Break Cryonics Down

The essence of analysis is to "break it down", to take apart vague wholes into clearer parts.  For the same reasons we make point lists to help us make tough job decisions, or ask people who sue for damages to name an amount and break it into components, we should try to break down these important social claims via simple calculations.  And the absence of attempts at this is a sad commentary on something. [Me last July]

Imagine you disagreed with someone about the fastest way to get from your office to Times Square NYC; you said drive, they said fly.  You broke down your time estimates for the two paths into part estimates: times to drive to the airport, wait at the airport, fly, wait for a taxi, ride the taxi, etc.  They refused to offer any component estimates; they just insisted on confidence in their total difference estimate. 

Similarly imagine some someone who disagree about which of two restaurants was better for a certain group, but wouldn't break that down into who would like or dislike what aspects of the two places.  Or imagine someone who claimed their business plan would be profitable, but refused to break this down into how many of what types of units would be sold when, or what various inputs would cost.  Or someone who said US military spending was worth the cost, but refused to break this down into which enemies were how discouraged from what sorts of damage by that last spending increment.

Such silent disputants reject our most powerful tool for resolving disagreements: analysis – breaking vaguer wholes into clearer parts.  Either they have not used this tool to test or refine their estimates, or they are not willing to discuss such parts with you.  I felt Tyler made this analysis-blocking move in our diavlog:

R: So is your skepticism about cryonics based on physics, or some more metaphysical relation of consciousness to physics, or skepticism about current brain science, or skepticism about the social institutions required to make this work?
T: I would say all of it, but just that its really hard … then I'm supposed to think that my head can be frozen and brought back with even a one in a million chance?  I'm just not there.
R: OK, so it's just sort of a pure surface implausibility; its not based on the structure or the pieces of the argument.
T: Absolutely, that's exactly right.
R: OK, so this does come back more to our initial hypothesis for our disagreement, which is I prefer to have some structure where I rely on the conclusions of the structure even if my intuitions will influence which pieces I choose for the structure, and you will more just have a reaction to the overall conclusion, …
T: You're always wanting to give me a number, what's the chance you'll be frozen, what's the chance you'll be thawed out, convert it to an expected utility equation; again I pretty quickly put it into the class of cases of extremely unlikely; payoff might be high, by Pascal's wager payoff might be murky … I don't think thinking about cryonics is silly at all. …
R: But at the end of the day your judgment is based on just a holistic evaluation that seems to say this has a really low chance, that isn't based on a structured calculation of the chances of each parts, its just overall it looks unlikely and thats it. 
T: I don't like that word just; your evaluation is as holistic as mine …

Here is my break down of the cryonics question.  (This one is from 20 years ago.)

Assume you die soon in a way a cryonics org can freeze you with its usual quality.  (If not, you don't have to pay for freezing.)  Regarding the year 2090, consider the conditional probability of each of the following steps, given the previous steps:

  1. Civilization still exists and has kept growing in technical capability.
  2. Your cryonics org and it successors have kept you continuously frozen.
  3. Someone is willing and allowed to pay modest costs to revive you.
  4. Brain science has workable input/output models of relevant brain cell types.
  5. Usual freezing quality preserved relevant model-needed details.
  6. Cheap scanning tech slices & 2D scans brains at model-needed spatial, chem resolution.
  7. Error correction codes reconstruct most connections across slices, fractures.
  8. Cheap computers can real-time sim entire scanned sets of connected cells.
  9. Sim life seems worth living enough that they don't prefer suicide.
  10. Such sims of you are as worthy as your kid of your identifying with them.

I choose the year 2090 because that is about twice as far in the future as cryonics orgs have already survived while preserving almost all patients, and it seems plenty of time for scan and sim techs to achieve required levels. 

Your chance of being usefully revived in 2090 as an em is roughly the product of these ten conditional probability terms.  Ten 90% terms gives a total chance of ~1/3.  Ten 80% terms gives a total chance of ~10%, except step 4 might be a 50% chance, for a total chance of ~6%, which seems about right to me. 

More precisely, if folks are reasonably smart about when to try to revive you, your total revival chance is something like a sum across all future times of such calculations, each one given no destructive failed prior attempt. 

If you make 50K$/yr now, and value life-years at twice your income, and discount future years at 2% from the moment you are revived for a long life, but only discount that future life based on the chance it will happen, times a factor of 1/2 because you only half identify with this future creature, then the present value of a 5% chance of revival is $125,000, which is about the most expensive cryonics price now.

GD Star Rating
loading...
Tagged as: ,
Trackback URL:
  • Kenny Evitt

    Maybe Tyler’s afraid of losing track of something important, something more easily accessible to his intuition, by using analysis of this kind. I understand him most clearly when he said “… I pretty quickly put it into the class of cases of extremely unlikely;” – I sense Tyler is playing devil’s advocate for the rationality of ignorance on this subject.

