so happens that Yvain’s (alleged) subjectively objective preferences are such that given the existence of a non-hangover branch the existence of a hangover branch is less desirable than if the hangover was substituted for death in that branch.
This is not entirely unrelated to the natural outcome of ‘average utilitarianism’. Under that value system the most simple solution is to go around killing the unhappiest person in the world until such time as the remaining people would be made on average less happy by the removal of the lowest happiness people. Sure, this is totally crazy. But if you have crazy preferences you should do crazy things.
Well, that was kind of rhetorical question; I want to make him be less into quantum-suicide by appeal to how (my guess) he’s not expecting MWI to save him from experiencing quantum-hangover.
edit: to explain, I myself believe death to be a real value from epsilon to 1 rather than a binary value, the epsilon being the shared memories remaining in other people, other people nearby who think like you, etc. and the 1 being the state where you are never forgetting anything at all, with living being somewhere close to 1, and amnesias somewhere between 0 and that. Having binary valued anything seem to result in really screwed up behaviours. (by other people nearby i mean within distances that are much smaller than your information content; the other people at distances that have as many bits in their relative coordinate as you have in your brain, it seems to me those really should be counted as substantially different even though I can’t quite pin point why)
It was a question that got a straight answer that is in accord with Yvain’s position. To the extent that such rhetorical questions get answers that do not require the rhetoric victim to change their mind, the implied argument can be considered refuted.
In practical social usage rhetorical questions can, indeed, often be used as a way to make arguments that an opponent is not allowed to respond to. Here on lesswrong we are free to reject that convention and so not only are literal answers to rhetorical questions acceptable, arguments that are hidden behind rhetorical questions should be considered at least as subject to criticism as arguments made openly.
Well, it is the case that humans often strive to have consistent goal systems (perhaps minimizing their descriptive complexity?), so while he can just say ‘because I defined my goal system to be this’ he is also likely to think and try to come up with some kind of general principle that does not have weird discontinuities at some point when the amnesia is too much like death for his taste.
edit: i think we are talking about different issues; i’m not making a point about his utility function, i’m making a point that he is expecting to become ‘him 20 points dumber and with a terrible headache’, who’s a rather different person, rather than to become someone in a galaxy far far away who doesn’t have the hangover and is thus more similar to him before the hangover.
Yvain does present a consistent goal system. It is one that may appear either crazy or morally abhorrent to us but all indications are that it is entirely consistent. If you were attempting to demonstrate to Yvain an inconsistency in his value system that requires arbitrary complexity to circumvent then you failed.
I think you’re misunderstanding me, see edit. The point I am making is not so much about his values, but about his expectations of subjective experience.
Yvain’s expectations of subjective experience actually seem sane to me. Only his values (and so expected decisionmaking) are weird.
Well, my argument is that you can propose a battery of possible partial quantum suicide set ups involving a machine that partially destroys you (e.g. you are anaesthetised and undergo lobotomy with varying extent of cutting, or something of this sort such as administration of a sublethal dose of neurotoxin). At some point, there’s so little of you left that you’re as good as dead; at some other point, there’s so much of you left that you don’t really expect to be quantum-saved. Either he has some strange continuous function inbetween, that I am very curious about, or he has a discontinuity, which is weird. (and I am guessing a discontinuity but i’d be interested to hear about function)