Re: Many Pasts? Not according to QM...
Le 12-juin-05, à 06:30, Jesse Mazer a écrit : My speculation is that p(y - x) would depend on a combination of some function that depends only on intrinsic features of the description of x and y--how similar x is to y, basically, the details to be determined by some formal theory of consciousness (or 'theory of observer-moments', perhaps)--and the absolute probability of x, since if two possible future OMs x and x' are equally similar to my current OM y, then I'd expect if x had a higher abolute measure than x' (perhaps x' involves an experience of a 'white rabbit' event), then p(y - x) would be larger than p(y - x'). To Jesse: You apparently completely separate the probability of x and x' from the similarity of x and x'. I am not sure that makes sense for me. In particular how could x and x' be similar, if x', but not x, involves a 'white rabbit events'. To Hal: I don't understand how an OM could write a letter. Writing a letter, it seems to me, involves many OMs. Evolution, and more generally computational history *is* what gives sense to any of the OM. What evolves is more real that the many (subjective or objective) bricks on which evolution proceeds and histories locally rely. To Russell: I don't understand what you mean by a conscious description. Even the expression conscious machine can be misleading at some point in the reasoning. It is really some person, which can be (with comp) associate relatively to a machine/machine-history, who can be conscious. Imo, only a person can be conscious. Even the notion of OM, as it is used in most of the recent posts, seems to me be a construction of the mind of some person. It is personhood which makes possible to attribute some sense to our many living 1-person OMs. Bruno http://iridia.ulb.ac.be/~marchal/
Re: Many Pasts? Not according to QM...
Le 12-juin-05, à 14:48, Stathis Papaioannou a écrit : Bruno Marchal writes: But the basic idea is simple perhaps: Suppose I must choose between a) I am 3-multiplied in ten exemplars. One will get an orange juice and 9 will be tortured. b) I am 3-multiplied in ten exemplars. One will be tortured, and 9 will get a glass of orange juice instead. OK. Now, with comp, strictly speaking the 1-uncertainty are ill-defined, indeed. Because the uncertainty bears on the maximal histories. Without precision I would choose b. But if you tell me in advance that all the 9 guys in b, who got the orange juice, will merge (after artificial amnesia of the details which differ in their experience), and/or if you tell me also that the one who will be tortured will be 3- multiplied by 1000, after the torture, this change the number of relative histories going through the 1-state orange-juice or tortured in such a way that it would be better that I choose a. Obviously other multiplication events in the future could also change this, so that to know the real probabilities, in principle you must evaluate the whole histories going through the states. To be sure, the reasoning of Stathis is still 100% correct with comp for what he want illustrate, but such probability calculus should not be considered as a mean to evaluate real probabilities. When you look at the math, this can be described by conflict between local information and global information. It is all but simple. Today I have only solve the probability 1 case, and it is enough for seeing how quantum proba could be justify by just comp. But even this case leads to open math questions. It is tricky in QM too. I was with you until you proposed the tortured copy in (a) be multiplied 1000-fold or the 9 orange juice copies in (b) be merged. I would *still* choose (a) in these situations. I look at it in two steps. The first step is exactly the same as without the multiplying/merging, so at this point (a) is better. If you had then proposed something like, the orange juice copies will then be tortured, then that would have made a difference to my choice. What you in fact proposed is that the absolute measure of the tortured copies be subsequently increased or the absolute measure of the orange juice copies be subsequently decreased. I would argue that changing the absolute measure in this way can make no possible first person difference; or, equivalently, that multiplying or reducing the number of instantiations of an observer moment makes no possible first person difference - it's all the one observer moment. Yes but this leads to paradoxes. It can be shown that all OM have the same measure in the running of the UD, or that there is no measure at all. The relative measure of OM2 relative to OM1 will be given by the density of computations going from OM1 to OM2. What does make a difference is the *relative* measure of candidate successor OM's, and it is crucial that this refers to the transition from one OM to the next. Strictly speaking I agree, but then I am taking the opportunity of the ambiguous nature of you statement. This is simply because that is how our minds perceive the passage of time and construct the illusion of a single individual who maintains his identity over time. I agree intuitively, but here I have a problem: for technical reasons I disbelieve in intuition at this stage. At this stage I cannot *not* take into account the reversal which makes the passage of time secondary from the way the 1-computations (the web of arithmetical dreams) are coherent. Of course this is probably highly counter-intuitive and that's why I turn on the math. I have said this recurrently on the list. The thought experiments are good for making us doubt about many prejudices. But to build a theory, at some point it is necessary to be utterly clear on what we assume or not, and to be open