Interview with Sam Finn

0:00 So, it's 5 p.m. of the Tricord de March, and I'm talking about San Fain. A portable thing is buttons get pressed that shouldn't be pressed, so that turned the microphone sensor way down low, so you don't pick up much of anything at all. Well let me see, the questions I was going to ask, well we kind of touched on some of the areas that I'm interested in, you know, data analysis and numerical relativity, both of which I wanted to bring up with you, but maybe it's best to start off by just asking you what brings you personally to working on gravitational waves and what brought you Well, I actually, that goes all the way back to when I was a grad student, and that was, I started as a grad student at Caltech in 82, and I had, before that, I had pretty well known that I was interested in working on gravity, kind of as an undergraduate. and was interested in astrophysical aspects of gravity. Places where gravity plays in a important role, but in astrophysics, I never really had an interest in quantum gravity or things like that. And when I interviewed as a graduate student here at Caltech, I ended up spending the day with Stan Whitcomb and Ron Drever, Bob Sparrow, Dana Anderson, pretty sure Dana, and Jekta Gersel, and Carl Caves, who was working with Kip at the time, but was really kind of doing gravity wave stuff, squeeze state stuff, and the like. And I was very intrigued with this notion of building a gravitational wave detector, a large interferometric gravitational wave detector.

2:30 It was not something that I ever thought of doing experimentally, because just my interests don't run there. I have great respect for the craft, and I've played with it at times, but I love listening to good classical music, but I'm not interested in playing. I mean, is there just some things you want to do and some things you don't? And as a graduate student working with Kip, Kip basically kind of points you in a direction, and he wasn't pointing anyone in that direction at that time. LIGO didn't really exist. But I always had in the back of my mind that, you know, gee, here is this really interesting thing, And as soon as I'm in a position where I can kind of pick my own problems, that I wanted to get involved with this in some ways. And I also had in mind the thought that this was kind of a future for the field, which is kind of in the back of my mind, with just, you know, kind of career prospects, and it just seemed to me that this was a future for the field, and this would be a place where, in addition to it being very interesting to me, it would also look like a place where I might hope to be able to find a job and continue doing what I was interested in. The thesis work that I did was on pulsations of relativistic stars, neutron star pulsations, quasi-normal modes of neutron stars, and the like. As I finished that up, it was very fortunate to me that Chuck Evans, who's now in North Carolina, he was a graduate student of Richard Mansner's, had worked with Jim Wilson at Livermore while while he was a post-doc of Larry Smars in Illinois. He was kind of Larry's post-doc, but I don't think that he ever actually lived in Illinois. He lived up at Livermore working with Jim. He had started here as a post-doc. John Hawley, also a numerical astrophysicist, who was also a graduate student of Larry Smars, who had been here for a year as a post-doc also.

5:00 and I ended up spending a lot of time hanging out with them and picking up kind of tricks of the numerical trade and kind of took that and for kind of the first thing that I was able to kind of as I was finishing up my thesis work and looking around for things that I could choose to do now I looked at the problem of supernovae which were at that time still being bandied around significant source of gravitational radiation. And there was a number of problems with the calculations, incidental to the problems of actually kind of... Well, I draw a distinction between the supernova problem, and now I'm going to use supernova in the technical sense that an astronomer does, which is something that relativists never do. And that is the light display. That's the light show. And then the core which is where the gravitational radiation comes from. And regardless of the problems that there are in getting the supernova, the results, if you look carefully at the numerical results of the core collapse simulations and the gravitational radiation from those, it was clear that there had to be some very large and difficult to quantify numerical errors in the calculation of the gravitational radiation. Not necessarily with the evolution of the hydrodynamics, but with the calculation of the radiation. And so I set about to try to clean that up and to do a better job of estimating the radiation from supernovae. And I took that problem with me to Cornell, which is where I went as a postdoc, working with Saul Tchaikovsky. And actually, of all the people at Cornell, Saul is the only one I never wrote a paper with. But I took that problem with me and basically working somewhat with Chuck Evans and found some new ways of dealing with the problems that in fact showed that the radiation, what had previously been thought to be 10 to the minus 4 of a solar mass being converted

7:30 into gravitational radiation was actually 10 to the minus 8 and that everything else was in fact numerical noise being amplified by the quadrupole formula. And so that was kind of My first project that was, in some sense, directly aimed at anticipating detection of gravitational waves, because that was my motivation for it, and that was what was targeted for it. In fact, in that paper, I put things in terms of signal-to-noise ratio in a proposed detector and tried to discuss it in those contexts and in those terms. So that was kind of how I ended up in the area. And that was still at a point when very, very few people were at all. I mean, it was Kip and Bernie and experimenters, and then I was kind of sticking my nose into the tent and looking for a place where I could sit. So this was a sit? Yeah, this was, I finished, I started working on the supernova problem probably around, it was December 86, January 87. It was right in there. I remember that because Saul offered me the postdoc position in, it was basically December, he called me. In fact, it was Christmas Day, he found me in the office, I think that's why he offered me the position. Yeah, that was Christmas Day in 86, and I started there in basically end of August of 87. And I had started working, just at that point I was starting to work on the supernova problem. There had been a seminar at Santa Barbara associated with or a conference or something that Doug Erdely had put together to address numerical problems, numerical hydrodynamics problems and problems with evolving these systems.

10:00 not from the point of view of gravitational radiation or anything, but from the point of view of making better estimates of accretion problems and the like. And people were talking about moving to 3D, even though they barely mastered 1D and were still struggling with 2. And I went there, along with Chuck and John, and was listening there and thinking about the supernova problem. In fact, it was at that conference that I kind of, in listening to people talk and thinking about different numerical things, I kind of formulated the problem and had my first ideas about what the numerical issues might be there. So it was right in that time frame. And the work probably, the work was finished in 89, roughly, and talked about it in several places in 90, 89 and 90. Actually, summer of 87, actually, was the first place I talked about it. It was at Bernie's Gravitational Wave. There was a workshop in Cardiff in July of 87. I don't have the book here it was a NATO ASI meeting in Cardiff in July of 87 that was the first conference where people got together and talked about gravitational wave data analysis, the first kind of primitive gathering and discussions going on. And so now I can remember some other people who were kind of milling around the field at that time. Massimo Tinto, who was a Bernie student. Jeff Livis, who was a student

12:30 of Ray's, Ray Weiss, and Andre Krolak. That was the first time I'd met Andre, and he was working with Bernie at the time on binary and spiral problems. At that time, were supernovas still the most important? They were not the favorite source at all. A lot of people, I mean, I wasn't unique in viewing the numbers with suspicion. And it was because of the suspicion that people had that they were casting around for sources for which that suspicion did not exist, because it was felt, and probably rightly so, money, you want to put your best cards down on the table. And supernovae just did not feel, there were just too many uncertainties about them. Right. So at what point did, to your mind, did binary neutrons to our coalescences become the sort of solution to that problem? Oh, I would say probably somewhere in around 85, I would say. I would say somewhere in around 85. I'd say that based upon the LIGO proposal certainly cited them, and that was right in there maybe a year after that, It meant that people were thinking about it. Bernie was clearly spending time on the binary and spiral problem, thinking deeply about it, because it was, I think, in 86 that he discussed the cosmological implications of observing binary and spiral and gravitational waves. And, I mean, so just based on those two things, wasn't really, Kip did not have any discussions about these issues in his, among his students