  • Tom Powers

    What is Robin’s solution to Pascal’s Wager? The existence of this problem suggests that there is something wrong with decision theory, which is basically what Robin uses, at least in certain situations similar to cryonics.

  • kurige

    I apologize if this has been covered in past discussions of cryonics in OB, but what if you just want to die when you die? Is this a “bad” philosophy for rationalist to adopt?

    On a slightly less serious note, here’s a very serious reason to reconsider.

  • Carl Shulman

    Kurige, that story seems to have been bogus.

  • dph

    I don’t know if this has been discussed, Robin’s point 3 attracts me for some additional thought…

    If you are depending on some future intelligence being interested enough to pay to revive you, and resources are limited in some fashion in this future, then my intuition is that some people have significantly lower probability of being thawed (would you revive Wally your great-great-great-grandfather’s next door neighbour or Alexandre Dumas) . The more people who sign up, the lower the odds are that you will be thawed.

    On the other hand, if there aren’t enough interesting people, society might not spend resources to investigate successful resuscitation methods, or some future fear monger might campaign to ban resuscitations (see stem-cells for a possible analogue).

    So my intuition says there is an optimal number of people, and that number is likely is a function of how interesting you will be in the future.

    Does this fact determine how aggressively you advocate for others to sign-up? Do you advocate people you deem likely to be less likely to be more interesting than you in some future?

  • ac

    “Assume you die soon in a way a cryonics org can freeze you with its usual quality. (If not, you don’t have to pay for freezing.)”

    Well, that’s a relief!

  • http://williambswift.blogspot.com/ billswift

    I actually think it more likely to be possible than Robin apparently does. I just don’t see any reason to bother. I want to live as long as I can, because I enjoy it, especially learning and making things. But I am not afraid of dying, when I die I’ll be dead and won’t be able to have regrets, or miss anything. So why bother with getting myself frozen, even if it was cheaper than it is, that is still money I can’t use for other things.

  • kebko

    I thought Tyler was just being a scamp in the video blog, but now that I see Robin’s defense, I am more persuaded of Tyler’s side. Firstly, I think it is reasonable to be skeptical of any series of 10 conditions where most of the probabilities are currently unknown or are 0. It is very easy to pull probability numbers out of a hat that seem reasonable, but with 10 items, the potential for error seems much more important than whatever probability it spits out. Thinking back 80 years in the past, if we imagined some similar contingency that was based on expectations of future technology, I think it is easy to see where things can go awry. While we have many technologies which past prognosticators would never have imagined, there would have been many contingencies on their list which they would be surprised to know never came about. So, for instance, mapping the human genome is extremely important, yet they would not have even known to predict it. But, they might have given a 90% probability to something like personal jet packs or universal pneumatic delivery systems or something. So, with a list of 10 contingencies, it seems likely that even some of the most reasonable-sounding ones are going to end up tripping it up. I can understand using a heuristic that simply writes off something with 10 steps of uncertainty. At the very least, it wouldn’t take much of an adjustment to put the present value a lot lower.
    Dr. Hanson’s views have fundamentally changed the way I see the world, but this argument seems specious.

  • http://www.ciphergoth.org/ Paul Crowley

    This is an argument that the cost to me is worth the benefit to me. I’d be interested to see an argument that the cost to the world is worth the benefit to the world.

  • Alejandro Dubrovsky

    While theoretically you don’t pay if you are not frozen, in practice most people at that point have paid already through their life insurance, so adding the following two factors seems to me to be a requirement:

    Dying in a vitrification-friendly manner (high likelihood for old people, medium for young people)
    Getting to CI/Alcor before your brain seeps out through your ears (low likelihood for US residents, very low for the rest)

  • http://profile.typepad.com/robinhanson Robin Hanson

    Tom, I don’t see much relation between this and Pascal’s wager.

    dph, the more are revived, the more chance one of them knows you and wants to revive you.

    kurige and bill, do you reject all medical treatment and forgo all safety precautions?

    kebko, most any business plan has at least ten things that can go wrong; should one assume any such plan has only a one in a million chance of success?

    Paul, the externality is positive; since fixed costs dominate, the more who sign the lower everyone’s costs are.

    Alejandro, most people know someone else they could usefully have the insurance money go to.

  • http://profile.typepad.com/6p011168fc9622970c Virge

    Robin, assume for the moment you’re living in 2090 and:
    – the technology that allows you to give runtime to a frozen head also allows you to create a duplicate sim of yourself;
    – hundreds of thousands of people have been frozen (almost none in the last couple of decades because one can either stay alive or upload to a sim);
    – for the last decade, resuscitation technology had been reliable and has given runtime to many thousands of those people;
    – all of those frozen who appeared to have any outstanding capabilities or perspectives have already been resuscitated as sims.

    If you had to choose which way to spend your money to benefit humanity (obviously now defined to include sims), which would be better:
    – resuscitate an ordinary, unremarkable person frozen 80 years ago, or
    – create a duplicate sim of yourself?