that the consequences of the theory are in contradiction with what intuition told us. After all this happened already more than one time with modern physics. I am not sure at all I can follow you when you describe how our minds perceive the passage of time. I have learn to accept that the notion of single individual is less an illusion than time, space and all physical modalities. But I know it is counter-intuitive and that is the reason I have eventually decided to interview lobian machine to take into account the lob-godel incompleteness (which is counter-intuitive at its roots). Sorry if this looks a little bit like an authoritative argument, but I can explain all the details if you are willing to cure your math anxiety As I said once, common sense is the only tool we have to go a little bit beyond ... common sense. Bruno http://iridia.ulb.ac.be/~marchal/
more torture
I have been arguing in recent posts that the absolute measure of an observer moment (or observer, if you prefer) makes no possible difference at the first person level. A counterargument has been that, even if an observer cannot know how many instantiations of him are being run, it is still important in principle to take the absolute measure into account, for example when considering the total amount of suffering in the world. The following thought experiment shows how, counterintuitively, sticking to this principle may actually be doing the victims a disservice: You are one of 10 copies who are being tortured. The copies are all being run in lockstep with each other, as would occur if 10 identical computers were running 10 identical sentient programs. Assume that the torture is so bad that death is preferable, and so bad that escaping it with your life is only marginally preferable to escaping it by dying (eg., given the option of a 50% chance of dying or a 49% chance of escaping the torture and living, you would take the 50%). The torture will continue for a year, but you are allowed one of 3 choices as to how things will proceed: (a) 9 of the 10 copies will be chosen at random and painlessly killed, while the remaining copy will continue to be tortured. (b) For one minute, the torture will cease and the number of copies will increase to 10^100. Once the minute is up, the number of copies will be reduced to 10 again and the torture will resume as before. (c) the torture will be stopped for 8 randomly chosen copies, and continue for the other 2. Which would you choose? To me, it seems clear that there is an 80% chance of escaping the torture if you pick (c), while with (a) it is certain that the torture will continue, and with (b) it is certain that the torture will continue with only one minute of respite. Are there other ways to look at the choices? It might be argued that in (a) there is a 90% chance that you will be one of the copies who is killed, and thus a 90% chance that you will escape the torture, better than your chances in (c). However, even if you are one of the ones killed, this does not help you at all. If there is a successor observer moment at the moment of death, subjectively, your consciousness will continue. The successor OM in this case comes from the one remaining copy who is being tortured, hence guaranteeing that you will continue to suffer. What about looking at it from an altruistic rather than selfish viewpoint: isn't it is better to decrease the total suffering in the world by 90% as in (a) rather than by 80% as in (c)? Before making plans to decrease suffering, ask the victims. All 10 copies will plead with you to choose (c). What about (b)? ASSA enthusiasts might argue that with this choice, an OM sampled randomly from the set of all possible OM's will almost certainly be from the one minute torture-free interval. What would this mean for the victims? If you interview each of the 10 copies before the minute starts, they will tell you that they are currently being tortured and they expect that they will get one minute respite, then start suffering again, so they wish the choice had been (c). Next, if you interview each of the 10^100 copies they will tell you that the torture has stopped for exactly one minute by the torture chambre's clock, but they know that it is going to start again and they wish you had chosen (c). Finally, if you interview each of the 10 copies for whom the torture has recommenced, they will report that they remember the minute of respite, but that's no good to them now, and they wish you had chosen (c). --Stathis Papaioannou _ Express yourself instantly with MSN Messenger! Download today - it's FREE! http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/
Re: more torture
I agree with everything you say in this post, but I am not sure that settles the issue. It does not change my mind on the preceding post where we were disagreeing; which was that IF I must choose between A) splitted between 1 finite hells and 1 infinite paradise B) Splitted between 1 infinite hell and 1 finite paradises where finite and infinite refer to the number of computational steps simulating the stories of thoise hells and paradises, THEN I should choose A. This is because all finite stories have a measure 0. Infinite stories, by their natural DU multiplications will have a measure one. But we are on the verge of inconsistency, because in practice there is no way to garantie anything like the finiteness of any computation going through our states (this is akin to the insolubility of the self-stopping problem by sufficiently rich (lobian) turing machine). The idea that I try to convey is that if I am in state S1, the probability of some next state S2 depends on the proportion, among the infinite stories going through S1 of those *infinite* stories