15:00 at that time. The weekly meetings and things, they didn't exist. That was something that started in somewhere around, I think somewhere around March or April of 80, 86, he started and meet with the students in a group. We kind of met with him irregularity, catch as catch can. In some cases, on average, I probably saw him for about an hour every two months as a graduate student. It's very, very different than the way he runs things now. But he didn't have any of his students working on anything related to LIGO at that point. So it was more that the individual students were working on? Well, he had some people working on, let me see, he had some people who were finishing up work related to the membrane paradigm stuff, Wymo Swen, Doug McDonald, Ian Redmount. Ian was also working kind of at the edges of the quantum field theory in curved space. He was working both with Kip and with Steve Frouchi, in addition to doing membrane paradigm stuff. Yekta had just finished up, and he had done some very nice stuff on multipole analyses and showing the equivalence of many different kinds of multipole analyses, including Kip's Symmetric Trace-Free Tensor Formulation. He was starting... Those were kind of the students ahead of me, the students who were behind me. he started working on more kind of quantum problems that was at the same time he got started on this wormhole business and Mike Morris was one or two years behind me who he co-authored the first wormhole

17:30 paper with And that kind of led into a whole several-year thing where he had a number of students and postdocs working on classical and quantum issues related to wormholes and the like. But his tradition has always been to kind of pick problems for his students. as students become more advanced they can come to him and work on something or the like but he likes to guide his students more directly so after working on the supernova that was sort of your entree that's right yeah that was kind of the first thing that I did there so I suppose by that stage interested both in the sort of numerical relativity stuff. That's right. That's right. And I became known for one reason or another as someone who was good with computers and numerical relativity, even though this was really just a quadruple formula type of thing. So, just as I was finishing up at Cornell and getting ready to move to Northwestern, Bush and Congress decided that they wanted to do something to really push high-performance computing. and thus was born. It was one of these initiatives that came not from the kind of scientific community up through the system, but actually came from the politicians and came down upon the scientific community, the HPCC High Performance Computing and Communication Initiative. And this was a big wad of money spread out over a number of different agencies,

20:00 DOD, DOE, NASA, the National Science Foundation, the National Institutes of Health, ARPA or DARPA. really significant increments of money over a number of years dedicated to funding research into problems involving high-performance computing or network communications, the NSFnet backbone across the country. And, you know, VBNS and all of the very high-performance networks going in in many places, had their starts in this project. One aspect of the HPCC, which was particularly drawn out, were what were referred to as grand challenges. And the idea was to fund for a long period of time, not on a year-to-year basis or two years, but over five years, to fund collaborative efforts, collaborative meaning not just within a discipline but across disciplines, bringing together scientists and computer scientists together to solve problems that could not be solved on the present generation of computers but would require the next generation. And the idea, which it's a good idea, was to basically push industry, saying, okay, there is money here to buy computers, to sell computers, but they have to be able to do these problems. and so you provide an incentive to industry in this way and the other thing that you do and this is something that I think was very insightful at the level of the government even if they didn't necessarily realize they were being insightful and I think is overlooked in the scientific community is that the tradition has always been that the computer engineers

22:30 build the best piece of hardware that they can and it comes out and it's virtually unusable on the problems that people have because the software is not there or because it was designed to be very, very fast by the standards or the benchmarks that were important to the people who were designing it but not necessarily fast according to the benchmarks people who needed those computers were going to put it to. And the idea here, and it worked, I think, to some extent, was to force industry to pay more attention to the user base to target their development more towards the customers, if you will. and I think it did have a modicum of success in that way although I couldn't give you any examples of that just something I've noticed as I've watched computers develop over the last several years is that the software tends to support them a little bit better for scientific applications than it did in the past But anyways, this was taking place in the kind of 1990-1991 time frame, and the NSF put out a call for proposals. however before it put out its announcement of opportunity Rich Isaacson at the NSF called together a bunch of people who were doing numerical relativity basically he called together and he asked to come to a meeting in Washington a group of people who he thought could make up a grand challenge team for numerical relativity, and I remember that I was traveling, and I got back, and there was a C-mail message asking me if I wanted to come to the meeting, and the meeting was basically the next day, and so I sent him back a note saying, you know, gee, I just got back. I'm sorry. I'm going to miss it, and I get a note back from him saying that you really don't want to miss this meeting, and well, you know, when the program officer at

25:00 The NSF tells you you don't want to miss a meeting. Well, you buy a ticket and you go. And so I did. And that was kind of the first organizational meeting of the Binary Black Hole Grand Challenge. And at the time, in fact, one of the problems that was out on the table and was seriously being discussed was the supernova problem. that is, core collapse to form a black hole or to form a neutron star to get the gravitational radiation out. And there was a lot of debate and discussion about whether or not we should tackle that problem or whether we should tackle binary in spiral and whether or not it should be a pure gravity problem or gravity plus matter. That is, do we want to treat black holes or do we want to treat neutron stars? And a lot of those questions revolved around political issues, personality issues, and science, kind of technical issues. Let me tell you who else was at the meeting. Okay, Rich Isaacson, Larry Smart, Jimmy York and Chuck Evans, Jeff Winokor, Stu Shapiro, and Saul Tokolsky, Richard Matzner, Okay, who am I missing, or is that it? Okay, that's Texas, Illinois, Cornell, North Carolina, Pittsburgh, and me. And that was the initial crew. That was the initial crew. Was Ed Seidel there? No, Ed wasn't there at that meeting. Ed wasn't there at that meeting. He was working at NCSA. But one of the things, in fact, that Larry pushed for,

27:30 I don't remember if it was after. Anyways, we ended up writing a proposal somewhat after that which focused on the binary black hole problem and submitted it and it was turned down the first time. One of the reasons it was turned down was because there was you'll notice that that was all a bunch of physicists we didn't have a cross disciplinary team And somewhere around the time that that proposal was submitted or just afterwards, Larry talked to us about getting Ed on as a co-investigator as well. And so that's when Ed joined as a co-investigator as opposed to kind of just being a post-doc type of thing. The original group were all people who were faculty or were about to become faculty in my case. And the proposal was rewritten, and we added in the new proposal Jeffrey Fox at Syracuse and also Jim Brown who was Computer Science at the University of Texas and there was an addition then of a substantial computer science component part of which grew into DAG H and that proposal flew the foundation and was funded. And so that would be, I think that that was, would that be September 93? That might be September 93. Is that right? Richard Master was telling me that the funding was running out in September of this year. That's right, that's right. The funding basically, yeah. It was very, very weirdly funded. I mean, the first year's funding was something like September 93 to October 93, and then the second year's funding was October 93 to October 94.

30:00 they gave us sometimes funding at the foundation can be very very odd you have a month to spend a year or a year to spend a month it can be very strange so that's basically how that started shortly after it was virtually our first phone call telephone conference after well maybe not our first But shortly after it started, that was at the same time, 1993, that Pablo Laguna started at Penn State. And Richard proposed in a telephone conference that we had Pablo as a principal investigator to the grant. And we did that, and so that's when Pablo joined. and Ed was already a co-investigator and then the only other person that we accreted as an investigator was Matt Shopto shortly after he joined the faculty he was post-docing with Richard at Texas he'd been post-docing there for several years He wasn't post-docing in 93, I don't think. I think he started shortly after that. But he joined the faculty there, and I think shortly after that we added him as a co-investigator. The whole business of co-investigator versus just being someone who's supported on the grant, is a very, very odd one that has political overtones that are entirely independent of money or credit or authorship on papers or anything like that. There is a certain cachet to being an official co-investigator. On the other hand, the people who wrote the proposal who are kind of held responsible for getting the work done.

