To answer this question we should talk to lots of neuroscientists. Here’s one:
There’s a burden of proof issue here: If there is a small group making a scientific claim that the larger scientific community finds ludicrous, skepticism should be the default position. I’m not aware of any peer-reviewed publication explicitly debunking cryonics. Probably the reason is that practicing lab scientists aren’t inclined to write up a refutation of an particular idea when all you need to see it’s bullshit is an undergraduate-level understanding of biology. So, since I can’t point you to a systematic refutation, I’ll give you this in the way of citation:
http://jcb.rupress.org/content/188/1/145.full
This is a technically impressive study, they get really pretty and informative EM results. Excepting minor advances in the few years since it was published, this is close to state of the art as far a vitrification of brain tissue goes. If what the cryonics huckster companies were offering provided THIS level of preservation in a whole brain, then maybe we could have an interesting conversation. Cryonics would still be hopeless and vapid for other reasons, but at least you could count on fine membrane structure in synapses being preserved.
But in order to GET such good preservation, they had to take a slice of brain 400 micrometers thick and 1mm in diameter and vitrify it at 2000 bar. The high pressure required would damage an entire brain, and this works specifically because you’re dealing with a small volume of tissue. Slicing an entire MOUSE brain and reconstructing it at this level is a major goal of connectomics, pretty far off. But I promise you (my lab does this kind of freezing for EM fairly routinely, it’s time-consuming and the sectioning is artifact-prone), what the huckster companies are offering to do to your head is NOT going to give the kind of preservation that maintains synaptic structure. Thinking that in the early 21st century you’re going to pay a company to freeze your brain in a way that it could be “reanimated” is insane, and shows a stunning naivete of the underlying biological complexity. The service you are paying for is “please destroy my already-dead brain in a way that involves chemicals and coldness”.
—David Ruhl on facebook:LessWrong
There was also a discussion with another neuroscientist, kalla724, here a year ago.
Do people know of other places where a neuroscientist who knows about vitrification gives their opinion?
“When all you need to see it’s bullshit is an undergraduate-level understanding of biology” is an extremely clear cue that the speaker does not understand the current state of the cryonics debate and cannot be trusted to summarize it. Anyone who does not specifically signal that they understand the concept of information-theoretic death as mapping many cognitive-identity-distinct initial physical states to the same atomic-level physical end state is not someone whose summaries or final judgment you can possibly reasonably trust. If they cite sources you might look at those sources, but you’re going to have to figure out on your own what it means.
Can you please elaborate on how and why sufficient understanding of the concept of information-theoretic death as mapping many cognitive-identity-distinct initial physical states to the same atomic-level physical end state helps to alleviate concerns raised by the author?
The basic idea of getting cryonics is that it offers a chance of massively extended lifespan, because there is a chance that it preserves one’s identity. That’s the first-run approximation, with additional considerations arising from making this reasoning a bit more rigorous, e.g. that cryonics is competitive against other interventions, that the chance is not metaphysically tiny, etc.
One thing we might make more rigorous is what we mean by ‘preservation’. Well, preservation refers to reliably being able to retrieve the person from the hopefully-preserved state, which requires that the hopefully-preserved state cannot have arisen from many non-matching states undergoing the process.
The process that squares positive numbers preserves perfectly (is an injection), because you can always in theory tell me the original number if I give you its square. The process that squares real numbers preserves imperfectly but respectably since, for any positive output, that output could have come from two numbers (e.g. 1^2=1=(-1)^2). Moreover, if we only cared about the magnitude (modulus, i.e. ignoring the sign) of the input, even squaring over real numbers would perfectly preserve what we cared about.
Similarly, there is a chance that the hopefully-preserved states generated by cryonics do/will be generated only by the original identity, or possibly some acceptably close identities. That we do not currently know if it is possible to retrieve acceptably close identities from hopefully-preserved states—or even if we did, how one would do so—does not necessarily make the probability that it is possible to do so in principle low enough that cryonics can be laughed off.
