There's some small amount of writing about limited lifespans and evolution, particularly about differences in lifespans between species. I don't know how much is reasonably based on research (probably modeling or similar) and how much is "just so" stories.
What I gather from this is that:
- if the death rate from causes other than aging is high, evolution favours "live fast, spawn copiously, and die young". There's no point evolving longevity if odds are you'll die young of predation, accident, disease, etc. etc.
- if the non-aging death rate is lower, evolution moves towards a longer lifespan, and often a later start to producing offspring. Maybe fewer offspring too, each with more parental investment. You can see this with (some) pairs of closely related species in environments with large differences in non-aging mortality.
- if an older individual has big reproductive advantages, such that spawning early nets you a small clutch, with each year's potential clutch increasing, you get the big mama fish phenomenon, and negligible aging - each year the fish survives, it's bigger, and can produce more offspring than the year before. Evolution strongly favours individuals of such a species not aging.
- as far as I know, 2 species have evolved menopause - females stop producing new offspring long before aging would be likely to kill them. I'm not satisfied that we fully understand this, except for the obvious - already-produced offspring must do better under this system, such that the menopausal female has more (great)grandchildren on average than a non-menopausal but otherwise similar female. Humans are one of these species, so this probably matters to understanding human aging.
One other point: humans have an enviably low non-aging death rate, at least in "advanced" countries. This ought to be creating evolutionary pressures towards longer breeding lifetimes. Except of course that most of us stop breeding well before out bodies can't do it, either for reasons of investment in existing children or simply because evolution hasn't made us into instinct-driven breeding robots - we have goals separate from yet more babies. (Some of us even work hard to force competitors to have babies they don't want!)
Michael Rose, an evolutionary biologist and gerontologist at UC Irvine has written much about evolution and aging. He also quadrupled the life span of fruit files by selective breeding over a few generations. Alas, that approach isn't going to help you or I. Michael insists that many current anti-aging researchers have a false paradigm but he does think we can make progress against aging.
My theory (i.e. guess) is that this is the most likely way that progress will be made against human aging. I.e. those of us who are alive now will get at best merely marginal improvements in total life span, or total disability-free life span. (Improving "life expectancy" is much easier - just prevent or delay more non-aging deaths.)
But later generations might be born, on average, with longer natural lifespans. And this doesn't even need conscious selective breeding.
Of course I could be wrong. This is not my field. But most of the potential advances I hear touted seem to me like people desperate for hope grasping at the flimsiest of straws.
Meanwhile, thanks for the pointer. I'm adding Michael Rose to my list of authors to look at.
Given that very old people don't have children, what do you see as the selective pressure that would lengthen lifespans? The only one I can see is that slower aging means you can have children a little longer and slowing aging should result in raising the maximum age.
You've put your finger right on the problem; there's no obvious selective pressure, and that means that any such pressures that exist can be presumed to be very small.
My gut feel is that medical interventions in late adulthood can't do much to increase lifespan - they'd have to be applied much earlier, perhaps in utero. That's probably a bit more plausible than getting significant increased lifespan by natural evolution.
Selective breeding would probably be a lot better - but are there enough people who desire long-lived descendants enough to pick their sperm/egg donors/spouses based on their ancestors' lifespans? I don't really think so.
However, if I were looking for such a mechanism, I'd start with the same logic as with the evolution of menopause. *If* having more living ancestors is good for descendants' fitness in a big enough way, those still-living ancestors don't need to continue producing children themselves.
Just-so story coming up: a great-grandparent who doesn't yet need assistance, and is in fact still at the peak of their abilities has been accumulating earnings for quite a while, and is still doing so. If they'd died 20 years earlier, their heirs would have got what was left after medical expenses, but instead they are making still more money - and giving a lot of it to their descendants. Those descendants can afford to marry younger, raise more kids younger, and provision them better, than if they'd merely received an inheritance. Net result: more great grandchildren, and more prosperous (= fit) ones besides.
Meanwhile, another less long-lived family is already assisting their failing parents or grandparents, and this is taking both energy and money away from their own children.
I'd use that just so story in a novel; I think it's plausible enough for folks to suspend disbelief. But I can't really believe it.
For people who would like to interact in real time, I host an online meetup on Saturday mornings as well as a very occasional realspace meetup, originally for readers of the blog Slate Star Codex. Details:
The whole idea of uploading consciousness to computers is pure techno-mysticism. It fundamentally misunderstands the concept at the most basic level. Consciousness is not a "thing" that can be transferred from platform to platform. It is a physiologic process dependent on the moment to moment working of your brain. “Uploading it” to a computer might result in a simulation of some sort, but it isn't "you", in the sense anyone cares about. Kind of like how a computer simulation of a hurricane doesn’t get anyone wet.
I’m no expert on this area but it seems a giant leap to assume an “uploaded” version of my self would be the same as my self. Even the best sci-fi that I’ve read on this topic has heavy conjecture (faith) in the story telling.
You do realize that your hurricane simulation point is decades old and has been argued to death? (I studied the argument in 1982 in reading Godel, Escher, Bach.) There are many smart people who disagree with you who cannot reasonably be called mystics. They don't believe that consciousness is a "thing". They think that it may be possible to recreate conscious experience on a different substrate. Do you need to simulate just the connectome? The details of synapses and neurotransmitters? The details of all the organelles? The atoms? The quarks?
I would agree that just *any* simulation of a mind is not necessarily going to be conscious. Just as LLMs often produce output that looks like an intelligent, self-aware mind but is not. But a non-biological instantiation of cognitive processes *might* be capable of the same emergent properties. It's problematic to use the term "simulation" because what I'm talking about is a recreation through a different physical substrate, not just something that is engineered to produce the same outputs. I don't know of anyone who thinks "uploading" is potentially possible who is a behaviorist.
To return to the example that David raised from Minsky and/or Moravec: Suppose we can swap out your neurons for synthetic components one-by-one or in small groups. (This is not SF; it is actual research going back at least to the early 1990s -- see Ted Berger's work on synthetic neurons to replace lost motor function in brain-damaged patients.) If you could detect no subjective difference as the replacement proceeded, you have an empirical refutation of the view that consciousness necessarily requires biology.
