Interview with Sam Finn (by phone)

0:00 If we hit this, then it seems as if we are being recorded for posterity. Okay. And I'll just, well, I'll begin by saying that, as I mentioned, I'm interested in Wilson Matthews and their work on the Dreadstar Binders and the Star Crushing Effect and so on. And just to get right to the main point, so I don't take up too much of your time, one of the things that I'm interested in in particular is the fact that, well, how much of an effect their results, if true, would have on signal extraction for LIGO. So I know some of the people who have written about it, for instance, A&N and his paper and some other guys at Caltech, have emphasized that it would seem like if the star crushing effect, for instance, were true, or if their low-frequency last stable-circular orbit were true, that it would make signal extraction of signals from binaries very difficult. And then other people I've talked to are of the opinion that it wouldn't be such a factor. So I was kind of wondering if it would be useful to get a different viewpoint from you. Okay, well, my view is that it wouldn't be at all. It would not make signal extraction any more difficult for binaries. And the central reason there is that the star-crushing, let's go back for a moment, absent the star-crushing effect, so if it's just point masses and the like, then what makes signal extraction via match templates so nice and so powerful is that you know the waveform, meaning that you know the waveform and it has a very nice structure. And what makes it very, the fact that you get a lot of amplification is because there are a large number of cycles and you can track them precisely. And the reason that you can track them precisely in this particular case is because essentially the advance of phase is very predictable. It's essentially a power law. Okay.

2:30 The same thing holds if the star-crushing effect is true, because over the time scales that one is talking about, the signal being in LIGO's bandwidth, it is still the case that the advance of phase is a power law function of time. The only difference is that instead of the power law function of time being time to the eight-thirds or time to the minus eight-thirds, as it is when the energy loss is dominated by gravitational radiation, now it's a different index but for the same reason that you can get for example the chirp mass very very accurately you can determine that index extremely accurately and in point of fact you do in fact determine that power law index through filtering right um so and and this this was apparent uh to me immediately and at the the aspirin conf at the aspen conference a couple of years ago when this kind of first made it splash right um i i i made that point in the sessions and other people uh you might want to talk to bruce allen If you haven't, he had exactly the opposite viewpoint on this, and he felt it was an absolute disaster, if it were true. But the key thing there is that in this particular case, and for this particular problem, the star-crushing effect, to say that the star-crushing effect would make it incredibly more difficult to find binary and spiral through match filtering, my personal opinion is that that statement is just flat-out false. Okay. One thing that some people suggested to me was that the point at which this collapse of one of the stars occurred would probably be associated with emission of neutrinos and so on with a loss of mass by the body,

5:00 so that that would kind of change your chirp mass at a certain point in the signal. And then you'd have trouble with your phase evolution because the power law would change from that point so that to match you'd have to sort of guess the point correctly at which this would occur. Do you see that as a real problem? Well, there are two issues there. One is the point at which the crushing occurs. There's more than two issues. A second issue is whether or not there is any significant mass loss involved at that point because we're not talking about the kind of collapse in formation of a black hole that occurs, for example, in an increase-induced collapse or in a supernova gravitational core collapse. This is a very different effect, if it is true. And so it's not at all clear. The issue of a neutrino burst and the like is also, it's not as clear-cut, because you're starting with neutron stars. And so you already have, neutron stars are not transparent to neutrinos. There is already significant neutrino trapping in neutron stars. There's a diffusion time for neutron stars to pass through. And so then the issue is when you get to the point where the crushing becomes an unstable thing as opposed to a slow adiabatic crushing, it's at that point that you suddenly become concerned that, well, the power law is going to change because now there's suddenly a tremendous outflux of neutrinos. However, if those neutrinos are trapped on the time scale it takes to collapse to a black hole, then there isn't an outflux of neutrinos. So one, there is the issue of whether or not there is anything happening immediately prior to the collapse. Now, following the collapse, you have a different situation. Right. Following the collapse, you have two black holes, or you have one black hole and one neutron star.