A monkey might be bamboozled by the sequence of square numbers written in Arabic numerals, but that would not prove that the rule could not be deduced in principle, or that information had been lost for human purposes. Similarly we might currently be unable to reverse vitrification or look under a microscope and retrieve the identity, but it is unfair to demand this level of proof, and it is annoying and frustrating in the same way as logical rudeness (even if technically it is not logically rude) when every few months another person smugly spouts this type of argument as a ‘refutation’ of cryonics and writes cryonicists off, and then gets upvoted handsomely. (Hence Eliezer losing patience and outright declaring that people who don’t seem to (effectively) understand this point about mappings don’t have a clue.)
Formalisations of these concepts arise in more obviously mathematical contexts like the study of functions and information theory, but it feels like neither of those should be necessary background for a smart person to understand the basic idea. But in all honesty, I think the inferential gap for someone who has not explicitly considered at least the idea of injections before is big enough that often people apply the absurdity heuristic or become scared to do something unconventional before the time it takes to cross that inferential gap.
I think there’s a good chance that there are neurodegenerative conditions that are currently irreversible but which many more would think worth working on than cryonics, simply because they associate cryonics with ‘computer nerd failure mode’ or apply the absurdity heuristic or because attacking neurodegenrative conditions is Endorsed by Experts whereas cryonics is not or because RationalWiki will laugh at them. Possible partial explanation: social anxiety that mockery will ensure for trying something not explicitly endorsed by an Expert consensus (which is a realistic fear, given how many people basically laugh at cryonicists or superficially write it off as ‘bullshit’). And yes, in this mad world, social anxiety really might be the decisive factor for actual humans in whether to pursue an intervention that could possibly grant them orders of magnitude more lifespan.
It seems from the comment that he does understand information theoretic views of death- after all he is talking about preserving the information via thin slices and EM scanning.
To get good vitrification results, enough to preserve the synaptic structure, his claim is that you have to take mm thin slices of brain and preserve them at nearly 2000 atmospheres of pressure. This is obviously not what current cryonics procedures do. They shoot you full of cryoprotectants (which destroys some chemical information in the brain at the outset), and then freeze the brain whole (which will lead to a lot of fracturing).
To attack his point, I think you’d need to claim that either:
he is technically wrong and current vitrification does decently preserve synaptic structures.
mechanically scrambling synaptic structures via cracking is a 1 to 1 process that preserves information.
(2) is the interesting claim, though I’d hardly trust his word on (1) since I didn’t see any especially alarming paragraphs in the actual paper referenced. I’m not an expert on that level of neurobiology (Anders Sandberg is, and he’s signed up for cryonics), but I am not interested in hearing from anyone who has not demonstrated that they understand that we are talking about doing intelligent cryptography to a vitrified brain and potentially molecule-by-molecule analysis and reasoning, rather than, “Much cold. Very damage. Boo.”
Unless someone spells out exactly what is supposed to destroy all cues of a piece of info, by explaining why two cognitively distinct start states end up looking like molecularly identical endstates up to thermal noise, so that we can directly evaluate the technical evidence for ourselves, all they’re asking us to do is trust their authoritative summary of their intuitions; and you’d be just plain dumb to trust the authoritative summary of someone who didn’t understand the original argument.
I’m trying not to be impatient here, but when I actually went to look at the cited paper and it said nothing at all about damage, it turned out this eminent authority’s original argument consisted merely of, “To read off great synaptic info with current big clumsy microscopes and primitive imaging processing, we need big pressures. Look at this paper involving excellent info and big pressures. Cryonicists don’t have big pressures. Therefore you’re dead QED.”
I suspect the basic underlying premise that causes this difference is that you believe that strongly superhuman AI will exist at the time of unfreezing, whereas most people either disbelieve that thesis or do not take it into account.
So say something about it. Your whole comment is an attack on 1, but regardless of his word on whether or not thing slice vitrification is currently the best we can do, we KNOW fracturing happens with current brain preservation techniques. Liquid nitrogen is well below the glass transition, so fracturing is unavoidable.
Why should we expect fracturing/cracking to be 1 to 1?
If you’re worried about the effects of cracking, you can pay for ITS. LN2 is only used because it is cheap and relatively low-tech to maintain.
