are you taking the positive externalities of cryonics into account?
I’m not taking them into account, but mostly because I think they’re small. The cryonics wants to be big article is only concerned with the storage costs, which I think are not a big issue. The big cost currently is in preparation for freezing, not the keeping cold. (Freezing and preparation is ~$100K while long term is only ~$100/year (so $10K if you can get 1% real return).
Hanson’s article brings up more issues, which I think is good.
Even if cryonics became massively popular, however, we’d still have several problems: we don’t know if we’re preserving all the information, we don’t know if it will ever be possible to extract the information, we might kill ourselves off first, it might never be cheap enough to revive a significant fraction of those frozen. These concern me enough that I think inducing others to sign up might actually be negative because it’s redirecting money from current spending to what I think is probably a waste.
the world of today, where cryonics is weird and marginalized
The way I see it becoming normal starts with adoption by the scientific minded rational atheists who are it’s most natural audience. This group is going to be (rightly) skeptical unless we can do a freeze/revive round trip on a person. (EDIT: I’ve backed off from this claim some downthread.) I don’t see us getting to a world where cryonics is normal just by signing up and encouraging friends to do so. What I see as most promising, actually, is the continuation of low temperature medical research that has no intended cryonics application.
The way I see it becoming normal starts with adoption by the scientific minded rational atheists who are it’s most natural audience. This group is going to be (rightly) skeptical unless we can do a freeze/revive round trip on a person.
I don’t think that is reasonable at all. Scientific minded rational atheists who want to be the gatekeepers on this matter should be basing their opinions on the entirety of the available relevant data, not (the equivalent of) whether video tapes of monkeys turning into humans can be produced.
Even if cryonics became massively popular, however, we’d still have several problems: we don’t know if we’re preserving all the information, we don’t know if it will ever be possible to extract the information, we might kill ourselves off first, it might never be cheap enough to revive a significant fraction of those frozen.
Frankly your analysis is one-sided. Uncertainty on this level should be making you move towards .5, not 0, you are seeing the absence of conspicuously conclusive evidence rather than conspicuously absent but expected evidence. That sort of observation increases randomness, it doesn’t count as disconfirmation.
We don’t know that the information is being lost (that’s kind of a key point here, and yes we do have ways of telling when information goes missing, actually the fact that we don’t know this is actually pretty good evidence to the contrary). We don’t have any particular reason to think we couldn’t extract it if it is not lost (while there are limits, the laws of physics do in fact allow a marvelous variety of things to be done). We probably won’t kill ourselves off if we’re careful (yeah there’s a reason we worry so much, and it’s because worrying works). Most expensive things not intrinsically resource-expensive tend to grow cheaper on a unit basis when there is enough demand to justify large scale infrastructure and cover initial development costs (I haven’t heard a proposal for why reviving humans would be resource-expensive, as opposed to primarily an R&D cost which really seems the more natural assumption).
What I see as most promising, actually, is the continuation of low temperature medical research that has no intended cryonics application.
It’s interesting that the guys who froze a rabbit kidney in M22 and reimplanted it successfully were in fact cryonicists (motivated by cryonics and funded by investors with cryonics interests), but kept their mouths shut about it in their published paper, making a big point instead about how this kind of research will be of great help for organ and tissue banking.
The trouble as I see it is that for cryonics to work optimally we need direct research on brain vitrification, so that we can minimize the damage that will need to be repaired. (We should strive as a rule of thumb to get it down to where mere engineered biology, as opposed to the more dramatic mechanosynthetic forms of nanotech, is sufficient.) Focusing on whole bodies and other organs is a suboptimal use of time for this goal except where it serves as a useful model for brains. Also banking of organs may become less of a concern from a general medical perspective over the next 50 years or so as we become better capable of regrowing tissue into functioning organs using scaffold and printing models, from a patient’s own DNA.
[base] opinions on the entirety of the available relevant data, not (the equivalent of) whether video tapes of monkeys turning into humans can be produced.
For evolution we have enough data that evolution comes out as the most likely theory given the data. Without freezing and thawing a brain, how do we get sufficient evidence that we’re storing everything necessary, such that “it’s all still there” becomes the most likely theory?