going also through S2. And all finite stories must be discounted. (It is not necessary I remain personally immortal in those infinite stories, the measure is given by the stories going through my states even if I have a finite 3-life-time in all of those stories). (btw, this entails also that comp implies at least infinite past and/or future for any universes supporting our present story). [Note that here I am going far ahead of what I can ask to the lobian machine, because our talk involves quantifiers on stories and that's very complex to handle. Well, to be sure I have till now only been able to translate the case of probability one, in machine term; but it is enough to extract non trivial information on the logic of observable proposition.] Bruno Le 13-juin-05, à 13:00, Stathis Papaioannou a écrit : I have been arguing in recent posts that the absolute measure of an observer moment (or observer, if you prefer) makes no possible difference at the first person level. A counterargument has been that, even if an observer cannot know how many instantiations of him are being run, it is still important in principle to take the absolute measure into account, for example when considering the total amount of suffering in the world. The following thought experiment shows how, counterintuitively, sticking to this principle may actually be doing the victims a disservice: You are one of 10 copies who are being tortured. The copies are all being run in lockstep with each other, as would occur if 10 identical computers were running 10 identical sentient programs. Assume that the torture is so bad that death is preferable, and so bad that escaping it with your life is only marginally preferable to escaping it by dying (eg., given the option of a 50% chance of dying or a 49% chance of escaping the torture and living, you would take the 50%). The torture will continue for a year, but you are allowed one of 3 choices as to how things will proceed: (a) 9 of the 10 copies will be chosen at random and painlessly killed, while the remaining copy will continue to be tortured. (b) For one minute, the torture will cease and the number of copies will increase to 10^100. Once the minute is up, the number of copies will be reduced to 10 again and the torture will resume as before. (c) the torture will be stopped for 8 randomly chosen copies, and continue for the other 2. Which would you choose? To me, it seems clear that there is an 80% chance of escaping the torture if you pick (c), while with (a) it is certain that the torture will continue, and with (b) it is certain that the torture will continue with only one minute of respite. Are there other ways to look at the choices? It might be argued that in (a) there is a 90% chance that you will be one of the copies who is killed, and thus a 90% chance that you will escape the torture, better than your chances in (c). However, even if you are one of the ones killed, this does not help you at all. If there is a successor observer moment at the moment of death, subjectively, your consciousness will continue. The successor OM in this case comes from the one remaining copy who is being tortured, hence guaranteeing that you will continue to suffer. What about looking at it from an altruistic rather than selfish viewpoint: isn't it is better to decrease the total suffering in the world by 90% as in (a) rather than by 80% as in (c)? Before making plans to decrease suffering, ask the victims. All 10 copies will plead with you to choose (c). What about (b)? ASSA enthusiasts might argue that with this choice, an OM sampled randomly from the set of all possible OM's will almost certainly be from the one minute torture-free interval. What would this mean for the victims? If you interview each of the 10 copies before the minute starts,
Re: Many Pasts? Not according to QM...
Bruno Marchal: To Jesse: You apparently completely separate the probability of x and x' from the similarity of x and x'. I am not sure that makes sense for me. In particular how could x and x' be similar, if x', but not x, involves a 'white rabbit events'. It's not completely separable, but I'd think that similarity would mostly be a function of memories, personality, etc...even if I experience something very weird, I can still have basically the same mind. For example, a hoaxer could create a realistic animatronic talking white rabbit, and temporarily I might experience an observer-moment identical to what I'd experience if I saw a genuine white talking rabbit, so the similarity between my current experience and what I'd experience in a white rabbit universe would be the same as the similarity between my current experience and what I'd experience in a universe where someone creates a realistic hoax. I don't think the first-person probabilities of experiencing hoaxes are somehow kept lower than what you'd expect from a third-person perspective, do you? Jesse
Re: Observer-Moment Measure from Universe Measure
Hi Brent, You didn't answer my last post where I explain that Bp is different from Bp p. I hope you were not too much disturbed by my teacher's tone (which can be enervating I imagine). Or is it because you don't recognize the modal form of Godel's theorem: ~Bf - ~B(~Bf), which is equivalent to B(Bf - f) - Bf, by simple contraposition p - q is equivalent with ~p - ~q, and using also that ~p is equivalent to p - f, where f is put for false. This shows that for a consistent (~Bf) machine, although Bf - f is true, it cannot be proved by the machine. Now (Bf f) - f trivially. So Bf and Bf f are not equivalent for the machine (although they are for the guardian angel). Bruno http://iridia.ulb.ac.be/~marchal/
Re: Many Pasts? Not according to QM...
