People sometimes say that it doesn’t really matter whether things like MWI are true (as opposed to the Copenhagen interpretation), since knowing whether it is correct or not wouldn’t affect your decision-making unless you are willing to kill yourself. I’ve been trying to come up with a scenario where you can exploit that knowledge without actually killing yourself and this is where I am at so far:
Say for the sake of argument that in a nuclear war big cities like London or New York have a much better chance of being nuked versus Sitka, Alaska or Swansea, Wales.
Furthermore, let’s say that you prefer being alive in a world where the Earth isn’t half-destroyed by a nuclear war versus one where it is.
Additionally, let’s say that you know MWI is correct in this scenario.
Under this paradigm If you live in London/New York you will be less likely to find yourself in a world where a nuclear war has erupted, as you will likely be dead in those worlds. However, if you live in Sitka/Swansea, you are more likely to survive some nuclear wars (as London/New York would be destroyed in more wars) and it is thus more likely to end up in a world where such an event has occurred.
By combining 1-4, if you have a choice between living in Sitka versus living in New York, everything else being equal, you should choose New York to decrease the probability of waking up in a post-apocalyptic world.
Alternative scenario under the same paradigm for LessWrogners—if you don’t want to live in a world where FAI research stagnates, you might want to move to the Bay Area, so you can decrease the probability, that you’ll end up in a world where most of MIRI and friends are all dead. (yeah, I know—finally a reason for rationalists to move to the Bay Area)
This fail-safe is only an actual net gain if you would rather die than see the aftermath of nuclear war, but don’t want to go through all the trouble of committing suicide manually.
The reasoning is fairly solid, though—looking back at the past, the lack of nuclear war is weaker evidence for human peacefulness if you grew up in Washington DC.
Here are 100 people. A few of them are really happy, but the rest are all fairly miserable. What to do? Kill all the miserable ones, then the average happiness shoots up!
That’s my view of quantum suicide, and just to spell it out, I think it’s pretty silly. That all 100 of the people are, in some sense, “you” does not change my view of it. Moving to New York in order to make sure that a nuclear war would kill you is like getting a DNR notice put on your bed in hospital. It may be a sensible thing to do, but if so is independent of MWI.
Note: Math on infinite sets works differently than math on finite sets. Anyway, if you have a good argument as to why measure matters here, I’d love to hear it.
I am familiar with infinite sets. As for measure, the entire QS argument is about the proportion of your measure in favourable scenarios. If you have a good argument as to why total measure does not matter but proportion of measure does, I’d love to hear it. Make sure that the answer is consistent with your implied view that total measure does matter on finite sets.
If we want to talk about majorities, most people think quantum mechanics is too complicated to worry about. Smaller fractions think that quantum mechanics means humans have souls, quantum mechanics will let you be in two places at once, and that quantum mechanics fix quantum cars.
In a rare display of majoritarianism, the “don’t worry about it” demographic is probably right. Egan’s law—it all adds up to normality. “Quantum killed” is just a more complicated way of saying “normally killed,” because quantum mechanics is how reality has always been.
If you had all the prerequisites, you would not need the argument. So I’ll give you a short overview of the point, and then just link you to some articles.
Quantum immortality is not fundamentally about quantum mechanics. It is about whether you can live forever by defining yourself as a person who doesn’t die. “You” can, but you can’t.
Now on to definitions: A Human’s Guide to Words on why we have them. The Metaethics Sequence on what we’re talking about when we say “I want to live forever.” At this point, if not earlier, you should be able to conceive of an agent that actually does believe in quantum immortality, and acts accordingly—and also an agent that doesn’t believe in quantum immortality, and acts accordingly.
Experiment: Attempt to define yourself as a rock, thus increasing your lifespan. Did it work? Why or why not?
Quantum immortality...is about whether you can live forever by defining yourself as a person who doesn’t die....Experiment: Attempt to define yourself as a rock, thus increasing your lifespan. Did it work?
The assumption of quantum immortality is that once some branches of “you” are gone, then you have to define “you” as the remaining branches. It’s about the impossibility of expanding the definition of “you”, not the possibility of expanding it.
I think there are some additional assumptions required to get you to Tenoke’s example, of dying in a nuclear blast not counting as a downside of living somewhere, because of quantum immortality.
Quantum immortality is not fundamentally about quantum mechanics. It is about whether you can live forever by defining yourself as a person who doesn’t die. “You” can, but you can’t.
That is just not correct, it is about how under some definitions of ‘you’, you don’t die, and some people use those definitions regardless of QM (pattern identity theory uses a definition compatible with quantum immortality for example).
The real issue with quantum immortality is whether measure matters, and as far as I know, this is an open question (although, I suspect there are plenty of good resources on the question which I haven’t seen)
If you have a choice between living in Sitka and living in New York, isn’t that a choice that splits the worlds too? So in one world you’d be living in Sitka and in another world you’d be living in New York. In general, it would seem like “choose X so that I am in this set of worlds and not in that set of worlds” doesn’t work—you’re always still in the set of worlds where you made the opposite choice.
