?

Log in

Thoughts on cryonic preservation & revival - Liam's write-only LJ
February 20th, 2010
05:29 pm

[Link]

Previous Entry Share Next Entry
Thoughts on cryonic preservation & revival
This is a rather long reply to a post from ciphergoth. The question being, is it plausible that, in future, we will be able to resurrect people from their head, cryonically frozen post-mortem?

I am keenly interested in the prospect of whole-brain emulation, which strikes me as potentially plausible, with reasonable probability. For one thing, I think that this is one of the more doable routes to AI - rather than trying to build a mind from scratch, to instead try to bootstrap it by attempting to reproduce the existing biological structures.

This being the case, it begs an obvious question, as it were: once (implicit "if" here & from now on) we have WBE, then the next big desirable leap would be scanning a biological brain and running the resultant dataset on an emulation. IOW, Kurzeilian "uploading". This strikes me as a consummation devoutly to be wished.

However, I feel - as a long-ago biology grad - that the prospects for taking a complete adult brain, scanning it & getting any data out of it that is worth uploading are virtually 0. I will come back to this in a moment.

But this is starting with a living, functioning brain - albeit possibly impaired by old age, disease or trauma. (Because why would one choose to do it if alive & healthy?)

However, the prospects of doing it from a dead brain seem to me to be far closer to 0, in a Zeno's-Paradox sort of way. Once one is outside that critical 4min window of an oxygen-deprived brain, I suspect that the remaining amount of useful information drops precipitately, with every passing minute, and after 2-3x that 4min window, I suspect there isn't enough left to be worthwhile. Given some hand-waving magical technology for interpreting memories absent the consciousness that recorded them - and of course we don't know if different consciousnesses record them in compatible or even comparable formats - one might be able to retrieve some memories from a dead brain, but a mind? I doubt it.


Now we are, ISTM, approaching really quite closely to 0. But taking the additional step of freezing the dead brain first, with all the damage that the process causes, and I think we can watch the decimal point recede into the distance as we're down into the 99.9 and lots more nines % probability that there is nothing left to recover.

I will skip the process of thawing the brain before recovery. From my reading, AIUI, people in coma, after cardiovascular incidents, chilled after falling in frozen rivers, etc., are often in a dormant but viable state; it is the process of recovery that is fraught. Again AIUI, it is the return of oxygen to the tissues in an uncontrolled fashion that often causes massive damage.

If we are merely trying to recover the mind that (might) be still stored in a frozen brain, then why thaw it? If there is a Plot Device™ magical brain scanning machine, then let us hypothesize one that can work on a frozen brain & avoid all the damage caused by thawing it.

So let us skip that part. We freeze the brain, very fast, somehow without causing microfractures, cellular rupture from ice crystal formation, etc. etc., and we manage to do this within 4min of death. Lot of big “ifs” in there. But if we can do that, then safer to find a way to sample the frozen brain without thawing.

So there are a succession of problems here:

[1] Getting to the person soon enough after death – I think preservation will need to be done within the same time period that revival would be.

[2] Freezing the brain inside that period, without damaging it – major implementation difficulty here.

[3] Implicit in #2, a preservation method that preserves the structures or patterns that encode the mind in question. Currently more or less impossible to say, as we don't know what structures or patterns those are.

[4] Sampling the result with fine enough resolution to retrieve useful amounts of info.

[5] Finding a way to process, implement or run the resultant dataset in a way that will effectively resurrect the consciousness.

That is one hell of a big list of implementation details there. I find them daunting, frankly.

Now, to skip back to the issues-with-scanning-a-living-brain thing I mentioned in ¶5.

I do not find this plausible with any technology that is usefully distinguishable from magic. You need not only an accurate synapse-level model, you also need measurements of the distribution of a whole slew of fairly fragile little molecules – dopamine, acetylcholine, adrenaline & noradrenaline, &c. &c. - not only within the cells but possibly also in the distribution between cells. One may also need information about the movements of these chemicals; this is a dynamic process we're talking about here, involving flows, gradients and so on, in the cytoplasm of the neurons and possibly in the synapses too. We simple do not know yet how most of this works, so we don't know what needs to be sampled or stored. In some cases we don't know the fine detail, in others, not even the outlines. What we do know is that it is a very complex dynamic system.