    Or to put it another way, can you quantify the expected relative effect of
    adding 1 clear-thinking rationalist with an up-to-date education and a known desire to improve the human condition,
    compared with
    adding 1 randomly selected person of unknown rationality, unknown morals, and an education that’s 80 years out of date.

  • Grant

    Tyler’s attitude towards breaking things down is something I’ve seen all over his blog as well. Occam’s razor suggests I should believe the simplest explanation: that Tyler’s chose his attitude because it is beneficial for his status as an academic, and may not have much do to with truth-seeking.

    Of course, a willingness to break down the probabilities doesn’t mean cryonics is worthwhile.

  • Paul

    “times a factor of 1/2 because you only half identify with this future creature”

    Can you really sweep all the metaphysical uncertainty under the rug with just a discount factor here?

    Unless we’re incredibly interesting historical figures, it seems like there isn’t much incentive for future people to actually revive us. Would it work to make a small, very secure investment now, payable upon revival to whoever brings us out, whenever they bring us out?

  • http://timtyler.org/ Tim Tyler

    “As worthy as your kid of your identifying with them”?

    You can get precisely that with kids. Freeze some sperm if you want to. Indeed, a clone would give you about double that! Reconstructing a diploid chromosome from a semen sample seems likely to be quite doable soon – and mitochondrial mutations are mostly just defects.

    From the perspective of a future society, an appropriately-educated clone would usually offer similar value to other types of reconstruction. It’s all much easier, cheaper, and more likely to happen.

  • frelkins

    I see here in some comments an attitude that acts as if people have no inherent value, as if love has no meaning, as if normal human attachment doesn’t exist, and as if they seem unsure that people are motivated by genetic relationships.

    dph, “Wally” may be nothing to you, but he is someone’s “great-grandfather;” as an “ordinary person,” he is deeply loved and valuable. In many cultures, your great-grandfather and grandfather are the most important members of your family!

    Distant relatives have great human value – before the wide spread of Christianity, forms of ancestor worship appear to have been most prevalent and of course still persist today. Genealogy remains a popular pastime.

    As for Tyler, he gave his answer clearly even as he squirmed out of dealing with the question honestly – signaling, he finally confessed. Thus I agree with Grant, he values his brief academic status more than his own life or the existence of his family. It seems a male monkey paradigm.

  • http://wealthywhitemale.blogspot.com Brian

    I guess my objection to cryonics is different from Tyler’s but equally squishy. I’d say my moral intuition tells me that, given the state of the world, this type of technology is likely to be available only to the wealthy and powerful. I’d rather focus on getting us out of the fix we’re in than working towards the preservation of a cohort I’ll never join*. Perhaps something like immortality is possible in the long run, I just think there are so many projects we’ve got to tend to first.

    There’s a deep tension in Robin’s thinking here: one minute he’s jonesing for personal reincarnation and the next he’s trashing the social value of localism/tribalism? This goes back to economists’ original sin, the de/reification of self interest (Robin, of course, adheres to his model well – global markets suit his class interests and the same goes for cryonics). But perhaps this is just another intuitional dispute – some see markets opening up opportunities for the dispossessed and others see them mostly preserving old power structures under a shellacking of democracy. Isn’t the establishment of institutions which produce brilliant and happy minds more vital to the good of humanity than the infinite preservation of a small fraction of those minds? Just recently I think a blogginghead mentioned Galton’s surprise…

    I find it shocking Robin can’t even conceive of a cogent critique of mainstream economics. Tyler comes up with the tribalists/nationalists but surely there are others harder to dismiss? How about an ecological critique of capitalism? Maybe you guys should argue with people with whom you actually disagree.

    Lastly, how will we know that cryonics will work out the way Robin and Tyler prefer? How will we know if consciousness is actually preserved?

    *what I mean is that if the technology is widespread enough for those I love to adopt it then I would too, but that I don’t see this scenario as likely within my lifetime and so i’d rather put my energy/money elsewhere

  • mathusalem

    omg 80%?
    breakdownthis:
    I.N.S.U.R.A.N.C.E. S.A.L.E.S.M.A.N.

  • http://www.theshitizens.com Grant

    “kebko, most any business plan has at least ten things that can go wrong; should one assume any such plan has only a one in a million chance of success?”

    Surely this is more bizarre than any proposed business plan, and on an extremely long time line. To suppose it’s a calculation between probabilities (and like many others, I’m new to this debate, but from where are you getting these probabilities?) and an ultimate value one might achieve at a small cost is one thing, but a business plan with elements under 50% for the next year is in trouble. I think the problem with your calculation is that you hold the probability steady from now until 2090. It seems that each probability would increase or decrease as we moved closer to the plausible technology and a) realized it was impossible or unpleasant, b) found the secret to ressurecting brains. If we’re supposed to calculate thesse odds, don’t we need to calculate the odds that they increase as well? I’m not paying it all now, based on these calculations, but a fee over time and so each payment should become necessarily refigured according to the possibility of the future in the future. Or am I missing something here?

