= 27d567bc2d48d9f40e9ccc20ea370b15)
Eliezer_Yudkowsky comments on Looking for opinions of people like Nick Bostrom or Anders Sandberg on current cryo techniques - Less Wrong
You are viewing a comment permalink. View the original post to see all comments and the full post content.
= 783df68a0f980790206b9ea87794c5b6)
Loading…= b715d02e85f7b3b4ae65ca3d3e450868)
Subscribe to RSS Feed
Powered by Reddit
= f037147d6e6c911a85753b9abdedda8d)
Comments (184)
To answer this question we should talk to lots of neuroscientists. Here's one:
There was also a discussion with another neuroscientist, kalla724, here a year ago.
Do people know of other places where a neuroscientist who knows about vitrification gives their opinion?
"When all you need to see it's bullshit is an undergraduate-level understanding of biology" is an extremely clear cue that the speaker does not understand the current state of the cryonics debate and cannot be trusted to summarize it. Anyone who does not specifically signal that they understand the concept of information-theoretic death as mapping many cognitive-identity-distinct initial physical states to the same atomic-level physical end state is not someone whose summaries or final judgment you can possibly reasonably trust. If they cite sources you might look at those sources, but you're going to have to figure out on your own what it means.
Can you please elaborate on how and why sufficient understanding of the concept of information-theoretic death as mapping many cognitive-identity-distinct initial physical states to the same atomic-level physical end state helps to alleviate concerns raised by the author?
The basic idea of getting cryonics is that it offers a chance of massively extended lifespan, because there is a chance that it preserves one's identity. That's the first-run approximation, with additional considerations arising from making this reasoning a bit more rigorous, e.g. that cryonics is competitive against other interventions, that the chance is not metaphysically tiny, etc.
One thing we might make more rigorous is what we mean by 'preservation'. Well, preservation refers to reliably being able to retrieve the person from the hopefully-preserved state, which requires that the hopefully-preserved state cannot have arisen from many non-matching states undergoing the process.
The process that squares positive numbers preserves perfectly (is an injection), because you can always in theory tell me the original number if I give you its square. The process that squares real numbers preserves imperfectly but respectably since, for any positive output, that output could have come from two numbers (e.g. 1^2=1=(-1)^2). Moreover, if we only cared about the magnitude (modulus, i.e. ignoring the sign) of the input, even squaring over real numbers would perfectly preserve what we cared about.
Similarly, there is a chance that the hopefully-preserved states generated by cryonics do/will be generated only by the original identity, or possibly some acceptably close identities. That we do not currently know if it is possible to retrieve acceptably close identities from hopefully-preserved states—or even if we did, how one would do so—does not necessarily make the probability that it is possible to do so in principle low enough that cryonics can be laughed off.
A monkey might be bamboozled by the sequence of square numbers written in Arabic numerals, but that would not prove that the rule could not be deduced in principle, or that information had been lost for human purposes. Similarly we might currently be unable to reverse vitrification or look under a microscope and retrieve the identity, but it is unfair to demand this level of proof, and it is annoying and frustrating in the same way as logical rudeness (even if technically it is not logically rude) when every few months another person smugly spouts this type of argument as a 'refutation' of cryonics and writes cryonicists off, and then gets upvoted handsomely. (Hence Eliezer losing patience and outright declaring that people who don't seem to (effectively) understand this point about mappings don't have a clue.)
Formalisations of these concepts arise in more obviously mathematical contexts like the study of functions and information theory, but it feels like neither of those should be necessary background for a smart person to understand the basic idea. But in all honesty, I think the inferential gap for someone who has not explicitly considered at least the idea of injections before is big enough that often people apply the absurdity heuristic or become scared to do something unconventional before the time it takes to cross that inferential gap.
I think there's a good chance that there are neurodegenerative conditions that are currently irreversible but which many more would think worth working on than cryonics, simply because they associate cryonics with 'computer nerd failure mode' or apply the absurdity heuristic or because attacking neurodegenrative conditions is Endorsed by Experts whereas cryonics is not or because RationalWiki will laugh at them. Possible partial explanation: social anxiety that mockery will ensure for trying something not explicitly endorsed by an Expert consensus (which is a realistic fear, given how many people basically laugh at cryonicists or superficially write it off as 'bullshit'). And yes, in this mad world, social anxiety really might be the decisive factor for actual humans in whether to pursue an intervention that could possibly grant them orders of magnitude more lifespan.