(This is one of the major reasons I find the prospect of resurrection from dead, frozen brains so implausible as to be massively improbable. I think by the time the meat is frozen, “you” are long gone.)

But let's say we had such a machine. To get this level of detail, you need to go inside the cells; it seems highly probable that the scanning process is therefore destructive. Very destructive.

Now if one had a Plot Device that could take the top off someone's head and start scanning at 1mm/sec, capturing all that ultramicroscopic detail – and given the volumes of data and the size of the structures, that is *fast* - then it's no use, because by the time you are 1-2cm in, the patient is dead, and the rest of the data is therefore useless.

In this thought experiment, in fact, it doesn't really matter how fast your machine is, because if you start at one point and work through – and if our machine goes at 1cm a second, and boy is this going to be one messy machine, a device for rapidly turning heads into a cloud of pinkish mist - then still, after you're a short way in, the rest of the brain collapses into a chaotic mess and there's no data left to sample. The machine will have a broadly similar effect to a dumdum bullet to the frontal lobes.

Essentially this seems to me like an insurmountable problem.

Even an imaginary machine that could destructively sample a whole brain in a fraction of a second – and I think it's a safe bet it would be destructive at such a speed – is going to work at such a speed that the waste heat would cook the brain being sampled and again destroy the data being acquired.

The size of the system and its fragility precludes effectively sampling a living system, and as discussed, I think that only a living brain is worth sampling; a dead one, I suspect, would not contain enough information to get a functioning mind out of it.


This leads, ISTM inevitably, to non-destructive sampling: to trying to design and build a system that can record the state of a whole conscious brain, while it is working, and record the info in such a way as to allow a functioning copy to be built.

Given the previously-discussed fine detail needed, then this is not going to be a magic cap you put on like a hat. The sampling probes are going to need to be right in there, up alongside the dendrites, sampling their activity over a period of time.

This seems to me to mandate nanotechnology, basically. You are going to have to grow the sensor system inside the brain & the end result will be of comparable complexity to the brain itself. It has to be.

(This leads to the mechanistic argument against all-knowing deities: any omniscient deity must be a system of equal or greater complexity to the universe in which it is omniscient, therefore, it must be bigger than the universe. But it has to know itself, too, so it has to be bigger than itself, which is clearly impossibe; therefore, omniscience is impossible. And if it's not omniscient and/or omnipotent, then it's not a deity.)

But one for a single human brain? That is imaginable.

Growing this sensor network will be slow, because if you do it fast, you'll cook the brain. So it probably has to be implanted in the person some time before – possibly years before. Possibly even at birth or soon after. Also, you need damned small sensor nets, as they have to fit into the gaps without disrupting the dendritic structures.

Given that - and at this point I think it's obvious you need nanotechnology to do it – then I find the prospect of getting enough data, and good data, to have a chance of replicating it.

Which just leaves the non-trivial problem of building something that can run it.

But if you have nanotech good enough to build the sensor, the computer to run the dataset is probably not a problem.

If enough data persisted in a dead, frozen brain, of course, one could grow the sensor net slowly in that. They're dead, they won't mind waiting. But I don't think enough data would persist; it is fairly logically clear that minds are dynamic systems & I don't think you can just turn off the hardware, stop it in place, then restart it again later. I would be very happy to be proved wrong on this point, but until neuroscience moves on a fair bit, we won't know.

[Deep breath]

Next comes the mundane issues:
[1] Can you trust the cryonics company to keep your head in good nick?

[2] When the technology exists to resurrect corpsicles, who is going to do this and why? Who pays? Why should they bother? What if they're in an overpopulated future (a very safe bet for the foreseeable one), why would they want more people in the form of resurrected dead? Perhaps they might bother bringing back genius scientists, great artists, and so on, but ordinary Joes? Why?