Most complex scientific procedures need a lot of testing to get right. We can’t test here. We might be lucky, but I do think we have only a ~20% chance of being lucky enough.
Uncertainty on this level should be making you move towards .5, not 0
Each of those steps needs to go right for cryonics to work. I’ve already moved my probabilities of success towards .5 to reflect uncertainty, but when you multiply them all together you still get a very small number.
we do have ways of telling when information goes missing
I’m not sure what you’re saying here. How would we know whether the current cryonics process were storing everything? The brain is very fragile, with lots of tiny connections, and it seems we could easily be disrupting them.
We don’t have any particular reason to think we couldn’t extract it if it is not lost
Sure. I put the impossibility of extraction given that we preserved everything at 5%.
We probably won’t kill ourselves off if we’re careful
I agree. I put us killing ourselves off at 10% and society falling apart (if we don’t kill ourselves first) at 40%. Together this gives a 46% chance of one of them happening. Do you think these odds are too pessimistic?
I haven’t heard a proposal for why reviving humans would be resource-expensive, as opposed to primarily an R&D cost which really seems the more natural assumption
Reviving someone means very powerful computers and lots of storage. We might figure out how to do this for some people, but still have it be too expensive for everyone. I do think this is unlikely, though: I put down 10% as the chances that it would be simply too expensive.
the guys who froze a rabbit kidney in M22 and reimplanted it successfully were in fact cryonicists (motivated by cryonics and funded by investors with cryonics interests), but kept their mouths shut about it in their published paper, making a big point instead about how this kind of research will be of great help for organ and tissue banking.
Interesting. I didn’t know this.
I was thinking more about the cooling we do for heart surgery.
For evolution we have enough data that evolution comes out as the most likely theory given the data. Without freezing and thawing a brain, how do we get sufficient evidence that we’re storing everything necessary, such that “it’s all still there” becomes the most likely theory?
A good starting point would be electron micrographs of the tissue. Another would be cellular viability assays based on cellular ion pumping action.
We can also rule out competing hypotheses of memory formation which would negate the premise by examining what happens to the memories of humans subjected to hypothermia. Even better, there are some tests on the connectome hypothesis we can plausibly run in the near future (IIRC this involves teaching music to some neurons, then scanning them and recreating the song in the computer).
Then there is the measurement of electrical activity in tissue samples. Organized whole-brain electrical activity should be adequate proof of biological (far more conservative than the standard of information-theoretic) survival. I expect this to be achieved in 20 years or so, given adequate funding and attention. I would expect perfect brain preservation to be a bit further on, probably closer to 50-year.
Most complex scientific procedures need a lot of testing to get right. We can’t test here. We might be lucky, but I do think we have only a ~20% chance of being lucky enough.
I agree that the process needs a lot of testing to get right. However the testing can take the form of tissue viability assays, it does not need to take the form of a whole body. That is far more ambitious than a whole brain and thus harder to perfect. Different organs have different optimal cooling rates and perfusion rates. We should only expect that particular test to be possible either when we are at the point of printing bodies and transplanting brains into them on demand, or when we have long since solved the problem of individual organ (including brain) preservation.
Another thing we have going for us is that technology of the future can almost certainly repair the brain to some degree. Even if we are limited to bio-analogous nanotech, such as genetically engineered microbes (and I don’t think this is the case where cryopreserved brains are concerned), we can still expect fairly extensive repair work to be possible.
Each of those steps needs to go right for cryonics to work. I’ve already moved my probabilities of success towards .5 to reflect uncertainty, but when you multiply them all together you still get a very small number.
They don’t all have to go right in the most dramatic 100% manner. For example, you probably do not need the brain to be able to remember more than 90% of the past self to be a valid form of survival, and you might not need more than 10%. To some degree this is a subjective values question. A clone-like individual with vague memories of your childhood may be worth spending a lot of money on, or it may not be… But it’s extremely low probability that you won’t preserve enough information for at least that kind of reanimation, whereas the likelihood that you won’t have any perceptible memory loss at all is also pretty low (for a cryopreservation performed today).