    I would also like to point out that I think cryonics is the worst possible location for this argument to take place. It has been often a foolhardy “industry” that has played on people’s fears (see This American Life). Which brings up another point I don’t think you’re properly taking into account. People who choose to prolong their lives this way, while it remains a fringe technology, have a need to exist into the future in someway in excess of their dead bodies. For them, I see an extra benefit to cryonics, while they’re dying, first of all, a relief that this death is not final, and could perhaps see a money savings you haven’t figured in if they avoid excessive end-of-life care. However, what they’re also paying for is additional “history” for lack of a better word; they want not just to be remembered but to contribute positively to the sense we have of them by committing more actions. If cryonics fails and their body becomes a rotten corpse in a leaking vat, exposed to the world as a fool who believed in ludicrious technology, they will again contribute to other’s memories of themselves, but in a pathetic and ugly. While this is a cost for nearly anyone with a well-developed sense of the future, it is particularly so for these people because of the high values they put on their future possibilities. No?

  • http://gondwanaland.com/mlog/ Mike Linksvayer

    I agree with Robin’s assessment (roughly .8 for each of the conditionals) and understand that he’s right about financial cost.

    However, I more or less agree with billswift’s rationale for not signing up. As to Robin’s query to bill — if there are safety or medical costs that skimping on would result in increased mortality but not morbidity or any other discomfort, please clue me in so I can start skimping.

    Furthermore, financial costs are not the only ones. I’ll sign up when and if cryonics becomes cheaper, more certain, and less weird.

  • spacenookie

    Step #10 seems to imply/presuppose that you should have as many children as you can afford.

    If your child has a 90% chance of being alive in 2090, but you only identify with him/her 50% as much as your own em, a child would be 9x as valuable as a 5% chance of revival, or $1,125,000. Not counting grandchildren.

  • Jeffrey Soreff

    Robin’s list of factors:

    1. Civilization still exists and has kept growing in technical capability.
    2. Your cryonics org and it successors have kept you continuously frozen.
    3. Someone is willing and allowed to pay modest costs to revive you.
    4. Brain science has workable input/output models of relevant brain cell types.
    5. Usual freezing quality preserved relevant model-needed details.
    6. Cheap scanning tech slices & 2D scans brains at model-needed spatial, chem resolution.
    7. Error correction codes reconstruct most connections across slices, fractures.
    8. Cheap computers can real-time sim entire scanned sets of connected cells.
    9. Sim life seems worth living enough that they don’t prefer suicide.
    10. Such sims of you are as worthy as your kid of your identifying with them.

    My list of factors:

    1. One has to actually receive a prompt (pre-autolysis) suspension
    2. Something like Drexler/Merkle nanotechnology has to be developed.
    3. Something capable of massively parallel microsurgery has to be developed from nanotechnology.
    4. The problem of backtracing movement of fractured neural tissue has to turn out to be easy to develop.
      (This one is a specialty application – if it takes more than a handful of person-years, we’re sunk.)
    5. Future medicine has to develop a cure for aging.
    6. The specific cryonics provider that one uses has to survive till the above conditions are met.
    7. One’s cryonics provider has to be motivated to actually thaw/repair its patients.

    My (2) is roughly your (1).
    My (6) is roughly your (2).
    My (7) is roughly your (3) [though your (3) allows broader potential revivers - point taken].
    More later (not sure if I can edit indefinitely…)
    I’m phrasing the odds in terms of a reconstruct-a-biological-brain,
    and you seem to be phrasing it in terms of reconstructing-into-an-upload.
    so the analysis and repair to an accurate neural topology is an equivalent
    precondition for both scenarios. I’d be happy to have either scenario
    succeed but describing the preconditions for both paths gets a bit more
    complex…

  • Russ Andersson

    To his credit Robin has proposed a model that attempts to quantify the probability of a successful thaw. And there has been some sensible and justified criticism of his model and approach. Fair enough.

    I’m curious, if Robin’s model is so bad and so flawed, why haven’t any of the critics been able to propose a better one?

    And if they can’t come up with a better approach, how can they say that Robin’s model is not the best assessment we have at the moment, even though it is flawed?

    Are you saying this issue can’t be meaningfully quantified at all, by anybody, or that it can be quantified but he is wrong?

    My basic position is that until somebody comes up with a better model, surely Robin’s model, by definition is the best one we have.