7:30 Okay? And so following the collapse, it is certainly the case that there will be a different power law, because now the objects have changed, and to the extent that this is an effect associated with matter, Yeah. then uh now your evolution is going to be is going to be different i don't see the um the abrupt change in power law as being a problem um particularly if well it's not a problem if it happens outside the ligo band yeah sure okay and to remember where that is wilson uh matthews and Wilson talked about this happening in and around, if I recall right, basically above 400 hertz. Yeah. Okay, is that what you recall as well? Yeah, I think so, about that, 500 hertz or something, above 400, yeah. Okay, now, the LIGO bandwidth is, as far as binary in spiral are concerned, over 90% of the signal-to-noise comes from below about 250 hertz. So that means that your signal-to-noise ratio, in a squared sense now, so I'm talking about rho squared, so your amplitude squared signal-to-noise ratio achieves 90% of its maximum value. if you knew the waveform in complete detail, by the time the in-spiral reaches to about 200 to 250 hertz. So the Wilson and Matthews effect then, coming at above 400 hertz, and I don't know how much you get in that last 200 hertz, but I do know it's a very rapidly falling function at that point, is almost certainly going to be less than 5 percent and maybe less than 2 percent in the squared sense of your signal-to-noise ratio. So, yes, the power law break may potentially be a problem. However, it's a problem that amounts to only a couple of percent or a percent in the signal-to-noise ratio. And that's not a problem. There are so many other issues involved that if you're talking about a source

10:00 it's the difference between a percent and the signal-to-noise ratio between you deciding whether you've seen it or whether you haven't seen it, then no one's going to believe that kind of decision anyway. Sure. Okay. One, I think, that I had in mind is supposing, I guess this depends on people's view of what the state of the discussion is about both in Matthews and whether people believe them right now, but supposing, you know, when you actually have signals available to you and you're trying to extract them in the detectors online and so on, if you have a case like this where some people say, well, the signal ought to behave in such a way, and others say the signal ought to behave in a different way, the theorists are saying that, is that in itself liable to be a big problem? Because, after all, you have a lot to be taking account of already without different models. Okay, so... Run this by me again. The question just got a little long. Yeah. What I'm saying is, supposing, for instance, two or three years down the road, people are still arguing back and forth about, say, the Wilson-Matthews effect or something like it. And that means that when you're trying to do your signal extraction, you're having to keep in mind not only all of the possible variations of type of source and so on, but also the model for your particular source, so that the theorists have different models in mind for what the signal from exactly a given source would be. Is that liable to be such a severe problem that it would really be necessary to have just, you know, to take one model and run with it, whatever you think is the one? Or is it possible to sort of leave it up to trying to find what type of signal you can detect, as it were? The key question there is, in some sense, does the parameter space expand so much that the problem becomes computationally intractable? And a second related question is when you expand the parameter space,

12:30 of course you also then are trying a number of different things, and so your threshold is going to have to go up a little bit. Just because you're looking at noise in more ways, and one of those ways is much more, you know, if you increase the number of ways you look at it, you're much more likely to see something that's high. For the first, I don't think, so the first being the issue of the expansion of the parameter space becoming a computationally intractable problem. The thing that you're looking at is you're looking at that power law index. Okay. And if you, the ways in which people have counted the number of filters and counted the computational problem have been extremely naive. um you can that is the uh if you look at the the owen stuff yes the assumption there is made that you have to um you know you you have to have your filters extremely closely spaced lest you miss anything. But in fact, that's not the case. If you followed that argument, then you would say you could turn that argument around into an argument about discretization in time. And you would say that, well, in order to reproduce a sine wave of, you know, frequency 2 units per unit time, frequency 2 hertz, you would need to have a sample rate of, actually, in order to do, you would need to have many more than two points per cycle of time per cycle. Whereas, in fact, you know, you can work it out very easily, but, you know, Nyquist did it for us.

15:00 that you only need two points per cycle in order to reproduce any, you know, if you've got a sample rate of two hertz, you can reproduce anything whose frequency content is below a hertz. Right. Okay. So in order to reproduce then something that has spatial structure on spatial frequencies, below, and let's use chirp mass as a unit, spatial frequencies below, you know, one per one-hundredth of a solar mass, I only need a sample rate that's twice that. I don't need a sample rate that is several hundred times that, a spatial sample rate. And so I don't need to sample my parameter space that finely. So that's one thing, and that will buy you basically in each dimension of the parameter space in which Owen laid things out, that buys you several orders of magnitude in the number of filters that you need. So you can reduce the number of filters by several orders of magnitude just from that. The second thing that you can do, and this was shown by Girondar and a student of his Mohanty, and it's something that you can see intuitively as well. I mean, they worked through it and actually did the detailed numbers and said how to do it. But you can see intuitively why they were led to this. And that is, what if you did things coarse, even coarser than that, but lower your threshold? So admit many more false alarms. And then for each false alarm, just go more finely in that neighborhood. Right. Okay. And then with a much higher threshold. Okay. The advantage there is twofold. One, it's the computational advantage. You're not running a huge number of templates on all the data. You're running a much smaller number of templates and only refining in the neighborhood of suspicious areas.