32:30 And so there are some people in the collaboration who are very uncomfortable with adding co-investigators because, well, they're still being held responsible. The new co-investigator is not being held responsible in the same way that the old ones are. On the other hand, they're getting the, you know, it's one of these responsibility authority questions. You know, the authority is being given to many more people than who are left holding the bag and who have the responsibility to deliver. And so they're just kind of mixed feelings about adding various people. But, of course, then there are career issues also as well. out of hand type of thing. But that was kind of the genesis of the Grand Challenge. The emphasis on LIGO was something that came later. Although Matzner was pushing the idea of a catalog of waveforms fairly early on. I would say at least half of the P.I.s did not feel that that was at all a realistic thing. In terms of whether it could be tamed or whether it was whether it could be done. I mean, there were a lot of people who were very uncomfortable with the idea of kind of saying, you know, walking up to the plate and saying we're going to hit a home run in five years. There were lots of arguments, some very loud ones, in fact, that I was involved in over the fact that we had set this one goal five years in the future, and we said that that's it. I mean, we literally put it that way, you know, binary black hole coalescence or bust. And I personally regarded bust as really much more probable. And I was arguing that we needed to set up a hierarchy of problems along the way so that we could show, you know, that, you know, we said we were going to do X, Y, Z, W, A, B, and C.

35:00 and, well, we got X, Y, Z, and W, and we didn't get A, B, and C, but we still got a big chunk of work done, and it was on the way to the problem and everything. And this is where kind of the politics of the choice of the problem came in, and that was that I think the idea was that, well, you know, this is a grand challenge with a big G and a big C, and it's, you know, we have to make it a grand challenge. And, you know, we can't make it kind of an engineering thing. It has to be something that is, you know, only Olympians do this. You know, you get the picture. What were the main reasons, do you think, why the binary backfold problem was chosen over supernova or binary nutrients? I think the binary black hole there were some technical arguments actually let me go back there was another kind of personality slash political issue that was there that I actually think was very important and that is that gravity is pure like a platonic ideal it's geometrical there's none of this messy matter business it's just it's the Einstein equations it's space time and we don't want to mess it up with matter matter dirties it that was definitely that was the first time in fact I had ever kind of heard that or seen that it wasn't said explicitly that way. And that somehow or another, that by having a simulation or by doing a problem where there was both gravity and matter, that somehow we were selling ourselves. And that the real thing to do is pure gravity, is pure relativity. And it's black holes, It's not neutron stars. That was a very, very strong... That kind of viewpoint came very, very strongly from Larry Smarr,

37:30 also from Richard, also from Richard Matzner. And, yeah, you can even kind of see it in their background. and there's a certain kind of political savviness to that as well you think about what does the public want to hear about what gets them excited it's black holes and it's Einstein it's not Chandrasekhar and white dwarfs and neutron stars so that played one role in the choice of the problems Another role was that in the choice of the problems was that the next step in the supernova problem was to do not something fairly simple like I had done, which was simple in terms of equation of state, but was to put in neutrinos, to put in radiation transport, to put in nuclear physics, to do a much more sophisticated problem. And none of the people in that room had any expertise in that kind of very dirty and nuts and bolts astrophysics. So just from the point of view of, you know, no one felt comfortable attacking that problem. There was a strong feeling that, now again, so now the supernova problem has got two strikes against it now. One is that it involves matter. Okay, even if you form a black hole, it still involves matter. and two is that this was just not the group of people who were competent to do it and they knew it the binary and spiral problem it was recognized as an important problem it could be done with black holes that is you didn't have to have any matter at all in it And you could say, there are lots of ways, then, that you could attach it.

40:00 You could talk about it as it's the simplest dynamical problem. Not quite, but in relative, it's the two-body problem. One body problem is a static black hole. And so it had a lot going forward in terms of importance as just a pure science problem. certainly a challenge and the like. There was still a strong feeling that that problem would be better done with neutron star in spiral than with black hole black hole in spiral. And that discussion hinged on a technical issue was that we didn't know how to handle singularities in a way that would allow us to evolve for a long time. This was before Excising the Horizon came about. I mean, people knew that that was what they had to do, but no one had yet figured out a way to do it. and there was a lot of feeling that that was not going to be an easy thing to do. I knew immediately that you couldn't evolve systems in the old ways with stretching the horizon and leaving it back before singularity formation or attaching it back at the throat. You could never evolve for long enough. Didn't know how to excise the horizon. On the other hand, people also knew that, well, you know, neutron stars, they don't have this problem. You know, the potential is nice and smooth through the neutron star. There's no singularity. You can evolve the whole space-time. And the structure of the neutron star is not important in these binaries as long as it's small enough. And so the gravitational radiation you're going to get out is the same. So from the point of view of doing the problem that's interesting for LIGO and interesting for gravitational radiation, you can do the same thing with the neutron star problem. Yes, you have matter. However, the hydrodynamics are not the hard part of the problem. Hydrodynamics are easy, and there's lots of experience with that.

42:30 I mean, Stu and Saul both had tremendous experience with that. And Larry had done stuff with that, and I had done stuff with that. And so that was not regarded as a problem, but it wasn't black holes. And that was what was the final, I think, the final issue, was that there was kind of a standoff amongst the big guns, you know, Stu and Saul and Larry and Richard Matzner. and Matzner ended up writing the proposal and he did it in such a way that he waited until the last minute. The first draft of the proposal was supposed to be circulated several months before it was due and it was only circulated a week before it was due so there was no time to change anything. Okay. And so that was kind of how we ended up. We would have ended up with that anyway, because Stu and Saul would have been the ones to write the proposal in the other direction, and they weren't ready they could have written the proposal first and circulated the first draft and controlled the discussion that way and they didn't so that was kind of how that particular problem came about there was even a discussion of not picking a problem, a single problem at all but that is making the grand challenge more of an investigations in and discussing a wide range of problems with a common thread. And so instead of advancing in kind of like a laser beam towards a particular distant goal, working across a broad front and perhaps not going as deeply, but going with greater breath. and that was pretty quickly dispensed with by the group and probably for good reason or probably for good effect

45:00 because although I didn't think so at the time, in retrospect I think that such a proposal would not have been seen as a collaborative effort individuals to do whatever they wanted to do collaboration was a very big part of the grand challenge so as you sort of implied earlier you feel that the main reason for getting sort of a big group of numerical relativity together was that this was something that had come down from a high that money was going to be made available for such big projects. That's right. Money was being made available and this was an opportunity to get that money to support work in this field, work people had been doing. I mean, it didn't it did not create groups. Those efforts had been ongoing and funded at varying levels through grants, and this was a way of funding that work, perhaps at a higher level, perhaps making greater progress on it than be done otherwise. And it brings more money into the program, a rising tide, you know, raises all boats. And so that freed money, perhaps, that Rich Isakson would have used to continue to fund numerical relativity to allow him to do other things. So, in your view, when students see the Grand Challenge experience as kind of a model for or what relativity might be moving towards, but more as a kind of one-off, possibly one-off thing? Well, it might. I think that particular experience was driven not by the things happening in the field,