It seems from the comment that he does understand information theoretic views of death- after all he is talking about preserving the information via thin slices and EM scanning.
To get good vitrification results, enough to preserve the synaptic structure, his claim is that you have to take mm thin slices of brain and preserve them at nearly 2000 atmospheres of pressure. This is obviously not what current cryonics procedures do. They shoot you full of cryoprotectants (which destroys some chemical information in the brain at the outset), and then freeze the brain whole (which will lead to a lot of fracturing).
To attack his point, I think you'd need to claim that either: 1. he is technically wrong and current vitrification does decently preserve synaptic structures. 2. mechanically scrambling synaptic structures via cracking is a 1 to 1 process that preserves information.
(2) is the interesting claim, though I'd hardly trust his word on (1) since I didn't see any especially alarming paragraphs in the actual paper referenced. I'm not an expert on that level of neurobiology (Anders Sandberg is, and he's signed up for cryonics), but I am not interested in hearing from anyone who has not demonstrated that they understand that we are talking about doing intelligent cryptography to a vitrified brain and potentially molecule-by-molecule analysis and reasoning, rather than, "Much cold. Very damage. Boo."
Unless someone spells out exactly what is supposed to destroy all cues of a piece of info, by explaining why two cognitively distinct start states end up looking like molecularly identical endstates up to thermal noise, so that we can directly evaluate the technical evidence for ourselves, all they're asking us to do is trust their authoritative summary of their intuitions; and you'd be just plain dumb to trust the authoritative summary of someone who didn't understand the original argument.
I'm trying not to be impatient here, but when I actually went to look at the cited paper and it said nothing at all about damage, it turned out this eminent authority's original argument consisted merely of, "To read off great synaptic info with current big clumsy microscopes and primitive imaging processing, we need big pressures. Look at this paper involving excellent info and big pressures. Cryonicists don't have big pressures. Therefore you're dead QED."
So say something about it. Your whole comment is an attack on 1, but regardless of his word on whether or not thing slice vitrification is currently the best we can do, we KNOW fracturing happens with current brain preservation techniques. Liquid nitrogen is well below the glass transition, so fracturing is unavoidable.
Why should we expect fracturing/cracking to be 1 to 1?
If you're worried about the effects of cracking, you can pay for ITS. LN2 is only used because it is cheap and relatively low-tech to maintain.
If you ask me it's a silly concern if we're assuming nanorepair or uploading. Cracking is just a surface discontinuity, and it forms at a point in time where the tissue is already in a glassy state where there can't be much mixing of molecules. The microcracks that form in frozen tissue is a much greater concern (but not the only concern with freezing). The fact that vitrified tissue forms large, loud cracks is related to the fact that it does such a good job holding things in place.
What's "ITS"? (Google 'only' hits for "it's") How much more expensive is it? Is it offer by Alcor and CI?
Short for Intermediate Temperature Storage.
Oh ok. Thank you.
I mean, it is either his authoritative summary or yours, and with all due honesty that guy actually takes care to construct an actual argument instead of resorting to appeals to authority and ridicule.
Personally I would be more interested in someone explaining exactly how cues of a piece of info are going to be reassembled and whole brain is going to be reconstructed from a partial data.
Proving that cryo-preservation + restoration does indeed work, and also showing the exact method as to how, seems like a more persuasive way to construct an argument rather that proving that your opponents failed to show that what you are claiming is currently impossible.
If cryonics providers don't have a proper way of preserving your brain state (even if they can repair partial damage by guessing), than I am sorry to say, but you are indeed dead.
I suspect the basic underlying premise that causes this difference is that you believe that strongly superhuman AI will exist at the time of unfreezing, whereas most people either disbelieve that thesis or do not take it into account.
I don't know why you'd say he gets it - he explicitly talks about re-animation. If he addresses the actual issue it would appear to be by accident.