17:30 And so the computational resources required are much lower. But the other thing that you gain is that your trials factor is much lower because the number of templates you're running is lower. And so, in fact, you can detect things that are slightly weaker this way because you haven't had to do as many trials, and so you are less concerned about, you have lower false alarm rate for this test. And so you can see further with fewer filters by doing a hierarchical search. And if you work those out, you find yourself, you know, you're no longer talking about 10 to the 5th or more filters. You're talking about many fewer filters that you're running. And so you can afford then, easily afford, to expand the parameter space because if you've decided that you can handle 10 to the 5th filters, but you only need 100 filters, well, so maybe you need to square that. Right. still a factor of 10 below the computational cost that you had originally budgeted. So it's for those reasons that I do not see that the expansion of the parameter space in this dimension would be a significant issue. So from your perspective, it would be fair to say then that you don't think that Wilson and Matthews the Wilson-Matthew results was a problem either from the point of view that if true, they would make life more difficult or from the point of view that if nobody was able to settle unambiguously one way or the other whether they're right or they're wrong it would still be sort of possible to continue on. I think it would be possible to continue on from a practical point of view.

20:00 You could definitely do the calculations and do the calculations that are wrong. You could analyze the data And in a way that was not prejudicial regarding their search. I mean, there's the other interesting question about when you want to talk about a threshold for detecting something. There's also kind of a subjective part of that threshold. The more uncertain you are about the character of the source, the more you want to push the threshold up, setting aside the issue of false alarms and false dismissals, because you want to nail down that uncertainty, and that's a separate issue. Sure. So kind of then related to what I was saying, as people do poor numerical work on binary systems and so on, do you kind of expect surprises that have been expected up to now of this type? but then is that something that one should be anticipating over the next few years as people do more numerical work on binaries? I mean, I hadn't expected the Wilson and Matthews stuff. It didn't, I mean, from the point of view of the analysis, it didn't disturb me after I'd sat and reflected on it for a while. The, but, you know, I hadn't expected, I thought that that was basically a settled issue, and now with ANA's, you know, kind of recent looking in detail, it looks like, in fact, that may have turned out to be a conceptual error that led to an incorrect calculation. Right. And so I still find myself back in the case that I don't expect for the purposes of LIGO for there to be any issue. There still remains the issue if you were talking about higher frequency detectors. So you were talking about detectors, for example, if the GRAIL project were to get built. Right. Okay, and do you know about GRAIL? Not much. I've heard the acronym before, but I can't remember anything about it. Okay, the GRAIL project is a proposal in the Netherlands to build a 100-metric ton copper-aluminum alloy sphere

22:30 that would operate at milli-kelvin temperatures as a classic acoustic resonant detector for gravitational waves. It would have actually an exceptionally wide bandwidth. It would be sitting down at around 500 hertz in terms of the kind of resonant frequency and the frequency where it's most sensitive, would be on the order of 100 hertz. Now, if you talk about binary and spiral for that detector, then you're in a very different regime. You're in a regime where the body-body interactions, where even kind of just sitting down with pencil and paper and doing the order of magnitude estimates, you can see the tidal effects start to become significant. If you do more detailed calculations like Dong Lai and Alan Weissman did, you see that there may be hydrodynamic effects that lead to instabilities so that the last stable circular orbit could conceivably, probably it's a bad phrase, But the last orbit could conceivably be in that regime, and so on. And so there's a regime where I think that there is a great deal of uncertainty. You're also in a regime where you know that the post-Newtonian approximation is almost certainly garbage. Or I shouldn't say garbage, but it's not going to provide you any guidance that you can rely on. and there are lots of ways that you could imagine evolution in that regime proceeding and those would have significant, those could all lead you in different directions for the issue of detection and recognizing what you detected. So there's a regime where I think there's much greater uncertainty