47:30 particularly LIGO, driven by money. However, that does not mean that it might not also represent, even if the causes were different, the effect might be very much the same as you find with what LIGO is doing to the field. LIGO is doing it now. It hasn't yet happened. It hasn't yet hit the theory community. It's going to, but it hasn't yet hit the theory community in the same way that it has come upon the experimental community, that there is going to be larger collaborations that are more centrally directed, and where the direction is going to come not necessarily from the intrinsic intellectual interest of a particular problem, but from its need in relationship to this experimental instrument, this observational instrument that is being built. To use an example from data analysis, People are right now spending a lot of time in a whole range of different areas. For example, let me choose an example. In particular is the searching for periodic sources of gravitational radiation. There is a tremendous intellectual interest in that problem. an all-sky search for a periodic source of gravitational radiation where you have no other observations. You're just going to see it in gravitational radiation. You can very quickly determine that for the kind of astrophysical, at the cherished beliefs level of astrophysics, that problem is intractable now,

50:00 and it will remain intractable for many, many years. well beyond LIGO 1. It will remain, it's intractable for LIGO 1, not simply because of LIGO 1's, well, not simply because of the computational problems involved, but also because of LIGO 1's very low sensitivity compared to this cherished belief level. Yet there is a tremendous effort going on there by people who think of themselves is doing data analysis relevant for LIGO. Okay. The, that work, there's going to come a point where that work people are going to say, you're perfectly free to do that work, but that won't buy you entree into LIGO because that is not relevant for LIGO. It may be relevant to some distant future, we have immediate and pressing problems. I don't mean we now myself. I'm now speaking as, you know, while I go. We have immediate and pressing problems, and if you want to be part of the team, you have to carry your weight. And that's an example of one area, and there are a number of other areas where there is work going on, and there's going to have to be a focusing on work, a focusing down of the community, the theoretical community, if they're at least of those people who are going to be part of the LIGO team, on to a narrower range of problems, which are not necessarily the ones that they find to be the most interesting or the most intellectually stimulating or challenging. And there are going to be choices made between, almost certainly between equally good directions to go. But there'll be choices made simply because there aren't the resources to pursue both, and you have to pursue one, and you want to make sure that it works. And so there are going to be choices that are going to be made. where it's not even going to be an issue of, well, this was clearly pressing now,

52:30 and that is, we can't imagine the future in which that is going to be relevant. There'll be things where, well, we just had to make a decision to cut this direction, and it's regrettable, but we have to do it, type of thing. And there are going to be shakeouts associated with that, And I don't think that the theory community recognizes that. A lot of people don't recognize it intellectually, and I think an even larger number of people, and even of the people who recognize it intellectually, I don't think that everybody kind of recognizes what that's going to kind of feel like on a gut level, an emotional level, when that happens to my work or when that happens to someone else's. But there's also an aspect that I don't think is appreciated, and that is that work is going to be very much programmatic in the sense that they're going to be deliverables. And it's not like an NSF grant where, you know, you write a proposal and they give you the money, and you write progress reports, and if something else looks particularly interesting that you find along the way, you're kind of free, more or less, to veer off in that direction as long as it's interesting and you're following your nose. And, you know, the proposal that you wrote three years ago doesn't necessarily bear any relationship to what came out at the end of three years. That's not going to be the way things have to run. People are depend upon what you're doing. Maybe you're building a piece of code. And, well, you don't have the freedom to say, well, I want to change what I take as input and change what I give as output. It's got to plug in in the same way that the plumbing has to fit in the building and the like. And I don't think that that's appreciated, the impact that that is going to have on people. And there'll be a lot of people who are interested in being involved now who, as time goes on, will find they don't like that style of work and they'll move out.

55:00 And there'll be other people who move in. And there'll be people who kind of grow up in this new environment, and it will be natural to them. But it's a transformation that's taking place. The LIGO science collaboration has just started to tackle the problem of of data analysis at the most recent meeting, which was the week before last, 12th, 13th, and 14th of March, or 13th, 14th, and 15th, I can't remember which, Thursday, Friday, Saturday. That meeting was in part devoted to the kind of first everybody get together who's interested and doing data analysis in the same room and start talking about what you're doing and talking about what you think are the things that need to be done in order for LIGO to be in a position to analyze data in 2002. And basically everybody listed out, well, the most important thing that needs to be accomplished is, and they would say, the research project that they're working on. Okay. And, you know, talk to Kent. You know, you can ask him. He sat in on those meetings. And, you know, he'll give you his viewpoint as well. But that's how I read it. And that's how I saw it kind of coming out at the end. But it's a first step. Now everybody's kind of met each other, seen what they're doing, and they're being forced to think about the fact, they're being forced to confront the fact that we can't do everything and that we're going to have to choose. And I don't think that yet everybody has really dealt with the fact that the choice is not going to be made on intellectual merit of an idea. but on the practical issue of is it important for LIGO 1, can it be delivered in time, can it be made to work, or is it more of a research problem, and the like. And those kind of things are coming down the line.

57:30 You mentioned it, and if you outline it, that there will be sort of a more centralized direction of research efforts So where do you think the, where do you see the sort of, if the decision-making process becomes a little bit more centralized, where do you see that being located? Is that sort of within the LSC or within LIGO itself? So, the LSC is really, it's bringing itself into being. And my hope is that its home will be in the LSC. And I think that it ought to be in the LSC. However, right now the LSC is still in the process of forming itself, and so it's not yet able to really do that. the LSC has yet as a group to actually make a hard decision not in this area, not in any area there are how familiar are you with the structure of the LSC? not very much, okay you become a member of the LSC well, when the LSC first formed the charter members of the LSC formed or became members of the LSC by basically coming to an understanding, a written understanding with the LIGO Laboratory, Barry, Barish, which was a written memorandum of understanding between your institution and the LIGO Laboratory in which you promised to do certain things, and those things were enumerated. And Barry iterated to make sure that at some level he felt those things were relevant to LIGO. And depending upon the area of interest, he may have made more,

1:00:00 or that may have been a narrower constraint data analysis area was a broader constraint, because LIGO had not yet focused itself, and because he recognized that the community hadn't yet formed itself in order to focus. I think in the experimental area, he was much more critical of the things that were proposed in the MOUs, and probably iterated more. so that group of people then formed the LIGO science collaboration it just brought itself into being and that was August of last year the collaboration set up a series of what are referred to as development groups there's a development group on seismic isolation and suspensions there is a development group and optics. There is a development group on advanced interferometer configurations. The people who are particularly interested in seismic isolations and suspensions obviously find themselves in that group and so on. And they are hashing out now, not issues related to the LIGO-1 design, which is pretty much determined. I mean, you can't change that the installation is beginning at this point. But what they're hashing out now are directions for future research leading to decisions for the enhancements to LIGO, which will come after the first science run. And those decisions on what enhancements will be made, the decision on that is going to be no later than the beginning of the science run, And so in the meantime, a lot of people have a lot of different ideas, and what they're trying to do right now is come up with a coherent program that will investigate reasonable directions and come up with a decision procedure that will allow them to make the hard decisions as to, well, we're going to do this and not that. we're going to do this and not that. But the hard decisions haven't been made yet.