I’m not sure what you’re saying here. How would we know whether the current cryonics process were storing everything? The brain is very fragile, with lots of tiny connections, and it seems we could easily be disrupting them.
The brain is warm, wet, and protein-based, and has to put up with a lot of chaos on a day to day basis. It is composed of wetware, not hardware. Thus it is reasonable to expect the data to be stored fairly redundantly and with the most important and oldest information in the toughest/thickest connections.
Many of the smaller connections are likely to be lost in a suboptimal preservation, as happens I believe with some forms of dementia. However this is one of the things we can get empirical feedback on by scanning dendrite preservation in small animals as we cryopreserve and thaw them.
Sure. I put the impossibility of extraction given that we preserved everything at 5%.
Sure, but it makes a big difference whether everything is preserved by say preserving viability in all the cells (a theoretical ideal currently achievable only in slices), versus say fixation, which locks the information in place but completely eliminates viability. All other things equal, a zero-viability mechanism has less probability of being survivable than a viability-preserving mechanism.
I put us killing ourselves off at 10% and society falling apart (if we don’t kill ourselves first) at 40%. Together this gives a 46% chance of one of them happening. Do you think these odds are too pessimistic?
Yes I think they are, given how long we have avoided either of these fates so far. Although I’d be cautious about downplaying them since the act of downplaying them could lead to them happening due to lack of caution. Literacy, science, and so forth have a very strong rachet effect on progress. I’d give the conjunction of the two (downfall of civilization and killing ourselves off) a 10% maybe, supposing relative tech stagnation over the next 200 years consistent with no reanimation. Less if tech doesn’t stagnate, since that is more correlated with literacy.
Another point worth raising at this point is that cryonicists have added incentive to look at extinction and civilization-wrecking events using near-mode reasoning. This could be a positive externality worth taking into consideration.
Reviving someone means very powerful computers and lots of storage. We might figure out how to do this for some people, but still have it be too expensive for everyone. I do think this is unlikely, though: I put down 10% as the chances that it would be simply too expensive.
The storage space needed (ongoing) should be no more than that of the brain itself, though I can see some being required temporarily for heavy extrapolations if we need to infer lots of details from badly preserved tissue. This should be correlated with worst case viable preservation (low to no cellular viability, lots of simulation and educated guesswork involved in the reanimation process).
If we get to the point of perfectly or near-perfectly preserved brains, it should be cheap in principle to implant them into printed bodies (possibly with artificial limbs, hearts, etc.) The reason to think simulation is the viable candidate is mainly an extrapolation of Moore’s Law which indicates this kind of storage space should be extremely cheap before we hit physical limits, and the fact that you’d need some hefty nano-equipment to scan on that level of detail anyway.
I was thinking more about the cooling we do for heart surgery.
This is hypothermic medicine, which is both important for cryonics and important to distinguish from cryonics (which is more a subset of cryobiology). The temperature levels involved are much closer to room temperature, not cryogenic. There is an interesting parallel to cryonics in that a patient might be cooled to hypothermic temperatures (similar to a hibernating animal), and if they could be stabilized there they would age more slowly, cancers would progress more slowly, etc.
You’ve convinced me that there’s actually a lot we can do to test how well current cryonics processes work that doesn’t require round trips on human brains.
If we do these tests and adjust the cryonics procedures in response to what we learn, my estimate for the chances that we’re not preserving what we need to will probably go down a lot.