  • http://profile.typepad.com/michaelvassar Michael Vassar

    I’d like data on what fraction of historical societies with hundreds
    of years of continual technological progress in their past still had
    80 years of similar speed or faster progress in their future (I am
    comfortable claiming no tech progress if population declines and
    industrial production declines over a 40 year period even if people
    invent some new tools as they always do). I’d guess it’s less than
    80%.
    I also doubt that 80% of corporations the size of cryonics orgs last
    80 years, especially if they are controversial.
    The probability of restoring ‘you’ without cryonics, as-la Kurzweil’s
    father, surely also has a non-zero probability. We are concerned with
    the difference of these probabilities.
    Tyler brought up immortality via eternal recurrence and the like but
    didn’t do much with the idea. It seems that it may have relevance to
    the cryonics debate, though that isn’t clear.

    All told, I think that one has to rig the components of your analysis
    to come out against cryonics and I think its reasonable to insist on
    doing analysis, but the case isn’t so clear cut that rigged numbers
    can’t deliver a negative conclusion without being individually
    non-credible.

    By the way, I really prefer the traditional term “upload” to “em”.

  • http://profile.typepad.com/robinhanson Robin Hanson

    Michael, are you honestly surprised the Cryo orgs have not even come close to collapsing after 40 years? You need to be a bit surprised I think if you expect little chance of surviving 80 more years.

  • steven

    The past didn’t have all the possible technological planet-killers that the future may well have; if past evidence is at all relevant, it’s as a very optimistic upper bound on the probability of no doom.

  • http://profile.typepad.com/bayesian Peter McCluskey

    Using the “break it down” approach to evaluate overall probabilities has a poor track record. According to Tetlock’s book Expert Political Judgment (starting around page 190), pundits using that approach overestimated probabilities more than those who didn’t (in part due to using the inside view instead of the outside view).

    Lists in business plans of 10+ things that can go wrong are rarely used to break down estimates of the business’s probability of success.

  • http://www.hopeanon.typepad.com Hopefully Anonymous

    Peter, superlatively excellent contribution to the thread.

  • Carl Shulman

    Michael,

    I think the idea of talking about ‘emulations’ reflects 1) an attempt to avoid incorporating controversial philosophical conclusions about continuity of identity into the definition of the term; 2) an attempt to connect the term to simpler brain modeling efforts 3) a sense that it sounds more ‘serious.’ However I agree that the more jargonish abbreviation ‘em’ loses most of the virtues of ‘brain emulation.’

  • Unnamed

    If the expected benefit of cryonics is only in the low hundred thousands, that suggests that cryocrastination isn’t such a bad idea, at least into your 40s or 50s. If you’re a 30-year-old male, you have about a 95% chance of surviving the next 20 years. So if being signed up for cryonics when you die has an expected benefit of $200,000, waiting 20 years only has an expected cost of $10,000 (and that’s assuming that you can’t change your mind and sign up if you get sick, and that you’ll die in a vitrifiable condition). Waiting lets you save on membership fees, see what’s happening with the relevant technology (will any of those probability estimates change in the next 20 years?), and learn more about different cryonics orgs before you commit to one. The biggest advantage of an early signup may be in helping to build the cryonics movement, rather than the direct personal benefit.

  • sternhammer

    Well, contra Mike Linksvayer, Robin doesn’t actually provide any estimates. He just says that IF they are all .9 then the result is …” I think Robin’s move of starting a breakdown is valuable, but there is a lot farther to go. Waving a hand at them and saying “they are all .8″ doesn’t seem any better than Tyler’s nonanswer to me.

    It would be nice to have forums on each one to debate the probabilities. I don’t know anything about 3 – 8, so I have nothing there, but some here might.

    9 and 10 are very abstract philosophical matters. I’m not sure they are the kind of empirical predictions we can put probabilities on. We can make up numbers for 1, but with a base n of 1 I don’t know how good our numebrs could be. Though I would also add a 1a that says “civilization does not have serious resource constraints that cause it to privilege the living over the claims of frozen heads.”

    #2 is something I know a little about. And I would cast the odds a lot less than .8.

    First of all, there are a lot of sub-risks there that should be broken out. Here’s my list, others might have stuff to add:

    2a. Mismanagement of funds. I don’t get the impression these Alcor guys are any kind of financial geniuses, since they put half the money into one commercial real estate building in Arizona right before the Arizona real estate market cratered. And they haven’t adopted a conservative strategy with the other half either. What’s the chance they will have any money left in 50 years? I’d guess less than .5.

    2b. Casualty. Fire destroys a lot of buildings. Or a truck hits the power lines that feed the tanks. There’s an earthquake. Someone might get this data from an insurance company, I’ll shortcut and call it .9 chance here, but if you correct me I’ll believe it.

    2c. Power outages. Lots of power outages don’t lead to loss of buildings, but might thaw a head. How long can a tank stay cold without electricity? We don’t even know that this hasn’t happened already. If your buddy’s head got thawed in a damaging way, and then they refroze it (or revitryfied it) how would you know? Arizona gets a lot of summer brown-outs due to AC usage. Again, I’m not an expert, I’ll call this .8.