25:00 and where I kind of expect surprises simply because there's a great deal of uncertainty. And, you know, people kind of pick the story. When there's a great deal of uncertainty, they generally pick a story that they like or that feels plausible to them, and there's going to be surprise because it can only go one way. So there is a regime where I think that there's going to be some surprises. potential for a lot of surprise. Since you mentioned Anna's work, was there other work that kind of convinced you one way or the other about the likelihood of Wilson and Matthews being right, or was it only recently with, would say, just specifically Anna's work that she sort of became convinced? Aina's work is the first work, in my mind, that really said that there's something wrong with, or that said that there is, here is something concretely wrong. The previous work, and that needs to be fixed before we can go back and re-examine this question. work said that, well, it can't be a first post-Newtonian effect, or we don't see an effect like this when we have a test mass orbiting a supermassive black hole or things like that. So what they did, what the previous work tended to do, and this I felt was very unfair about the way it was phrased, was it said that the problem, Wilson and Matthews, or let me back up, Matthews and Wilson said we see a star-crushing effect. Okay. And then you had authors saying that they said they see a first post-Newtonian effect and we're showing that that first post-Newtonian effect doesn't exist. Or they say we're seeing this kind of effect, and we say it doesn't exist. What they said is that they see a star-crushing effect. And the central issue then is to determine whether or not that effect exists.

27:30 And you can do that by excluding all the possibilities, but you don't pick one possibility and then knock it down and say that you have shown that it doesn't work. Aina found that they'd used the wrong equations in one key place, and that that could lead to an effect whose magnitude scaled in the way that they saw it scale. Right. And so that's a very concrete thing that addresses the problem. Right. to a problem. And so, you know, prior to that, my own feeling was that it's interesting and it needs to be pursued because we need to know where it goes. And I still see that. I mean, I want to see them make that change. And I want to see them run again. And I want to know, is it there or is it gone? Right. Sure. And that's where I am on the issue. Right. Well, I was just going to quickly, and not to take up too much more of your time, since you mentioned about the meeting at Aspen, which may be in 97. Yeah, I think it was either 96 or 97. Yeah, probably 97, because this is 99 already. Right, yeah. So a number of people kind of talked to me about that, and people like Stu Shapiro and Alan Weidman, for instance, were saying that they were kind of influenced because here the experimentalists were listening to, LIGO people were listening to Wilson and Matthews and sort of going, oh, hey, what's up with this? And that they were kind of motivated to react to that. And I was wondering what your sense of the impact of their presentations at the meeting was. Well, it was really interesting because, I mean, I was aware on the periphery that they were doing this and they had been saying something. And I think a number of other people were, not necessarily the LIGO people, but other people on the theory side were kind of vaguely aware on the periphery that they were doing this.

30:00 but there had been no connect between their results and its impact on something like LIGO. And at the Aspen meeting, it was just, it was astonishing. I mean, you know, Jim gave his presentation, and it was like a dozen different light bulbs went on in the same room at once, just among a whole group of people about, you know, Jesus, this is important. And this can really affect things. Right. And what made it, you know, what aspect of the timing or the venue or the, you know, the talks that had gone before and the frame of mind that people were in contributed to that, I don't know. And it was very interesting to kind of see it and to be in the middle of it in that regard. The other thing that you raised was the issue of the LIGO people. I mean, they very quickly recognized the issue, and perhaps because a number of them are experimentalists and they don't have a particular, you know, theoretical animus or the same kind of... They're not wedded in the same ways to the, you know, kind of to the theory. Yeah, sure. They're much more, you know, very simplistically, but they see their enterprises as testing theories. Right. And so they're much more receptive to the idea and much more interested in the idea that the current wisdom, that there are alternatives to the current wisdom.

32:30 And so I think that that contributed to the reception that it received and to the interest as well. I wonder if you have a sense of something that interested me because a lot of the theorists that I talked to who were there one of the ways that they put the reception it received to me was well that the experimentalists LIGO, Virgo type of people who were there were all worried on hearing this result and saying oh this is going to be really bad and that you know theorists kind of then reacting to that At the same time, I haven't had a chance to talk to many experimentalists who were there, but one impression that I get then is more that, at least in hindsight, they weren't particularly perturbed in the sense of feeling like you, that this is going to be a disaster for LIGO. It's just an interesting thing that has to be taken account of. I wish you had been in my office when you said what the theorists said, because I was shaking my head. I would say that there's a little bit of rewriting of history there. The greatest opposition, the most violent opposition, the greatest doomsayers, were all theorists. the way you characterized the way the experimentalists you said perhaps in hindsight the way I remember it was that they were interested in it as, I'll use the phrase I used before as an alternative and this is interesting and it's important and we want to understand this you know, be able to take a look at it with the instrument that we're building, and so on. And it was the theorists who were, you know, gravely worried and talking about this is, you know, this is the end of our ability to detect anything, and so on. And all of that was coming from the theorists.