[3] In what form would you get resurrected, or want to be? As a flesh human in a cloned body? As an awareness in a robot body? As a purely simulated awareness in a virtual space?

[4] Assuming the hypothetical future society does resurrect people if it can, what for? What if you're a cheaper equivalent to an AI, managing stock databases for a subjective millennium? Would you want that?

There are so many ifs and buts in the way, I do not see it as a useful option. Personally I would rather bet on uploading of live people while I still am one.

Which allows me to end on a song, with only a minor rewrite of the lyrics...
I'm your prime cut of meat, I'm your choice
I wanna be uploaded...

Tags: , , , ,

(126 comments | Leave a comment)

Comments
 
[User Picture]
From:ciphergoth
Date:February 20th, 2010 05:39 pm (UTC)
(Link)
I'd like to reply to your points in detail, but just as a general enquiry: if you found that one of the assumptions you've made in your argument in this post was contradicted by existing science, would it cause you to take cryonics more seriously?
[User Picture]
From:lproven
Date:February 20th, 2010 06:03 pm (UTC)
(Link)
Absolutely, yes, and I am keen to know more, especially if I'm wrong or out-of-date.

And in the nature, as someone else put it, of Pascal's Wager, well, if you can afford to do it, then definitely do; there is no real downside.
[User Picture]
From:zotz
Date:February 20th, 2010 05:50 pm (UTC)
(Link)
Something that occurred to me a while after reading the good Mr Crowley's thoughts on the matter seems to be confirmed (FSlimitedVO) by thee ackurssed Witchipedia:

Research has suggested that long term memory storage in humans may be regulated by DNA methylation.

That's a real oh-shit problem in my view - getting all the methyl groups right on the dna of at least a good proportion of someone's neurons strikes me as showstopper. DNA methylation not being involved in brain functioning would make the data extraction much easier.
[User Picture]
From:lproven
Date:February 20th, 2010 06:09 pm (UTC)
(Link)
That sort of detail is an area I know next-to nothing about, but I am currently doing some reading in the broad area...

Wouldn't surprise me if you're bang on, though!
[User Picture]
From:arkady
Date:February 20th, 2010 05:52 pm (UTC)
(Link)
Growing this sensor network will be slow, because if you do it fast, you'll cook the brain. So it probably has to be implanted in the person some time before – possibly years before. Possibly even at birth or soon after. Also, you need damned small sensor nets, as they have to fit into the gaps without disrupting the dendritic structures.

It would be most logical to implant the neural sensor net at birth, so it grows and develops as the brain does, becoming a nanotechnology mirror of the flesh-and-blood brain.

Which pretty much rules out preserving and bringing back the genius scientists, great artists etc - there is no way of telling who will be a great artist, who will be a genius scientist - and who will have the mental stability to reach their potential. Though implantation with a neural net offers a whole new exciting world of nanotreatments for mental illnesses, disorders and epilepsy - with a whole slew of moral dilemmas inherent in meddling with what makes a person's mind and personality.

I suspect any such immortality is likely to come as a convenient byproduct of medical treatments with other aims in mind. It's not going to come as a result of direct research into cryonics and immortality; given the increasing problems of world overpopulation, the governments of the Western world are not going to be funding research that will add to that overpopulation.
[User Picture]
From:lproven
Date:February 20th, 2010 06:07 pm (UTC)
(Link)
Absolutely, yes.

The only cavil is that if you do this for everybody, or more or less everybody, and store all the data just in case, then possibly, /post facto/, one can decide who was a great genius, unappreciated in their time, and bring 'em back later on. This presupposes that data storage continues to get cheaper & cheaper, which is, I think, one of the safest bets in the whole general area under discussion.
[User Picture]
From:grendelkhan
Date:March 29th, 2010 04:37 pm (UTC)

Have you read "Learning To Be Me"?

