Conference opening remarks (contd.)

0:00 Well, are a function of a position, therefore they commit to the position. Therefore, position should be, is a good candidate for such a point of observable. But before I get there, let's look at this whole thing that we've talked about so far from the point of view of information, because I think that's a very revealing way of looking at things. Let's look at the information transfer in course of the interaction between two quantum systems. First of all, I define the interaction in defined by Shannon, Boltzmann, whatever, that is, trace, minus trace, roll, log, roll. Question? You can, this part is certainly completed. You can add another thing. E prime, plus Hermitian constant. It seems like the underlying idea is that the environment makes the measurement better, and usually something like that can be found in mathematics to make it better, but only if the observing apparatus is set up in a correct way. Do you think you have a criterion that doesn't talk about position? Yes, yes. I've looked to see whether you could... Well, let's come back to it when I talk about position, because I won't be, I'll be taking, yeah, but let's see transparency. But I'm not hiding anything. One can add as you are perfectly correct. There's no information I should have added permission from you. And everything would be going on the way it is. Information. So information I define the way it's always defined. Minus trace rho of rho plus logarithm of the total number of dimensions of the Hilbert space. The information transfer is the thing that I want to talk about. The way it happens in this example that I talked about.

2:30 First of all, the subscript system is our spin, apparatus is our atom. Initially, the density matrix of the system and the density matrix of the apparatus before the interaction between the two took place are just projection operators. The information about each of them is complete. That's as much as you can pack of the information that two-dimensional Hilbert space, log of the number of dimensions. The total information in that system initially is just the traits of rho, log, rho of the whole joint density matrix and again you are in four-dimensional Hilbert space, you are dealing with a pure state in the four-dimensional Hilbert space, so you are going to expect log four after the dust settles and sure enough that's equal to the sum of the information in the two subsystems. Now, let's look at the correlated system apparatus state and let's see what happened to the information there. First of all, if we trace either of them, the density of matrix of each of them is a unit, proportion of unity. You have no information whatsoever about in which state does the apparatus of the system prefer to be. So the information about either of them is zero. There must be four bits of information hidden. Now where is this information? This information, of course, is in correlation information or mutual information.

5:00 That's precisely the difference between the knowledge about the fact that the whole system is still in a pure state and the knowledge about the fact that each of the subsystems is in a perfect mixture. Which does not have a corresponding feature in classical information theory, and it's quite, it's one of the counterintuitive features about non-separable correlations, is that one can pack all the information into the correlations and have no information whatever the subsystems. There is an inequality that one can prove in classical information theory which is violated by this statement. The pure state density matrix, the one that has all the information, has this form of diagonal terms, are as big as the diagonal terms, and that makes that form and that factor there as big makes it pure. There is a complete information present in it, state is still pure, but information about any of the subsystems is zero. In the language of information theory... Or just accounting for entropy. In order to accomplish the second stage of the measurement, we need to go from the pure state density matrix to the mixed state density matrix. So we somehow have to get rid of these off-diagonal terms. It's easy to see that if one just brute force throws them off the transparency, what will happen to the information in the process is we will have to get rid of log2. So, in order to accomplish the reduction of the wave path, the second stage of the measurement, we have to decrease the information in the system or about the system by appropriate amount. Now, to dwell a little bit on this, the way one is using the environment is to dump this excess information, the environment by becoming correlated.

