I enjoyed the post (enough for a vote up!) but I find myself wishing it had stopped at #5.
6 is mostly correct but has significant edge cases (even if you subscribe to MWI, probabilities pop up when dealing with tiny things). Something like “Probabilities exist in minds” is a much more agreeable statement than “Probabilities don’t exist elsewhere,” and has the same framing benefits.
7 just flat out bothers me. Many Worlds is just an interpretation, a flavor- it shares the exact same math with all other flavors of quantum mechanics. I agree with Eliezer that it’s a far more agreeable flavor than Copenhagen- but those aren’t the only two flavors available. And if you are making predictions based on your flavor preferences, something went wrong somewhere. I cannot see how your tastes when it comes to QM should impact whether or not you sign up for cryonics with the currently existing firms offering cryonic services.
I didn’t get the impression that MWI mattered to cryonics. The connection from the Quantum physics sequence to cryonics that I got was “This atom is essentially the same as that atom, Replacing all your atoms wouldn’t change ‘you’. ” And related to that, that your atoms could be computer simulated and you’d still be you.
That’s a very reasonable interpretation, but it’s orthogonal to why I’m bothered.
If the argument is “my objection to cryonics was I wasn’t convinced a remade me would be me, but as soon as I realized the configuration was important and not the pieces inside the configuration, that toppled my last objection,” then I don’t have an issue with that.
What it looked like to me was “I am convinced of Eliezer’s viewpoint” instead of “I believe Eliezer’s arguments are correct in the domain that they are argued.” The linked argument that cryonics is reasonable is an argument that cryonics is possible, not an argument that signing up with Alcor or CI actually increases your likelihood of being awoken in the future. The linked argument is necessary but not sufficient for the action stated.
That came to the forefront of my mind because Eliezer’s declaration that MWI is “correct” could mean two things- either MWI is the single truest / best flavor of QM, which I do not think he is qualified to state, or MWI gives the right answers when you ask it relevant questions, just like Copenhagen. Eliezer can rightly say MWI is more satisfactory than Copenhagen, but when you go further and make plans based on multiverses that you would not make if you were just planning for a singular future, that is a giant red flag.
MWI agrees with Copenhagen in all currently reasonably accessible experimental regimes. But it is not just a flavor—it allows for the possibility of “uncollapse” after an observation by delicate recoherence. (Though after such a demonstration the Copenhagenite could just say that the collapse was inferred too soon.)
I agree that the question “Has this system collapsed?” is a bad question, and people shouldn’t be interested in it. (That’s the main reason I don’t like Copenhagen; it invented that question and still considers it relevant.)
The real question is “if we set up a bunch of these systems in identical conditions, what distribution of results do we expect?”. The reason I am not optimistic about MWI ‘beating’ Copenhagen with such an experiment is that any physical process that “uncollapses” an observation is readily understandable by both MWI and Copenhagen. The Copenhagenite would just say “well, the system collapsed here, and then you uncollapsed it there, and then you recollapsed it in this last place” and come up with the same answer for the final state as the MWI believer.
Wave function collapse deletes eigenvalues. ‘Uncollapse’ would have to put them back, in which case you have to keep track of those eigenvalues, in which case you never deleted them in the first place, in which case no collapse occured. So the Copenhagen interpretation can deal with uncollapse, so long as nothing ever collapses.
Illustration: if we use the example of the experiment here, the Copenhagenite would just point to the steps between measurement 2 and measurement 3 that reverse measurement 2 and say “look, to do this you need to put your subject in either |+> or |-> according to what’s still in your memory, and so the collapse at measurement 2 is entirely separate from what the result of measurement 3 will be.”
The only difference between the two physicists will be their vocabulary- one will have the unfortunate word “collapse” and the other will have the unfortunate word “multiverse”- but they’ll agree on the final result.
OK, the example linked is defective, in that there are two different operations that get the same result when the machine reverses its x-axis measurement. The first is the time-reversal of the measurement operation; the second is the recreation of the state created by measurement 1. You seem to be saying that the Copenhagenite would assume the latter.
Here is a modification of the experiment that tests the idea of collapse more severely. Instead of preparing an electron in a |+z> or |-z> state, I prepare an entangled pair of electrons with opposite z-axis spin (a spin anti-correlated pair). I now give one electron to the machine intelligence, which measures its spin in the x-axis, and then applies the time-reversal of the measurement, restoring the electron’s original state and erasing its memory of the x-axis state. It then passes the electron back to me, and I measure the two electrons’ z-axis spins.