They don’t all have to go right in the most dramatic 100% manner
Most of the steps I’m on the negative side for are pretty much binary: either they succeed and you proceed to the next step or they fail and you’re done. Looking at all the ones that I think there’s over a 10% chance of failure, and labeling those where you might get a partial reconstruction:
You die suddenly or in a circumstance where you would not be able to be frozen in time: 0.1 (partial)
Some law is passed that prohibits cryonics before you die 0.1 (absolute)
The cryonics people make a mistake in freezing you 0.1 (partial)
The current cryonics process is insufficient to preserve everything 0.8 (partial)
All people die 0.1 (absolute)
Society falls apart 0.4 (absolute)
Some time after you die cryonics is outlawed 0.2 (absolute)
All cryonics companies go out of business 0.4 (absolute)
The cryonics company you chose goes out of business 0.1 (absolute)
The technology is never developed to extract the information 0.6 (absolute)
No one is interested in your brain’s information 0.4 (absolute)
It is too expensive to extract your brain’s information 0.4 (absolute)
The technology is never developed to run people in simulation 0.4 (absolute)
Running people in simulation is outlawed 0.2 (absolute)
No one is interested running you in simulation 0.3 (absolute)
It is too expensive to run you in simulation 0.1 (absolute)
Other 0.2 (both)
given how long we have avoided either of these fates so far
Humanity is far more powerful and capable than it’s been for most of it’s history. It will probably get more that way.
cryonicists have added incentive to look at extinction and civilization-wrecking events using near-mode reasoning. This could be a positive externality worth taking into consideration.
I’d be surprised if promoting cryonics beats straight up x-risk awareness advocacy.
The storage space needed (ongoing) should be no more than that of the brain itself
You may be right. If we can get a compact representation this might not be that big at all. At 20 billion neurons with, maybe 100 connections per neuron, four bytes per connection we have ~7TB of information. This is way less than the processing requirements. The WBE Roadmap thinks we might need anywhere from 50 TB to 10^9 TB depending on the level we need to emulate at (p79).
I’m not taking them into account, but mostly because I think they’re small. The cryonics wants to be big article is only concerned with the storage costs, which I think are not a big issue. The big cost currently is in preparation for freezing, not the keeping cold. (Freezing and preparation is ~$100K while long term is only ~$100/year (so $10K if you can get 1% real return).
Hanson’s article brings up more issues, which I think is good.
Even if cryonics became massively popular, however, we’d still have several problems: we don’t know if we’re preserving all the information, we don’t know if it will ever be possible to extract the information, we might kill ourselves off first, it might never be cheap enough to revive a significant fraction of those frozen. These concern me enough that I think inducing others to sign up might actually be negative because it’s redirecting money from current spending to what I think is probably a waste.
The way I see it becoming normal starts with adoption by the scientific minded rational atheists who are it’s most natural audience. This group is going to be (rightly) skeptical unless we can do a freeze/revive round trip on a person. (EDIT: I’ve backed off from this claim some downthread.) I don’t see us getting to a world where cryonics is normal just by signing up and encouraging friends to do so. What I see as most promising, actually, is the continuation of low temperature medical research that has no intended cryonics application.
I don’t think that is reasonable at all. Scientific minded rational atheists who want to be the gatekeepers on this matter should be basing their opinions on the entirety of the available relevant data, not (the equivalent of) whether video tapes of monkeys turning into humans can be produced.
Frankly your analysis is one-sided. Uncertainty on this level should be making you move towards .5, not 0, you are seeing the absence of conspicuously conclusive evidence rather than conspicuously absent but expected evidence. That sort of observation increases randomness, it doesn’t count as disconfirmation.
We don’t know that the information is being lost (that’s kind of a key point here, and yes we do have ways of telling when information goes missing, actually the fact that we don’t know this is actually pretty good evidence to the contrary). We don’t have any particular reason to think we couldn’t extract it if it is not lost (while there are limits, the laws of physics do in fact allow a marvelous variety of things to be done). We probably won’t kill ourselves off if we’re careful (yeah there’s a reason we worry so much, and it’s because worrying works). Most expensive things not intrinsically resource-expensive tend to grow cheaper on a unit basis when there is enough demand to justify large scale infrastructure and cover initial development costs (I haven’t heard a proposal for why reviving humans would be resource-expensive, as opposed to primarily an R&D cost which really seems the more natural assumption).
It’s interesting that the guys who froze a rabbit kidney in M22 and reimplanted it successfully were in fact cryonicists (motivated by cryonics and funded by investors with cryonics interests), but kept their mouths shut about it in their published paper, making a big point instead about how this kind of research will be of great help for organ and tissue banking.