(Link)
There's a short story by Greg Egan called "Learning To Be Me" from 1990, which seems to explore a very similar idea.
[User Picture]
From:ciphergoth
Date:February 20th, 2010 06:12 pm (UTC)
(Link)
First, many thanks for sharing your thoughts as a stand-alone blog post. This is exactly what my blog post about this was intended to encourage. And it meets three of my criteria at the end of the article!

So I'm hoping you'll take the fourth step, of explicitly discussing the arguments set out by cryonics advocates on the subjects you discuss - you're clearly much better placed to evaluate them than I am. For example, you say:
However, the prospects of doing it from a dead brain seem to me to be far closer to 0, in a Zeno's-Paradox sort of way. Once one is outside that critical 4min window of an oxygen-deprived brain, I suspect that the remaining amount of useful information drops precipitately, with every passing minute, and after 2-3x that 4min window, I suspect there isn't enough left to be worthwhile.
I'm guessing you're thinking of ischemic damage? This issue is discussed in detail in Scientific Justification of Cryonics Practice which is the document to rebut if you're interested in this subject. Fahy's cited study on frozen and vitrified rat hippocampal slices which looked for ischemic damage in the most vulnerable regions with an electron microscope, and Safar's study of cats subjected to a full hour of global cerebral ischemia would seem to be relevant here. Does that evidence indicate to you that you've overestimated the rate of information destruction through ischemic damage?
Now we are, ISTM, approaching really quite closely to 0. But taking the additional step of freezing the dead brain first, with all the damage that the process causes, and I think we can watch the decimal point recede into the distance as we're down into the 99.9 and lots more nines % probability that there is nothing left to recover.
This is a little vague - can you discuss exactly what form of freezing damage you're most concerned by? Obviously the toxicity of cryoprotectants isn't a concern for WBE unless they actually damage information.
We freeze the brain, very fast, somehow without causing microfractures, cellular rupture from ice crystal formation, etc. etc., and we manage to do this within 4min of death.
Cryonics organisations claim that their current freezing and vitrification process completely avoids ice crystal formation. Some fractures are certain; for WBE, you're betting on the computer being able to match up the two sides of the fracture; it isn't obvious to me that that's an intractable jigsaw puzzle.
[User Picture]
From:thette
Date:February 20th, 2010 08:49 pm (UTC)
(Link)
After reading that article and looking at his sources, I have to say it reads as if it's written by someone who really, really want his pet theory to be true. The mixture of old data, suspect publications, animal data, almost-but-not-quite unrelated real advances, fudged middle steps ("and then a miracle happens") and boundless optimism makes me distrust it all. I definitely wouldn't trust him with my body or my brain.

(It's also not applicable to us living in countries where cardiac arrest isn't the definition of legal death.)
[User Picture]
From:ciphergoth
Date:February 20th, 2010 06:13 pm (UTC)
(Link)
[1] Getting to the person soon enough after death – I think preservation will need to be done within the same time period that revival would be.

[2] Freezing the brain inside that period, without damaging it – major implementation difficulty here.

[3] Implicit in #2, a preservation method that preserves the structures or patterns that encode the mind in question. Currently more or less impossible to say, as we don't know what structures or patterns those are.

[4] Sampling the result with fine enough resolution to retrieve useful amounts of info.

[5] Finding a way to process, implement or run the resultant dataset in a way that will effectively resurrect the consciousness.
[1] and [2] I've discussed above - it currently looks to me as though existing cryonics best practice will suffice.

[3] Again, I refer you to "Scientific Justification", in particular this bit. If Best is overreaching here that would be an excellent observation to post.

[4], perhaps this shows my ignorance, but I just find this really hard to worry about. Seeing the advances in various kinds of scanning technology over the decades, it feels as though the safe assumption today is that at least for a frozen subject that we can slice up we'll eventually reach the required resolution, unless there's some fundamental limit that we already know about that we're in danger of hitting. Of course there could be a fundamental limit that we don't know about, but that doesn't seem like the right side to bet on.