If the machine intelligence’s measurement had caused a collapse, the anti-correlation would be erased. But in fact everything we know about quantum mechanics says that the electrons should remain anti-correlated.
I now give one electron to the machine intelligence, which measures its spin in the x-axis, and then applies the time-reversal of the measurement, restoring the electron’s original state and erasing its memory of the x-axis state.
I don’t see why the Copenhagenite can’t make the exact same objection here. Perhaps it would be clearer if you gave an example of how one would perform the time reversal of a measurement? If I have a z spin up electron, and I put it through a Stern-Gerlach device and find it is now a x spin up electron, how do I go back to a z spin up electron?
The link doesn’t make it explicit, but a reversible machine intelligence which can actually reverse a measurement is a quantum computer. In this context, a measurement occurs when the AI purposefully entangles its computing elements with the electron. The AI can now choose whether to let the information it gains leak out of it or not. Provided it does not allow the entanglement between the electron and the outside world to increase, it can choose to unentangle its state from that of the electron. In the simplest case, where it does not allow the rest of its mind to become entangled with the part of itself that it is using as a measurement apparatus, all it need do is run the inverse of the unitary transform that it used to entangle the apparatus with the electron. However, it can theoretically do quite a bit more. It can use the information in other computations, and then carefully carry out an operation that restores the original state of the electron and turns the results it obtains into superpositions.
Humans don’t have such fine-grained control over where they shuffle quantum information, nor can they keep themselves from becoming entangled with their environment. Using macroscopic devices to register phosphorescence is right out.
It seems to me that this makes assumptions about entanglement and disentanglement which I find suspect (but I am not an expert on entanglement, so they may hold). It doesn’t appear to be “choosing” to unentangle its state from the electron- we’re assuming that the information it generates through entanglement is not leaked to the outside world, and that the information can be thrown away and the system returned to where it was before. If it’s making a choice, it seems that that choice would cause information leak.
If those assumptions hold, I don’t see why they hold for just MWI. That is, I believe it may be possible to get to a final situation where you have your initial configuration despite the fact that your apparatus poked the system- but I don’t think that gives you any meaningful information differentiating the flavors of QM.
Quantum Information cannot be thrown away. Nor can it be copied. Information is conserved. *Apart, perhaps, from Copenhagen collapse). Information can be made difficult to retrieve by e.g. entanglement with the environment, specifically propagating modes that take it beyond your control, but it’s still “in principle” there.
Is there a meaningful difference between “propagating modes that take it beyond your control” and “throwing it away”? In my mind, the first is a much longer restatement of the second, but I apologize that it was unclear. (Here, you’re throwing it back into the electron, not the outside world, but the idea is the same from the computer’s point of view.)
Yes, they have very different effects. Throwing it into the electron allows recoherence in principal. Throwing it into the environment makes that impossible.
As stated you can’t. In the MWI picture, you are split into two one who has measured it x+, the other x-. Both must send it back to have it recohere, and they must at the same time erase their measurement of which way it went—really anything that distinguishes those two branches. They can record the fact that they did measure it, as this is the same in the two branches.
A person obviously can’t just “forget”, and will at best leak information into the environment, encoded as correlations in the noise of heat. A (reversible) computer, on the other hand, works quite well for doing this.
They use four supports, all of which collapse under examination (I don’t number them the way they do, because they seem confused about what are separate supports):
Though it makes the same predictions about our world as de Broglie-Bohm, they have different philosophical implications. Believe in something because of math, not philosophy.
They list three predictions made MWI, all of which are already disproved or nonsense:
If memory is reversible, it’s not memory because thermodynamic fluctuations make it unreliable. Beyond that confusion, the crux of this argument is whether or not a spin measurement can be reversed- if so, it should work for any flavor, and not depend on whether or not you also erase what’s in memory.
Their discussion of quantum gravity serves to make MWI not more plausible, as it supposedly requires quantum gravity, while other flavors function whether gravity is quantum or classical.