Bruno Marchal wrote: Bruno Marchal: To Jesse: You apparently completely separate the probability of x and x' from the similarity of x and x'. I am not sure that makes sense for me. In particular how could x and x' be similar, if x', but not x, involves a 'white rabbit events'. It's not completely separable, but I'd think that similarity would mostly be a function of memories, personality, etc...even if I experience something very weird, I can still have basically the same mind. For example, a hoaxer could create a realistic animatronic talking white rabbit, and temporarily I might experience an observer-moment identical to what I'd experience if I saw a genuine white talking rabbit, so the similarity between my current experience and what I'd experience in a white rabbit universe would be the same as the similarity between my current experience and what I'd experience in a universe where someone creates a realistic hoax. I don't think the first-person probabilities of experiencing hoaxes are somehow kept lower than what you'd expect from a third-person perspective, do you? Perhaps I misunderstood you, but it seems to me, that in case you ask me to compute P(x - y) (your notation), it could and even should change that prediction result. In particular if the rabbit has been generated by a genuine hoaxer I would predict the white rabbit will stay in y, and if the hoaxer is not genuine, then I would still consider x and x' as rather very dissimilar. What do you think? This follows *also* from a relativisation of Hall Finney's theory based on kolmogorov complexity: a stable white rabbit is expensive in information resource. No? Well, note that following Hal's notation, I was actually assuming y came before x (or x'), and I was calculating P(y - x). And your terminology is confusing to me here--when you say if the hoaxer is not genuine, do you mean that the white rabbit wasn't a hoax but was a genuine talking rabbit (in which case no hoaxer is involved at all), or do you mean if the white rabbit *was* a hoax? If the latter, then what do you mean when you say if the rabbit had been generated by a genuine hoaxer--is the white rabbit real, or is it a hoax in this case? Also, when you say you'd consider x and x' as very dissimilar, do you mean from each other or from y? Remember that dissimilar is just the word I use for continuity of personal identity, how much two successive experiences make sense as being successive OMs of the same person, it doesn't refer to whether the two sensory experiences are themselves dissimilar or dissimilar. If I'm here looking at my computer, then suddenly close my eyes, the two successive experiences will be quite dissimilar in terms of the sensory information I'm taking in, but they'll still be similar in terms of my background memories, personality, etc., so they make sense as successive OMs of the same person. On the other hand, if I'm sitting at my computer and suddenly my brain is replaced with the brain of George W. Bush, there will be very little continuity of identity despite the fact that the sensory experiences of both OMs would be pretty similar, so in my terminology there would be very little similarity between these two OMs. As for the cost of simulating a white rabbit universe, I agree it's more expensive than simulating a non-white rabbit universe, but I don't see how this relates to continuity of identity when experiencing white rabbits vs. not experiencing them. Jesse
Re: more torture
Hi Quentin, concerning finite/infinite number of steps, it seems to me that it is always possible to have a computation that will take an infinite number of steps to arrive at a particular state, since for any state, there exists an infinity of computational histories which go through it, so it seems to me that some of them needs infinite steps... Do I miss something ? Yes. My fault. I was not clear enough. First it is obvious that any states in the execution of the DU has only a finite 3-computations. But, as you say, any states belong to an infinite set of computations, and this justifies that from the first person point of view we can have infinite past. I should have mention the 1-3 difference. Apology. (But I would not say that some state *needs* an infinite 3-computation, that one would not even be generated by the DU. Regards, Bruno Sincerely, Quentin Anciaux Le lundi 13 juin 2005 à 15:37 +0200, Bruno Marchal a écrit : I agree with everything you say in this post, but I am not sure that settles the issue. It does not change my mind on the preceding post where we were disagreeing; which was that IF I must choose between A) splitted between 1 finite hells and 1 infinite paradise B) Splitted between 1 infinite hell and 1 finite paradises where finite and infinite refer to the number of computational steps simulating the stories of thoise hells and paradises, THEN I should choose A. This is because all finite stories have a measure 0. Infinite stories, by their natural DU multiplications will have a measure one. But we are on the verge of inconsistency, because in practice there is no way to garantie anything like the finiteness of any computation going through our states (this is akin to the insolubility of the self-stopping problem by sufficiently rich (lobian) turing machine). The idea that I try to convey is that if I am in state S1, the probability of some next state S2 depends on the proportion, among the infinite stories going through S1 of those *infinite* stories going also through S2. And all finite stories must be discounted. (It is not necessary I remain personally immortal in those infinite stories, the measure is given by the stories going through my states even if I have a finite 3-life-time in all of those stories). (btw, this entails also that comp implies at least infinite past and/or future for any universes supporting our present story). [Note that here I am going far ahead of what I can ask to the lobian machine, because our talk involves quantifiers on stories and that's very complex to handle. Well, to be sure I have till now only been able to translate the case of probability one, in machine term; but it is enough to extract non trivial information on the logic of observable proposition.] Bruno Le 13-juin-05, à 13:00, Stathis Papaioannou a écrit : I have been arguing in recent posts that the absolute measure of an observer moment (or observer, if you prefer) makes no possible difference at the first person level. A counterargument has been that, even if an observer cannot know how many instantiations of him are being run, it is still important in principle to take the absolute measure into account, for example when considering the total amount of suffering in the world. The following thought experiment shows how, counterintuitively, sticking to this principle may actually be doing the victims a disservice: You are one of 10 copies who are being tortured. The copies are all being run in lockstep with each other, as would occur if 10 identical computers were running 10 identical sentient programs. Assume that the torture is so bad that death is preferable, and so bad that escaping it with your life is only marginally preferable to escaping it by dying (eg., given the option of a 50% chance of dying or a 49% chance of escaping the torture and living, you would take the 50%). The torture will continue for a year, but you are allowed one of 3 choices as to how things will proceed: (a) 9 of the 10 copies will be chosen at random and painlessly killed, while the remaining copy will continue to be tortured. (b) For one minute, the torture will cease and the number of copies will increase to 10^100. Once the minute is up, the number of copies will be reduced to 10 again and the torture will resume as before. (c) the torture will be stopped for 8 randomly chosen copies, and continue for the other 2. Which would you choose? To me, it seems clear that there is an 80% chance of escaping the torture if you pick (c), while with (a) it is certain that the torture will continue, and with (b) it is certain that the torture will continue with only one minute of respite. Are there other ways to look at the choices? It might be argued that in (a) there is a 90% chance that you will be one of the copies who is killed, and thus a 90% chance that you will escape the torture, better than your chances in (c). However, even if
Re: Many Pasts? Not according to QM...
