Fwd: Cryonics Chances
John Clark
"If we used chemical bonds to fix proteins in place, we could store cryonics patients at much closer to room temperature, and then they’d not need to be stored or managed by centralized orgs that could fail. Full-body patients could be buried as mummies in caskets, while brain-only patients could be squirreled away in small 1300cc containers. They might be stored in secret locations, perhaps in isolated permafrost, perhaps with pointers to those locations stored and protected cryptographically, only to be revealed to an advanced civ later. Yes, fixation makes restarting biology harder, but that isn’t much of an obstacle to brain emulation, which seems to me pretty sure to be the first feasible revival tech."
From: Robin Hanson from Overcoming Bias <overcom...@substack.com>
Date: Wed, Jul 24, 2024 at 9:08 PM
Subject: Cryonics Chances
To: <johnk...@gmail.com>
Attending an event on cryonics this last weekend tempts me to revisit the topic. And given the crazy tiny number of folks who have signed up for it (~4K), compared to the vast numbers of people (e.g., ~30% of my followers, & ~10% of my undergrads) who say that they think the product makes sense for them, the big cryonics marketing puzzle is: why aren’t far more people customers? The unexplained ratio (~5% of 8B / 4K) is roughly a factor of ~10,000! To explore a rational choice framework for this question, I did “break it down” polls:
Those steps:
I also asked about:
For each of these steps, here are the medians and sigmas for lognormal fits to responses for each step. (Sigmas are in e units.)
Note that the product over median step chances (0.099%) is close to median of the total chances (0.063%), making those totals look believable. Note also that the sigma of the total, though big, is much less than the sum of step sigmas, suggesting big correlations across step estimates. Finally, note that the key marketing puzzle remains, ~14% say total chance is >5%, yet only two parts in a million of the world are actual customers. Even so, these polls can plausibly help us see where people’s doubts are concentrated. We should ask not just where are doubts the biggest, but more importantly, where might we cut doubts the easiest? For example, the last two steps (7,8), though often mentioned as issues, are the least worrisome to respondents. So we should mostly ignore them. In contrast, the first two steps (1,2) have median estimated failure rates of ~69%, far higher than we see in actual cryonics data. So educating folks on these stats seems the easiest way to cut customer doubts. (Cryonics orgs: care to post authoritative stats on these params?) Two middle steps (4,5) also seem promising; respondents estimate a median ~81% failure chance that advanced future civs able to revive them would choose not to do so, and a ~72% failure chance that no civ will ever arise near Earth with very advanced tech abilities. These seem quite wrong to me, and it shouldn’t be that hard to explain why to potential customers. If we could raise customer success estimates by a factor of two on each of four steps (1,2,4,5), that should raise the median estimated total chance of success by a factor of 16, up to 1.0%, which is in the ballpark of justifying cryonics as a cost-effective purchase! Though cryonics is a tiny industry, it has disproportionately large (though still absolutely small) associated R&D activity, mostly supported by donations, not customers. And this research is mostly targeted at step 3, the one with the second lowest median chance of 8.5%. However, far less research is targeted at step 6, with the lowest median chance of 2.7%. That is, research focuses much more on increasing the odds of saving enough info in the freezing process, and much less on ensuring that orgs will preserve frozen patients for long enough. Yet we seems to have an obvious way to greatly increase the step 6 chance, at the cost of a modest cut of the step 3 chance: fixation. If we used chemical bonds to fix proteins in place, we could store cryonics patients at much closer to room temperature, and then they’d not need to be stored or managed by centralized orgs that could fail. Full-body patients could be buried as mummies in caskets, while brain-only patients could be squirreled away in small 1300cc containers. They might be stored in secret locations, perhaps in isolated permafrost, perhaps with pointers to those locations stored and protected cryptographically, only to be revealed to an advanced civ later. Yes, fixation makes restarting biology harder, but that isn’t much of an obstacle to brain emulation, which seems to me pretty sure to the first feasible revival tech. And I suspect most of these respondents low median estimate of 2.7% for step 6 didn’t take into account the bad news of a likely several centuries long innovation fall due to falling population, bad news that should greatly lower the chances of cryonics orgs maintaining their structure and plans for long enough. Yes, I have big doubts that the key cryonics marketing puzzle is best explained by this sort of rational choice framework. And I’ll try to post later on other kinds of explanations. Even so, this rational choice analysis suggests some plausible marketing strategies. First, try a lot harder to show potential customers that the chances of failure on steps 1,2,4,5 are higher than they think. Second, switch to fixation, so that patents could be stored near room temperature. If step 3’s chance falls from 8.5% to 5%, while step 6’s chance rises from 2.7% to 50%, that’s a gain of a factor of 11. Combined with a factor of 16 gain on steps 1,2,4,5 gives a new median success chance of 11%, which seems high enough to attract lots more customers. |
John Clark
>> I also agree with Hanson chemical fixation is a very promising alternative to liquid nitrogen that deserves much more emphasis:
> I don't think it makes a lot of difference. The difference between zero and what it costs to keep a neuro patient stored is about $200/year.
> Being recovered from suspension probably takes nanotechnology.
> For example, the expensive and extremely toxic chemical osmium tetroxide is routinely used for stabilization of lipids in preparation for electron microscopy.