Even “choose X so as to increase my measure in this set of worlds and decrease my measure in that set of worlds” won’t really work. By definition, the whole set of worlds encompasses all choices. You can’t choose X so as to affect something about your situation with respect to the worlds. What would that mean anyway? Would there be a meta-set of sets of worlds, and there’s a branch of the meta-set where you chose X and increased the measure of part of the set, and another branch of the meta-set where you didn’t choose X and didn’t increase the measure of the same part?
‘Died from nuke in London’ is a vague thing, but even if you choose a boundary by which to delineate it, that partition does not carve reality at the joints. There would probably be intermediate states of some measure where you suffer but live for some time after the initial blast, and probably be states of lesser measure where you are unharmed or even gain from the blast.
It might still turn out that the measure/probability/whatever of the suffering states is low enough that it’s still worth leaving yourself open to being nuked, though. Like, maybe P(Alive|Nuked)=10^-5 and that’s an acceptable proportion of worlds where you live through the nuking since you do successfully die in the vast majority of worlds. But if you’re not sure, all else being equal, you’d want to pick the place within London that would be most likely to kill you if a bomb hit, rather than some obscure, heavily-sheltered place on the outskirts where you’ll just about live and suffer after the blast.
In general, it’s pretty suspicious if death seems to be the fundamental deciding factor in such matters (e.g. whether to live in London, whether MWI vs. Copenhagen matters); the reason people suspect so is anthropic—we can’t observe (parts of) worlds where we are dead. But since ‘I’ and ‘dead’ are going to be fuzzy notions that do not carve reality at the joints, we should not expect them to be fundamental to decision-making in multiverses, even if they are efficient shorthands. It’s important to bear this in mind with MWI and anthropics.
(yeah, I know—finally a reason for rationalists to move to the Bay Area)
Unless, I believe that all possible worlds are real, then I’ll make the opposite choice(Sitka/Swansea), as I’d rather see half the world be obliterated instead of dying. If I believe that all possible worlds are equally real then no decision I make matters, so I wouldn’t even bother finishing this sen
People sometimes say that it doesn’t really matter whether things like MWI are true (as opposed to the Copenhagen interpretation), since knowing whether it is correct or not wouldn’t affect your decision-making unless you are willing to kill yourself. I’ve been trying to come up with a scenario where you can exploit that knowledge without actually killing yourself and this is where I am at so far:
Say for the sake of argument that in a nuclear war big cities like London or New York have a much better chance of being nuked versus Sitka, Alaska or Swansea, Wales.
Furthermore, let’s say that you prefer being alive in a world where the Earth isn’t half-destroyed by a nuclear war versus one where it is.
Additionally, let’s say that you know MWI is correct in this scenario.
Under this paradigm If you live in London/New York you will be less likely to find yourself in a world where a nuclear war has erupted, as you will likely be dead in those worlds. However, if you live in Sitka/Swansea, you are more likely to survive some nuclear wars (as London/New York would be destroyed in more wars) and it is thus more likely to end up in a world where such an event has occurred.
By combining 1-4, if you have a choice between living in Sitka versus living in New York, everything else being equal, you should choose New York to decrease the probability of waking up in a post-apocalyptic world.
Alternative scenario under the same paradigm for LessWrogners—if you don’t want to live in a world where FAI research stagnates, you might want to move to the Bay Area, so you can decrease the probability, that you’ll end up in a world where most of MIRI and friends are all dead. (yeah, I know—finally a reason for rationalists to move to the Bay Area)
Almost no one who believes in MWI believes that it makes a difference even then.
This fail-safe is only an actual net gain if you would rather die than see the aftermath of nuclear war, but don’t want to go through all the trouble of committing suicide manually.
The reasoning is fairly solid, though—looking back at the past, the lack of nuclear war is weaker evidence for human peacefulness if you grew up in Washington DC.
Well, not really if you see branching in MW the way that most people seem to.
What is that way?
Here are 100 people. A few of them are really happy, but the rest are all fairly miserable. What to do? Kill all the miserable ones, then the average happiness shoots up!
That’s my view of quantum suicide, and just to spell it out, I think it’s pretty silly. That all 100 of the people are, in some sense, “you” does not change my view of it. Moving to New York in order to make sure that a nuclear war would kill you is like getting a DNR notice put on your bed in hospital. It may be a sensible thing to do, but if so is independent of MWI.
Note: Math on infinite sets works differently than math on finite sets. Anyway, if you have a good argument as to why measure matters here, I’d love to hear it.
I am familiar with infinite sets. As for measure, the entire QS argument is about the proportion of your measure in favourable scenarios. If you have a good argument as to why total measure does not matter but proportion of measure does, I’d love to hear it. Make sure that the answer is consistent with your implied view that total measure does matter on finite sets.
