Re: relevant probability distribution
Russell wrote: > Saibal Mitra wrote: > > > > > > Russell wrote: > > > > > > I take "consciousness" to be that property essential for the operation > > > of the Anthropic Principle. The universe is the way it is because we > > > are here observing it as conscious beings. > > > > > > The first problem this raises is why does the anthropic principle > > > work? - one can conceive of being immersed in a virtual reality which > > > is totally inconsistent with our existence as conscious observers, for > > > example. > > > > > That must be explained by the unlikeliness of such a situation. Why would > > anyone simulate me living happily on the surface of Venus? They could have > > taken any possible person > > I'm not sure how you could advance that argument. Bostrom's recent paper > indicates it is rather likely... Well, Bostrom's argument evaporates when you bring in the SIA. > > Nevertheless a clear logical argument in favour of the AP would be > most welcome. I will explain below that the measure of a universe should be the product of an intrinsic measure and the number of observers. This last factor leads implies the SIA. > > > > > However, let us accept the AP. After all, it has passed observational > > > test with flying colours. We should also expect that we should be an > > > example of the most likely form of consciousness. > > > > > > The second problem is raises is that if ameobae are conscious, then > > > why aren't we amoebae? There are many more amoebae on the planet than > > > there are human beings. I can well accept that dolphins and > > > chimpanzees (for instance) _could_ be conscious, since there are > > > vastly greater numbers of humans around today than there are of these > > > other species, but there is something special that we have that amoeba > > > (or even ants, lets say) don't have. > > > > > > Not sure about ant nests (Hofstadter style). Anyone got a good > > > estimate of the number of extant ant nests vis a vis human population? > > > > > > > I think that one should first define oneself as a particular program, and > > then look at where and how often that program is actually running. Amoebas > > are incapable of running me. Maybe artificial intelligent agents pose more > > of a problem. Why am I not a robot, that can copy himself as many times as > > he pleases? > > > > No the issue concerns any conscious "program", rather than any > particular one. The fact that there are vastly more amoeba than homo > sapiens tends to argue against amoebae being consious. I agree with that. > > B.t.w. Ken Olum made an interesting remark in his paper in which he > > advocates the Self Indicating Assumption (SIA) (see arxiv.org). If > > universes with more observers are more likely than universes with less > > observers, then why don't we live in a universe in which the number was > > pre-programmed to be some ridicolously large number, say N = 10^10? > > (Ken gave a different example). He concludes that apparently such universes > > must be unlikelier by a factor of at least N, to compensate for the factor N > > coming from the number of observers. This fits in nice with the idea that > > more complex programs should have lower measure. You can see that the > > measure of a program must decrease faster than 2^(-p) where p is the length > > of the program. > > > > Saibal > > > > > > I'm not entirely sure where this SIA measure really comes from. Consider a modification of Bostrom's God's coin toss experiment. Let's assume that God creates two universes. One with 10 observers and another with 10^10 observers. Each observer is numbered 1...N, where N is the number of observers. The intrinsic measure of both Universes is the same. However, I would say that I am 10^9 times more likely to find myself in the second universe. If you assume that God chooses his observers perfectly randomly, then this must be true. If I then look at my number, and find that it is 7, then both universes are equally likely again. So the flaw in the Doomsday argument is to say that on the one hand: ''10^10 observers or 10 observers, it doesn't matter. The two universes are equally likely a priory''. But then on the other hand when you observe your number: ''I have number 7, so the second universe is disfavored by a factor 10^9''. This is contradictory. >From > the multiverse, we get that each observer moment is weighted by a > factor that decreases exponentially with time. Hence we can predict by > the classic SSA argument the we should expect to find ourselves in a > period of time when the population outweighs previous epochs by at > least an exponential factor, and that future populations will grow > slower than exponentially. > > However, all this seems irrelevant to the conscious amoeba question. It is relevant to some other related problems. SSA together with SIA make your predictions independent of your definition of observer. You can narrow the definition of observer down to include only
Re: relevant probability distribution
On 15 Jun 2002, at 14:27, Russell Standish wrote: > > No the issue concerns any conscious "program", rather than any > particular one. The fact that there are vastly more amoeba than homo > sapiens tends to argue against amoebae being consious. > This remind me of Jack Vance novels "Alastor". One of the characters is the king that rules over a vast area of the galaxy. He likes to travel incognito among his subjects, and he often ask himself the question: "There is billions of men and only one king. How is it possible that it happens that I am the king ?" Do your position about this is that subjects are not conscious, only kings? >From a third-person point of view (the reader of the novel), the question is simple. There is billions of subjects, and they can all ask themselves "Why I am me and not someone else ?". The problem is we have only a first-person point of view on our universe (or on the "everything"). We must use our imagination, to do thought experiments, to get a third-person point of view. Matthieu. -- http://matthieu.walraet.free.fr
Re: relevant probability distribution
Saibal Mitra wrote:
>
>
> Russell wrote:
> >
> > I take "consciousness" to be that property essential for the operation
> > of the Anthropic Principle. The universe is the way it is because we
> > are here observing it as conscious beings.