If something as dangerous to handle and expensive as osmium tetroxide were needed then that could be a show stopper for chemical fixation being an alternative to cryonics, but I don't see why it would be needed. Osmium tetroxide is primarily not a stabilizer but a stain, it works well as a contrast agent in electron microscopes because heavy metals like Osmium scatter lots of electrons. However, how information is extracted from my 3 pounds of frozen or chemically fixed brain is not my problem, it is the problem of beings who live in an age of advanced Nanotechnology. My only concern is that the information remains intact inside that 3 pounds of grey goo, I don't know exactly how it will be extracted but I doubt it will be by electron microscopes.
> Unlike the cryobiologist, the chemical fixation researcher cannot reverse fixation and test for viability.
> The cryobiologist does not have to confine himself to this fate because he can attempt to measure viability in the brain
Obviously during revival at every step you'd like to know if you're doing it right and are on the right track, but again I don't see why cryonics would be better at this than fixation.
> or even the whole organism.
Preserving any part of the body with either method except for the brain seems completely pointless to me.
> Let us assume, for the sake of the argument, that the chemopreservation advocate has identified a number of fixatives (and other treatments) that are sufficient for complete ultrastructural preservation of the brain. The next question is going to be: how stable will chemopreservation be over time? This is a very important point for the technical feasibility of chemopreservation.
Yes, that is an important point. With Cryonics, unless we're talking about millions of years and as long as things remain cold (but will it?), pretty much all the damage that is going to be done has been done by the time the brain reaches liquid nitrogen temperatures. And I'm not worried about damage caused during thawing because that won't be done with existing technology, assuming it's even thawed at all and it probably won't be; the information will probably be read out by disassembling the brain from the outside in while it remains in solid form.
I don't know if chemical fixation would remain as stable over the centuries as freezing, my hunch is that cryonics has a small edge over fixation in this regard but I could be dead wrong, maybe it's a big edge. And I don't want to be dead.
> It is not only necessary to demonstrate that all chemicals can be introduced by perfusion fixation without perfusion artifacts
Both methods are imperfect so it is only necessary to demonstrate that fixation produces fewer artifacts than Cryonics or that the artifacts produced are easier to identify as artifacts to make it the superior technology.
> In my opinion, the prospect of autolysis is much worse because when biomolecules break up into their constitutive parts, and go into solution,
True, but if fixation is done correctly there won't be any fluid for things to move in.
> there is a risk that essential parts of the brain will not be fixed, as a result of inadequacies of the protocol, perfusion artifacts, or long term degradation. It is at this point where classic cryopreservation really shines. Even tissue that is not protected from ice formation as a consequence of perfusion impairment will still be "fixed" through low temperatures.[...]
That is another very good point. If the cryo-preservative doesn't reach a certain part of the brain things might not be hopeless because at least it still gets frozen so you still might be able to get information out of it if your technology is good enough, but if the chemical fixative doesn't reach part of the brain things are far far more serious.
But the smaller the biological sample you're trying to infuse with cryo-preservative or chemical fixative the easier it is, so both methods might be improved if before any chemical was infused or any freezing done a dozen or so thin cuts were made to slice the brain into smaller pieces. The cuts could be made very thin indeed, 30 nanometers or about 100 atoms thick. Yes you would be destroying some tissue but if the technology is good enough to repair all the damage caused by freezing or chemical fixation then I don't think they'd have much trouble figuring out what is supposed to be in that very narrow gap.
> [...] there is little hope of inferring the original structure of the brain.
Yes, the important thing is that things stay put, or at least if they must move the flow should not be turbulent so you can figure out where the parts were before they moved. If things are turbulent then a small change in initial conditions will lead to a huge change in outcome and you'll never figure out where things are supposed to go. I don't see why turbulence would occur in chemical fixation and fortunately (see below) it doesn't look like it would happen during the freezing of a brain either (I'm not interested in what happens during unfreezing, that's a problem for advanced nanotechnology, I just want to be sure the information is still inside that frozen lump of tissue). That's why I think Cryonics has a pretty good chance of working at least from a technical viewpoint, whether the brain will actually remain at liquid nitrogen temperatures until the age of nanotechnology and whether anybody will think we're worth the bother of reviving is an entirely different question.
Fluid flow stops being smoothly Laminar and starts to become chaotically turbulent when a system has a Reynolds number between 2300 and 4000, although you might get some non chaotic vortices if it is bigger than 30. We can find the approximate Reynolds number by using the formula LDV/N. L is the characteristic size we're interested in, we're interested in cells so L is about 10^-6 meter. D is the density of water, 10^3 kilograms/cubic meter. V is the velocity of the flow, during freezing it's probably less than 10^-3 meters per second but let's be conservative, I'll give you 3 orders of magnitude and call V 1 meter per second. N is the viscosity of water, at room temperature N is 0.001 newton-second/meter^2, it would be less than that when things get cold and even less when water is mixed with glycerol as it is in cryonics but let's be conservative again and ignore those factors. If you plug these numbers into the formula you get a Reynolds number of about 1. 1 is a lot less than 2300 so it looks like any mixing caused by freezing would probably be laminar not turbulent, so you can still deduce the position where things are supposed to be.
John K Clark