7:30 With the already correlated system apparatus tandem, sacks up some of the excess information into this correlation, so afterwards when we trace out the environment, we have precisely the sort of density matrix we wanted to have in the first place. Now afterwards, if we recognize the final result, look at the apparatus, etc., our information will increase again by delta i for one bit. First, there is a correlation that happens without any change in information. Second, reduction of the weight factor. Information is being dumped. Third, we eventually come up and look at the apparatus, we gain the information. But the real problem is not this last recognition we think we know how to understand it, we've been dealing with these sorts of things in classical domain. We are just realizing what was the outcome. The real mystery is how to accomplish the second stage, how to get rid of this information. And what I'm proposing is you can dispose of it by letting extra system take the excess part out. Why do you call that spurious? What is spurious about? Spurious in the sense that it's uncomfortable. If there was, if there was... That's right. That's your own sense. Is the idea basically that the phased information runs off into... So far away and if I'm doing a local measurement, I can't measure the correlations having to do with that phasor information. So as far as the local measurement is concerned, it's just if a reduction is taking place. But is that the idea? Then you put it slightly differently but I think you've got it exactly on the mind. The idea is that you can do first stage of the measurement twice. And if you are not looking at whoever did the first stage of the measurement for the second time, it will keep some of the information, and for your practical purposes, you will not be able to recover it. The outcome of the measurement, as recorded in the preferred pointer basis, is going to spread because of the form of the interaction of gossip that goes out into the environment. As soon as this gossip has spread out, you've lost this crucial couple of bits of the information.

10:00 And the outcome of the measurement has become fixed, unless you are prepared to run out there and chase them out. There is nothing that can prevent you from doing this. Schrodinger equation is still valid. It's just that the place where they are hiding is somewhat unusual. Why do you lose just that much and no more? Is that put in by hand or you have a mechanism? Why doesn't it all leak out? Well, that's just due to this information, to this form of the interaction I'm talking about. But you put that in by hand. The interaction, yes, I've put in there by hand, but I better build my apparatus in such a way that it is built in by hand, because if it was sort of arbitrary interaction, then it would just mess up. In that sense, it's a reformulation of the problem, not a solution. Well, he's going to tell us why it is the position, I guess, that does this. If I understand you correctly, then I would think that you would be ending this talk by negating your zero pieces because the universe is the system that doesn't have enough space and environment. There's no place for the correlations to go up. It's information sink. And therefore precisely none of this could apply to a different part of the universe. You are perfectly right. Point number seven, the universe has no environment. So we clearly know about this. However, point number eight, some of the degrees of freedom inside the universe can use the others as the environment. You can't do it for the whole universe, no question about it. Now, one more thing that let me just allude to, more in passing rather than focus on, is that in a sense I think part of the story about second law and about entropy is buried in this discussion. I would call the information that's not in the systems but in the cracks between them, in the correlations, a lost information. And that lost information one could associate with entropy. In fact, if one looks at the so-called projection operator formalism of deriving master equations, etc., that's precisely what happens. One splits out dynamical irreversible equations in such a way that you... We choose deliberately to ignore the information about correlations. And that's not something that started with quantum mechanics. That's something that was done by Boltzmann when he derived the Boltzmann equation, basically.

12:30 You just focus yourself on the first stage of B-B-G-K-Y hierarchy and say, well, forget the rest. Wouldn't you have to prove that a Maxwell vehement couldn't be covered? If I'd have... I mean, except at the expense of more information for us. Wouldn't you have to do that? Let me make two points. I don't think it's a part of the story. I don't think it's a whole story. I think the other part of the story has to do with the fact that the information as defined here does not really always correspond to our intuition about what is random. I think there is another, and I maybe get to 15 minutes later on to talk about this. Information is not just related to probabilities. There is some other measure of randomness which has nothing to do with probabilities and has nothing to do with ensembles, which has to do with this. And there, Maxwell's demon enters again. Yeah, I agree with you. I mean, there's second law and so on. So... Patel, you have a little more time because you've been interrupted. Take another about 50 minutes more. Okay. A lot of discussion has been during your talk. So, one more point, which also has an information-theoretic flavor. Redundancy in amplification. Suppose instead of a single atom sitting up there on top, I put in 10 to the 23 of them. A number that usually sort of gets thrown around is one possible microscopy. They are all sitting, say, on top of the upper trajectory, and all of them do the same thing, that is, you know, get picked up. What's the difference now? I mean, I just changed the number, so there's no difference. In the basis, up-down, this wave function has just two components, okay? In other words, if I lose half of them and look at the other half, I still know what was the outcome. If I lose 99.9% of them and look at whatever's left over, I still know what was the outcome with 100% probability. Now suppose I try to do the trick of rotating, the Einstein-Propagandized trick, into a different basis. Well, nothing can prevent me from doing it. So here I do it, I substitute the other bases for the atoms, and I've done it. And by going instead of two components of the wave function, I end up with four for three atoms, for three atoms.