> >
> > The first problem this raises is why does the anthropic principle
> > work? - one can conceive of being immersed in a virtual reality which
> > is totally inconsistent with our existence as conscious observers, for
> > example.
> >
> Tha must be explained by the unlikeliness of such a situation. Why would
> anyone simulate me living happily on the surface of Venus? They could have
> taken any possible person
I'm not sure how you could advance that argument. Bostrom's recent paper
indicates it is rather likely...
Nevertheless a clear logical argument in favour of the AP would be
most welcome.
>
> > However, let us accept the AP. After all, it has passed observational
> > test with flying colours. We should also expect that we should be an
> > example of the most likely form of consciousness.
> >
> > The second problem is raises is that if ameobae are conscious, then
> > why aren't we amoebae? There are many more amoebae on the planet than
> > there are human beings. I can well accept that dolphins and
> > chimpanzees (for instance) _could_ be conscious, since there are
> > vastly greater numbers of humans around today than there are of these
> > other species, but there is something special that we have that amoeba
> > (or even ants, lets say) don't have.
> >
> > Not sure about ant nests (Hofstadter style). Anyone got a good
> > estimate of the number of extant ant nests vis a vis human population?
> >
>
> I think that one should first define oneself as a particular program, and
> then look at where and how often that program is actually running. Amoebas
> are incapable of running me. Maybe artificial intelligent agents pose more
> of a problem. Why am I not a robot, that can copy himself as many times as
> he pleases?
>
No the issue concerns any conscious "program", rather than any
particular one. The fact that there are vastly more amoeba than homo
sapiens tends to argue against amoebae being consious.
> B.t.w. Ken Olum made an interesting remark in his paper in which he
> advocates the Self Indicating Assumption (SIA) (see arxiv.org). If
> universes with more observers are more likely than universes with less
> observers, then why don't we live in a universe in which the number was
> pre-programmed to be some ridicolously large number, say N = 10^10?
> (Ken gave a different example). He concludes that apparently such universes
> must be unlikelier by a factor of at least N, to compensate for the factor N
> coming from the number of observers. This fits in nice with the idea that
> more complex programs should have lower measure. You can see that the
> measure of a program must decrease faster than 2^(-p) where p is the length
> of the program.
>
> Saibal
>
>
I'm not entirely sure where this SIA measure really comes from. From
the multiverse, we get that each observer moment is weighted by a
factor that decreases exponentially with time. Hence we can predict by
the classic SSA argument the we should expect to find ourselves in a
period of time when the population outweighs previous epochs by at
least an exponential factor, and that future populations will grow
slower than exponentially.
However, all this seems irrelevant to the conscious amoeba question.