    2d. Change in market conditions. What if electricity becomes much more expensive? Where is the energy market going? I can easily imagine power getting expensive faster than the endowment grows, especially in conjunction with 2a. I call it .9.

    2e. Fraud and embezzlement. What makes us think that what happens on Wall Street with money every day couldn’t happen here? There is no agency oversight, no legal recourse, and no market that really cares. And the victims can’t complain. As I mentioned in the last thread, I think the risk of this is huge. I say .4 chance of surviving it.

    So I get .1296 for step 2 alone. OK, that’s higher than my off-the-cuff guess of .05. Others please do suggest your own numbers.

    I wonder if these organizational and legal factors are not getting enough attention in the calculation of many OB devotees because they are boring and not-science-fiction-y. But interestingly, these are factors that cryonics fans could do something about: all these risks are meliorable, if you face up to them.

  • http://www.theshitizens.com Grant L. (sorry 8:46 AM Grant, didn’t notice you earlier.)

    “I’m curious, if Robin’s model is so bad and so flawed, why haven’t any of the critics been able to propose a better one?”

    This argument supposed Robin’s model is the best model to this point. I don’t think that’s clear at all, and how could we measure it if it were? Generally, I would measure it against my own valuations and having found cryonics thoroughly unconvincing (not to mention, having an overly large population of crackpots who believe other, even less convincing things) and I find it faulty. If you find cryonics justifiable or rational, then your measure is probably more sympathetic, but it’s an old, absurd trope that the only substantive critique replaces the existing model with a more accurate one.

  • Patri Friedman

    I like Grant’s brief mention of cryonics vs. techniques to briefly extend one’s life. Here in the US, a large portion of medical costs occur at the end of someone’s life, trying to extend their life a little bit longer. If I had a choice between an extra week and a shot at immortality, I’d take the latter. Maybe that’s a place to be pitching cryonics, to insurance companies as a cheaper alternative to making customers happy when they are about to die, compared to buying extremely expensive marginal units of life. And to the customers themselves.

  • http://pancrit.org Chris Hibbert

    Brian objected that “this type of technology is likely to be available only to the wealthy and powerful.” Can you name any technology that has become available to the wealthy and powerful and remained out of the reach of the middle class (in relatively free societies)? At any point, you may be able to point to a few technologies that are currently rare or expensive, but the normal course of things is for everything valuable to be more affordable and more ubiquitous.

    And he closes with “How will we know if consciousness is actually preserved?” Consciousness isn’t magic, but there’s also no simple gauge to detect it. We distinguish between the mental levels of dogs, chimpanzees, and people. Deciding whether or not emulations or revived people are conscious will be more a question of how aware they are of the world they inhabit and of themselves than of detecting whether they retain some ephemeral spark communicated from the preserved person.

  • peco

    Here are what I think are the lowest plausible (<5% chance of being lower) probabilities for #1-10:

    1. 80%
    2. 50%
    3. 30%
    4. 30%
    5. 85%
    6. 15%
    7. 10%
    8. 70%
    9. 90%
    10. 60%

    -> 0.00017 or 0.02%

  • peco

    The final probability is extremely low. Compared to 1/6 (nine 90% terms and one 50% term), it is ~1000 times lower, which tells me that the probability is very hard to estimate usefully this way (there are reasonable estimates both significantly above and significantly below 5%).

  • Jeffrey Soreff

    @peco – mostly agreed, though I think of 10 as being more of a
    philosophical point than as part of the question of whether cryonics
    succeeds. I think 1 could be lower than 80%: We depend on a particular
    kind of progress, something like Moore’s law continuing.
    Computational progress might go the way of the SST, or turn into a
    backwater like nuclear power or manned spaceflight became for many years.
    6 is a potentially big hurdle. People use ultramicrotomes but
    they don’t generally try to section an entire brain to synapse-level
    (~20nm!) precision. Even with Drexler/Merkle nanotechnology, this is
    a special-purpose project, possibly peculiar to cryonics. The same is
    true of writing software to backtrack cooling fractures. I think that we
    are basically guessing at the probability that we are lucky and these
    tasks are inherently easy, given enough cheap nanotech and CPUs to throw
    at them.

    I very much agree that “the probability is very hard to estimate usefully this way” on a similar calculation…

  • http://profile.typepad.com/6p011168f25f92970c Ralph

    Alas, Robin, you have erred in your analysis.

    Before you can multiply multiple items to get an overall probability, you must first show they are (at least to first approximation) independent.

    The items you list are, for the most part, highly correlated.

    As an example, take the first two items from your list:

    1 Civilization still exists and has kept growing in technical capability.
    2 Your cryonics org and it successors have kept you continuously frozen.

    If your cryonics organization has succeeded in keeping you cryopreserved, then we can reasonably conclude that civilization continues to exist.