1:02:30 And people are still working on kind of coming to grips with how are we going to make the hard decisions in a reasonable way so that the decisions are made on kind of programmatic needs and not kind of who yells the largest or who throws the most weight type of thing. And the same thing is happening in the lasers and optics group. The advanced interferometers configuration group is kind of a different group because they're thinking much further. Their horizon is much further out there. Advanced interferometer configurations won't be an issue for the first enhancements to LIGO. You'll notice that in that list there were no data analysis groups set up. The data analysis groups were set up at this most recent meeting. Ray tried to create three different groups. There was a group on diagnostics, instrument diagnostics, which was mostly oriented towards hardware. There was a group that he tried to set up that was focused on statistical detection confidence and parameter estimation and statistical measures of confidence. And he tried to set up a group then that was focused on astrophysical interpretation and algorithm development. And the distinctions, particularly between those last two groups, We're not good ones to make for several reasons. It's hard to imagine separating the algorithms from the statistical tests. I mean, you don't do those in isolation. And then also for the other reason is that most of the theorists who are the only ones who have the time to worry about that right now come from an astrophysical bias. of course, is that everybody ended up in one group, and no one was thinking about the statistical confidence issues. And so there needs to be a rethinking of how the data analysis groups are set up, or whether there are groups, plural, or how that work is going to shake

1:05:00 down. But my hope is that the LIGO science collaboration through the spokesperson and the executive committee that the spokesperson has, which are basically going to be the leaders of the development groups and other people that the spokesperson finds to be, you know, kind of particularly, give particularly sage advice on these things, will be formulating kind of the direction and herding, if you will, the other people. That the leadership will come from the spokesperson and from the leaders of the groups. And the leaders of the groups eventually will be elected by the membership, by the interested people, but that is where I think the authority will have to rest in kind of the ultimate decision-making responsibilities. If a group leader does it right, then he or she won't ever actually have to make a hard decision. The group will make it. takes a bit of work to set up a situation in which that happens that's that I think is going to be a year or so down the line though because this this group doesn't have any experience yet in real experience in working together in this way or in making decision. You know, it's not been a trial by fire. Yeah, sure. Touching back to what you said earlier about the context of the Grand Challenge, about the difficulty where some people have the responsibility and then others have the authority about their responsibility, is there and if that enters here too, since the charter members are the ones who made the Memorandum of Understandings? members join the, after the charter membership period, the way, these MOUs were only good for six months, and they have to be renewed every six months, which means a new negotiation of what you're going to do for the project, reports on how well you succeeded in meeting your goals

1:07:30 of the last time, of your last MOU, and so on. The way, if someone now wants to join the LSC, the way they do it is they have to negotiate an MOU with the laboratory. However, the Collaboration Council, which basically consists of one member from each participating institution, so if there's an institution like Caltech where there are several different groups who might be participating in the LIGO science collaboration, Caltech has one representative plus one more representative for every multiple of five people there. And this forms the Collaboration Council. A proposal for a new group or institution to join must involve an MOU and then a positive vote by two-thirds of the Collaboration Council. And so they have to contract to do something real, and it has to be seen both by the laboratory, who they would end up signing the MOU with, and the Collaboration Council as valuable. so membership is not as easy to come by now that it's found itself it involves not just satisfying bearish but also the collaboration itself must look at it and say that this is something we need done we think this is a valuable group of people and it's not something that someone else is already doing we haven't been faced with that yet don't expect that we'll be faced with that for at least a year So an MOU between an institutional group and LIGO, does that, that obviously involves undertaking certain projects and so on, does it somehow entitle the people in the group to funds connected with LIGO or is it just something that they would then apply to the NSF and say we're working on this? LIGO does not provide any funds.

1:10:00 So your MOU is with LIGO, and the fact that you have an MOU with LIGO cuts some weight with the NSF in the sense that there are funds that are being made available in the foundation to support LIGO-related research. And, of course, you have a better claim on those funds or to be considered for those funds if you have a specific MOU with the laboratory and are a member of the science collaboration and that the science collaboration has indicated that this work is kind of particularly critical. However, you still have to apply to the foundation separately for that, and it's reviewed. Your work is reviewed as your work. Now, there is a twist to the way in which proposals for LIGO-related work are being reviewed by the foundation. Do you know how the foundation normally reviews a proposal? You send in a proposal, which has to follow certain guidelines, including the length of the science part and the like, budget, etc. And then it goes out usually to about four people who might be in the U.S., they might be abroad, who the program officer feels are competent to read and judge the merits of that work. and then those reviews are received back by the program officer who then pools them with the other reviews on other proposals and looks at what the program officer feels are the goals of the program and what's particularly important and makes judgments, makes decisions. That still happens with proposals that would fall under this category of LIGO-related research and development or LIGO-related science. There is an additional thing, though, that goes on with those, and that is that those proposals that purport to be for LIGO-related research are seen by the laboratory.

1:12:30 and the laboratory is asked to write a technical assessment of those proposals. Different than a review, they're asked to assess the practicality of the work at a certain level, but more important, what resources that would require of the laboratory. In addition, the proposals are given to the LIGO Program Advisory Committee, which is kind of an advisory committee to Barry Barish, to the director of the project. However, it consists of people outside of the project for the most part. and they are asked to read the proposals and to make their own independent judgment about them and to provide essentially a brief review of what they think is meritorious in terms of relevant for LIGO and important for LIGO and what they think is not meritorious along those lines. And that information is also given to the NSF and also forms part of the information that they have for making their decision. And so that's something different than the normal review process. And that's happened now twice. And it will continue to be an ongoing thing, I think, for a number of years. I think the foundation likes that mechanism. Well, you mentioned earlier that you felt that there were certain areas in data analysis which were going to be needed, but I know you were most important. Where do you think those areas are? Well, one of the things that I think is, well, it's tough.

1:15:00 I haven't found yet a really good way to articulate it. There is, I think it's important, for LIGO 1, I think that it's important for a number of reasons to have something available that looks for binary and spiral. the way in which it's done and I don't mean now mass filtering or something else but I mean the way in which the data flows through the system is something that's very very much up for grabs in order to keep up with the data so here's something that's very important that people have not been thinking about and that is that it is critically important that we keep up with the data no matter what kind of statistical tests or computation schemes we have, that's kind of a hard bottom line. One must keep up with the data. That almost certainly means that there are going to have to be some kind of pre-selections made. There's going to have to be a staged data analysis. In some context, you call it a hierarchical search. Although, once you're doing a hierarchical search where you're looking on a coarse grain and then concentrating on interesting areas, on a fine grain, you can ask the question of whether or not that first coarse grain level necessarily needs to use the same techniques that you use at higher levels. And the right way to do that needs some careful thought. That's an area that really no one has been looking at or thinking about. It's a way, we're not used to thinking about the problem that way. The second, and so in terms of a source, well, for many reasons we need to be looking for binary unspiral, and if for none other reason than the fact that that was singled out in the proposal,

1:17:30 and has been kind of the flag that we have been waving over our castle for the last decade. And so we have to do that. I think we have to also look, in the old saw, you have to look where the light's the brightest also. And so that means that periodic sources. We don't have necessarily a lot of confidence that we will see something there, but we ought to do a targeted search on some select pulsars, you know, particularly things like Vela, which glitches, and so on, It may be aspherical, there may be something there. And to the extent that we can, it's not clear that we can, a targeted search at the galactic center where you actually look, you're targeted in space but not necessarily in frequency. So you look over a fairly wide range of frequency at the galactic center where there might be a lot. And you might get lucky. Um, the, um, I think that we need to do something like, we need to be, we need to be doing something that gives us the best opportunity to notice serendipity. like what I was kind of talking about earlier today, not necessarily having anything to do with the things that I pointed to, but other than just the general thing, is that I think we really have to have some kind of search or search techniques going on as able to notice those things that we have not anticipated as is possible. And so I've almost kind of covered the bases of kind of...