If you ask me it’s a silly concern if we’re assuming nanorepair or uploading. Cracking is just a surface discontinuity, and it forms at a point in time where the tissue is already in a glassy state where there can’t be much mixing of molecules. The microcracks that form in frozen tissue is a much greater concern (but not the only concern with freezing). The fact that vitrified tissue forms large, loud cracks is related to the fact that it does such a good job holding things in place.
I mean, it is either his authoritative summary or yours, and with all due honesty that guy actually takes care to construct an actual argument instead of resorting to appeals to authority and ridicule.
Personally I would be more interested in someone explaining exactly how cues of a piece of info are going to be reassembled and whole brain is going to be reconstructed from a partial data.
Proving that cryo-preservation + restoration does indeed work, and also showing the exact method as to how, seems like a more persuasive way to construct an argument rather that proving that your opponents failed to show that what you are claiming is currently impossible.
If cryonics providers don’t have a proper way of preserving your brain state (even if they can repair partial damage by guessing), than I am sorry to say, but you are indeed dead.
I would not have updated much at all based on the David Ruhl comment. Cryobiology is a young field, and the details about what we can and can’t do with current preservation technology change every couple of years. Also, I think almost all bio-scientists who say current preservation techniques are insufficient to preserve you are underestimating the likelihood and the powers of machine superintelligence, and in general haven’t studied superintelligence theory hardly at all. Skepticism about cryonics working should, I think, mostly come from the other parts of the conjunction.
You lose some data at the borders between slices, and you’d have to do it extremely quickly, but I’d expect it to work much better than the current method.
I think he means “create a functional human you, while primarily sourcing the matter from your old body”. He’s commenting that slicing the brain makes this more difficult, but it sounds like the alterations caused by current vitrification techniques make it impossible either way.
We shouldn’t make what thing we want dependant on what is harder or easier to do, and in any case if one of these is possible, the other is too. Some more centuries of technology development don’t mean much when you’re suspended.
It doesn’t change what we want, but it does change how likely we are to get it. Waiting additional centuries increases the probability of catastrophe significantly, be it a power outage or a nuclear war, as well as making it correspondingly harder to reintroduce yourself to society. And we don’t actually have any reason to believe that if one is possible then the other is too—perhaps human technology will never get to the point of being able to map the brain perfectly and the only path to resurrection is with the same tissue, perhaps the tissue will be beyond repair and uploading will be the only viable option. Both are plausible.
To answer this question we should talk to lots of neuroscientists. Here’s one:
There was also a discussion with another neuroscientist, kalla724, here a year ago.
Do people know of other places where a neuroscientist who knows about vitrification gives their opinion?
“When all you need to see it’s bullshit is an undergraduate-level understanding of biology” is an extremely clear cue that the speaker does not understand the current state of the cryonics debate and cannot be trusted to summarize it. Anyone who does not specifically signal that they understand the concept of information-theoretic death as mapping many cognitive-identity-distinct initial physical states to the same atomic-level physical end state is not someone whose summaries or final judgment you can possibly reasonably trust. If they cite sources you might look at those sources, but you’re going to have to figure out on your own what it means.
Can you please elaborate on how and why sufficient understanding of the concept of information-theoretic death as mapping many cognitive-identity-distinct initial physical states to the same atomic-level physical end state helps to alleviate concerns raised by the author?
The basic idea of getting cryonics is that it offers a chance of massively extended lifespan, because there is a chance that it preserves one’s identity. That’s the first-run approximation, with additional considerations arising from making this reasoning a bit more rigorous, e.g. that cryonics is competitive against other interventions, that the chance is not metaphysically tiny, etc.
One thing we might make more rigorous is what we mean by ‘preservation’. Well, preservation refers to reliably being able to retrieve the person from the hopefully-preserved state, which requires that the hopefully-preserved state cannot have arisen from many non-matching states undergoing the process.
The process that squares positive numbers preserves perfectly (is an injection), because you can always in theory tell me the original number if I give you its square. The process that squares real numbers preserves imperfectly but respectably since, for any positive output, that output could have come from two numbers (e.g. 1^2=1=(-1)^2). Moreover, if we only cared about the magnitude (modulus, i.e. ignoring the sign) of the input, even squaring over real numbers would perfectly preserve what we cared about.