Of course, with a non-biological substrate, it will probably be easier to modify it to change the way cognition works, but that's a distinct line of thought.
You can think this through logically, as well. Whatever strengths or limitations we experience in our thinking would be a matter of our brain's physical abilities. If we find it easy or hard to remember a past event, or figure out a math question, or ponder philosophy, those are all reflections of our brains - not consciousness. Living as/inside a computer will fundamentally change how even our thoughts work. Even assuming we could transfer consciousness, our fundamental experiences would be vastly different inside our thoughts, let alone from lack of physical senses and a physical form.
I agree that if I was uploaded to a computer, or to a robot very different from a human, I would be having different experiences which would change me. But it is equally true that if I moved to a very different country or started associating with a very different set of people that would change me.
I don't think that is relevant to the question of whether, immediately after upload, I would be the person I was before.
There's always this assumption that an uploaded mind will have no body. I find that puzzling. It seems quite possible that (especially while our minds remains structured much like they were when biologically-instantiated) we will need a body or bodies to make sense of the world. Our sense of body may differ -- why not have multiple bodies for different purposes? Perhaps frequently-used sensors would be felt as part of your extended body?
As the non-biological platform enabled us to change our cognitive and sensory architecture, our experiences would change. We would likely have many more senses that operate in a wider spectrum than the few that humans have. We may also have far greater emotional self-understanding due to new pathways from the cognitive to the emotional centers. The latter weren't very useful for most of our evolution; in fact they were dangerous -- if you took time to ponder your feeling of fear instead of just reacting and running, you would be another creature's meal.
I suspect that without the feedback indicating a physical body, any kind of sense of selfhood would be impossible. Our spatially discrete embodiment is not an incidental feature.
I'm sure some kinds of feedback like that are possible, but I think the fundamental problem is deeper. We can get confused when we talk about "consciousness". That can encompass a large variety of physiologic processes. It is the moment to moment workings of those processes that are what we call consciousness: it is the wetware, and the wetware needs to be moving in a particular way. If the wetware stops, the show stops. I suspect consciousness is not fundamentally a matter of information, nor is it algorithmic, but rather is physical and substrate-dependent.
So, just as a human being is a machine made up of trillions of smaller semi-autonomous machines, each performing a billion-ish reactions per second, which is how we get the often stupefying complexity of the human mind; if we designed non-organic machines with analogous complexity, perhaps we would conclude that those entities have a form of consciousness, but it would be of a fundamentally different character than ours, since we're made of different stuff.
You mention being afraid of death. Although I have cryonics arrangements and hope for aging to be conquered in my time, I do not fear death. I wonder if we have different views or if you are using "fear" differently. I don't fear death because death is a state of nothingness. I won't be bored, or scared, or frustrated. I won't be. There is literally nothing to be afraid of. (Was it Epictetus who first said that?) What I fear is two things: 1. The dying process. 2. Not being alive to enjoy learning, growing, exploring, developing, loving, and all those good things. You could say I fear those things being taken away but that seems different from fearing death itself.
P.S. Congratulations on the 50th anniversary of The Machinery of Freedom. I read it in July 1982 -- one of the first three books that stirred my brain with libertarian ideas.
Sure, but "afraid of dying" is used to mean both the dying process and being dead. In context, he might easily mean "being dead" since all the rest of the post is about avoiding being dead. But only David can say what he meant...
Keep in mind that lots of us greatly enjoy reading your work. Dying will end that!
If you should decide to get cryopreserved, I will make this commitment: If I am repaired and revived before you, I will (if reasonably possible and if you don't reject the commitment) be there to help you get restarted. Such "life pacts" as they have been called, help address one major objection to cryonics: "I won't know anyone."
If I get preserved and am revived in less than fifty years I will have living grandchildren, very likely living children. What is your guess about how long the necessary medical progress will take?
Also, unless society has changed quite a lot — entirely possible, given technological changes such as AI — I will probably still have marketable skills, very likely even fans of my writing.
Two hundred years later would be a very different situation.
I can make a guess but it's not worth much. First, it seems likely that cryonauts will be revived at different dates, depending on (a) when they were cryopreserved and the level of technology at the time, and (b) depending on the conditions under which they were cryopreserved.
But let's say we're talking about someone cryopreserved in the next ten years or so and under good conditions -- little delay between clinical death and the start of procedures. Usually I will guess at between 50 years and 150 years. Less than 50 years seems highly implausible -- but that could change if we get much more capable AI that accelerates biomedical knowledge and capabilities. I don't think LLM's are going to do that. Part of my minimum timeframe is due to the need to solve the problem of aging before we revive anyone. (With the possible exception of young people who died of something now curable.)
It seems unlikely to me that it would be more than 100 to 150 years, but with so many unknowns and unknown unknowns I feel uncomfortable even guessing. I know that is an unsatisfying answer but at least it's honest!
>What about the possibility that I am wrong, some religion right, about the nature of consciousness? The modern scientific world view, which I generally share, rejects it. But quite a lot of the people who have ever lived, including quite a lot now living, many of them educated people in modern developed societies, accept it.
This seems like a sub-optimal heuristic. Imagine applying it to to e.g. vaccines. "A lot of people now living, many of them educated people in modern developed societies accept anti-vax arguments."
Okay. Let's assume 10% of such people are anti-vaxxers. Should we assign a 10% probability to the anti-vax proposition?
Maybe if that's all the information we had, but crucially, it isn't! We can look at the evidence for vaccine efficacy ourselves. More meaningful that simply polling people on *what* they believe, we can see *why* they tend to believe things. What are the main anti-vax arguments, and how well do they fit the data? What are the main arguments in favor of vaccines, and how well do they fit the data?
Once that has been evaluated, then the raw number of people who believe the respective conclusions, or even the respective arguments, should be of relatively little value.
Similarly, if the only information we had access to was a poll on the number of percentage of "educated people in modern developed societies" who believe in a religious proposition, then perhaps using that would be our best heuristic. But again, that isn't the situation that faces us. We can hear people's arguments for the proposition and evaluate those.