35:00 The theorists were the ones who were the greatest doomsayers. I don't recall any experimentalists at all saying that, oh, well, we may as well just hang it up. So as you would say, it was more the theorists that were talking to me that might have been looking back and rewriting it in hindsight. Yeah, you know, and I probably shouldn't say rewriting it in hindsight because that suggests a motive or an intent, but that's not how I saw it, and that's not how I remember it. Sure. Well, it's interesting to get that because I must say, insofar as I had an opinion of what I sensed was a sort of a slight difference in recollection between the people I talked to, It sounded to me like that psychologically I couldn't imagine a bunch of experimenters reacting that way to a disagreement between theorists. They're much healthier. But, yeah, I mean, and not all the theorists necessarily were saying that it was... Oh, no, no, no. Yeah, that's true. But the people who seemed to be the most perturbed about it were, in my recollection, they were all theorists. And there was a great disconnect between, there was a very widespread. There were people who didn't see that it was necessarily a problem. them, they saw it as an opportunity, and they were the experimental crowd, and then there were people who saw it as a, I think at a certain level it evolved into they saw it as a threat, although not at that meeting. Kind of a threat to the established structure that needs to be defended against. You mean more on the established theoretical picture, you mean? Yeah, that's right.

37:30 Well, that was certainly an aspect of that seemed to come through in what I was talking to people about, that they were saying, well, we have to answer this in terms of otherwise everything we said is maybe called into question or that sort of a feeling. And one thing that some of the theorists that I talked to did put to me was that maybe they put it in a way that, well, they put it anyway, that a lot of the experimentalists were reacting like, well, if this is bad for match filtering, then that's fine by us because we're not too keen on match filtering anyway. So I suppose that theorists who were talking to me were sort of saying, well that so we have to kind of defend this type of signal analysis as well that there was there was that although i tended to hear that a little bit later um i mean i think you know in it is it's definitely true that the experimental community by which now i'm gonna i'm gonna draw that has actually done some signal analysis in the past. And so it's not necessarily the high-energy community, but some of the people who have come from Stanford, people who do control systems and the like, the people who have Axion searches and radio astronomy, leery of matched filtering. The people who have had practical experience with it, with real instruments, are much more leery of matched filtering and are not particularly wedded to it as an analysis technique, as the theoretical community that has gotten itself involved in data analysis. I've only picked that up overtly after that, at least as far as I can remember. Although, in retrospect, I look at a lot of the things that have gone on, and I recognize that in a lot of cases, you know, they tended to sit politely and listen,

40:00 really listen. You know, we would get up and we would give our talks about, well, I showed this using matched filtering as a model and I showed that using matched filtering as a model and we could do the other thing using matched filtering as a model. And they would sit and they would listen politely and then they would go off and do their work as if we hadn't shown anything. And I understand this, you know, previously I didn't understand because I didn't understand why they were ignoring us to the degree that they were. And now I understand that they were ignoring us because they were looking at it from their viewpoint and their experience, and they're saying that this looks too much like a theoretical construction to me, or my experience is that this is not robust and that instruments rarely behave in a way that allows you to get any particular gain out of doing it this way and you need less sensitive methods that are more robust. And we'll worry about talking to the theorists and bringing them along when it's time, when we see a need. And so that's, it's been recently that I've understood some of, come to have that view and understood it that way. But at that time, I don't recall anyone on the experimental side expressing that, well, I'm not particularly keen on match filtering anyway, so that aspect doesn't bother me. I will tell you, someone you might want to talk to in this connection is Dan Debray at Stanford. Because he was at that meeting. He has a lot of experience in control systems and signal processing. He knows match filtering. He knows all of these other things. But he knows it from a real practical hardware point of view. He builds things and he makes them run. And he was, like I say, he was at that meeting, and he was listening very carefully.