1:20:00 There's a very targeted binary and spiral, a very clear source. there is looking at the best periodic source in the best way that one can imagine doing given our resources in a very nebulous way I think it's important that we have something going, running all the time that is sensitive to generic bursts whether that is a kind of model this kind of damp sinusoid model or whether it's some other crazy cost correlation scheme or the like, or something else entirely. And then we should also just be on the off chance looking at a stochastic background. The stochastic background I throw in there mostly because that is computationally, you can do that on a workstation. So that really doesn't drive the data analysis problem. That is, if it's extremely low bandwidth, and you buy an old Sun 2 workstation, or, you know, you take someone's HP calculator and you dedicate it to the task, and that's fine. So the things that I think that are clearly important are being able to figuring out the right way to handle the data flow so that you take this fire hose and you very quickly divert most of it keep just the parts that kind of have a very high probability of being most interesting. But then also with that, covering all of the source bases, not necessarily in terms of specific sources, but in terms of the kind of generic burst source, periodic source, stochastic source type of thing. I think we really need to... It doesn't have to be the best, and it doesn't have to have all sources in every category, but I think we would be remiss if we didn't have at least one example analysis, kind of the best example analysis we could come up with going in all the different directions.

1:22:30 There are a number of other things that I think we need to grapple with as well, but I can't articulate them right now. I mean, there's kind of a general sense of unease about certain things, but I can't quite put it into words. before we started to a certain tension between the experimentalists and the theorists over the question of who really is the responsibility for data analysis and we discussed how it seems to be somewhat unusual that in this field it looks as if the theorists will be called upon to take on the biggest burden in dealing with data analysis issues in what ways does this tension manifest itself or has it done or is it easy to see it coming itself in the way in which a natural thing that you would think of doing would be to cut your teeth on data analysis by using observations from the 40 meter. Okay. There is an extremely strong resistance to making any of that available to theorists walking in. And the reason for that is that a lot of people have poured their entire professional lives into that, and they should have the opportunity then to reap the benefits. I mean, one of the groups, I won't say who, when Ray Weiss was going around and surveying all the people who had expressed an interest in different aspects of data analysis, one of the groups said that, well, we're going to focus on the astrophysical interpretation and not do any of the kind of other stuff. And, you know, there's a reaction of, you know, well, gee, the rest of us aren't in this because we like noise. There's this sense that if you want to ride the horse, you've got to take your turn shoveling in the barn. And that applies with even greater force

1:25:00 to the experimentalists who have been sweating this for years and building the instrument. And then you get these people like me here and aren't down with a wrench at the 40 meter and aren't working on lasers or cleaning optics or doing any of the grunt work to make the instrument go, we're already talking about looking at the data stream and all the wonderful things we're going to do with it. And, well, we may have kind of the best motives in the world, or not, but we may have the best motives in the world. We're trying to chip in. Some of us are completely incompetent in a lab. You would pay us to keep away. And we're trying to help because the project's understaffed and because you've got this big data problem looming and so on. But on the other hand, you know, their response is, you know, well, you know, Jesus, I'm not doing this because I like threading notes on bolts. Yeah. And it's a very legitimate concern. You know, you don't want to get to the end of the day and then pass the baton on to someone who says, you know, thanks a lot, buddy, see you later. Yeah. Now, there's also concern, a very legitimate concern that comes from a different point of view, and that is that you can't do data analysis. I mean, a pure theorist cannot do data analysis. Data is never that clean. It's never that ideal. You have to have the people who built the instrument, who know the instrument, who know its creaks and its groans, and who know kind of the ins and outs of it. They have to be intimately involved to interpret the data to say, yeah, that may look like a gravitational wave signal, But if you look down here at this control loop, it's doing something a little bit funny.

1:27:30 Well, it's doing this funny, and it causes this to do something funny. And yeah, it can cause exactly that effect over here. Theorists can't do that. Even another experimenter who hasn't worked on that instrument before can't do it, although they at least know what questions to ask. and at least they'll stay awake at night worrying about things that a theorist won't even imagine. And so there's also a concern that comes up in this that you've got all these theorists coming in to do the data analysis, but who's going to ride herd on them and who's going to do the reality checks when all the experimentalists are really tied up building the damn instrument. And so that's a tension within the project about, on the one hand, they kind of recognize they need to free up people's time, but on the other hand, they realize they're understaffed and they can't. You know, they don't want another Weber-type incident to happen. I'll take a break, I need to go to the bathroom. Okay, so we're off again. So you were saying that there was this important issue with the fact that the theorists, the what's in the data. Yeah, that's right, to interpret the data and to understand it. And actually, there's something that is related to this that Harry, in fact, told me about. It was something that he had done early on in looking at different kind of experimental apparatuses are built. I mean, you know, you can go to the journal and you can read some description of how to build an apparatus, as he related the story to me. And then you go and build it and you find it doesn't work. And the only way to make it work is to go and to talk to the people who actually built it and work with them in their lab. And there are all these little tricks and, oh, yeah, that happens and then you've got to do this. And there are all these little things that don't make it into the paper because you can't articulate them.

1:30:00 and because they're seen at some level as details. And it's not quite the same issue, but it's very closely related to that. There are parts of the experimentalist's craft that they come to understand the data in ways that you will not find in any textbook. They get a feeling for the instrument such that they can look at it and kind of look at the data and tell when it's working right and when it's not. And you can't teach that in any textbook sense. It's a feeling that comes from working with it. Theorists have the same thing in an area that you're working in. You know, you can kind of look at a problem, or you can look at someone's solution to some problem, and without looking at it in detail, sometimes you can just say, oh, something looks wrong there. something looks fishy there. And, you know, for a student or something to work it out, they'd have to go through it in painstaking detail. But you develop an eye for certain things. Yeah, it's an excellent point. How do you see, or can you see any way present that the dilemma can be worked out, that if the experimentalists are simply too busy to do the data analysis and the theorists have to take the load? How are the experimentalists going to coordinate with what the theorists are doing? I don't know. I don't have a solution to the problem. I think it's a real, I think it's a big problem. And I just do not, I do not know how to deal with it. Part of it is that I don't know, I don't know all of the things that are going on in the LICO project. and maybe it's entirely possible that there are some things that are going on in the project that are not essential, not on the critical path and it might be better to just not do those and as a result free up people to work more on the data analysis. And I don't know. I have no way of knowing that. And that's a possibility. Hiring more people is not exactly a real possibility, because budgets are fixed. And people are expensive.