There are other reasons, too, why the measure of what "quite a lot of the people who have ever lived, including quite a lot now living, many of them educated people in modern developed societies, accept" may not be the most meaningful, which I noted here: https://daviddfriedman.substack.com/p/the-puzzle-of-consciousness/comment/13929720.
The problem with your argument is that you are greatly overestimating the confidence I am entitled to have in my conclusions. Most of our beliefs depend at least in large part on second hand information. I have observed at least four situations in my lifetime where the accepted belief, the orthodoxy among educated people, was, in my present judgement, mistaken. That makes me reluctant to treat my conclusions, ultimately based on a different orthodoxy, as known with certainty. The observation that other intellectually competent people disagree should make me consider that they might be right.
Indeed, the interesting puzzle, which I think Robin Hanson has discussed, is why I should view my opinion as more likely to be true that the opinion of someone else as intelligent as I am and with access to the same information, why I should be of my own opinion.
I observe a world where, in a variety of contexts, most obviously the political, a large fraction of the population is confident that their view is true, even obvious, and a similar fraction similarly confident of a different view.
In this case, these arguments are reinforced by my inability to use the scientific view to explain the nature of consciousness, a point I discussed in an earlier post.
Considering how large the population of people in the world is that consider a religious afterlife to be accurate, it's clear that David has already modulated his estimate down extensively to get to 10% chance.
Also, considering that we are looking at something with no experimental observations (dead people with access to the afterlife no longer being able to share their results - whether they are simply dead or in the afterlife), I'm not sure how you would propose that we make that determination scientifically? People clearly do that kind of thing all the time, but it's not scientific - by definition since we cannot know the results of any kind of controlled experiment. Anecdotes from people claiming to have died and experienced some part of the afterlife are pretty common, but these are about as far from controlled experiments as you can get. Similarly for people who have died and been brought back that had no such experience.
To repost a former comment of mine to the linked piece regarding consciousness and an immortal soul:
>It seems likely that the mind depends on the brain and dies when the body dies. The effect on the mind of things happening to the brain short of dying — alcohol or other drugs in the bloodstream or a blow to the skull — supports that conjecture. I know of no positive evidence against it. But as long as the nature of consciousness is a mystery to me, I cannot be certain.
More trivially, one could say that as long as anything remains unknown in the universe, that we cannot be certain of anything, since that unknown thing in the universe might affect anything. But this is trivially true about anything and therefore does not reflect any actual uncertainty with a given proposition.
Regarding the specific proposition at hand, I don't see why the existence of questions about consciousness in general should be a reason to consider the proposition that consciousness is not dependent on physical hardware and processes, given the stated evidence that it is.
This seems equivalent to saying "It seems likely that Tylenol depends on physically interacting with the body to work, so one person taking a Tylenol could not cure a different person's headache. I know of no positive evidence against it. But as long as the mechanism of action of Tylenol is a mystery to me, I cannot be certain." [Note that this is about *my* knowledge of the MOA of Tylenol - it is not dependent on someone else knowing the MOA.]
Am I missing some reason why the point about consciousness is different?
Yes. It isn't just that I don't know how to explain consciousness, it's that I have no idea how to explain it, how to fit it into my general world view. Nor, so far as I can tell, does anyone else. The behavior of Tylenol is probably understood by someone, and I understand enough about it to fit it into my world view.
I haven't given tis real thought, since at 76 I'm pretty unlikely to get a chance to get much of an extended life span, but among my concerns are that I am not only spending on my needs and wants, but trying to continue to accumulate resources to pass on to my descendants. I have children that I love very much, and grandchildren and great-grandchildren that I don't "know" well enough to actually love them, but whom I care about a great deal.
So, how much of my current resources should I spend in an effort to live longer is a tough question. No doubt I should continue to try to maintain my current pretty good health (for more reasons than just living longer), but what is worth spending additional resources on, as compared to being able to bequeath them to my descendants?
My father always joked about two things in his personal life. The first was that he had been living "on borrowed time" since he joined the Marine Corps in December of 1942. The second was that he wanted to "run out of breath and money at the same time." He only really meant the first. He strove mightily to leave behind enough money to assure my mother's situation for however long she lived. I don't think I'm on his kind of borrowed time, but my considerations for my wife and descendants are real.
I'm 78. On the other hand, I had a test for biological markers of age a decade or two ago and it came out about ten years younger than my chronological age. That's consistent with my general behavior and appearance — people in realspace commonly underestimate my age.
So I may have a little longer than my age would suggest.
The rest of your point I agree with, but at this point in life I'm well enough off so it isn't a serious issue for me.
I've had a pretty weird life, but my general health has always been very good. But it did mean I started building wealth pretty late. Still, I appreciate where I got to. I'd like to hang around a bit longer.
One way to think about it is to ask yourself: Will my younger relatives -- who are more likely to live to a time when aging has been beaten -- begrudge me using my resources to have a shot at living again? In many cases, they WILL, so the real question perhaps should be: *Should* they? Should they be more grudging than they would about you buying a second (or nicer) home or car or a bunch of expensive vacations?
Luckily, my family doesn't seem much like that. I'm reasonably sure they ones old enough to be rational are content to let me choose whichever path I choose.
I bet you a nickel consciousness is just electron flow and that preserving the meaningful structures you define as "you" plus a continuity of consciousness would just require cabling your brain to a computer and synchronizing computational processes. This is a dumb action bet and doesn't represent epistemic confidence. Make it $50.
I would certainly consider Minsky's process for "uploading" to be a means of survival. Ray Kurzweil offered the same scenario and I found it plausible, though I don't think we are going to have enough computational power to make it commonly available in my longest possible lifespan.
I wouldn't consider the more usual versions to be survival. At best I would consider them to be reproduction.
Here's what Daniel Dennett calls an "intuition pump":
Suppose that we have a technological means of scanning your entire body, and assembling a perfect duplicate. We do so, on the other side of a room, so the two can look at each other. Now we take out a shotgun and shoot you through the skull—by "you" I mean the original, the one that was scanned. Do you believe that, in that instant, you will suddenly be seeing through the eyes of the DDF on the other side of the room, watching in shock as the original DDF, now headless, collapses?