I don’t have a very good argument for either side, which is why I am not discounting one or the other.
If we want to talk about majorities, most people think quantum mechanics is too complicated to worry about. Smaller fractions think that quantum mechanics means humans have souls, quantum mechanics will let you be in two places at once, and that quantum mechanics fix quantum cars.
In a rare display of majoritarianism, the “don’t worry about it” demographic is probably right. Egan’s law—it all adds up to normality. “Quantum killed” is just a more complicated way of saying “normally killed,” because quantum mechanics is how reality has always been.
Good job on arguing against majoritarianism, but you haven’t provided any arguments as to why the view that I attributed to a majority is wrong.
If you had all the prerequisites, you would not need the argument. So I’ll give you a short overview of the point, and then just link you to some articles.
Quantum immortality is not fundamentally about quantum mechanics. It is about whether you can live forever by defining yourself as a person who doesn’t die. “You” can, but you can’t.
Links: You could learn some quantum mechanics. Then look into where the relative state interpretation (MWI) comes from, by reading Everett’s quite accessible paper. Key thing that you will understand after this: probability is a measure, and norm-squared measure is all there is. Look into the foundations of VNM decision theory, but maybe also temper it by reading Savage’s decision theory. Now you should understand how quantum mechanics fits into VNM decision theory by providing a measure. At this point it all adds up to normality—you make the same decisions using any interpretation of quantum mechanics.
Now on to definitions: A Human’s Guide to Words on why we have them. The Metaethics Sequence on what we’re talking about when we say “I want to live forever.” At this point, if not earlier, you should be able to conceive of an agent that actually does believe in quantum immortality, and acts accordingly—and also an agent that doesn’t believe in quantum immortality, and acts accordingly.
Experiment: Attempt to define yourself as a rock, thus increasing your lifespan. Did it work? Why or why not?
The assumption of quantum immortality is that once some branches of “you” are gone, then you have to define “you” as the remaining branches. It’s about the impossibility of expanding the definition of “you”, not the possibility of expanding it.
I think there are some additional assumptions required to get you to Tenoke’s example, of dying in a nuclear blast not counting as a downside of living somewhere, because of quantum immortality.
That is just not correct, it is about how under some definitions of ‘you’, you don’t die, and some people use those definitions regardless of QM (pattern identity theory uses a definition compatible with quantum immortality for example).
The real issue with quantum immortality is whether measure matters, and as far as I know, this is an open question (although, I suspect there are plenty of good resources on the question which I haven’t seen)
If you have a choice between living in Sitka and living in New York, isn’t that a choice that splits the worlds too? So in one world you’d be living in Sitka and in another world you’d be living in New York. In general, it would seem like “choose X so that I am in this set of worlds and not in that set of worlds” doesn’t work—you’re always still in the set of worlds where you made the opposite choice.
Even “choose X so as to increase my measure in this set of worlds and decrease my measure in that set of worlds” won’t really work. By definition, the whole set of worlds encompasses all choices. You can’t choose X so as to affect something about your situation with respect to the worlds. What would that mean anyway? Would there be a meta-set of sets of worlds, and there’s a branch of the meta-set where you chose X and increased the measure of part of the set, and another branch of the meta-set where you didn’t choose X and didn’t increase the measure of the same part?
See the free will sequence; it assumes one deterministic world but it easily generalizes to many deterministic negligibly-interacting worlds.
‘Died from nuke in London’ is a vague thing, but even if you choose a boundary by which to delineate it, that partition does not carve reality at the joints. There would probably be intermediate states of some measure where you suffer but live for some time after the initial blast, and probably be states of lesser measure where you are unharmed or even gain from the blast.
It might still turn out that the measure/probability/whatever of the suffering states is low enough that it’s still worth leaving yourself open to being nuked, though. Like, maybe P(Alive|Nuked)=10^-5 and that’s an acceptable proportion of worlds where you live through the nuking since you do successfully die in the vast majority of worlds. But if you’re not sure, all else being equal, you’d want to pick the place within London that would be most likely to kill you if a bomb hit, rather than some obscure, heavily-sheltered place on the outskirts where you’ll just about live and suffer after the blast.
In general, it’s pretty suspicious if death seems to be the fundamental deciding factor in such matters (e.g. whether to live in London, whether MWI vs. Copenhagen matters); the reason people suspect so is anthropic—we can’t observe (parts of) worlds where we are dead. But since ‘I’ and ‘dead’ are going to be fuzzy notions that do not carve reality at the joints, we should not expect them to be fundamental to decision-making in multiverses, even if they are efficient shorthands. It’s important to bear this in mind with MWI and anthropics.
lolol
How would your decision be different if you evaluate possible worlds as opposed to Everett branches?
Unless, I believe that all possible worlds are real, then I’ll make the opposite choice(Sitka/Swansea), as I’d rather see half the world be obliterated instead of dying. If I believe that all possible worlds are equally real then no decision I make matters, so I wouldn’t even bother finishing this sen