    If your cryonics organization has kept you cryopreserved, then it is also very likely that your item 3 is true: “Someone is willing and allowed to pay modest costs to revive you.” In particular, your cryonics organization should be willing to revive you provided that the costs of doing so are less than the present value of the costs of keeping you cryopreserved in perpetuity.

    The other items listed also suffer from severe correlations, meaning the product of their probabilities is not admissible.

    Producing a list of independent factors is harder than it looks (Steve Harris also got this wrong when he ventured a similar analysis many years ago, see Will Cryonics Work?
    Examining the Probabilities
    . Mike Perry also pointed out this error, and provided a better analysis).

    If we want to look at independent risks, we might proceed as follows:

    1) If information theoretic death occurs at any time, then you die.
    1a) Information theoretic death might occur before you are cryopreserved (you might be lost at sea, or suffer a neurological disease that destroys the brain).
    1b) Information theoretic death might occur at the time you are cryopreserved (you have a heart attack and aren’t found for a week).
    1c) Information theoretic death might occur if you are prematurely thawed (civilization collapses, cryonics is outlawed, the universe comes to an end, etc).
    1d) Information theoretic death might occur because of the slow accumulation of damage.
    2) Technologies that are feasible in principle are never developed and applied.

    That’s it.

    Even here, item (2) is not independent because, strictly speaking, if we never develop and apply the technology to revive you then either 1(c) or 1(d) will occur.

    Item 1(a) does not seem to pose a major risk. Most people suffer legal death because of a heart attack or cancer. Further quantification of the risk of neurological deterioration (which, in the context of cryonics, might result in some degree of amnesia) prior to a declaration of legal death would be of interest.

    For further consideration of 1(b), the risk that information theoretic death occurs when you are cryopreserved, see Cryonics, Cryptography, and Maximum Likelihood Estimation. While severe cases (unattended death followed by many days of deterioration prior to discovery) would likely result in information theoretic death, shorter periods (likely including multiple hours) should pose a relatively small risk (though there is much disagreement about this). Short periods (cryopreservation under favorable conditions) would seem unlikely to pose a significant risk.

    Item 1(c) can reasonably be summed up as “Your cryonics organization fails, for whatever reasons, to carry out its mission.” There are a wide range of estimates and considerable disagreement. Alcor was founded in 1972, and its continued existence and growth is a favorable indicator. Pragmatically, the more people who support cryonics the less likely that this failure mode will occur.

    Item 1(d) is unlikely to occur in less than a few thousand years (see Merkle’s cryonet posting on this subject and the comments by a cryobiologist).

    Item 2, the failure to develop the required technology despite the passage of multiple centuries, seems unlikely. More specifically, given our current understanding of physics and chemistry we should, in principle, be able to answer this question reliably. Published analyses strongly support the feasibility of both nanotechnology and nanomedicine. Feynman’s class talk There’s Plenty of Room at the Bottom is required reading for anyone interested in this subject.

  • http://profile.typepad.com/robinhanson Robin Hanson

    Ralph, I cited the Harris attempt, and I explicitly talked about conditional probabilities given prior list items; so independence isn’t required. I don’t think I’m disagreeing with you, though I note you don’t offer any probability numbers.

  • Kevin Dick

    Robin, I have a question. The way I read your post, you aren’t discounting during the time from death until reanimation. Why? It seems like I really care a whole lot less about dollars (as a proxy for utility) in the far future.

    If you apply even a modest discount to the dollars (or utility) of an infinite series starting some time substantially far in the future, the present value remains pretty small. You have to posit a geometric average growth rate over the expected time frame that is greater than the discount rate. Actually, it has to be quite a bit more to make up for the low probability of success.

    I’m wondering if there’s some explanation of this somewhere that I missed.

  • sternhammer

    So, Robin, sorry I misunderstood you before. You ARE claiming that of your ten highly disparate items, answerable with very different modes of analysis, all have the same predicted probability (except step 4).

    Wow. That’s a pretty big coincidence.

    You chided Tyler for not offering more analysis. Do you want to offer more detailed thoughts for each of those claims? Or is making a list and saying “80% for each” the kind of detailed breakdown you mean?

    Of course, the more detail you offer, the more chance you give people to notice and point out disagreements with your reasoning. A prophet/guru wouldn’t want that. Though a scientist might.

  • http://deconstructingzaniness.blogspot.com Pavan Ramkumar

    1. If ‘sims’ become successful, there doesn’t seem to be a historical (i.e. non-emotional) need to resurrect anyone. So chances are, dph’s argument about exceptional characters being most likely to be resurrected falls flat.

    2. Would ‘em’s be socially productive entities? If not, what percentage of tax money would go into supporting these ‘em’s? What if a resurrected ‘em’ committed suicide? How would the law in 2090 address that?