15:00 Now the distinguishing feature about being correlated with the state in versus state out, actually took me sort of 15 minutes to realize what was going on, whether the number of minuses is even or odd. So, if I have 10 to the 23 of these guys, and I lose one of them, I'm done. I mean, that's it. No recovery. And basically, to come back to the business of environment, there's really no distinction between this information spreading this way, between plenty of different systems, or spreading that way, between plenty of different systems. It's more which information has gotten spread. It has gotten recorded redundantly in the environment. So you're saying that a macroscopic detector would take the place of your sort of two-state detector plus the entire one? Yes. It will choose what's a preferred basis in an identical way. I mean, if you don't care about the interactions but only about the final state, you wouldn't be able to tell one from the other. The point is the same. You lose some of the stuff, you know, on any other basis, but the preferred one was done for. So, the redundancy, it's again, the story comes down to information. It's crucial. And it's the same redundancy that was here. It's the information that was leaked from the apparatus, about a specific observable of the apparatus, the classical observable of the apparatus, is the information that we call the redundancy, the gossip spreads, systems interact with each other, that's the information that was leaked, therefore that's the only information that can keep us ready. So much for the information theoretic part, let me get back to the environment business and let's stop talking about the apparatus. We know we are not really getting to the apparatus, we are really getting with the business of what's classical, what's quantum. So let's talk about position as a pointer.

17:30 Let's take a quantum particle of mass m and let's choose a harmonic oscillator environment, that's the only one we can calculate. Let's not delude ourselves. The total Hamiltonian, let's write it down, is the Hamiltonian of the harmonic oscillator of the single particle, is a coupling Hamiltonian, x pointer variable of the particle coupled to position variables or whatever. Now, if you want to model the real universe, you would presumably want to get, after the dust settles and you've somehow arrived at something that looks like classical. An equation which looks like a Langevin equation for a classical Brownian particle. Okay, so something that is being kicked around by a fluctuating force which has a delta function correlation in an appropriate limit and something which moves in a potential. I shouldn't say a general potential because the only place where I can derive it fairly convincingly is a kind of harmonic oscillator. We place around this viscosity coefficient, so if you push something it will keep on slowing down, and let's just, yeah, that's the normal. This condition is sort of very obscure, but let's not worry too much about it. And let me also say that the standard method of dealing with this problem is to take whatever you have here, a bunch of harmonic oscillators, And try to somehow derive an effective master equation, as it's called, for the particle in the environment. By the way, how do we know what is the system and what are the harmonic oscillators? You have a totally quadratic Hamiltonian and you can decompose it into harmonic modes.

20:00 And then the question is, why isn't the x that the measurement is made on rather than one of those normal modes which behave totally as three systems? How do we know what is the system? Because we can form generalised coordinates on our configuration space and think about the system as something described by a different generalised coordinate. The brutal answer is, I don't know. I'm not addressing it. On the other hand, I know that Dave Deutsch, and I hope that Don Page may be able to comment on this a little bit, Without imposing the vision into the system right from the very beginning, we spent with Dave a couple of weeks trying to do his formalism for two spins and then find out what are the systems that we never succeeded in because it leads to such terrible linear equations right off the bat that one is sunk from the very beginning. So to try to understand what's going on, I start with this. But I appreciate your question. I mean, it's really looming there. So there is some a priori input which you are putting in, and that a priori input is what are the systems? Exactly. Yes. I guess my question is just a variant of that. Why would you expect a classical number to get a viscosity out of this when you start with a perfectly versatile system? Oh, because I would like to... I don't expect to. I would like to. I know that particles are like this. And it so happens that if I define my coefficients in this Hamiltonian correctly, I do get it. Yes, that's right. Thank you. Incidentally, Bill and I are involved for years in trying to do this calculation in a more rigorous way than was done so far and using a quantum field as the environment. I think part of the problem is that we like to visit each other so much that we don't want the project to end. So what happens now? One starts with this Hamiltonian. Everything is perfectly reversible. There's no reason for anything irreversible to happen. What one is doing now is one starts with a definite initial state. That is, one plugs in a particle which is initially not correlated with the environment.