Cheers
A/Prof Russell Standish Director
High Performance Computing Support Unit, Phone 9385 6967, 8308 3119 (mobile)
UNSW SYDNEY 2052 Fax 9385 6965, 0425 253119 (")
Australia[EMAIL PROTECTED]
Room 2075, Red Centrehttp://parallel.hpc.unsw.edu.au/rks
International prefix +612, Interstate prefix 02
Re: relevant probability distribution
Russell wrote: > > I take "consciousness" to be that property essential for the operation > of the Anthropic Principle. The universe is the way it is because we > are here observing it as conscious beings. > > The first problem this raises is why does the anthropic principle > work? - one can conceive of being immersed in a virtual reality which > is totally inconsistent with our existence as conscious observers, for > example. > Tha must be explained by the unlikeliness of such a situation. Why would anyone simulate me living happily on the surface of Venus? They could have taken any possible person > However, let us accept the AP. After all, it has passed observational > test with flying colours. We should also expect that we should be an > example of the most likely form of consciousness. > > The second problem is raises is that if ameobae are conscious, then > why aren't we amoebae? There are many more amoebae on the planet than > there are human beings. I can well accept that dolphins and > chimpanzees (for instance) _could_ be conscious, since there are > vastly greater numbers of humans around today than there are of these > other species, but there is something special that we have that amoeba > (or even ants, lets say) don't have. > > Not sure about ant nests (Hofstadter style). Anyone got a good > estimate of the number of extant ant nests vis a vis human population? > > One possibility is that there is some kind of measure function that > rates our consciousness as far more likely to be occupied than an > amoeba's, however I'm personally sceptical of this. Consciousness seem > to be so much of an either/or thing... I think that one should first define oneself as a particular program, and then look at where and how often that program is actually running. Amoebas are incapable of running me. Maybe artificial intelligent agents pose more of a problem. Why am I not a robot, that can copy himself as many times as he pleases? B.t.w. Ken Olum made an interesting remark in his paper in which he advocates the Self Indicating Assumption (SIA) (see arxiv.org). If universes with more observers are more likely than universes with less observers, then why don't we live in a universe in which the number was pre-programmed to be some ridicolously large number, say N = 10^10? (Ken gave a different example). He concludes that apparently such universes must be unlikelier by a factor of at least N, to compensate for the factor N coming from the number of observers. This fits in nice with the idea that more complex programs should have lower measure. You can see that the measure of a program must decrease faster than 2^(-p) where p is the length of the program. Saibal
Re: relevant probability distribution
Saibal Mitra wrote:
>
> So, I am not saying that only certain programs are conscious and others not.
> I am really saying that if you the universe is running (in some
> approximation) a certain program in my head. That program defines me. If you
> run that program on a computer, that computer would have my consciousness,
> i.e. that computer would be me. Since most members of this list (except for
> Russell?) believe that our universe itself a program, you could say that in
> some sense it is conscious. Most of us think, however, that our universe's
> program is very simple. A retarded amoeba would probably be more
> intelligent.
>
I take "consciousness" to be that property essential for the operation
of the Anthropic Principle. The universe is the way it is because we
are here observing it as conscious beings.
The first problem this raises is why does the anthropic principle
work? - one can conceive of being immersed in a virtual reality which
is totally inconsistent with our existence as conscious observers, for
example.
However, let us accept the AP. After all, it has passed observational
test with flying colours. We should also expect that we should be an
example of the most likely form of consciousness.
The second problem is raises is that if ameobae are conscious, then
why aren't we amoebae? There are many more amoebae on the planet than
there are human beings. I can well accept that dolphins and
chimpanzees (for instance) _could_ be conscious, since there are
vastly greater numbers of humans around today than there are of these
other species, but there is something special that we have that amoeba
(or even ants, lets say) don't have.
Not sure about ant nests (Hofstadter style). Anyone got a good
estimate of the number of extant ant nests vis a vis human population?
One possibility is that there is some kind of measure function that
rates our consciousness as far more likely to be occupied than an
amoeba's, however I'm personally sceptical of this. Consciousness seem
to be so much of an either/or thing...