  • Anonymous

    Robin: If you make 50K$/yr now, and value life-years at twice your income, and discount future years at 2%
    I think for most humans utility calculations and time discounting don’t work that way (just stating the matter of affairs, not arguing whether or not one should try to override them with econ.analysis).

    sternhammer: Or a truck hits the power lines that feed the tanks.
    2c. Power outages. Lots of power outages don’t lead to loss of buildings, but might thaw a head. How long can a tank stay cold without electricity?

    They don’t liquefy nitrogen on-site, rather buy it from a commercial provider to refill every few weeks. According to Alcor a tank left unattended will boil off in 90 days or more.

    peco: 0.02%
    Please name another immortality wager that you estimate has these or better chances of success, costs as much or less and is not mutually exclusive with cryonics so that I can also examine if I should put it on my to-buy list.

  • http://profile.typepad.com/robinhanson Robin Hanson

    Peter, I delayed responding to you until I could review that chapter again. It found that people asked to directly estimate aggregate events did better than people asked to estimate component events, even when they were forced to normalize their estimates. A striking finding, but just not the same as estimating a event by itself versus via component conditional probabilities.

  • sternhammer

    Anonymous,

    Thanks, I didn’t know that. That’s helpful.

    So I would update my estimate of surviving issue #2 from .1296 to .162.

    SInce I have a hard time estimating 3-8 I think I will follow Peco for those, since he seems to have independently estimated each one. That gets me to .00013, without considering the risk of civilization collapsing or the abstract “what is existence?” ones in 9 & 10.

  • jgm

    Robin, even if we thought that the probability of successful cryopreservation were 100%, it seems unlikely to be cost-effective. If you are right about emulations, the future doesn’t need us: it will be cheaper to add additional life-years to the population without thawing us. Compared to cryonics, investments in existential risk reduction and even in global health would seem to have a lower cost per life-year saved/created. Do you believe that cryonics is the best use of your money and, if so, why?

  • afu

    The likelihood of an anoxic brain injury is incredibly high at death. This makes me doubt if the probability of reconstructing the same consciousness from a frozen rises above 0. There is way to much noise in the system to make even make it worthwhile to try.

  • http://amckenz.googlepages.com Andy McKenzie

    afu: Agreed, but surely there is some potential way to reverse these effects using a computer algorithm? Look for the connections that have been altered in these sorts of situations in the past and make the necessary adjustments before turning the upload “on”.

    As far as I’m concerned, #4 and #6 are the bottlenecking steps. What kind of resolution is necessary? That was one of the main questions considered in the “roadmap”: http://www.fhi.ox.ac.uk/Reports/2008-3.pdf, and they obviously couldn’t answer it adequately.

  • http://profile.typepad.com/robinhanson Robin Hanson

    Tyler posts that this comment thread is excellent. :)

  • SUR

    Someone might consider undertaking a more rigorous, better conditioned analysis of cryonics’ NPV. Introducing and then taking the expectation on stochastic processes, while perhaps also permitting modest risk aversion that biases the NPV value upward taken as utility. (Two NPVs respectively.)

    Something relatively short, and not so highly conditioned such that its’ results remain minimally meaningful, supposing this is achievable with given information and time horizons.

    Finally, this CBA-esque calculation ought to be motivated so that a reader can vaguely appreciate the author’s own methods and biases.

    Noise may overwhelm signal as others’ have claimed, but at the margins we may be better informed by the exercise.

  • Zac

    Robin’s 10 conditional probability items make sense to me. I just think he is far too optimistic about some of the probabilities. I say:

    1. 99%, 2. 80%, 3. 95%, 4. .1%, 5. 5%, 6. 5%, 7. 10%, 8. 80%, 9. 90%, 10. 100%, Total: .0000135432%, around one-in-one-billion, which seems about right to me

    “Brain science has workable input/output models of relevant brain cell types.” In the words of Tyler, I’m just not there. And “Usual freezing quality preserved relevant model-needed details” seems pretty outlandish to me, too, but I say 5% to be deferential.

    @jgm: You are basically saying that conjecture 3 has a low probability, but I think it is one of the most likely of the claims. If revival becomes possible, and the cost of revival is low, I can think of many reasons why people would be revived.

  • http://mark.reid.name Mark Reid

    Robin’s breakdown of the probability of being usefully revived after cryonics seems a lot like the type of reasoning used in the Drake equation to estimate the number of extraterrestrial civilisations that we may encounter.

    Does anyone here (Robin?) find the cryonics breakdown plausible but not the Drake equation?

  • Luke

    Morally speaking, if you delay the advancement or adoption of a lifesaving technology, you are killing people. If you actually know about it, you are killing people deliberately.

    So if you think cryonics is possible potentially, but we aren’t there yet, you still have a moral obligation on grounds of common decency to support rapid research and adoption of cryonics.

    So don’t give me that “I’m a skeptic” garbage. This is and always has been about saving lives, not just our own but everybody’s.

  • Pingback: Sequence Thinking vs. Cluster Thinking | The GiveWell Blog