Oops sorry. I did misunderstood you. Thanks for the clarification. I agree with your preceding post to Hal now. Bruno Le 13-juin-05, à 16:23, Jesse Mazer a écrit : Bruno Marchal wrote: Bruno Marchal: To Jesse: You apparently completely separate the probability of x and x' from the similarity of x and x'. I am not sure that makes sense for me. In particular how could x and x' be similar, if x', but not x, involves a 'white rabbit events'. It's not completely separable, but I'd think that similarity would mostly be a function of memories, personality, etc...even if I experience something very weird, I can still have basically the same mind. For example, a hoaxer could create a realistic animatronic talking white rabbit, and temporarily I might experience an observer-moment identical to what I'd experience if I saw a genuine white talking rabbit, so the similarity between my current experience and what I'd experience in a white rabbit universe would be the same as the similarity between my current experience and what I'd experience in a universe where someone creates a realistic hoax. I don't think the first-person probabilities of experiencing hoaxes are somehow kept lower than what you'd expect from a third-person perspective, do you? Perhaps I misunderstood you, but it seems to me, that in case you ask me to compute P(x - y) (your notation), it could and even should change that prediction result. In particular if the rabbit has been generated by a genuine hoaxer I would predict the white rabbit will stay in y, and if the hoaxer is not genuine, then I would still consider x and x' as rather very dissimilar. What do you think? This follows *also* from a relativisation of Hall Finney's theory based on kolmogorov complexity: a stable white rabbit is expensive in information resource. No? Well, note that following Hal's notation, I was actually assuming y came before x (or x'), and I was calculating P(y - x). And your terminology is confusing to me here--when you say if the hoaxer is not genuine, do you mean that the white rabbit wasn't a hoax but was a genuine talking rabbit (in which case no hoaxer is involved at all), or do you mean if the white rabbit *was* a hoax? If the latter, then what do you mean when you say if the rabbit had been generated by a genuine hoaxer--is the white rabbit real, or is it a hoax in this case? Also, when you say you'd consider x and x' as very dissimilar, do you mean from each other or from y? Remember that dissimilar is just the word I use for continuity of personal identity, how much two successive experiences make sense as being successive OMs of the same person, it doesn't refer to whether the two sensory experiences are themselves dissimilar or dissimilar. If I'm here looking at my computer, then suddenly close my eyes, the two successive experiences will be quite dissimilar in terms of the sensory information I'm taking in, but they'll still be similar in terms of my background memories, personality, etc., so they make sense as successive OMs of the same person. On the other hand, if I'm sitting at my computer and suddenly my brain is replaced with the brain of George W. Bush, there will be very little continuity of identity despite the fact that the sensory experiences of both OMs would be pretty similar, so in my terminology there would be very little similarity between these two OMs. As for the cost of simulating a white rabbit universe, I agree it's more expensive than simulating a non-white rabbit universe, but I don't see how this relates to continuity of identity when experiencing white rabbits vs. not experiencing them. Jesse http://iridia.ulb.ac.be/~marchal/
Re: more torture
At 06:00 AM 6/13/2005, Stathis Papaioannou wrote: I have been arguing in recent posts that the absolute measure of an observer moment (or observer, if you prefer) makes no possible difference at the first person level. A counterargument has been that, even if an observer cannot know how many instantiations of him are being run, it is still important in principle to take the absolute measure into account, for example when considering the total amount of suffering in the world. The following thought experiment shows how, counterintuitively, sticking to this principle may actually be doing the victims a disservice: You are one of 10 copies who are being tortured. The copies are all being run in lockstep with each other, as would occur if 10 identical computers were running 10 identical sentient programs. Assume that the torture is so bad that death is preferable, and so bad that escaping it with your life is only marginally preferable to escaping it by dying (eg., given the option of a 50% chance of dying or a 49% chance of escaping the torture and living, you would take the 50%). The torture will continue for a year, but you are allowed one of 3 choices as to how things will proceed: (a) 9 of the 10 copies will be chosen at random and painlessly killed, while the remaining copy will continue to be tortured. (b) For one minute, the torture will cease and the number of copies will increase to 10^100. Once the minute is up, the number of copies will be reduced to 10 again and the torture will resume as before. (c) the torture will be stopped for 8 randomly chosen copies, and continue for the other 2. Which would you choose? To me, it seems clear that there is an 80% chance of escaping the torture if you pick (c), while with (a) it is certain that the torture will continue, and with (b) it is certain that the torture will continue with only one minute of respite. RM writes. . . Here is my criteria: There are those who suggest that there is only one electron in the universe, but that it travels forward and backward in time, thus making multiple copies of itself. If the individual percipient would eventually have to experience the pain and suffering of all whom he had affected--or caused to experience pain and suffering, then the most selfish, altruistic *and* sensible choice would be (c). Rich Miller