Their discussion of linearity is flat-out bizarre. Paraphrased: ‘We’re pretty damn sure that QM is linear, but if it weren’t and MWI were true, aliens would have teleported to our dimension, and that hasn’t happened yet.’ Why they think that is evidence for MWI is beyond me- using Bayesian logic, it strictly cannot increase the probability of MWI.
I don’t think the intention was to offer these as evidence for MWI. The evidence for MWI is that it has one less postulate (and therefore is “simpler”). They’re just showing what MWI rules out. That these predictions are different correctly justifies saying “MWI is not just an interpretation”.
It is best not to use quotation marks—unless you are actually quoting—or otherwise make it very clear what you are doing. The resulting self-sabotage is too dramatic.
I read that—and your incorrect comments about reversible memory—and concluded that you didn’t know what you were talking about.
At your suggestion, I’ve revised my comment to make clear that I’m paraphrasing my interpretation of their comment instead of quoting it directly.
I was least sure about my reversible memory objection, and was considering placing a disclaimer on it; however, I feel I should stand by it unless given evidence that my understanding of information entropy is incorrect. My statement is in accord with Landauer’s Principle, which I see is not known to be true (but is very strongly suspected to be). There appears to be a fundamental limit that their trend is bucking up against, and so I feel confident saying the trend will not continue as they need it to.
Even if we shelve the discussion of whether or not memory can be reversible, the other objection- that any process which reverses a measurement can be understood by both MWI and Copenhagen- demolishes the usefulness of such an experiment, as none of the testable predictions differ between the two interpretations.
I don’t see the relevance- the description of the experiment linked purports to hinge on the reversibility of information erasure. It sounds like both of us agree that’s impossible.
(It actually hinges on whatever steps they take to ‘reverse’ the measurement they take, which is why it’s not an effective experiment.)
Reversible computer designs people actually consider building do a small bit of irreversible computation copying end results of the reversible computations into irreversible memory before rolling back the reversible computation. Perfectly reversible computations are a bit useless since they erase their results when they start rolling backwards.
I enjoyed the post (enough for a vote up!) but I find myself wishing it had stopped at #5.
6 is mostly correct but has significant edge cases (even if you subscribe to MWI, probabilities pop up when dealing with tiny things). Something like “Probabilities exist in minds” is a much more agreeable statement than “Probabilities don’t exist elsewhere,” and has the same framing benefits.
7 just flat out bothers me. Many Worlds is just an interpretation, a flavor- it shares the exact same math with all other flavors of quantum mechanics. I agree with Eliezer that it’s a far more agreeable flavor than Copenhagen- but those aren’t the only two flavors available. And if you are making predictions based on your flavor preferences, something went wrong somewhere. I cannot see how your tastes when it comes to QM should impact whether or not you sign up for cryonics with the currently existing firms offering cryonic services.
I didn’t get the impression that MWI mattered to cryonics. The connection from the Quantum physics sequence to cryonics that I got was “This atom is essentially the same as that atom, Replacing all your atoms wouldn’t change ‘you’. ” And related to that, that your atoms could be computer simulated and you’d still be you.
That’s a very reasonable interpretation, but it’s orthogonal to why I’m bothered.
If the argument is “my objection to cryonics was I wasn’t convinced a remade me would be me, but as soon as I realized the configuration was important and not the pieces inside the configuration, that toppled my last objection,” then I don’t have an issue with that.
What it looked like to me was “I am convinced of Eliezer’s viewpoint” instead of “I believe Eliezer’s arguments are correct in the domain that they are argued.” The linked argument that cryonics is reasonable is an argument that cryonics is possible, not an argument that signing up with Alcor or CI actually increases your likelihood of being awoken in the future. The linked argument is necessary but not sufficient for the action stated.
That came to the forefront of my mind because Eliezer’s declaration that MWI is “correct” could mean two things- either MWI is the single truest / best flavor of QM, which I do not think he is qualified to state, or MWI gives the right answers when you ask it relevant questions, just like Copenhagen. Eliezer can rightly say MWI is more satisfactory than Copenhagen, but when you go further and make plans based on multiverses that you would not make if you were just planning for a singular future, that is a giant red flag.
MWI agrees with Copenhagen in all currently reasonably accessible experimental regimes. But it is not just a flavor—it allows for the possibility of “uncollapse” after an observation by delicate recoherence. (Though after such a demonstration the Copenhagenite could just say that the collapse was inferred too soon.)