The trouble as I see it is that for cryonics to work optimally we need direct research on brain vitrification, so that we can minimize the damage that will need to be repaired. (We should strive as a rule of thumb to get it down to where mere engineered biology, as opposed to the more dramatic mechanosynthetic forms of nanotech, is sufficient.) Focusing on whole bodies and other organs is a suboptimal use of time for this goal except where it serves as a useful model for brains. Also banking of organs may become less of a concern from a general medical perspective over the next 50 years or so as we become better capable of regrowing tissue into functioning organs using scaffold and printing models, from a patient’s own DNA.
For evolution we have enough data that evolution comes out as the most likely theory given the data. Without freezing and thawing a brain, how do we get sufficient evidence that we’re storing everything necessary, such that “it’s all still there” becomes the most likely theory?
Most complex scientific procedures need a lot of testing to get right. We can’t test here. We might be lucky, but I do think we have only a ~20% chance of being lucky enough.
Each of those steps needs to go right for cryonics to work. I’ve already moved my probabilities of success towards .5 to reflect uncertainty, but when you multiply them all together you still get a very small number.
I’m not sure what you’re saying here. How would we know whether the current cryonics process were storing everything? The brain is very fragile, with lots of tiny connections, and it seems we could easily be disrupting them.
Sure. I put the impossibility of extraction given that we preserved everything at 5%.
I agree. I put us killing ourselves off at 10% and society falling apart (if we don’t kill ourselves first) at 40%. Together this gives a 46% chance of one of them happening. Do you think these odds are too pessimistic?
Reviving someone means very powerful computers and lots of storage. We might figure out how to do this for some people, but still have it be too expensive for everyone. I do think this is unlikely, though: I put down 10% as the chances that it would be simply too expensive.
Interesting. I didn’t know this.
I was thinking more about the cooling we do for heart surgery.
A good starting point would be electron micrographs of the tissue. Another would be cellular viability assays based on cellular ion pumping action.
We can also rule out competing hypotheses of memory formation which would negate the premise by examining what happens to the memories of humans subjected to hypothermia. Even better, there are some tests on the connectome hypothesis we can plausibly run in the near future (IIRC this involves teaching music to some neurons, then scanning them and recreating the song in the computer).
Then there is the measurement of electrical activity in tissue samples. Organized whole-brain electrical activity should be adequate proof of biological (far more conservative than the standard of information-theoretic) survival. I expect this to be achieved in 20 years or so, given adequate funding and attention. I would expect perfect brain preservation to be a bit further on, probably closer to 50-year.
I agree that the process needs a lot of testing to get right. However the testing can take the form of tissue viability assays, it does not need to take the form of a whole body. That is far more ambitious than a whole brain and thus harder to perfect. Different organs have different optimal cooling rates and perfusion rates. We should only expect that particular test to be possible either when we are at the point of printing bodies and transplanting brains into them on demand, or when we have long since solved the problem of individual organ (including brain) preservation.
Another thing we have going for us is that technology of the future can almost certainly repair the brain to some degree. Even if we are limited to bio-analogous nanotech, such as genetically engineered microbes (and I don’t think this is the case where cryopreserved brains are concerned), we can still expect fairly extensive repair work to be possible.
They don’t all have to go right in the most dramatic 100% manner. For example, you probably do not need the brain to be able to remember more than 90% of the past self to be a valid form of survival, and you might not need more than 10%. To some degree this is a subjective values question. A clone-like individual with vague memories of your childhood may be worth spending a lot of money on, or it may not be… But it’s extremely low probability that you won’t preserve enough information for at least that kind of reanimation, whereas the likelihood that you won’t have any perceptible memory loss at all is also pretty low (for a cryopreservation performed today).
The brain is warm, wet, and protein-based, and has to put up with a lot of chaos on a day to day basis. It is composed of wetware, not hardware. Thus it is reasonable to expect the data to be stored fairly redundantly and with the most important and oldest information in the toughest/thickest connections.
Many of the smaller connections are likely to be lost in a suboptimal preservation, as happens I believe with some forms of dementia. However this is one of the things we can get empirical feedback on by scanning dendrite preservation in small animals as we cryopreserve and thaw them.