Cheers
A/Prof Russell Standish Director
High Performance Computing Support Unit, Phone 9385 6967, 8308 3119 (mobile)
UNSW SYDNEY 2052 Fax 9385 6965, 0425 253119 (")
Australia[EMAIL PROTECTED]
Room 2075, Red Centrehttp://parallel.hpc.unsw.edu.au/rks
International prefix +612, Interstate prefix 02
Re: relevant probability distribution
Hello Joe, > > I have difficulty with the concept of many distinct programs, each > representing an individual conscious entity. My understanding of modern physics > is that the concept of an isolated individual is essentially obsolete, in that > nothing can be defined without relation to everything else. As a result, surely > the underlying "program" for each must be similarly connected, so that in fact > an individual physical object is simply a concentration of processes operating > in one part of the program? Indeed. Our universe operates via simple laws, and, as a consequence, all the complexity that it generates must be highly compressible. That implies that everything becomes related at the macroscopic level. > The significance of this is that the paradox of intelligent objects > doesn't arise at all. I work on the assumption that your program is synonymous > with universal awareness (the abstract form of consciousness), and that > intelligence would be the result of local information-processing systems. > Partly because of the view of everything being inter-related, I'm uncomfortable > with a sharp, intelligent/non-intelligent distinction, and have no problem with > a mechanical object expressing a very low degree of "intelligence". Indeed, > anything which responds to stimuli could be seen in this way, including a rock > undergoing thermal expansion. However, an object can only become self-aware > once the processing centre is reasonably complex, and based on sufficient > local inputs to define a boundary to the region of the observer; this, I guess, > would be the manifestation of a closed (or at least self-referent) processing > loop within the program. > As I understand your view, it by-passes the paradox by introducing > arbitrariness, and any approach of this type seems to me to result in more > problems. At what point in evolution did an organism first become intelligent? > Do we then assume that a qualitatively different faculty was introduced? If so, > how? These sorts of questions seem to be the result of over-reductionism, of > separating gradations into artificial categories. (Of course, being a > palaeontologist, I spend much of my time doing just that, but never mind!) Some interesting thoughts here. Actually I don't think we disagree at all! I also think that one has to look at the system in question and look at the information processing capabilities. E.g. an ants nest could be found to be conscious... So, I am not saying that only certain programs are conscious and others not. I am really saying that if you the universe is running (in some approximation) a certain program in my head. That program defines me. If you run that program on a computer, that computer would have my consciousness, i.e. that computer would be me. Since most members of this list (except for Russell?) believe that our universe itself a program, you could say that in some sense it is conscious. Most of us think, however, that our universe's program is very simple. A retarded amoeba would probably be more intelligent. Now, one of the reason why I introduced these ''personal programs'' is that one needs an objective way to tell who is who. I will write more about this (in particular about how to resolve the quantum suicide problem) when I have more time (probably in a few weeks from now). Saibal
Re: relevant probability distribution
Hi Saibal, > > As I said I agree with you. But do you really mean a measure defined > > on a set of computer programs, or a set of computer program *states*? > I think that you can derive one from the other. I have thought about this > before, and I now think that the observer should associate himself with a > (to himself unknown) program, or better, a set of programs, that could > generate him. > > E.g. there exists a program that only calculates me and nothing else. This > program e.g. could compute me in an infinite dream. Many such (very complex) > programs must exist. I think that these programs define our identities (or > vice versa, but then not uniquely). Now, if conscious objects correspond to > programs then you don't have the paradox that any clock or lookup table has > intelligence. The fact that I don't live in my own personal universe, but > that my universe is generated by a simpler one, suggests that simpler > programs have larger probabilities. > > If you now have an a priory probability over the set of all programs, you > can compute (in principle) the probability that I will observe a certain > outcome if I perform a certain experiment. At least you can formulate this > question in a mathematical unambiguous way. I have difficulty with the