Similarly, there is a chance that the hopefully-preserved states generated by cryonics do/will be generated only by the original identity, or possibly some acceptably close identities. That we do not currently know if it is possible to retrieve acceptably close identities from hopefully-preserved states—or even if we did, how one would do so—does not necessarily make the probability that it is possible to do so in principle low enough that cryonics can be laughed off.
A monkey might be bamboozled by the sequence of square numbers written in Arabic numerals, but that would not prove that the rule could not be deduced in principle, or that information had been lost for human purposes. Similarly we might currently be unable to reverse vitrification or look under a microscope and retrieve the identity, but it is unfair to demand this level of proof, and it is annoying and frustrating in the same way as logical rudeness (even if technically it is not logically rude) when every few months another person smugly spouts this type of argument as a ‘refutation’ of cryonics and writes cryonicists off, and then gets upvoted handsomely. (Hence Eliezer losing patience and outright declaring that people who don’t seem to (effectively) understand this point about mappings don’t have a clue.)
Formalisations of these concepts arise in more obviously mathematical contexts like the study of functions and information theory, but it feels like neither of those should be necessary background for a smart person to understand the basic idea. But in all honesty, I think the inferential gap for someone who has not explicitly considered at least the idea of injections before is big enough that often people apply the absurdity heuristic or become scared to do something unconventional before the time it takes to cross that inferential gap.
I think there’s a good chance that there are neurodegenerative conditions that are currently irreversible but which many more would think worth working on than cryonics, simply because they associate cryonics with ‘computer nerd failure mode’ or apply the absurdity heuristic or because attacking neurodegenrative conditions is Endorsed by Experts whereas cryonics is not or because RationalWiki will laugh at them. Possible partial explanation: social anxiety that mockery will ensure for trying something not explicitly endorsed by an Expert consensus (which is a realistic fear, given how many people basically laugh at cryonicists or superficially write it off as ‘bullshit’). And yes, in this mad world, social anxiety really might be the decisive factor for actual humans in whether to pursue an intervention that could possibly grant them orders of magnitude more lifespan.
It seems from the comment that he does understand information theoretic views of death- after all he is talking about preserving the information via thin slices and EM scanning.
To get good vitrification results, enough to preserve the synaptic structure, his claim is that you have to take mm thin slices of brain and preserve them at nearly 2000 atmospheres of pressure. This is obviously not what current cryonics procedures do. They shoot you full of cryoprotectants (which destroys some chemical information in the brain at the outset), and then freeze the brain whole (which will lead to a lot of fracturing).
To attack his point, I think you’d need to claim that either:
he is technically wrong and current vitrification does decently preserve synaptic structures.
mechanically scrambling synaptic structures via cracking is a 1 to 1 process that preserves information.
(2) is the interesting claim, though I’d hardly trust his word on (1) since I didn’t see any especially alarming paragraphs in the actual paper referenced. I’m not an expert on that level of neurobiology (Anders Sandberg is, and he’s signed up for cryonics), but I am not interested in hearing from anyone who has not demonstrated that they understand that we are talking about doing intelligent cryptography to a vitrified brain and potentially molecule-by-molecule analysis and reasoning, rather than, “Much cold. Very damage. Boo.”
Unless someone spells out exactly what is supposed to destroy all cues of a piece of info, by explaining why two cognitively distinct start states end up looking like molecularly identical endstates up to thermal noise, so that we can directly evaluate the technical evidence for ourselves, all they’re asking us to do is trust their authoritative summary of their intuitions; and you’d be just plain dumb to trust the authoritative summary of someone who didn’t understand the original argument.
I’m trying not to be impatient here, but when I actually went to look at the cited paper and it said nothing at all about damage, it turned out this eminent authority’s original argument consisted merely of, “To read off great synaptic info with current big clumsy microscopes and primitive imaging processing, we need big pressures. Look at this paper involving excellent info and big pressures. Cryonicists don’t have big pressures. Therefore you’re dead QED.”