RE: more torture
Jesse Mazer writes: If you impose the condition I discussed earlier that absolute probabilities don't change over time, or in terms of my analogy, that the water levels in each tank don't change because the total inflow rate to each tank always matches the total outflow rate, then I don't think it's possible to make sense of the notion that the observer-moments in that torture-free minute would have 10^100 times greater absolute measure. If there's 10^100 times more water in the tanks corresponding to OMs during that minute, where does all this water go after the tank corresponding to the last OM in this minute, and where is it flowing in from to the tank corresponding to the first OM in this minute? I would propose to implement the effect by duplicating the guy 10^100 times during that minute, then terminating all the duplicates after that time. What happens in your model when someone dies in some fraction of the multiverse? His absolute measure decreases, but where does the now-excess water go? Hal Finney
RE: more torture
Hal Finney wrote: Jesse Mazer writes: If you impose the condition I discussed earlier that absolute probabilities don't change over time, or in terms of my analogy, that the water levels in each tank don't change because the total inflow rate to each tank always matches the total outflow rate, then I don't think it's possible to make sense of the notion that the observer-moments in that torture-free minute would have 10^100 times greater absolute measure. If there's 10^100 times more water in the tanks corresponding to OMs during that minute, where does all this water go after the tank corresponding to the last OM in this minute, and where is it flowing in from to the tank corresponding to the first OM in this minute? I would propose to implement the effect by duplicating the guy 10^100 times during that minute, then terminating all the duplicates after that time. What happens in your model when someone dies in some fraction of the multiverse? His absolute measure decreases, but where does the now-excess water go? In my model, death only exists from a third-person perspective, but from a first-person perspective I'm subscribing to the QTI, so consciousness will always continue in some form (even if my memories don't last or I am reduced to an amoeba-level consciousness)--the water molecules are never created or destroyed. For what would happen when an observer is duplicated from a third-person perspective, it might help to consider the example I discussed on the 'Last-minute vs. anticipatory quantum immortality' thread at http://www.escribe.com/science/theory/m4841.html , where a person is initially duplicated before a presidential election, and then depending on the results of the election, one duplicate is later copied 999 times. All else being equal, I'd speculate that the initial 2-split would anticipate the later 999-split, so that 999 out of 1000 water molecules of the first observer would split off into the copy that is later going to be split 999 times, so before this second split, OMs of this copy would have 999 times the absolute measure of the copy that isn't going to be split again. I'm not absolutely sure that this would be a consequence of the idea about finding a unique self-consistent set of absolute and conditional probabilities based only on a similarity matrix and the condition of absolute probabilities not changing with time, but it seems intuitive to me that it would. At some point I'm going to try to test this idea with mathematica or something, creating a finite set of OMs and deciding what the possible successors to each one are in order to construct something like a similarity matrix, then finding the unique vector of absolute probabilities that, when multiplied by this matrix, gives a unit vector (the procedure I discussed in my last post to you at http://www.escribe.com/science/theory/m6855.html ). Hopefully the absolute probabilities would indeed tend to anticipate future splits in the way I'm describing. So if this anticipatory idea works, then any copy that's very unlikely to survive long from a third-person perspective is going to undergoe fewer future splits from a multiverse perspective (there will always be few branches where this copy survives though), so your conditional probability of becoming such a copy would be low, meaning that not much of your water would flow into that copy, and it will have a smaller absolute measure than copies that are likely to survive in more branches. Jesse
Re: more torture
Because no such thing as free will exists one has to consider three different universes in which the three different choices are made. The three universes will have comparable measures. The antropic factor of 10^100 will then dominate and will cause the observer to find himself having made choice b) as one of the 10^100 copies in the minute without torture. Saibal - Defeat Spammers by launching DDoS attacks on Spam-Websites: http://www.hillscapital.com/antispam/ - Original Message - From: Stathis Papaioannou [EMAIL PROTECTED] To: everything-list@eskimo.com Sent: Monday, June 13, 2005 01:00 PM Subject: more torture I have been arguing in recent posts that the absolute measure of an observer moment (or observer, if you prefer) makes no possible difference at the first person level. A counterargument has been that, even if an observer cannot know how many instantiations of him are being run, it is still important in principle to take the absolute measure