1:32:30 And there's also the problem that already as it is, if you look at the staffing of the LIGO project, it's dominated by people not from gravity, not from experimental gravity, in fact. And that's in large part because there was not a large enough pool of people to support the construction of the instrument to begin with. You know, at Caltech right now, in the LIGO project, the only person from experimental gravity at Caltech, whose background was in that at some level, is Stan Whitcomb. All of the other people are no longer part of the project. Ron Drever is not part of the project. Bob Sparrow has left. There's even been a tremendous turnover of the people who were brought in shortly after the project got started from outside. You know, so there's not a pool, there's not really a pool of people there. And the people who are here, a lot of them come from high-energy physics. And they grew up in a culture that is where theorists are not part of experiments. but on the other hand they do have phenomenologists which is something that gravity doesn't really have right now it's just kind of beginning barely to build up such a group of people and also they have bigger workforces

1:35:00 that is the projects the big experimental projects and collaborations that they build are bigger and there are more people to handle the different jobs. And there is also a model. There was always the last experiment. We're kind of more in the situation where, if you could imagine E.L. Lawrence building the first cyclotron, maybe he couldn't build it in his lab, but he needed 100 people to do it. Well, that might have been an analogous situation. It was the first one, and it was needed a huge group, only for them they were able to build up more slowly to the state of being, and we have to kind of make a big step all at once. Pat Brady was talking to me a bit about the LSC meeting and mentioning that there was some time devoted to the question, you know, how to decide when you've got a result or a detection that you can announce. And this sort of touches on to what we were discussing, as you mentioned, you know, probably one thing that is prevalent in the field is a desire to avoid things that happened in the past, such as, for instance, with the Webber Congress. So the impression I got is that at this stage, actually, although that's an issue that's alive in people's minds, that they brought it up at this point, that it's actually not one that there's a consensus has developed. That was the first time that in any organized or open setting, people have discussed it. And Ray brought it up particularly because we have to start talking about it now because it's going to take a long time before, not to settle on anything, but rather people need to start thinking about it and rolling it around in their minds and arguing with each other about it so that when the time comes, not necessarily that we'll have any kind of hard and fast rules, but that we'll at least have put some deep thought into it collectively as a group and understand where everyone is. And a very important thing that's going to come up again and again and again,

1:37:30 but we're at a very, very basic level on this question. What would you say that certain range of the discipline is? Well, okay, on the most conservative side, one feeling that was expressed was that I don't care how strong an event is, if I don't see two of them, it's not real. So, you know, for example, if you were to be fortunate enough to catch a nearby in-spiraling binary that gave you a whopping big signal, you wouldn't count that until you had seen a second in-spiraling binary. that is a fairly conservative point of view. Point of view that Ray expressed, and I would say a lot of people were happy with, but not everyone, was this view that anything that we see has to be completely consistent across all of the operating detectors. That is, it couldn't be that it was statistically significant in... Well, there were actually arguments over exactly what that meant. Some people wanted to say that, way they understand it and the way they're happy with it is that it means that if that Ligo Hanford all by itself saw binary in spiral and Ligo Livingston saw binary in spiral and both Hanford saw it with the right proportionality and Livingston and Virgo and everybody individually individually sought. There's another kind of way of looking at that that says that, well, maybe everyone individually didn't see it at a threshold that they would set appropriate if they were the only detector in the world.

1:40:00 but rather that this kind of network coherently sought in a manner that was fully consistent. So that would mean in some sense lower thresholds but associated with those lower thresholds are confirmation by other instruments and so on. There are other people who were saying that well LIGO should be able to make a decision by itself without relying on Virgo. you know there's another group of people which is the this is a very European thing, I mean there's a very very clear division between the way the American scientific community views this question and the way the European community views this question there's another group and this was expressed particularly by Bernie is that well if there is some anomalous event in the and there's no obvious reason to rule it out, why can't we just publish it and say we found this anomalous event and not say that we found gravity waves? And the North American contingent, The American contingent is very unhappy with that point of view. That's not a point of view that they can accept. And part of it has to do with history, but I think part of it has to do with just a kind of a different way in which science is done in a different way in which the attitude, I don't know, maybe the question is one of the way in which one invests kind of personal, or takes personal responsibility for a result as opposed to community responsibility for a result. There's American contention, that it's our responsibility to say what this is, and not just to set it

1:42:30 out there as an interesting thing to look at, and for the community to decide. We have to make the decision. We have to take the responsibility. But then conversely, we also than take possession of it. It's ours. And that may, I suspect that that has something to do with it that maybe ties up with kind of American individualism and the like. As opposed to something more peculiar to Americans, as opposed to the gravitational wave community in the United States and the effect of Joe Weber on it. And that's interesting what you say about the division between the Europeans and the Americans. Would you say that there's any similar tendency for the experimentalists versus theorists, say, to have a typical opinion? Well, I mean, the same thing shows up in the bar community. I mean, it's particularly highlighted in the bar community there. The Rome group is much more willing to find things and publish them than, for example, the LSU group. The LSU group has been very, very careful with the analysis of their data. They don't want to be in a position where they publish anything that they refer to in any term that might be loaded, event or the like, unless they are kind of willing to be very clear about what they mean and what they say this means. And the Rome group attitude is much more that, well, we need to be open to possibilities, and if we saw it, we saw it. And it's our responsibility to report it without necessarily saying what it is or whether we believe it. And the same thing, and that also involves Blair and the Australian Bar effort as well.

1:45:00 And so that tension exists. exists there as well. Right. But within, like they say, the LSC, would you see a distinction between the kind of attitudes towards what you publish or what you claim between, any difference between what the theorists and what you plan to think about the experimentalists? Well, the most conservative viewpoints came from the experimentalists. Stan pushed the idea that if you don't see two, you haven't seen one. And he was the one who brought that up. Barry Bearish also chimed in, that in high energy, that you don't publish one event. And then he hedged, and he said that, well, that has happened before, but it has to be really compelling and extremely important. So a recent example, of course, is the discovery of the top quark. That was published as no top quark was seen. And rather, there was a statistically significant excess of events after all of the vetoes and the triggers were in there that there were more events that could be explained. There were more events there than could be explained as background. Things that were not tops and would get through all of their vetoes and triggers. and they really agonized over that and in the end published that first detection, if you will, as evidence for the existence of the top quark. Okay, so yeah, they hedged the words, but they did publish.

1:47:30 However, it is not the culture of the field to do that, and they had to agonize over it. And I guess there was one other example that was given, and I think Barry said the Omega meson. But that's, he said, in his entire professional life, which spans 30 years, and so the policy which is the experimental field, is that you don't just publish an event. A difference here, then, in that field, is that you can always make more events by running your accelerator longer, or at higher luminosity. And so it becomes the distinction then between an experiment and an observation, where we, in this field, we can't create gravitational wave events. we have to take what nature delivers. And so, correspondingly, there is a, I think that there should be a different standard based upon the fact that we are more at the whim of nature. And corresponding to the different standard, there's also a different understanding of confidence. That is, when a high-energy physicist says that they saw this particle, they have incredibly compelling evidence for it and they've checked it every way from Sunday because they can. And when an astronomer says that they saw something, now taking the other point of view, well, there are a lot more uncertainties. And it's understood in each field. that there is more uncertainty here than there is there. And I think that this field, setting aside the possible distortions brought on it by the Weber experience, I think that this field is going to end up having a standard that falls somewhere in between.

1:50:00 But that that's going to require adjustment on the part of the people who are coming into the field from high energy physics and carrying with them all of their experience and they don't necessarily recognize that they're in a different neighborhood and that they can't do all the things that are second nature to them to do. And it's certainly going to require a lot of thinking on the part of the theorists who are involved and who've never had to face these issues before. And even perhaps for the people, say, within the field for a long time, the experimentalists in the field for a long time, because they're dealing with their historical baggage. They're dealing with their historical baggage. That's absolutely right. We all have to deal with that historical baggage, whether we like it or not. and there's also yet another aspect of people who have been in the field for a long while and that is that this is the first instrument where we, going in, we believe it to be marginally sensitive enough to detect events every other instrument was always viewed as a prototype and itself, except in very unusual circumstances, unable to see events, not sensitive enough. And so that skews the way you look at the data. Now you're looking at it. It's no longer an exercise, an intellectual exercise, but this is no longer a charade. It's no longer a rehearsal, and that puts a lot more... there's a lot more on the line. One of the things I've heard some people suggest is that the experimentalists are uneasy about placing any reliance on match filtering for the first detection, for early detections.