I don't see that it makes any difference if the DDF that was scanned is destroyed in the process of scanning, rather than afterward.
If DDF1 and DDF2 were telepathically linked, that might lead to a different conclusion. But I don't think there is convincing evidence of telepathy.
I know this isn’t what you were looking for David but I think it’s pretty unlikely you are going to be the first person to avoid death. So it might be worthwhile to try to relieve your fear instead. If you can do that your remaining time may be richer. Note: I’m a little younger than you - about seven years - and I’m not thrilled about my life ending either.
I take articles from MAPS with a large grain of salt but this seems more likely to be successful than cryogenics or uploading your consciousness to a machine.
A quibble. If you have three independent 10% chances of living, I don’t think the accumulated probability is 30%. You die if none of them come up, the probability of which is 0.9³ = 0.729. So the probability of living is 0.271 (27.1%). However, given the inaccuracy of the three initial estimates, you can laugh this off.
You are assuming the three things are independent. I can't both live another hundred years because aging has been solved and because I have been successfully suspended — if age reversal happens ten years from now I won't die of old age and be frozen. If either of those happens, the reason I still exist in fifty years won't be the immortality of the soul.
I should have made it clear that my probabilities are not about aging, suspension, or the soul in general but about each being the reason I am still around in more than fifty years.
I think it remains true that, for you to die of old age within a normal time, none of these things must happen, which is what my calculation was based on. However, admittedly I’m no expert on probability theory, so I shouldn’t press the point strongly. Perhaps we should also take into account the possibility that you may die of something other than old age—which could happen to any of us at any time, regardless of life-extension technology. Now there’s a cheerful thought…
There is another possibilty here (see Nick Bostrom's simulation arguements, they're persuasive) and that's the idea that we're already simulations living in a simulation.
In that world, when the body dies, the mind might already be backed up automatically, or not.
If yes, the question would be whether you get respawned or not, if yes, wow, if no, you'd never know.
George Hotz said, "the diffence between an NPC and a PC is that a PC knows it's a PC," so we might be just NPCs and death is death, done, finished, sorry, game over.
He gave an interesting talk at the Austin SBSW (I think) conference some years ago on "How to Hack" the simulation. Interesting, but no final conclusions.
I think we're in a simulation, but whether we'll find out, I suspect the odds are slim.
I don't think the simulation hypothesis solves the consciousness puzzle. I'm pretty sure that WoW NPC's are not conscious and have no idea how one could program ones that were.
Natural selection drove us to conscious through means we don't understand, but it's still just ones and zeroes tied to a physical processor (unless the revealed religions are correct and we have souls -- but that would seem argue even more forcefully that an NPC could be conscious, like a demon or an angel).
“the physical reality that we more or less understand.”
This phrase is making a questionable assertion here. Why believe we “more or less” understand physical reality, given that we don’t yet know the true laws of physics?
We understand it well enough to build airplanes that fly, nuclear reactors that produce power, lots of other things that depend on our understanding of physical reality. "More or less" because there is still much we don't understand.
The question of exactly how much we don’t understand matters a lot! Given that everything you experience is mediated through consciousness, you have more evidence of consciousness than anything else. Yet we don’t have an evidentiary model of consciousness or even any clue of where to start! It seems to me we _think_ we understand the laws of physics because of the incredible things we can do. But framing the situation as, we really don’t know, opens the door for a lot more room for possible optimism.
You didn't mention the "other" category. The obvious, most likely to occur and to succeed option is that AGI will invent a way of making us immortal. We can be agnostic to what exactly it will look like. The important thing is that AGI is built, and is aligned. In particular, if it is not aligned, any other option on your list (except for the soul argument) is meaningless.
Fair point. I don't see the large language models as implying that human level AI is just around the corner and superhuman not far behind, although some do. They are not thinking, they are mining a vast database of the results of human thinking, and there isn't a body of superhuman thinking for them to mine.
So my guess is that human level AI will be a while, superhuman still further off, hence that the latter is unlikely to show up in my "natural" lifetime. I include the possibility of superhuman AI destroying us, which I discussed briefly in _Future Imperfect_, as one of the reasons that cryonic suspension might not lead to revival.
It seems experience with feral children shows pretty conclusively that language is required for any advanced thought, i.e. ideas, plans, solutions, "what-ifs," and the like.
When you have an idea, pre- "putting into words" what does that look like if it's not based in a language?
Your mind might not have to say "economics" to know you're thinking of economics, but that's only because you've read about economics (or listened to people talk about it).
A feral child isn't thinking about economics "economics" although, sure, they might feel hunger and see a stick and conceptualize that they could use the stick to knock an apple down from the tree.
But without words or experience, how would they conceptualize a ladder?
Whereas with language, someone without experience or words for ladder, could be made to understand a ladder and it's use for getting apples.
An interesting aside is that AGI, if it doesn't extinct us, and instead serves us, could both emulate the brain inside a computer (virtual world) and figure how to extend life and/or replace the body (with silicon parts or just a clone or perhaps even a better body made from AI sequenced DNA to be whatever you want).
The question I wonder is what a brain emulation will be able to feel, if anything. Our brains feel emotions due to more than just processing information, but instead are effected by our horomones and other physical factors besides just information (thinking, thoughts, reasoning).
What would "living in a computer simulation" be like.....perhaps we already know as we're in one now. Perhaps not.
I’m with Woody Allen on this one. I’m not afraid of dying. I just don’t want to be there when it happens.
Good joke, but also kind of true. I’m not afraid of being dead but the actual process of going through that door seems like it could be unpleasant.
There's some small amount of writing about limited lifespans and evolution, particularly about differences in lifespans between species. I don't know how much is reasonably based on research (probably modeling or similar) and how much is "just so" stories.
What I gather from this is that:
- if the death rate from causes other than aging is high, evolution favours "live fast, spawn copiously, and die young". There's no point evolving longevity if odds are you'll die young of predation, accident, disease, etc. etc.