22:30 A bunch of harmonic oscillators and one looks at what happens. Okay? And one looks at what happens by throwing away at some stage the state of the environment. That can be done in many different ways. Usually it's done sort of right off the bat and without much ado. And what one ends up with after the draft settles is something that's called master equation for a density matrix particle. So here is a master equation. The master equation has a couple of components. The top line is all really the initial Hamiltonian of the harmonic oscillator. I mean, that's a, that's a, top line is a von Neumann equation. But there are two additions, two additions which are not time-reversible. And they are not time-reversible not because the underlying dynamic is not time-reversible, but because we've chosen to ignore. Rather we've chosen to choose a certain specific initial state, okay? So these two extra components are one of them responsible for viscosity. Eta is the viscosity coefficient. Eta divided by 2m will give you the inverse of the relaxation time. If you push the particle it's going to slow down and eta over 2m decides whether the hole will be a screeching hole or it's going to take forever to plow into the medium. Which has the same parameter, surprisingly enough, as eta, sorry, as a viscosity, the same parameter eta, plus temperature, plus Planck's constant, and this piece takes care of the off-diagonal correlations, off-diagonal in the basis, in the position basis. So, to look at it piecewise again, first of all, in classical limits... We would get from this equation, if we set potential, we have to take a free particle, okay, so forget about the potential, the simplest things. If we take a classical particle, we would get slowing down, the particle is slowing down with the rate given by minus gamma. But the real thing that we would like to worry about and understand is what happens if we put this particle in superposition in a non-classical state, superposition of two gaussians but which are separated by a certain distance, zeta.

25:00 So we know how to write a wave function of this gadget. We know how to write the density matrix given the wave function. One and two are these two gaussians. Everything we are dealing with is a superposition. The density matrix has four peaks, two of them on the diagonal and two of them on the corresponding corners of the square in the x-prime representation. And now we can look at the rate, well now we can get rid of the trivial time dependence, that is move to something like, I don't know, Tomonaga, it's called representation. And just calculate the rate at which the diagonal and off-diagonal terms evolve. Am I being very... No, no. And one more thing that I would like to stress is that this number that appears here, kT over h bar squared, is one of the many definitions of thermal de Broglie wavelengths. So it's a number that we know and love from the first year of statistical mechanics. So what happens now if we do the calculation? If we do the calculation, or let me put the equation for our answer here, if we do the calculation, what will happen is that the rate at which this off-diagonal coefficient is going to evolve will depend on the separation of the wave packets. So if the dust settles, when the dust settles and you just look at the rates at which diagonal and off-diagonal coefficients evolve, you come to the conclusion that you can express the de-correlation rate in terms of the relaxation rate. Now that's a very interesting conclusion because the standard argument...

