5 ms·
I am glad progress is being made here, but until we can avoid the destruction that occurs in the last 24 hours of expected death (eg cancer), or the damage that
by danieltillett 11y ago
I am glad progress is being made here, but until we can avoid the destruction that occurs in the last 24 hours of expected death (eg cancer), or the damage that occurs with unexpected death (eg heart attack or trauma) from sitting at room temperature for hours after death, then all we are going to be preserving is grey mush.
- ianpurton 11y agoIf people learn to take regular backups this wouldn't be a problem.
- danieltillett 11y agoIf only this were possible :(
- crimsonalucard 11y agoWill those backups be you? What if one of those backups activated on accident? Would you kill it? Or should it kill you?
- SCHiM 11y agoDepending on your outlook on life, if the backup is you, don't you already know the answer?
- imaginenore 11y agoYou're assuming this "grey mush" can't be recovered. We don't actually know that. A sufficiently advanced AI should be able to recover a person from way less.
- danieltillett 11y agoThe problem is the information is encoded in the arrangements of the neural connections. Once your brain has reached the grey mush stage all those connection are lost.
- htns 11y agoAren't neurons relatively sizable? I don't see why they would totally "decay" in a short time frame, but maybe it's because I'm ignorant.
- davidgerard 11y agoYou're assuming the phrase "a sufficiently advanced AI" answers anything at all. Presumably you're assuming if the information is there at all - if the necessary data hasn't been scrambled beyond the noise floor of the scrambling process - then there's something for magic (because you're really talking about magic here) to work with. So, please (a) set out your claim with precision (b) back up your claim. * What is the information you need to recover? * To what degree is it scrambled? * What of it is scrambled below the noise floor of the process? * How do you know all this? (wrong answer: "here's a LessWrong/Alcor page." right answer: "here's something from a relevant neuroscientist.") For comparison: even a nigh-magical superintelligent AI can't recover an ice sculpture from the bucket of water it's melted into. It is in fact possible to just lose information. So, since you're making this claim, I'd like you to quantify just what you think the damage actually is.
- atrus 11y agoSure it can. Look at a photograph, make a mould and find a freezer :P
- M8 11y agoIf it's a godlike AI it can simulate humankind in reverse and resurrect everyone's consciousness before each person's death.
- eli_gottlieb 11y agoThat is absolute bunk that defies the basic principles of computational complexity and information theory in almost every possible way.
- SCHiM 11y agoI'm quite sure I've read somewhere that information cannot be lost in the absolute sense, lost to us: yes, lost irrevocably and irrefutably: no. In that sense, 'a sufficiently advanced AI' is not magic, because when people say that they definitively have something in mind, at least the people I often discuss this with do. In short: if you're smart(fast, precise, determined) enough to look at the individual molecules of a puddle of brain-goo. And if you can infer the way it has collapsed by ray tracing those molecules back through how they collided with each other/the walls of your mold then it should be possible to reconstruct the spatial form of the brains at least. That's a pretty big IF obviously, but equally obviously not impossible. If only you can look deep/far/fast enough. If you want to be theoretical about it, then yes. There is probably an upper bound on how smart/big an AI mind can possibly be. And thus there is a limit on how much information it can extract from arbitrary systems. So I agree with your assertion that there is information that even the smartest of all AIs cannot possibly reconstruct, but I'm not sure that the brain is such a structure. Any justification about why/how 'a sufficiently advanced AI' could come about is more questionable. Many knowledgeable people are making guesses based on our current understanding of intelligence/computation/AI, and then extrapolating. The paradoxical thing is that on the one hand AI-doomsday speakers tell us no to anthropomorphise (for good reasons) with the motives of an AI, but on the other hand apply human reasoning/understanding to predict such machines/patterns.
- M8 11y agoSuch AI will only be able to recover a gradient set of possible personalities depending on the percentage of missing data. Something like 1000000 of slightly different permutations of you.
- Udo 11y agoThere has to be enough information left over to reconstruct a person in "high enough fidelity" (whatever we may decide that means in the future). A brain just sitting there for hours after death at room temperature isn't idea - however, there is some good news in the area. It turns out that the most destruction happening to a brain after an ischemic episode is actually due to a cascade triggered by eventual re-perfusion. Since the dead brain is never re-perfused, this cascade is never triggered. Cellular decay after death makes a biological re-animation infeasible, but speaking as someone who did prepare a lot of neuro slides at uni, it takes more than a few hours for the structure itself to decay heavily, so we should be good for a scan/upload scenario. Pertaining to the article, the first hours after death are nowhere near as problematic to the brain's information content as the plastination procedure they're using! Brain trauma before death is another matter. Since we don't have the capability to create backups or checkpoints of our neuronal structure, what's physically destroyed is simply lost beyond recovery. However, for example in aggressive brain cancers, a functional copy of the person might still be recovered in principle even if their neocortex was severely compromised, as long as the actual data is still there.
- jimrandomh 11y agoI agree. But I also think this work (and future work like it) will help with that; having convincing proof that a preservation process works will make it easier to get obstacles to prompt preservation out of the way.