- if the non-aging death rate is lower, evolution moves towards a longer lifespan, and often a later start to producing offspring. Maybe fewer offspring too, each with more parental investment. You can see this with (some) pairs of closely related species in environments with large differences in non-aging mortality.
- if an older individual has big reproductive advantages, such that spawning early nets you a small clutch, with each year's potential clutch increasing, you get the big mama fish phenomenon, and negligible aging - each year the fish survives, it's bigger, and can produce more offspring than the year before. Evolution strongly favours individuals of such a species not aging.
- as far as I know, 2 species have evolved menopause - females stop producing new offspring long before aging would be likely to kill them. I'm not satisfied that we fully understand this, except for the obvious - already-produced offspring must do better under this system, such that the menopausal female has more (great)grandchildren on average than a non-menopausal but otherwise similar female. Humans are one of these species, so this probably matters to understanding human aging.
One other point: humans have an enviably low non-aging death rate, at least in "advanced" countries. This ought to be creating evolutionary pressures towards longer breeding lifetimes. Except of course that most of us stop breeding well before out bodies can't do it, either for reasons of investment in existing children or simply because evolution hasn't made us into instinct-driven breeding robots - we have goals separate from yet more babies. (Some of us even work hard to force competitors to have babies they don't want!)
Thanks. Interesting.
Michael Rose, an evolutionary biologist and gerontologist at UC Irvine has written much about evolution and aging. He also quadrupled the life span of fruit files by selective breeding over a few generations. Alas, that approach isn't going to help you or I. Michael insists that many current anti-aging researchers have a false paradigm but he does think we can make progress against aging.
My theory (i.e. guess) is that this is the most likely way that progress will be made against human aging. I.e. those of us who are alive now will get at best merely marginal improvements in total life span, or total disability-free life span. (Improving "life expectancy" is much easier - just prevent or delay more non-aging deaths.)
But later generations might be born, on average, with longer natural lifespans. And this doesn't even need conscious selective breeding.
Of course I could be wrong. This is not my field. But most of the potential advances I hear touted seem to me like people desperate for hope grasping at the flimsiest of straws.
Meanwhile, thanks for the pointer. I'm adding Michael Rose to my list of authors to look at.
Given that very old people don't have children, what do you see as the selective pressure that would lengthen lifespans? The only one I can see is that slower aging means you can have children a little longer and slowing aging should result in raising the maximum age.
You've put your finger right on the problem; there's no obvious selective pressure, and that means that any such pressures that exist can be presumed to be very small.
My gut feel is that medical interventions in late adulthood can't do much to increase lifespan - they'd have to be applied much earlier, perhaps in utero. That's probably a bit more plausible than getting significant increased lifespan by natural evolution.
Selective breeding would probably be a lot better - but are there enough people who desire long-lived descendants enough to pick their sperm/egg donors/spouses based on their ancestors' lifespans? I don't really think so.
However, if I were looking for such a mechanism, I'd start with the same logic as with the evolution of menopause. *If* having more living ancestors is good for descendants' fitness in a big enough way, those still-living ancestors don't need to continue producing children themselves.
Just-so story coming up: a great-grandparent who doesn't yet need assistance, and is in fact still at the peak of their abilities has been accumulating earnings for quite a while, and is still doing so. If they'd died 20 years earlier, their heirs would have got what was left after medical expenses, but instead they are making still more money - and giving a lot of it to their descendants. Those descendants can afford to marry younger, raise more kids younger, and provision them better, than if they'd merely received an inheritance. Net result: more great grandchildren, and more prosperous (= fit) ones besides.
Meanwhile, another less long-lived family is already assisting their failing parents or grandparents, and this is taking both energy and money away from their own children.
I'd use that just so story in a novel; I think it's plausible enough for folks to suspend disbelief. But I can't really believe it.
For people who would like to interact in real time, I host an online meetup on Saturday mornings as well as a very occasional realspace meetup, originally for readers of the blog Slate Star Codex. Details:
http://www.daviddfriedman.com/SSC%20Meetups%20announcement.html
The whole idea of uploading consciousness to computers is pure techno-mysticism. It fundamentally misunderstands the concept at the most basic level. Consciousness is not a "thing" that can be transferred from platform to platform. It is a physiologic process dependent on the moment to moment working of your brain. “Uploading it” to a computer might result in a simulation of some sort, but it isn't "you", in the sense anyone cares about. Kind of like how a computer simulation of a hurricane doesn’t get anyone wet.
I’m no expert on this area but it seems a giant leap to assume an “uploaded” version of my self would be the same as my self. Even the best sci-fi that I’ve read on this topic has heavy conjecture (faith) in the story telling.
You do realize that your hurricane simulation point is decades old and has been argued to death? (I studied the argument in 1982 in reading Godel, Escher, Bach.) There are many smart people who disagree with you who cannot reasonably be called mystics. They don't believe that consciousness is a "thing". They think that it may be possible to recreate conscious experience on a different substrate. Do you need to simulate just the connectome? The details of synapses and neurotransmitters? The details of all the organelles? The atoms? The quarks?
I would agree that just *any* simulation of a mind is not necessarily going to be conscious. Just as LLMs often produce output that looks like an intelligent, self-aware mind but is not. But a non-biological instantiation of cognitive processes *might* be capable of the same emergent properties. It's problematic to use the term "simulation" because what I'm talking about is a recreation through a different physical substrate, not just something that is engineered to produce the same outputs. I don't know of anyone who thinks "uploading" is potentially possible who is a behaviorist.
To return to the example that David raised from Minsky and/or Moravec: Suppose we can swap out your neurons for synthetic components one-by-one or in small groups. (This is not SF; it is actual research going back at least to the early 1990s -- see Ted Berger's work on synthetic neurons to replace lost motor function in brain-damaged patients.) If you could detect no subjective difference as the replacement proceeded, you have an empirical refutation of the view that consciousness necessarily requires biology.
Of course, with a non-biological substrate, it will probably be easier to modify it to change the way cognition works, but that's a distinct line of thought.
Yes I'm familiar with the arguments— no claims to originality here.
I agree it is ultimately an empirical question.