I suspect the basic underlying premise that causes this difference is that you believe that strongly superhuman AI will exist at the time of unfreezing, whereas most people either disbelieve that thesis or do not take it into account.
So say something about it. Your whole comment is an attack on 1, but regardless of his word on whether or not thing slice vitrification is currently the best we can do, we KNOW fracturing happens with current brain preservation techniques. Liquid nitrogen is well below the glass transition, so fracturing is unavoidable.
Why should we expect fracturing/cracking to be 1 to 1?
If you’re worried about the effects of cracking, you can pay for ITS. LN2 is only used because it is cheap and relatively low-tech to maintain.
If you ask me it’s a silly concern if we’re assuming nanorepair or uploading. Cracking is just a surface discontinuity, and it forms at a point in time where the tissue is already in a glassy state where there can’t be much mixing of molecules. The microcracks that form in frozen tissue is a much greater concern (but not the only concern with freezing). The fact that vitrified tissue forms large, loud cracks is related to the fact that it does such a good job holding things in place.
What’s “ITS”? (Google ‘only’ hits for “it’s”) How much more expensive is it? Is it offer by Alcor and CI?
Short for Intermediate Temperature Storage.
Oh ok. Thank you.
I mean, it is either his authoritative summary or yours, and with all due honesty that guy actually takes care to construct an actual argument instead of resorting to appeals to authority and ridicule.
Personally I would be more interested in someone explaining exactly how cues of a piece of info are going to be reassembled and whole brain is going to be reconstructed from a partial data.
Proving that cryo-preservation + restoration does indeed work, and also showing the exact method as to how, seems like a more persuasive way to construct an argument rather that proving that your opponents failed to show that what you are claiming is currently impossible.
If cryonics providers don’t have a proper way of preserving your brain state (even if they can repair partial damage by guessing), than I am sorry to say, but you are indeed dead.
I don’t know why you’d say he gets it—he explicitly talks about re-animation. If he addresses the actual issue it would appear to be by accident.
Thanks for this. This has substantially reduced my estimate of the odds of cryonics working.
I would not have updated much at all based on the David Ruhl comment. Cryobiology is a young field, and the details about what we can and can’t do with current preservation technology change every couple of years. Also, I think almost all bio-scientists who say current preservation techniques are insufficient to preserve you are underestimating the likelihood and the powers of machine superintelligence, and in general haven’t studied superintelligence theory hardly at all. Skepticism about cryonics working should, I think, mostly come from the other parts of the conjunction.
From approximately what to approximately what?
You can’t use “Sufficient connectomics are far off” as a counterargument to cryonics, with its passing the technological buck to the future and all.
The important part of the claim is that current vitrification is nowhere near sufficient to preserve your brain’s information.
Might it be enough if the brain was cut into slices and those stored with the process mentioned above, ignoring excessive cost?
You lose some data at the borders between slices, and you’d have to do it extremely quickly, but I’d expect it to work much better than the current method.
You’d lose any ability to do direct revival in that case.
What does “direct revival” mean? If the slices were properly reconnected, the function of the brain should be unchanged.
I think he means “create a functional human you, while primarily sourcing the matter from your old body”. He’s commenting that slicing the brain makes this more difficult, but it sounds like the alterations caused by current vitrification techniques make it impossible either way.
That criterion doesn’t make sense as per No Individual Particles and Identity Isn’t In Specific Atoms .
Unless you expect that revival of existing tissue will be a much easier path than assembly of completely new tissue. That’s a plausible assumption.
We shouldn’t make what thing we want dependant on what is harder or easier to do, and in any case if one of these is possible, the other is too. Some more centuries of technology development don’t mean much when you’re suspended.
It doesn’t change what we want, but it does change how likely we are to get it. Waiting additional centuries increases the probability of catastrophe significantly, be it a power outage or a nuclear war, as well as making it correspondingly harder to reintroduce yourself to society. And we don’t actually have any reason to believe that if one is possible then the other is too—perhaps human technology will never get to the point of being able to map the brain perfectly and the only path to resurrection is with the same tissue, perhaps the tissue will be beyond repair and uploading will be the only viable option. Both are plausible.