[5] I think the Church-Turing thesis applies - again there are huge practical problems, but I can't identify one that really seems like it might stand against a century of technological improvement.
(This is one of the major reasons I find the prospect of resurrection from dead, frozen brains so implausible as to be massively improbable. I think by the time the meat is frozen, “you” are long gone.)
You're asserting information-theoretic death? By what process is the information destroyed?

I agree that trying to scan a living, non-frozen brain certainly seems like a much greater challenge; I wouldn't want to bet on it being impossible forever, but it's clearly much, much further away. However that doesn't directly touch on the arguments about cryonics.

I'm staying off the non-technical issues for now, so I'll stop there. Once again, many thanks for making this blog post - I've linked to it from my blog.
[User Picture]
From:autopope
Date:February 20th, 2010 06:23 pm (UTC)
(Link)
I think a reference to Hans Moravec's gedankenexperiment on mind/body dualism is probably worthwhile at this point, if for no other reason than that he suggests an actual protocol for uploading a non-dead (at first) mind (and expands on it at some length in "Robot").
[User Picture]
From:ciphergoth
Date:February 20th, 2010 06:16 pm (UTC)
(Link)
BTW please let me know if there's a part of your argument relevant to the technical feasibility of future reanimation given current cryonics practice that I've failed to address - thanks!
[User Picture]
From:lproven
Date:February 21st, 2010 07:35 pm (UTC)
(Link)
I am going to have to get back to you on this after about 2-3 days' worth of reading, at the current rate of rapid Firefox tab multiplication.

However, I think there is a general problem here that is going to make it very hard for me or anyone to come up with the sort of specific refutation that you are looking for.

In general, this is that you seem to want (FWIW, I think reasonably enough) a specific point-by-point refutation or rebuttal of a set of assertions and claims of potential future techniques, methods and knowledge.

More specifically:

One one hand, currently, it's hard to get really detailed informed speculation on the nature of memory, the timespan of its preservation post-mortem, the possibility for its artificial retrieval, etc., because nobody yet knows exactly how memory is stored or accessed, and very little indeed is known about the nature of consciousness, its locality, function, etc.

When little is known, it is certainly possible to say "one day, we might do X", and it is virtually impossible to gainsay this. One can only speak in generalisations. You are unsatisfied with these, which I can entirely understand, but that is all that it possible to have right now. I don't think any hypothetical neuroanatomist could refute a comment that claimed that in future a machine will be invented to read memories out of frozen dead brain cells. Nobody knows.

Secondly, you seek specific, detailed refutations of what are actually quite vague, general claims of what may, possibly, be possible with future technology. This too is essentially impossible to do. What one can do is say that, probably, this will be possible or that will not; for instance, I would say that, in general terms, I think it will eventuate, more or less, that what is possible with nanotech is what is possible in living tissues, no more, no less. Darwinian evolution is an insanely good optimisation mechanism and it's been operating in this instance for ~3½ æons. I suspect that no amount of human ingenuity will fundamentally transcend what living cells do. This means, on the nanoscale, liquid-phase chemically-driven machines which operate quite slowly.

Or, more broadly, houses growing out of engineered coral-like or bone-like substances, but not fleets of microscopic robots repairing dead bodies.

Overall, I entirely sympathise with what you're looking for, but the sort of informed, authoritative commentators who /could/ potentially write such refutations won't, because it's not scientific. Those who can or will write such refutations are not specialists, they are, at best, educated amateurs, and their refutations will be irritatingly woolly and I think you'll find them unconvincing.

Climate science is actually a relatively well-explored area, with active research, lots of widespread involvement & much political interest. This means that it's possible to find some very high-quality commentary & analysis. Nonetheless, much controversy remains, at heart because lots of not-that-well-informed amateurs find what the experts are saying to be, literally, unacceptable & unbelievable.

Cryonics is in the same boat, but with a minority interest & much less debate. Many people find what the experts are saying unacceptable or unbelievable, but there isn't yet much evidence either way & so what experts there are, are staying well out of it.