into account, for example when considering the total amount of suffering in the world. The following thought experiment shows how, counterintuitively, sticking to this principle may actually be doing the victims a disservice: You are one of 10 copies who are being tortured. The copies are all being run in lockstep with each other, as would occur if 10 identical computers were running 10 identical sentient programs. Assume that the torture is so bad that death is preferable, and so bad that escaping it with your life is only marginally preferable to escaping it by dying (eg., given the option of a 50% chance of dying or a 49% chance of escaping the torture and living, you would take the 50%). The torture will continue for a year, but you are allowed one of 3 choices as to how things will proceed: (a) 9 of the 10 copies will be chosen at random and painlessly killed, while the remaining copy will continue to be tortured. (b) For one minute, the torture will cease and the number of copies will increase to 10^100. Once the minute is up, the number of copies will be reduced to 10 again and the torture will resume as before. (c) the torture will be stopped for 8 randomly chosen copies, and continue for the other 2. Which would you choose? To me, it seems clear that there is an 80% chance of escaping the torture if you pick (c), while with (a) it is certain that the torture will continue, and with (b) it is certain that the torture will continue with only one minute of respite. Are there other ways to look at the choices? It might be argued that in (a) there is a 90% chance that you will be one of the copies who is killed, and thus a 90% chance that you will escape the torture, better than your chances in (c). However, even if you are one of the ones killed, this does not help you at all. If there is a successor observer moment at the moment of death, subjectively, your consciousness will continue. The successor OM in this case comes from the one remaining copy who is being tortured, hence guaranteeing that you will continue to suffer. What about looking at it from an altruistic rather than selfish viewpoint: isn't it is better to decrease the total suffering in the world by 90% as in (a) rather than by 80% as in (c)? Before making plans to decrease suffering, ask the victims. All 10 copies will plead with you to choose (c). What about (b)? ASSA enthusiasts might argue that with this choice, an OM sampled randomly from the set of all possible OM's will almost certainly be from the one minute torture-free interval. What would this mean for the victims? If you interview each of the 10 copies before the minute starts, they will tell you that they are currently being tortured and they expect that they will get one minute respite, then start suffering again, so they wish the choice had been (c). Next, if you interview each of the 10^100 copies they will tell you that the torture has stopped for exactly one minute by the torture chambre's clock, but they know that it is going to start again and they wish you had chosen (c). Finally, if you interview each of the 10 copies for whom the torture has recommenced, they will report that they remember the minute of respite, but that's no good to them now, and they wish you had chosen (c). --Stathis Papaioannou _ Express yourself instantly with MSN Messenger! Download today - it's FREE! http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/
No torture
"No tortue". Now, sit and contemplate if you felt a difference when, after reading message after message with the opposite words in it, and then suddenly you see "No tortue". -- ___Sign-up for Ads Free at Mail.com http://www.mail.com/?sr=signup
Re-Observer-Moment Measure from Universe Measure
Bruno Marchal wrote: Godel's theorem: ~Bf - ~B(~Bf), which is equivalent to B(Bf - f) - Bf, Just a little aside a la Descartes + Godel: (assume that think and believe are synonymous and that f = you are) B(Bf - f) - Bf can be rendered as: If you believe that if you think that you are therefore you are, then you think you are. That's what Descartes thought! :-)George
Re: more torture
IMO belief in the ASSA is tantamount to altruism. The ASSA would imply taking action based on its positive impact on the whole multiverse of observer-moments (OMs). We have had some discussion here and on the extropy-chat (transhumanist) mailing list about two different possible flavors of altruism. These are sometimes called averagist vs totalist. The averagist wants to maximize the average happiness of humanity. He opposes measures that will add more people at the expense of decreasing their average happiness. This is a pretty common element among green political movements. The totalist wants to maximize the total happiness of humanity. He believes that people are good and more people are better. This philosophy is less common but is sometimes associated with libertarian or radical right wing politics. These two ideas can be applied to observer-moments as well. But both of these approaches have problems if taken to the extreme. For the extreme averagist, half the OMs are below average. If they were eliminated, the average would rise. But again, half of the remaining OMs would be below (the new, higher) average. So again half should be eliminated. In the end you are left with the one OM with the highest average happiness. Eliminating almost every ounce of intelligence in the universe hardly seems altruistic. For the extreme totalist, the problem is that he will support adding OMs as long as their quality of life is just barely above that which would lead to suicide. More OMs generally will decrease the quality of life of others, due to competition for resources, so the result is a massively overpopulated universe