concept of many distinct programs, each representing an individual conscious entity. My understanding of modern physics is that the concept of an isolated individual is essentially obsolete, in that nothing can be defined without relation to everything else. As a result, surely the underlying "program" for each must be similarly connected, so that in fact an individual physical object is simply a concentration of processes operating in one part of the program? The significance of this is that the paradox of intelligent objects doesn't arise at all. I work on the assumption that your program is synonymous with universal awareness (the abstract form of consciousness), and that intelligence would be the result of local information-processing systems. Partly because of the view of everything being inter-related, I'm uncomfortable with a sharp, intelligent/non-intelligent distinction, and have no problem with a mechanical object expressing a very low degree of "intelligence". Indeed, anything which responds to stimuli could be seen in this way, including a rock undergoing thermal expansion. However, an object can only become self-aware once the processing centre is reasonably complex, and based on sufficient local inputs to define a boundary to the region of the observer; this, I guess, would be the manifestation of a closed (or at least self-referent) processing loop within the program. As I understand your view, it by-passes the paradox by introducing arbitrariness, and any approach of this type seems to me to result in more problems. At what point in evolution did an organism first become intelligent? Do we then assume that a qualitatively different faculty was introduced? If so, how? These sorts of questions seem to be the result of over-reductionism, of separating gradations into artificial categories. (Of course, being a palaeontologist, I spend much of my time doing just that, but never mind!) All the best, Joe - Department of Earth Sciences University of Cambridge Downing Street Cambridge CB2 3EQ Phone: ( +44 ) 1223 333400 Fax: ( +44 ) 1223 333450
Re: relevant probability distribution
Hello Bruno: - Oorspronkelijk bericht - Van: "Bruno Marchal" <[EMAIL PROTECTED]> Aan: "Saibal Mitra" <[EMAIL PROTECTED]>; "everything" <[EMAIL PROTECTED]> Verzonden: dinsdag 4 juni 2002 19:50 Onderwerp: Re: JOINING posts > Hi Saibal, > > At 0:26 +0200 1/06/2002, Saibal Mitra wrote: > >I think that statistical physics, and especially renormalization group > >techniques, are essential if one wishes to derive the physics that we > >observe from abstract concepts like a measure defined on a set of computer > >programs. > > As I said I agree with you. But do you really mean a measure defined > on a set of computer programs, or a set of computer program *states*? I think that you can derive one from the other. I have thought about this before, and I now think that the observer should associate himself with a (to himself unknown) program, or better, a set of programs, that could generate him. E.g. there exists a program that only calculates me and nothing else. This program e.g. could compute me in an infinite dream. Many such (very complex) programs must exist. I think that these programs define our identities (or vice versa, but then not uniquely). Now, if conscious objects correspond to programs then you don't have the paradox that any clock or lookup table has intelligence. The fact that I don't live in my own personal universe, but that my universe is generated by a simpler one, suggests that simpler programs have larger probabilities. If you now have an a priory probability over the set of all programs, you can compute (in principle) the probability that I will observe a certain outcome if I perform a certain experiment. At least you can formulate this question in a mathematical unambiguous way. > In my setting it is the latter although it can and must eventually lead > to a measure on consistent *sequences* of computer program states. > (This is reminiscent of the passage made by Isham going from the Quantum > logics of states to the quantum logics of histories). > > The states must be considered also as seen by the machines themselves, > should I add. cf the 1-person/3-person distinction (which I attempt to > capture by the variant of self-reference logics. > > I said we meet but of course we are not yet at the Stanley-Livingston Junction, > but let us say we each begin to appear on our respective horizon :) Maybe, one can already try to do calculations without knowing the correct probability distribution. Many different models have the same critical behaviour. This is called universality. The effect arises, because different models at criticality can sometimes be mapped to the same fixed point under the renormalization group map. I think that to apply such techniques further assumptions need to be made. Like e.g. in statistical mechanics one makes the assumption that the system under investigation is ergodic. Regards, Saibal