27:30 The only thing used by people who are talking about isolated, closed quantum systems is, well, look, I mean, the system is not relaxing. We know that the telltale sign of being coupled to the environment is some sort of relaxation, so, you know, if it's not relaxing on any appreciable timescale, it's being coupled to the environment, therefore forget about any appeal to the environment for the solution of your classical quantum dichotomy. And the point about these two rates is that the correlation rate or the coherence rate is given by the relaxation rate, but multiplied by the difference in locations of these two wave facets divided by the thermal de Broglie wavelengths. Now to give you a number, something classical, let's take mass of this thing that we were dealing with to be one gram. Let's take the separation to be one centimeter. First of all, you haven't given us the expression. Yeah, no, eta is, yeah, eta is . I know, but there was a Hamiltonian you wrote down. It's the sum of the coupling squared times the weight factor omega. You like superpositions. Okay, eta shows up here on the bottom. And it's given in terms of the, it's given in terms of this Hamiltonian and this coupling. Well, this is a connection. I don't have a system yet. Sorry, yes. But there is one that I want to write down. Let me put it differently. This is a better way. If you have a fluctuating force, a fluctuating force is given by this. I think that's physically more relevant than this down here. You're putting in numbers now. It seems to be you need it.

30:00 No, I don't. Well, I'm not going to put these numbers. I'm going to put different numbers. Put your worries aside. Okay? And all I worry about is how much this relaxation rate and the correlation rate differ. Okay? So, suppose I have a mass of one gram, separation of one centimeter, classical but not excessive, and room temperature to be comfortable, which gives me the probability. Okay? Between the relaxation and the correlation timescales comes out to be 10 to the 27. In other words, if I make sure that my system is really isolated, relaxes on a Hubble timescale, I can't do any better. Then the correlation will happen on a timescale of 10 to the minus 10 seconds. It's a disparity between these two timescales. Sure, if I could isolate it better than that and be comfortable with it, I could start going on Schrodinger cats and stuff. And it's more this coefficient that shows up here than anything else. What is the remark about the cat? If I had a Schrodinger cat, which if I tap on the ear will ring for a Hubble time, then I could... You have to tap on the ear. Cat with a quality coefficient, that's what it's called, in Weber bars. Now, if you hit the Weber bar in order for it to be a good gravity detector and not introduce correlation with the environment, it has to have a quality coefficient of 10 to the 9, which means it has to go to 10 to the 9 oscillations. This is coupled with a Q. Oh, I see. What is the relevance of this? No, it's a question of when can you think that the system is isolated. How well would you have, how far would you have to push the relaxation rate down in order for it to be isolated for the purpose of this argument from the environment. Okay, so let me, let's see. Now one more thing about the...

32:30 One more thing about the equations that I wrote down as a master equation is written by brute force method by saying that the temperature is bigger than any other quantity that shows up on the same page of the derivation. And the one worry that one had is that one put a bit more dynamics. Some of the quantum paradoxes might come up and threaten us again, and this was the motivation behind what Beranu and I were doing, and our conclusion is that, if anything, the situation looks even more hopeless from the point of view of Being able to detect traces of quantum behavior if one puts in an exact treatment, because basically instead of being hit steadily by a constant viscosity coefficient, one gets zapped right away with a delta function looking thing, which saps the correlations on a fine scale comparable to the cutoff, upper frequency cutoff, the expressions for viscosity and such are divergent at the upper frequency. Yeah, which hits the system with the cut of, with the energy corresponding to the cut of coefficient, basically gets rid of all the correlation on a comparable time scale. So, let me see, two more transparencies, very brief, okay? First, a very brief summary. Amplification, spreading of correlations are the same thing, the resulting point is . Stage two is, in a sense, stage one to some power. Decoration timescale for any sensible system which we would care to call classical are so extremely short compared with relaxation timescales that we don't have to worry about them. Now, French primitivist summary, we call Le Douanier, the French primitivist painter. Classical domain, quantum domain, division between the two. Bohr's idea was that classical domain, well, we know what things are there, most of them are no surprise to us, sun, planets, us, sometimes there is a surprise, quantum fluids. Quantum domain, electrons, photons, atoms, sometimes a one-ton thing like a Weber bar flips in. Quantum bill of rights, interfere if you can.