with everyone leading terrible lives. This again seems inconsistent with the goals of altruism. In practice it seems that some middle ground must be found. Adding more OMs is good, up to a point. I don't know if anyone has a good, objective measure that can be maximized for an effective approach to altruism. Let us consider these flavors of altruism in the case of Stathis' puzzle: You are one of 10 copies who are being tortured. The copies are all being run in lockstep with each other, as would occur if 10 identical computers were running 10 identical sentient programs. Assume that the torture is so bad that death is preferable, and so bad that escaping it with your life is only marginally preferable to escaping it by dying (eg., given the option of a 50% chance of dying or a 49% chance of escaping the torture and living, you would take the 50%). The torture will continue for a year, but you are allowed one of 3 choices as to how things will proceed: (a) 9 of the 10 copies will be chosen at random and painlessly killed, while the remaining copy will continue to be tortured. (b) For one minute, the torture will cease and the number of copies will increase to 10^100. Once the minute is up, the number of copies will be reduced to 10 again and the torture will resume as before. (c) the torture will be stopped for 8 randomly chosen copies, and continue for the other 2. Which would you choose? For the averagist, doing (a) will not change average happiness. Doing (b) will improve it, but not that much. The echoes of the torture and anticipation of future torture will make that one minute of respite not particularly pleasant. Doing (c) would seem to be the best choice, as 8 out of the 10 avoid a year of torture. (I'm not sure why Stathis seemed to say that the people would not want to escape their torture, given that it was so bad. That doesn't seem right to me; the worse it is, the more they would want to escape it.) For the totalist, since death is preferable to the torture, each person's life has a negative impact on total happiness. Hence (a) would be an improvement as it removes these negatives from the universe. Doing (b) is unclear: during that one minute, would the 10^100 copies kill themselves if possible? If so, their existence is negative and so doing (b) would make the universe much worse due to the addition of so many negatively happy OMs. Doing (c) would seem to be better, assuming that the 8 out of 10 would eventually find that their lives were positive during that year without torture. So it appears that each one would choose (c), although they would differ about whether (a) is an improvement over the status quo. (b) is deprecated because that one minute will not be pleasant due to the echoes of the torture. If the person could have his memory wiped for that one minute and neither remember nor anticipate future torture, that would make (b) the best choice for both kinds of altruists. Adding 10^100 pleasant observer-moments would increase both total and average happiness and would more than compensate for a year of suffering for 10 people. 10^100 is a really enormous number. Hal Finney
Re: Many Pasts? Not according to QM...
On Mon, Jun 13, 2005 at 11:45:52AM +0200, Bruno Marchal wrote: To Russell: I don't understand what you mean by a conscious description. Even the expression conscious machine can be misleading at some point in the reasoning. A description could be conscious in the same way that with computationalism, a program might be conscious. With computationalism, a certain program is considered conscious when run on an appropriate UTM. However, as you showed in chapter 4 of your thesis it is not necessary to actually run the program on a physical machine. Church-Turing thesis and arithmetical platonism (my all description strings condition fulfills a similar role to arithmetical platonism) are enough. Furthermore, if the conscious program _is_ a UTM in its own right, it can run on itself (actually this is pretty much what my reading of what the Church-Turing thesis is). This obviates having to fix the UTM. Perhaps this is the route into the anthropic principle. This is a model of a conscious description, under the assumption of computationalism. Perhaps this model can be extended to not-computationalism, where a description is conscious if it is able to interpret itself as conscious. I do not have problem with observers being capable of universal computation as a necessary precondition here, should it be necessary. Finally, there is the possibility that a concrete observer (the noumenon) exists somewhere, and that conscious descriptions are merely the anthropic shadow of the observer being observed by itself. It is really some person, which can be (with comp) associate relatively to a machine/machine-history, who can be conscious. Imo, only a person can be conscious. Isn't this the definition of person? Or do you define personhood by something else. Even the notion of OM, as it is used in most of the recent posts, seems to me be a construction of the mind of some person. It is personhood which makes possible to attribute some sense to our many living 1-person OMs. Bruno http://iridia.ulb.ac.be/~marchal/ -- *PS: A number of people ask me about the attachment to my email, which is of type application/pgp-signature. Don't worry, it is not a virus. It is an electronic signature, that may be used to verify this email came from me if you have PGP or GPG installed. Otherwise, you may safely ignore this attachment. A/Prof Russell Standish Phone 8308 3119 (mobile) Mathematics0425 253119 () UNSW SYDNEY 2052 [EMAIL PROTECTED] Australiahttp://parallel.hpc.unsw.edu.au/rks International prefix +612, Interstate prefix 02 pgpMbp4kmkYGl.pgp Description: PGP signature