I fear this will no be resolved for several decades yet, at least.
[User Picture]
From:watervole
Date:February 20th, 2010 07:12 pm (UTC)
(Link)
This kind of post is why I read LJ rather than Twitter.
[User Picture]
From:ciphergoth
Date:February 20th, 2010 07:57 pm (UTC)
(Link)
For me, Twitter is part of how I find this kind of post...
[User Picture]
From:lproven
Date:February 20th, 2010 10:58 pm (UTC)
(Link)
I think you said that last time I posted a long 'un. :-)

While I do take your point, Paul's is also a valid 1. & they're not mutually exclusive!

P. S. Thanks! ;-)
[User Picture]
From:reddragdiva
Date:February 20th, 2010 08:10 pm (UTC)
(Link)
I assume you've read our lovely RationalWiki article? Lots to chew through there, and on the talk page ...

I was neutral-to-positive until I actually looked into it (prompted by Paul's posts) and rapidly concluded it was - and here's a useful new word - pseudotechnology. Arguing with cryonics advocates on the talk page has some small entertainment value as well.

Don't forget, the answer to every object is "nanobots."
[User Picture]
From:reddragdiva
Date:February 20th, 2010 08:16 pm (UTC)
(Link)
Every objection.
[User Picture]
From:lovingboth
Date:February 20th, 2010 09:57 pm (UTC)
(Link)
[1] On the timescales I think even the optimists are talking about, I wouldn't bet on the survival of the USA, never mind an American company.

[2] I can imagine historians / anthropologists wanting to.
[User Picture]
From:lproven
Date:February 20th, 2010 10:43 pm (UTC)
(Link)
Actually, yes, good points both.

The "time capsule" argument /had/ occurred to me - honest! - but I'm not sure it's something I'd particularly want for myself. But hey, better'n the alternative!

P. S. Happy birthday!
[User Picture]
From:reddragdiva
Date:February 20th, 2010 11:10 pm (UTC)
(Link)
Note that one cryonics facility already went bust (leading to the thawing of the frozen corpses) and cryonics advocates are themselves concerned that the existing two are small and not financially robust.
[User Picture]
From:waistcoatmark
Date:February 21st, 2010 10:18 am (UTC)
(Link)
I would have thought that if you can build a nano-level set of sensors capable of aptruring the entire brain, you're capable of building a computer that could "run" the result.

Certainly in 2-D mircro-processor(1) land the technology required to build a sensor net on top of a CPU that could detect the charge levels on each transistor/capacitor while that CPU is running is way in advance of the technology to build the CPU.

(1) shouldn't we be calling them nano-processors these days?
[User Picture]
From:lproven
Date:February 21st, 2010 01:11 pm (UTC)
(Link)
:¬) Good points, well made!
[User Picture]
From:steer
Date:February 21st, 2010 09:22 pm (UTC)
(Link)
I said this to ciphergoth and I'll say the same to you. Our knowledge of what makes consciousness is not really more advanced than when we were cavemen. We have no clue as to what is necessary for consciousness. Is emulation of a brain in software sufficient to cause consciousness? Would any material running "that algorithm" be equivalent to my consciousness (whether it be software or a vast "choose your own adventure book)? Is there some piece of physics missing which explains why consciousness resides where it does and why, as far as we know, it is only an attribute of humans and perhaps some animals.

For this reason we do not know at all what it takes to reconstruct consciousness. Perhaps, as some have speculated, a sufficiently advanced civilisation could reconstruct the consciousness of anyone who ever lived by tracking back their effects on the present. Perhaps, on the other hand, ten minutes after death, too much information is lost. Until we know actually what causes consciousness, for me, any speculation on the usefulness of cryo will be as scientific as down the pub "I reckon what they should do is" as an estimate of economics.
[User Picture]
From:reddragdiva
Date:February 22nd, 2010 09:15 pm (UTC)
(Link)
I'd personally assume the rough-and-ready Turing test humans have generally used, i.e. "it's human if it fights back hard enough that we will negotiate with it better assuming it's human." This avoids requiring testing the unprovable about whether the entity in question has the right to use the word "I."
Liam's Google Profile Powered by LiveJournal.com