You can think this through logically, as well. Whatever strengths or limitations we experience in our thinking would be a matter of our brain's physical abilities. If we find it easy or hard to remember a past event, or figure out a math question, or ponder philosophy, those are all reflections of our brains - not consciousness. Living as/inside a computer will fundamentally change how even our thoughts work. Even assuming we could transfer consciousness, our fundamental experiences would be vastly different inside our thoughts, let alone from lack of physical senses and a physical form.
I agree that if I was uploaded to a computer, or to a robot very different from a human, I would be having different experiences which would change me. But it is equally true that if I moved to a very different country or started associating with a very different set of people that would change me.
I don't think that is relevant to the question of whether, immediately after upload, I would be the person I was before.
There's always this assumption that an uploaded mind will have no body. I find that puzzling. It seems quite possible that (especially while our minds remains structured much like they were when biologically-instantiated) we will need a body or bodies to make sense of the world. Our sense of body may differ -- why not have multiple bodies for different purposes? Perhaps frequently-used sensors would be felt as part of your extended body?
As the non-biological platform enabled us to change our cognitive and sensory architecture, our experiences would change. We would likely have many more senses that operate in a wider spectrum than the few that humans have. We may also have far greater emotional self-understanding due to new pathways from the cognitive to the emotional centers. The latter weren't very useful for most of our evolution; in fact they were dangerous -- if you took time to ponder your feeling of fear instead of just reacting and running, you would be another creature's meal.
I suspect that without the feedback indicating a physical body, any kind of sense of selfhood would be impossible. Our spatially discrete embodiment is not an incidental feature.
Why wouldn't the feedback to a computer with sensors and manipulators have the same effect?
I'm sure some kinds of feedback like that are possible, but I think the fundamental problem is deeper. We can get confused when we talk about "consciousness". That can encompass a large variety of physiologic processes. It is the moment to moment workings of those processes that are what we call consciousness: it is the wetware, and the wetware needs to be moving in a particular way. If the wetware stops, the show stops. I suspect consciousness is not fundamentally a matter of information, nor is it algorithmic, but rather is physical and substrate-dependent.
So, just as a human being is a machine made up of trillions of smaller semi-autonomous machines, each performing a billion-ish reactions per second, which is how we get the often stupefying complexity of the human mind; if we designed non-organic machines with analogous complexity, perhaps we would conclude that those entities have a form of consciousness, but it would be of a fundamentally different character than ours, since we're made of different stuff.
My 2 cents at least.
You mention being afraid of death. Although I have cryonics arrangements and hope for aging to be conquered in my time, I do not fear death. I wonder if we have different views or if you are using "fear" differently. I don't fear death because death is a state of nothingness. I won't be bored, or scared, or frustrated. I won't be. There is literally nothing to be afraid of. (Was it Epictetus who first said that?) What I fear is two things: 1. The dying process. 2. Not being alive to enjoy learning, growing, exploring, developing, loving, and all those good things. You could say I fear those things being taken away but that seems different from fearing death itself.
P.S. Congratulations on the 50th anniversary of The Machinery of Freedom. I read it in July 1982 -- one of the first three books that stirred my brain with libertarian ideas.
No he does not. He wrote, "I am afraid of dying" not that he is afraid of death. He does mention avoiding death, but there is a world of difference.
Sure, but "afraid of dying" is used to mean both the dying process and being dead. In context, he might easily mean "being dead" since all the rest of the post is about avoiding being dead. But only David can say what he meant...
I also like being alive, and dying ends that.
Keep in mind that lots of us greatly enjoy reading your work. Dying will end that!
If you should decide to get cryopreserved, I will make this commitment: If I am repaired and revived before you, I will (if reasonably possible and if you don't reject the commitment) be there to help you get restarted. Such "life pacts" as they have been called, help address one major objection to cryonics: "I won't know anyone."
Thanks.
If I get preserved and am revived in less than fifty years I will have living grandchildren, very likely living children. What is your guess about how long the necessary medical progress will take?
Also, unless society has changed quite a lot — entirely possible, given technological changes such as AI — I will probably still have marketable skills, very likely even fans of my writing.
Two hundred years later would be a very different situation.
I can make a guess but it's not worth much. First, it seems likely that cryonauts will be revived at different dates, depending on (a) when they were cryopreserved and the level of technology at the time, and (b) depending on the conditions under which they were cryopreserved.
But let's say we're talking about someone cryopreserved in the next ten years or so and under good conditions -- little delay between clinical death and the start of procedures. Usually I will guess at between 50 years and 150 years. Less than 50 years seems highly implausible -- but that could change if we get much more capable AI that accelerates biomedical knowledge and capabilities. I don't think LLM's are going to do that. Part of my minimum timeframe is due to the need to solve the problem of aging before we revive anyone. (With the possible exception of young people who died of something now curable.)
It seems unlikely to me that it would be more than 100 to 150 years, but with so many unknowns and unknown unknowns I feel uncomfortable even guessing. I know that is an unsatisfying answer but at least it's honest!
>What about the possibility that I am wrong, some religion right, about the nature of consciousness? The modern scientific world view, which I generally share, rejects it. But quite a lot of the people who have ever lived, including quite a lot now living, many of them educated people in modern developed societies, accept it.
This seems like a sub-optimal heuristic. Imagine applying it to to e.g. vaccines. "A lot of people now living, many of them educated people in modern developed societies accept anti-vax arguments."
Okay. Let's assume 10% of such people are anti-vaxxers. Should we assign a 10% probability to the anti-vax proposition?
Maybe if that's all the information we had, but crucially, it isn't! We can look at the evidence for vaccine efficacy ourselves. More meaningful that simply polling people on *what* they believe, we can see *why* they tend to believe things. What are the main anti-vax arguments, and how well do they fit the data? What are the main arguments in favor of vaccines, and how well do they fit the data?
Once that has been evaluated, then the raw number of people who believe the respective conclusions, or even the respective arguments, should be of relatively little value.
Similarly, if the only information we had access to was a poll on the number of percentage of "educated people in modern developed societies" who believe in a religious proposition, then perhaps using that would be our best heuristic. But again, that isn't the situation that faces us. We can hear people's arguments for the proposition and evaluate those.