42:30 Because I remember I was sitting next to him at one of the discussion sessions that we had in which this came up. This issue of the Matthews and Wilson came up and was a major, it basically took up the entire evening of roundtable discussion. And I don't remember him saying anything at all about that. And I was sitting next to him, and I was listening to him and trading comments with him. But a year later at Aspen, so now this would be the 98 Aspen meeting, I definitely remember him saying at a couple of points, there were one or two talks. Basically, there were no theorists at the 98 Aspen meeting. I was the only one. who could be counted in that group who was there. And there were a number of talks, there were still a number of talks on data analysis issues or the touched upon data analysis issues by some of the experimentalists. And they were saying, in talking to themselves, there were no theorists in the room, as I said, they were talking about analysis techniques that had nothing to do with match filtering and making comments about, you know, never particularly trusted match filtering or don't like that, and this is a different way, this is the way we looked at our Axion data, this is the way we looked at our this data, or that data. And Dan was encouraging that discussion and chiming in with his own views. And so that was the first time that I really saw it overtly stated at that meeting. But you might ask, because Dan was at the 97 meeting, and was listening to and participating at some level in the discussions, he's an experimentalist who you might talk to. That would be interesting, yeah, because it's something that I am interested in. So do you think that there's kind of a debate ahead that sort of still has to be made about the virtues of match filtering and how it can be used? I can't see clearly enough. I can't see clearly enough ahead. At a certain level, I don't think that the experimentalists really care to get into a debate over it. And so if the resources are available to do it, they're perfectly happy to let the theorists go ahead and do it.

45:00 I think where the real debate is going to occur is when someone says they saw something. Let me see, there are two places where a debate is going to occur. One is where someone says they saw something and it's match filtering. And then you're going to see the entire experimental crowd really bear down on how well is this characterized and have you taken account of this and have you taken account of that and you haven't looked at this and you haven't looked at that and they're really going to beat on it. And it wouldn't surprise me if they're going to demand that it also be seen by looking at the data in other ways. That wouldn't surprise me. The other place where I think you're going to struggle with the debate is people who are less wedded to match filtering coming from the experimental side are coming from the theory side and who are looking at things differently now, start to look at the data or want to look at the data in ways that are not matched filtering for some of these sources. And if they see something that is perhaps ambiguous in matched filtering or not there, then you're going to have a lot of the theorists very disturbed. the people who are wedded to match filtering because it wasn't seen in this way of looking at it and we know that this way of looking at it is optimal therefore if you saw it and we didn't you're doing something wrong and so I can see the possibility of that coming up yeah, kind of either way if one method gets it and another doesn't be an issue there well so to get back finally then to an aspect that kind of I guess we've been talking about how the experimentalists look at it, how the theorists look at it insofar as one can tell these things another thing that people I talked to spoke about was sort of the difference in outlook between

47:30 you know numerical people like Wilson and Matthews and say like most of the people at Caltech and how difficult it was to kind of well, you know, for them to really talk to each other without talking past each other sort of thing and I suppose maybe your perspective is one that you can kind of see both points of view maybe how much do you think that's a problem in an issue like this where a numerical result says this effect and then the analytic people are saying well okay but we're not able to come up with it with any of our methods that we're using right now and since the two types of ways of doing things are so different it's very difficult to compare across them so you know a lot of you know people kept saying to me well okay basically so-and-so just doesn't really understand what we're saying so that's all we can say is there any kind of way around that or is it just that has to be worked out? I mean, will it only be worked out sort of when you have more numerical people doing, replicating what Wilson Matthews did? Or is it possible for the analytic people or the numerical people to convince each other in your experience? I think it's possible. I think it's difficult. And I think, and here I have to say that the analytic people have been much more problematical in this regard. The analytic people, they don't have a lot of respect for someone who doesn't do analytical work. You know, there's this very clear sense in, hierarchy in the physics community theorists are on top and if you're not smart enough to become a theorist you become an experimentalist experimentalists are dumb right okay um and it's it's ingrained into you and that you know among the theorists the ones who do analytic work with paper and pencil are smarter than the ones who do numerical work