Sure, but it makes a big difference whether everything is preserved by say preserving viability in all the cells (a theoretical ideal currently achievable only in slices), versus say fixation, which locks the information in place but completely eliminates viability. All other things equal, a zero-viability mechanism has less probability of being survivable than a viability-preserving mechanism.
Yes I think they are, given how long we have avoided either of these fates so far. Although I’d be cautious about downplaying them since the act of downplaying them could lead to them happening due to lack of caution. Literacy, science, and so forth have a very strong rachet effect on progress. I’d give the conjunction of the two (downfall of civilization and killing ourselves off) a 10% maybe, supposing relative tech stagnation over the next 200 years consistent with no reanimation. Less if tech doesn’t stagnate, since that is more correlated with literacy.
Another point worth raising at this point is that cryonicists have added incentive to look at extinction and civilization-wrecking events using near-mode reasoning. This could be a positive externality worth taking into consideration.
The storage space needed (ongoing) should be no more than that of the brain itself, though I can see some being required temporarily for heavy extrapolations if we need to infer lots of details from badly preserved tissue. This should be correlated with worst case viable preservation (low to no cellular viability, lots of simulation and educated guesswork involved in the reanimation process).
If we get to the point of perfectly or near-perfectly preserved brains, it should be cheap in principle to implant them into printed bodies (possibly with artificial limbs, hearts, etc.) The reason to think simulation is the viable candidate is mainly an extrapolation of Moore’s Law which indicates this kind of storage space should be extremely cheap before we hit physical limits, and the fact that you’d need some hefty nano-equipment to scan on that level of detail anyway.
This is hypothermic medicine, which is both important for cryonics and important to distinguish from cryonics (which is more a subset of cryobiology). The temperature levels involved are much closer to room temperature, not cryogenic. There is an interesting parallel to cryonics in that a patient might be cooled to hypothermic temperatures (similar to a hibernating animal), and if they could be stabilized there they would age more slowly, cancers would progress more slowly, etc.
You’ve convinced me that there’s actually a lot we can do to test how well current cryonics processes work that doesn’t require round trips on human brains.
If we do these tests and adjust the cryonics procedures in response to what we learn, my estimate for the chances that we’re not preserving what we need to will probably go down a lot.
Most of the steps I’m on the negative side for are pretty much binary: either they succeed and you proceed to the next step or they fail and you’re done. Looking at all the ones that I think there’s over a 10% chance of failure, and labeling those where you might get a partial reconstruction:
You die suddenly or in a circumstance where you would not be able to be frozen in time: 0.1 (partial)
Some law is passed that prohibits cryonics before you die 0.1 (absolute)
The cryonics people make a mistake in freezing you 0.1 (partial)
The current cryonics process is insufficient to preserve everything 0.8 (partial)
All people die 0.1 (absolute)
Society falls apart 0.4 (absolute)
Some time after you die cryonics is outlawed 0.2 (absolute)
All cryonics companies go out of business 0.4 (absolute)
The cryonics company you chose goes out of business 0.1 (absolute)
The technology is never developed to extract the information 0.6 (absolute)
No one is interested in your brain’s information 0.4 (absolute)
It is too expensive to extract your brain’s information 0.4 (absolute)
The technology is never developed to run people in simulation 0.4 (absolute)
Running people in simulation is outlawed 0.2 (absolute)
No one is interested running you in simulation 0.3 (absolute)
It is too expensive to run you in simulation 0.1 (absolute)
Other 0.2 (both)
Humanity is far more powerful and capable than it’s been for most of it’s history. It will probably get more that way.
I’d be surprised if promoting cryonics beats straight up x-risk awareness advocacy.
You may be right. If we can get a compact representation this might not be that big at all. At 20 billion neurons with, maybe 100 connections per neuron, four bytes per connection we have ~7TB of information. This is way less than the processing requirements. The WBE Roadmap thinks we might need anywhere from 50 TB to 10^9 TB depending on the level we need to emulate at (p79).
That would be convenient :) Then I don’t have to do anything.
Thanks for the well argued article and comment.