There are other reasons, too, why the measure of what "quite a lot of the people who have ever lived, including quite a lot now living, many of them educated people in modern developed societies, accept" may not be the most meaningful, which I noted here: https://daviddfriedman.substack.com/p/the-puzzle-of-consciousness/comment/13929720.
The problem with your argument is that you are greatly overestimating the confidence I am entitled to have in my conclusions. Most of our beliefs depend at least in large part on second hand information. I have observed at least four situations in my lifetime where the accepted belief, the orthodoxy among educated people, was, in my present judgement, mistaken. That makes me reluctant to treat my conclusions, ultimately based on a different orthodoxy, as known with certainty. The observation that other intellectually competent people disagree should make me consider that they might be right.
Indeed, the interesting puzzle, which I think Robin Hanson has discussed, is why I should view my opinion as more likely to be true that the opinion of someone else as intelligent as I am and with access to the same information, why I should be of my own opinion.
I observe a world where, in a variety of contexts, most obviously the political, a large fraction of the population is confident that their view is true, even obvious, and a similar fraction similarly confident of a different view.
In this case, these arguments are reinforced by my inability to use the scientific view to explain the nature of consciousness, a point I discussed in an earlier post.
Considering how large the population of people in the world is that consider a religious afterlife to be accurate, it's clear that David has already modulated his estimate down extensively to get to 10% chance.
Also, considering that we are looking at something with no experimental observations (dead people with access to the afterlife no longer being able to share their results - whether they are simply dead or in the afterlife), I'm not sure how you would propose that we make that determination scientifically? People clearly do that kind of thing all the time, but it's not scientific - by definition since we cannot know the results of any kind of controlled experiment. Anecdotes from people claiming to have died and experienced some part of the afterlife are pretty common, but these are about as far from controlled experiments as you can get. Similarly for people who have died and been brought back that had no such experience.
To repost a former comment of mine to the linked piece regarding consciousness and an immortal soul:
>It seems likely that the mind depends on the brain and dies when the body dies. The effect on the mind of things happening to the brain short of dying — alcohol or other drugs in the bloodstream or a blow to the skull — supports that conjecture. I know of no positive evidence against it. But as long as the nature of consciousness is a mystery to me, I cannot be certain.
More trivially, one could say that as long as anything remains unknown in the universe, that we cannot be certain of anything, since that unknown thing in the universe might affect anything. But this is trivially true about anything and therefore does not reflect any actual uncertainty with a given proposition.
Regarding the specific proposition at hand, I don't see why the existence of questions about consciousness in general should be a reason to consider the proposition that consciousness is not dependent on physical hardware and processes, given the stated evidence that it is.
This seems equivalent to saying "It seems likely that Tylenol depends on physically interacting with the body to work, so one person taking a Tylenol could not cure a different person's headache. I know of no positive evidence against it. But as long as the mechanism of action of Tylenol is a mystery to me, I cannot be certain." [Note that this is about *my* knowledge of the MOA of Tylenol - it is not dependent on someone else knowing the MOA.]
Am I missing some reason why the point about consciousness is different?
Yes. It isn't just that I don't know how to explain consciousness, it's that I have no idea how to explain it, how to fit it into my general world view. Nor, so far as I can tell, does anyone else. The behavior of Tylenol is probably understood by someone, and I understand enough about it to fit it into my world view.
I haven't given tis real thought, since at 76 I'm pretty unlikely to get a chance to get much of an extended life span, but among my concerns are that I am not only spending on my needs and wants, but trying to continue to accumulate resources to pass on to my descendants. I have children that I love very much, and grandchildren and great-grandchildren that I don't "know" well enough to actually love them, but whom I care about a great deal.
So, how much of my current resources should I spend in an effort to live longer is a tough question. No doubt I should continue to try to maintain my current pretty good health (for more reasons than just living longer), but what is worth spending additional resources on, as compared to being able to bequeath them to my descendants?
My father always joked about two things in his personal life. The first was that he had been living "on borrowed time" since he joined the Marine Corps in December of 1942. The second was that he wanted to "run out of breath and money at the same time." He only really meant the first. He strove mightily to leave behind enough money to assure my mother's situation for however long she lived. I don't think I'm on his kind of borrowed time, but my considerations for my wife and descendants are real.
I'm 78. On the other hand, I had a test for biological markers of age a decade or two ago and it came out about ten years younger than my chronological age. That's consistent with my general behavior and appearance — people in realspace commonly underestimate my age.
So I may have a little longer than my age would suggest.
The rest of your point I agree with, but at this point in life I'm well enough off so it isn't a serious issue for me.
I've had a pretty weird life, but my general health has always been very good. But it did mean I started building wealth pretty late. Still, I appreciate where I got to. I'd like to hang around a bit longer.
One way to think about it is to ask yourself: Will my younger relatives -- who are more likely to live to a time when aging has been beaten -- begrudge me using my resources to have a shot at living again? In many cases, they WILL, so the real question perhaps should be: *Should* they? Should they be more grudging than they would about you buying a second (or nicer) home or car or a bunch of expensive vacations?
Luckily, my family doesn't seem much like that. I'm reasonably sure they ones old enough to be rational are content to let me choose whichever path I choose.
I bet you a nickel consciousness is just electron flow and that preserving the meaningful structures you define as "you" plus a continuity of consciousness would just require cabling your brain to a computer and synchronizing computational processes. This is a dumb action bet and doesn't represent epistemic confidence. Make it $50.
I would certainly consider Minsky's process for "uploading" to be a means of survival. Ray Kurzweil offered the same scenario and I found it plausible, though I don't think we are going to have enough computational power to make it commonly available in my longest possible lifespan.
I wouldn't consider the more usual versions to be survival. At best I would consider them to be reproduction.
Here's what Daniel Dennett calls an "intuition pump":
Suppose that we have a technological means of scanning your entire body, and assembling a perfect duplicate. We do so, on the other side of a room, so the two can look at each other. Now we take out a shotgun and shoot you through the skull—by "you" I mean the original, the one that was scanned. Do you believe that, in that instant, you will suddenly be seeing through the eyes of the DDF on the other side of the room, watching in shock as the original DDF, now headless, collapses?