50:00 and so on. Now, and that makes it very difficult for these groups to communicate because particularly the analytic people, the people who place themselves on the top of the hierarchy, they doubt that you know what you're talking about, and so they don't want to spend a lot of time listening. and this is independent of the Matthews and Wilson stuff. Right. You see this, it's independent of the particular issues and the people and the personalities involved. Right. And so I think that it's always going to be a problem as long as that's the culture. It's a problem in the theorists talking to the experimentalists, the experimentalists don't sense this and resent it. And some really don't think that the numerical people don't sense it and resent it. And it leads to difficulties in that regard as well. My own experience in getting to know these different communities better is that, and I'll talk particularly about the experimental community, The experimentalists who I've met working on LIGO, you know, kind of the key people, the people like, well, let's set aside Ray Weiss because Ray is just, he's so far out on the tail of the distribution that he's just an anomaly in his command of things. But, you know, Peter Salson, Barry, Stan Whitcomb, Gary Sanders, and so on, they know a hell of a lot more about the theory than the theorists know and understand about the experiment. And similarly, the top-notch numerical people know a hell of a lot more about the nitty-gritties of the analytic work and in a lot of cases have a much greater physical understanding of it than the analytic people have of the numerical work.

52:30 But there is a kind of closed community, a kind of priesthood that has established itself, and that makes it very difficult to communicate, and it detracts then from the ability of the theoretical community to lend assistance in the numerical enterprises and in the experimental enterprise. And I don't see how that is going to be overcome while that culture is there. They can still talk to each other, though, because ultimately these things get resolved not on the basis of one community interacting with another community, but these things always get resolved on the basis of one or two people interacting with one or two people. And individuals make up a community, but they're not the community. So that's why I sense there's a problem, but that while I also sense that they can still talk to each other. Because when they do finally talk to each other, Aina has, you know, we'll look at, go back, let's not talk about Aina. Let's talk about some work that Stu Shapiro did. Right. Okay. All right, well, Stu's a slightly different case because he's both firmly in the numerical camp and firmly in the theoretical camp. Okay. But as a result, when he does something that says that, well, you know, maybe things are not so crazy with Matthews and Wilson, Okay, that gets the attention. That gets a different kind of attention in the analytic community and the theoretical community than if another numerical person or a person who is seen as more numerically inclined says that, well, you know, I've done the following calculation and I see a way in which maybe it's not so crazy. So the same work would be received in very different ways. Right.

55:00 Right. And so in your view, I mean, the largest part of the communication gap, or the most significant, rather, part of it is simply due to the fact that the top-down thing and the people who are more on top have no kind of motivation or desire to listen or express themselves. In the language appropriate to the other groups. They don't have the motivation to because they don't give the result. Unless the result confirms their prejudice, they don't believe it. Or they are much less likely to believe it. and they don't want to, in some sense, lower themselves to getting into the details of how this thing works numerically and what are some of the issues. There were some papers that attempted from a theoretical perspective to debug the code and to suggest, you know, where the errors were without looking at the equations that they were solving or anything like that, but setting up a toy problem and saying, well, you know, our toy problem suggests that they did the following thing wrong. And, you know, that's not a way in which you help. Right. Yeah. It's not probably going to win any friends. That's right. That's right. It doesn't move things forward. And that was the sense in which it, you know, that's the sense in which it kind of turns into a battle and the sense in which I was saying earlier that there was like a threat. And like some of the papers were as if you were repulsing an enemy. Mm-hmm. Okay. Well, that's a lot of help. Thank you very much, Sam. And I hope I haven't taken up too much of your time, but it was very interesting talking to you. Yeah. And I'll probably be calling you again at some point, or I might get a chance to actually drop by at some time during the year.

57:30 I'm still trying to figure out my travel plans, but I'm hoping to get to the U.S. sometime later in the year. Yeah, well, that would be great. Okay. State college is kind of out of the way, but if you have a reason to pass through D.C. or New York or Cornell or something like that, then, you know, absolutely. Not so far, yeah. We even set you up with a seminar. I'd be interested to hear about what you're up to and what you're doing. Okay. I understand you have a new baby. I do, yeah. He's just four months old today. Thank you very much. So, yeah, he's four months old today, and we're having a great time. Great. Yeah. Okay. All righty. So thanks a lot, Sam, and thanks for your time, and see you again soon, hopefully. Will do. Say hello to all my friends for me. I will. Okay. Bye-bye. Bye, Sam. Bye. Thank you.