I agree that the question “Has this system collapsed?” is a bad question, and people shouldn’t be interested in it. (That’s the main reason I don’t like Copenhagen; it invented that question and still considers it relevant.)
The real question is “if we set up a bunch of these systems in identical conditions, what distribution of results do we expect?”. The reason I am not optimistic about MWI ‘beating’ Copenhagen with such an experiment is that any physical process that “uncollapses” an observation is readily understandable by both MWI and Copenhagen. The Copenhagenite would just say “well, the system collapsed here, and then you uncollapsed it there, and then you recollapsed it in this last place” and come up with the same answer for the final state as the MWI believer.
Wave function collapse deletes eigenvalues. ‘Uncollapse’ would have to put them back, in which case you have to keep track of those eigenvalues, in which case you never deleted them in the first place, in which case no collapse occured. So the Copenhagen interpretation can deal with uncollapse, so long as nothing ever collapses.
Illustration: if we use the example of the experiment here, the Copenhagenite would just point to the steps between measurement 2 and measurement 3 that reverse measurement 2 and say “look, to do this you need to put your subject in either |+> or |-> according to what’s still in your memory, and so the collapse at measurement 2 is entirely separate from what the result of measurement 3 will be.”
The only difference between the two physicists will be their vocabulary- one will have the unfortunate word “collapse” and the other will have the unfortunate word “multiverse”- but they’ll agree on the final result.
OK, the example linked is defective, in that there are two different operations that get the same result when the machine reverses its x-axis measurement. The first is the time-reversal of the measurement operation; the second is the recreation of the state created by measurement 1. You seem to be saying that the Copenhagenite would assume the latter.
Here is a modification of the experiment that tests the idea of collapse more severely. Instead of preparing an electron in a |+z> or |-z> state, I prepare an entangled pair of electrons with opposite z-axis spin (a spin anti-correlated pair). I now give one electron to the machine intelligence, which measures its spin in the x-axis, and then applies the time-reversal of the measurement, restoring the electron’s original state and erasing its memory of the x-axis state. It then passes the electron back to me, and I measure the two electrons’ z-axis spins.
If the machine intelligence’s measurement had caused a collapse, the anti-correlation would be erased. But in fact everything we know about quantum mechanics says that the electrons should remain anti-correlated.
I don’t see why the Copenhagenite can’t make the exact same objection here. Perhaps it would be clearer if you gave an example of how one would perform the time reversal of a measurement? If I have a z spin up electron, and I put it through a Stern-Gerlach device and find it is now a x spin up electron, how do I go back to a z spin up electron?
The link doesn’t make it explicit, but a reversible machine intelligence which can actually reverse a measurement is a quantum computer. In this context, a measurement occurs when the AI purposefully entangles its computing elements with the electron. The AI can now choose whether to let the information it gains leak out of it or not. Provided it does not allow the entanglement between the electron and the outside world to increase, it can choose to unentangle its state from that of the electron. In the simplest case, where it does not allow the rest of its mind to become entangled with the part of itself that it is using as a measurement apparatus, all it need do is run the inverse of the unitary transform that it used to entangle the apparatus with the electron. However, it can theoretically do quite a bit more. It can use the information in other computations, and then carefully carry out an operation that restores the original state of the electron and turns the results it obtains into superpositions.
Humans don’t have such fine-grained control over where they shuffle quantum information, nor can they keep themselves from becoming entangled with their environment. Using macroscopic devices to register phosphorescence is right out.
It seems to me that this makes assumptions about entanglement and disentanglement which I find suspect (but I am not an expert on entanglement, so they may hold). It doesn’t appear to be “choosing” to unentangle its state from the electron- we’re assuming that the information it generates through entanglement is not leaked to the outside world, and that the information can be thrown away and the system returned to where it was before. If it’s making a choice, it seems that that choice would cause information leak.
If those assumptions hold, I don’t see why they hold for just MWI. That is, I believe it may be possible to get to a final situation where you have your initial configuration despite the fact that your apparatus poked the system- but I don’t think that gives you any meaningful information differentiating the flavors of QM.
Quantum Information cannot be thrown away. Nor can it be copied. Information is conserved. *Apart, perhaps, from Copenhagen collapse). Information can be made difficult to retrieve by e.g. entanglement with the environment, specifically propagating modes that take it beyond your control, but it’s still “in principle” there.