I don't see that it makes any difference if the DDF that was scanned is destroyed in the process of scanning, rather than afterward.
If DDF1 and DDF2 were telepathically linked, that might lead to a different conclusion. But I don't think there is convincing evidence of telepathy.
I know this isn’t what you were looking for David but I think it’s pretty unlikely you are going to be the first person to avoid death. So it might be worthwhile to try to relieve your fear instead. If you can do that your remaining time may be richer. Note: I’m a little younger than you - about seven years - and I’m not thrilled about my life ending either.
I take articles from MAPS with a large grain of salt but this seems more likely to be successful than cryogenics or uploading your consciousness to a machine.
https://maps.org/bulletin-psychedelics-are-poised-to-change-how-we-age-and-die/
A quibble. If you have three independent 10% chances of living, I don’t think the accumulated probability is 30%. You die if none of them come up, the probability of which is 0.9³ = 0.729. So the probability of living is 0.271 (27.1%). However, given the inaccuracy of the three initial estimates, you can laugh this off.
You are assuming the three things are independent. I can't both live another hundred years because aging has been solved and because I have been successfully suspended — if age reversal happens ten years from now I won't die of old age and be frozen. If either of those happens, the reason I still exist in fifty years won't be the immortality of the soul.
I should have made it clear that my probabilities are not about aging, suspension, or the soul in general but about each being the reason I am still around in more than fifty years.
I think it remains true that, for you to die of old age within a normal time, none of these things must happen, which is what my calculation was based on. However, admittedly I’m no expert on probability theory, so I shouldn’t press the point strongly. Perhaps we should also take into account the possibility that you may die of something other than old age—which could happen to any of us at any time, regardless of life-extension technology. Now there’s a cheerful thought…
There is another possibilty here (see Nick Bostrom's simulation arguements, they're persuasive) and that's the idea that we're already simulations living in a simulation.
In that world, when the body dies, the mind might already be backed up automatically, or not.
If yes, the question would be whether you get respawned or not, if yes, wow, if no, you'd never know.
George Hotz said, "the diffence between an NPC and a PC is that a PC knows it's a PC," so we might be just NPCs and death is death, done, finished, sorry, game over.
He gave an interesting talk at the Austin SBSW (I think) conference some years ago on "How to Hack" the simulation. Interesting, but no final conclusions.
I think we're in a simulation, but whether we'll find out, I suspect the odds are slim.
I don't think the simulation hypothesis solves the consciousness puzzle. I'm pretty sure that WoW NPC's are not conscious and have no idea how one could program ones that were.
Argument from incredulity.
Natural selection drove us to conscious through means we don't understand, but it's still just ones and zeroes tied to a physical processor (unless the revealed religions are correct and we have souls -- but that would seem argue even more forcefully that an NPC could be conscious, like a demon or an angel).
Do you see no difference between yourself and a p-zombie?
“the physical reality that we more or less understand.”
This phrase is making a questionable assertion here. Why believe we “more or less” understand physical reality, given that we don’t yet know the true laws of physics?
We understand it well enough to build airplanes that fly, nuclear reactors that produce power, lots of other things that depend on our understanding of physical reality. "More or less" because there is still much we don't understand.
The question of exactly how much we don’t understand matters a lot! Given that everything you experience is mediated through consciousness, you have more evidence of consciousness than anything else. Yet we don’t have an evidentiary model of consciousness or even any clue of where to start! It seems to me we _think_ we understand the laws of physics because of the incredible things we can do. But framing the situation as, we really don’t know, opens the door for a lot more room for possible optimism.
"it is better to go to a house of mourning than to a house of feasting... because the living will take it to heart".
It's a sign of your usual humility that you tangle with this topic.
You didn't mention the "other" category. The obvious, most likely to occur and to succeed option is that AGI will invent a way of making us immortal. We can be agnostic to what exactly it will look like. The important thing is that AGI is built, and is aligned. In particular, if it is not aligned, any other option on your list (except for the soul argument) is meaningless.
Fair point. I don't see the large language models as implying that human level AI is just around the corner and superhuman not far behind, although some do. They are not thinking, they are mining a vast database of the results of human thinking, and there isn't a body of superhuman thinking for them to mine.
So my guess is that human level AI will be a while, superhuman still further off, hence that the latter is unlikely to show up in my "natural" lifetime. I include the possibility of superhuman AI destroying us, which I discussed briefly in _Future Imperfect_, as one of the reasons that cryonic suspension might not lead to revival.
Leads one to wonder if a newborn is thinking.
A toddler?
A teenager, well, we know those creatures don't think....
When does "thinking" begin for humans?
Toddlers at least, perhaps earlier.
Thinking doesn't require language. I quite often have an idea then have to figure out how to put it into words.
It seems experience with feral children shows pretty conclusively that language is required for any advanced thought, i.e. ideas, plans, solutions, "what-ifs," and the like.
When you have an idea, pre- "putting into words" what does that look like if it's not based in a language?
Your mind might not have to say "economics" to know you're thinking of economics, but that's only because you've read about economics (or listened to people talk about it).
A feral child isn't thinking about economics "economics" although, sure, they might feel hunger and see a stick and conceptualize that they could use the stick to knock an apple down from the tree.
But without words or experience, how would they conceptualize a ladder?
Whereas with language, someone without experience or words for ladder, could be made to understand a ladder and it's use for getting apples.
An interesting aside is that AGI, if it doesn't extinct us, and instead serves us, could both emulate the brain inside a computer (virtual world) and figure how to extend life and/or replace the body (with silicon parts or just a clone or perhaps even a better body made from AI sequenced DNA to be whatever you want).
The question I wonder is what a brain emulation will be able to feel, if anything. Our brains feel emotions due to more than just processing information, but instead are effected by our horomones and other physical factors besides just information (thinking, thoughts, reasoning).
What would "living in a computer simulation" be like.....perhaps we already know as we're in one now. Perhaps not.