35:00 Classical law and order do not interfere. Newton's law, second law of thermodynamics, and what we are really talking about is a border territory. And what defines the bonus? About ten minutes for... Well, this is one of the way of looking at it. What I would say is, if I can justify a density matrix before I look, a density matrix which has lost some information, before I look at it, then I can say that the system is ordered in a well-defined state and I'm just acquiring information about it. Now, if I go and, you know, seriously look at all the steps... What I would be saying is I happen to be in the branch of the universe, etc., etc., which contains this victim to be in such and such place. I don't think I can do it without talking about it this way. I'm uncomfortable talking about it this way. Because I think that to some degree the answer is too simple. I think there is a bigger message there. It might be right. I mean, I have no objection, you know, in principle. But I have a feeling that the problem is a hint about a bigger solution than just his answer. I think you're wrong. I have no argument beyond that. So, you know. This is coming back to your zero principle. Yes, yes, yes. And to the fact that we should agree or not. Can I remind people to please make your name? People who haven't had a chance to talk yet. Don't make me, could have a one minute, please, yes, yes, don't, I just want to say that there are... I think Wojtek has this very interesting proposal for how to define sort of a preferred basis. And just to say that, you know, there were a couple of competing ones. Well, David Deutsch had one in which he attempted to define how you divide up the whole thing between the two systems. And that got rather complicated. I guess about all, I'm trying to see, I'm not even sure I remember the details of it, but roughly speaking, it was if you had some apparatus with a finite dimension, and suppose you said we're going to say the apparatus has one dimension,

37:30 And the other system has another dimension. Now we want to break up the full system such that if you took the full density matrix, you traced it over one thing to get the density matrix of the first object, you traced it over the first object to get the density matrix of the second object, took the outer product of these two and said that was your density, I mean that's not the density matrix, but if you took that and now evolved this infinitesimally with the Hamiltonian, would it remain in this diagonal form? And that was his assumption that you make the choice between dividing up the system into the two parts such that if you throw away the correlations, then under an infinitesimal evolution, the correlations wouldn't come back. Now, okay, as Wojtek said, that's complicated to implement, and I'm not too sure, you know, about that. But once you've had that division, then... Then David Deutch's definition of the preferred basis was that the preferred basis just diagonalizes the density matrix of your apparatus, so that was his thing. Now, Wojtek's is the one that such that it commutes with this interaction Hamiltonian. I came up with a third which I call information basis which is basically then when Wojtek summarized in his this redundancy thing well okay so this is basic I didn't realize what he is so this is the same thing as what he had if you just make a measurement where you couple the system well I call it an instrument just for technical differences If you couple it with one instrument, then you get this thing, and the density matrix is that, and of course if A is equal to B, then it's, you know, purely diagonal, there's no preferred basis. But whereas if you couple it to two parts, if the instrument has two parts, then you couple it, and then the density matrix for the instrument has this form, and in this preferred basis, it's only got two terms. Whereas if you use the different faces for these, I mean, if you had the pointing in something, well, pointing in and out, in the paper and out of the paper, then there would be four terms there. Okay, now just to come up with an ad hoc definition of a preferred thing, I said, well, suppose at the end you have your instrument and you choose, it has several subsystems, maybe n equals 10 to the 23, now you choose the basis of the subsystem just to minimize what I call a diagonal entropy, that is, you throw away the off-diagonal terms of the density matrix in whatever basis I choose, and then I just calculate the entropy of that, and then I'll minimize that, or in other words, maximize the information.