Is there a meaningful difference between “propagating modes that take it beyond your control” and “throwing it away”? In my mind, the first is a much longer restatement of the second, but I apologize that it was unclear. (Here, you’re throwing it back into the electron, not the outside world, but the idea is the same from the computer’s point of view.)
Yes, they have very different effects. Throwing it into the electron allows recoherence in principal. Throwing it into the environment makes that impossible.
As stated you can’t. In the MWI picture, you are split into two one who has measured it x+, the other x-. Both must send it back to have it recohere, and they must at the same time erase their measurement of which way it went—really anything that distinguishes those two branches. They can record the fact that they did measure it, as this is the same in the two branches.
A person obviously can’t just “forget”, and will at best leak information into the environment, encoded as correlations in the noise of heat. A (reversible) computer, on the other hand, works quite well for doing this.
Right, and after several such experiments it would become apparent that the Copenhagenite doesn’t know how to predict when collapse happens.
It isn’t according to The Everett FAQ’s: Q16 Is many-worlds (just) an interpretation?
Have you read that and considered it convincing?
They use four supports, all of which collapse under examination (I don’t number them the way they do, because they seem confused about what are separate supports):
Though it makes the same predictions about our world as de Broglie-Bohm, they have different philosophical implications. Believe in something because of math, not philosophy.
They list three predictions made MWI, all of which are already disproved or nonsense:
If memory is reversible, it’s not memory because thermodynamic fluctuations make it unreliable. Beyond that confusion, the crux of this argument is whether or not a spin measurement can be reversed- if so, it should work for any flavor, and not depend on whether or not you also erase what’s in memory.
Their discussion of quantum gravity serves to make MWI not more plausible, as it supposedly requires quantum gravity, while other flavors function whether gravity is quantum or classical.
Their discussion of linearity is flat-out bizarre. Paraphrased: ‘We’re pretty damn sure that QM is linear, but if it weren’t and MWI were true, aliens would have teleported to our dimension, and that hasn’t happened yet.’ Why they think that is evidence for MWI is beyond me- using Bayesian logic, it strictly cannot increase the probability of MWI.
I don’t think the intention was to offer these as evidence for MWI. The evidence for MWI is that it has one less postulate (and therefore is “simpler”). They’re just showing what MWI rules out. That these predictions are different correctly justifies saying “MWI is not just an interpretation”.
It is best not to use quotation marks—unless you are actually quoting—or otherwise make it very clear what you are doing. The resulting self-sabotage is too dramatic.
I read that—and your incorrect comments about reversible memory—and concluded that you didn’t know what you were talking about.
At your suggestion, I’ve revised my comment to make clear that I’m paraphrasing my interpretation of their comment instead of quoting it directly.
I was least sure about my reversible memory objection, and was considering placing a disclaimer on it; however, I feel I should stand by it unless given evidence that my understanding of information entropy is incorrect. My statement is in accord with Landauer’s Principle, which I see is not known to be true (but is very strongly suspected to be). There appears to be a fundamental limit that their trend is bucking up against, and so I feel confident saying the trend will not continue as they need it to.
Even if we shelve the discussion of whether or not memory can be reversible, the other objection- that any process which reverses a measurement can be understood by both MWI and Copenhagen- demolishes the usefulness of such an experiment, as none of the testable predictions differ between the two interpretations.
If it helps, this seems relevant: http://en.wikipedia.org/wiki/Reversible_computing
Landauer’s Principle doesn’t seem particularly relevant—since in reversible computing there is no erasure of information.
I don’t see the relevance- the description of the experiment linked purports to hinge on the reversibility of information erasure. It sounds like both of us agree that’s impossible.
(It actually hinges on whatever steps they take to ‘reverse’ the measurement they take, which is why it’s not an effective experiment.)
It seems relevant to the comment that “if memory is reversible, it’s not memory”. Reversible computers have reversible memory.
Reversible computer designs people actually consider building do a small bit of irreversible computation copying end results of the reversible computations into irreversible memory before rolling back the reversible computation. Perfectly reversible computations are a bit useless since they erase their results when they start rolling backwards.
You can erase some of their results without erasing others, of course.
Nobody says you have to run a reversible computer backwards.
A big part of the point is to digitise heat sinks and power management. For details about that, see here.