40:00 Okay, that's just an ad hoc definition in order to say what this information basis is. I suppose the physics or the motivation for saying this might be a good basis is that if I look around the room and I see somebody, presumably where I see him is going to get stored in one memory and then I look just a little later and it's going to get stored in another memory and another memory and another memory. And whereas if I put it in one memory, I might be able to store a superposition of where he is. I put it in another memory, maybe I could, but that will, I will get a whole lot more terms and I will get a whole lot of inconsistent terms if I tried to store the information about something other than roughly the position of somebody. So, in other words, It's just a conjecture, and psychologists would have to analyze this to see what makes sense, that our brains can only be conscious if the different memory units in our brain have agreement, because we're sort of checking back and forth between the different memory units. And therefore, in some sense, eigenstates of consciousness are going to be states in which there's more or less agreement between these different memory units, and that's the sense in which the eigenstates of consciousness may be somewhat close to this information basis. But I could conclude by saying, in principle, I think in the Everett interpretation, one does not need any preferred bases. People often say the Everett interpretation is the one where you do need preferred bases to say what the worlds are, but I would say that the Everett interpretation, you're not, you're never collapsing the wave function, so you don't have to have preferred bases. It's just like coordinate system in general relativity. But there could be some coordinate systems that are preferred to make it easier. And if you use something like the information basis or the basis of eigenstates of your consciousness, it may be much more easy to then, you know, answer conditional probability questions. I mean, well, okay, this is standard. Maybe I don't leave this up. But I mean, I think everything, if you have an NC matrix, everything you ought to be able to calculate by... There are many conditional probabilities where you project the density matrix onto the condition and then project it onto the outcome and I think in as much as you're allowed to ask any conditional probability statement, you ought to be, there's not any preferred basis, but there may be ones, some of these questions are of more interest to us and in that sense there may be preferred basis because the basis of eigenstates of B, for example, might be preferred.

42:30 Definitions, if you want to get a precise definition of what a preferred basis is, I think there's no absolutely preferred basis, but some of these are useful for different things. And then the point would be, even though they have different, the precise definition is different. Deutch is saying the density matrix is diagonal. Wojtek is saying it commutes with the interaction Hamiltonian. Mine is saying that you minimize the diagonal entropy. The point is, those should all roughly agree for a good instrument. Then if it's not going to work, maybe it won't agree. Do you see any problems left with the interpretation of quantum mechanics? I wouldn't if I were more complacent about adopting my own interpretation. I think we can calculate a lot in the way we can. There are no problems anymore with how the universe splits into branches, etc. I think there is probably quite a bit more that's going to emerge from a better understanding of the role of information analysis. I've sort of formulated things in terms of the information and I have not, but I don't think it's a final word. I think, let me put it this way. If we, if one, one can be happy if one adopts the new world interpretation and not worry about anything more. But I think there is going to be a bigger message and I think it's a place where it's going to emerge and the way it's going to modify our view of the business will come from the information side and from the entropy side and from the relation to the second law. I don't think it's incomplete. I think it's unsatisfying. And unsatisfying is a personal statement. Ok, so it was a personal question. Well, I think... My feeling is that... First of all, my feeling is that if we finally understand what quantum gravity is really all about, it's gonna modify both. Quantum gravity and... both gravity and... I have no, please don't ask me for justification, but I have a feeling that perhaps one can gain an insight on how it can happen or what are the things that are going to give by trying to understand better the role of information and pops up in so many, I'm not being, I'm not being fair. I'm, I'm, you know, rambling and headway.

45:00 But isn't what Carroll said a problem? I mean, after all, he's discontinuing. After all, when we look at the chairs, the old tech won't like to say, why do we see the chairs? Why do we see them? No, no, I mean, why the position of observables is a good observable? That's what Polchek would like to use. But he explained it, provided we already singled out the position x among all possible, as an observable in the configuration space. And then... Everything is okay, but why the world is singled out at the first place? Why don't you know it already? Oh, he doesn't know that? I thought he didn't know that. I thought that's what... I think the principle of division of systems is somewhat up here. But let me put it somewhat differently. Let me come back to Jim's question. Let me put it somewhat differently. I'm somehow uncomfortable with the, you know, assertions that we are always going to perceive one branch. I mean, why are we going to always perceive one branch of the universe? I think the answer to some degree is... First of all, the problem is about why does the universe look the way it does when we look at it, in spite of the fact that there is all these other branches all over the place. Why are we generally only perceiving all in one branch? Why do you say one? Why do I say one? Why one?