1:52:30 Is that a sense that was also... That wasn't expressed there. That wasn't expressed there. Although I have heard expressed in some other places. Vern Sandberg is a very good experimentalist. Comes from the physics community right now. He was on the LIGO, he's on the LIGO Program Advisory Committee. He's someone that Barry regards as a very sharp and a very good person to give advice. He stated at one of the Program Advisory Committee meetings, I was on the Program Advisory Committee in my capacity as the chair of the LIGO research community. He stated at one of the meetings in one of our sessions that this was after we had seen a presentation of the LIGO data analysis system that, you know, damn it, he thought that if gravitational waves were detected, it sure as hell wasn't going to be matched filtering. That did it. he views that as kind of a theorist's game now that's coming from his background I don't know how I'm just reporting now what I heard Dan Debray at Stanford experimentalist, real controls very, very sharp. His feeling is that matched filtering does not capture the reality of the data. That is, all the little things that go on, and that the data almost always involves correlations with other instrument monitors and control systems and you can't just look at the data with matched filtering and get kind of useful information out. You have to look at it in the context of all of the, you know, what's going on in all the control loops and everything like that. The other thing that both of them

1:55:00 are thinking about is they're thinking that one of the things that makes match filtering so, and this is something that I'm kind of drawing out of their comments if they haven't stated explicitly, is that one of the things that makes match filtering so sensitive is this assumption that you know what the signal looks like and they question you really know it that well, meaning not necessarily the theory, but that you know the response function of the instrument well enough so that you know if your calibrations in frequency is only good to 10% across the bandwidth of the instrument, well, then your match filter No matter how well you know the waveform, you don't know how it affects the instrument, except to that degree. And when you admit to yourself that possibility, then suddenly match filtering is no longer so much more appealing than other ways of doing things. In addition, you run across the problem that when you assume that you know what you're looking for, you are excluding things that you weren't necessarily looking for. You know, match filtering is not the best way to look for supernovae. And so if you focus too much on what you are looking for, you blind yourself. to things that are even not necessarily that far removed from it. And so there's a sense from that community that you want to look at non-parametric tests of the data, things that are less or not model-dependent at all. And I don't know. I've been listening to them, and these are very, very sharp people, I've been thinking very, very hard about the things that they're saying, and I don't know, for myself, how I'm balancing them yet.

1:57:30 But these are not people to be dismissed. They know a lot more than I do. As a theorist, would you see it as a worry that without optimal filtering, you're going to lose a sufficient fraction of signal to noise the threshold that the theory would indicate? I worry about that. I also worry about the flip side. And that is that if it requires that level of assurance that I know how everything works in order to get that last little bit that match filtering gives me, then I also worry about whether or not I really know things that well can I really claim to be at the end? In a sense, it's kind of how far out on a limb do I feel comfortable crawling? And yeah, maybe MASH filtering allows me to crawl a little bit further out on the limb, but do I still feel as safe? Especially from the point of view of false positives? Especially from the point of view of false positives, from the point of view of... yeah basically just you know instrumental things things that go bump in the night but now in the instrument that i don't know and no matter you know i can instrument the thing uh my detector you know uh every way from here to sunday and i'm still not gonna catch everything or be able to interpret the things that I see. And so there's kind of a subjective level beyond the pure mathematics of Gaussian noise or non-Gaussian noise by any model you want to impose. There's also a subjective safety level that you probably want to back off. I've never used this analogy before, but I'll use it now. And that is, you know, when engineers build an airplane, they build in an extra factor an extra safety factor so you know the wings can support this stress and we know they know all the details of the metals and the rivets and everything that goes into it in excruciating

2:00:00 detail and they have tremendous experience with previous plane designs and metal fatigue and everything and they've got it all worked out and then they throw in an extra factor of two. It adds more expense. It's not called for in the formulae, but that extra safety margin, they throw that in, in large part for a subjective sense of security. And I think that something like that has to apply here as well. I mean, we're going to need to throw in some kind of safety margin to take care of all the things that we haven't thought about or to take care of our worries. There will be arguments over what that safety margin should be. Well, yes, the points you make are very interesting, and I think what it puts into my mind is that with match filtering, you're sort of putting putting hardwiring the theory into the data extraction method and therefore potentially insulating yourself in a programmatic way from the vital input that you'd need from the experimentalist to say, no, that's really something else it's going to be difficult to see how the experimentalist can tell what that was in his instrument if it was a bump in the instrument if you've put it through match filtering. That's right. There are many more channels that have to be looked at. when you start to include all of those, well, suddenly you start to become more concerned. Well, suddenly match filtering no longer seems to buy you the advantages that it does in a pure problem. There are other issues as well, and I'll mention this one just because it's one that's been bugging me now for a while, and I know you've thought about these issues before. I'll give you one to think about in terms of mesh filtering now. You know, LIGO's noise curve has got these very, very sharp resonant features that come from the violin modes of the suspension wires.

2:02:30 It's thermally excited. It's a calcium random noise. long correlation time. Correlation time is minutes, maybe even hours in some cases for some of those lines. Q's of 10 to the 6, 10 to the 7 at 100 hertz is 10 to the 5th seconds. Correlation time. Okay, so very long correlation times. Now, let's do a match filtering experiment binary end spiral, the signal length from, say, 40 hertz to 200 hertz, about 20 seconds less. Over 20 seconds, that line is coherent. There is no way to distinguish that violin mode, what that violin mode is doing, from an absolutely rock-steady stationary oscillator. Why should I be able then to treat the noise as normal? I mean, when I derive the Wiener optimal filter, I make the assumption that it's stochastic, that it's a normal process. And so what I seem to be coming to the conclusion is that really somewhere buried in there is an assumption when I derive an optimal filter, when I derive the inner optimal filtering, there is an assumption that the signal is much longer than a correlation time, than the longest correlation time in the noise. That's what seems to be, to me, to be buried in there somewhere. What that means for LIGO is that I have to either abandon optimal filtering or I have to remove those features from the noise spectrum through filtering to basically notch out or in some way distribute the power in that mode, whiten it. Now to whiten a feature that is that

2:05:00 narrow will mean basically taking, the only way to whiten it is going to be to expand it in time. That is, the filter is going to have to have a very, very narrow feature in it, which means that the impulse response of the filter is going to be very long. the correlation time in the line. And so that means that, well, I don't exactly know what it means, but it worries me a great deal because even if I want to use optimal, Wiener optimal filtering, I'm suddenly now very worried about whether or not the results I get out are, I can interpret the way I would like to because the signals are generally going to be of much shorter duration than a correlation time of the noise. that's just, I thought I would throw that in. Well, it's clear, of course, it's been clear that you know, the theorists have been talking point about Gaussian noise. Whenever you hear the experimentalist talk, it's... That's right. That's yet another issue. Is that the noise is not going to be Gaussian. Whatever we do is going to have to accommodate that. And so I suppose I'm wondering if this is a problem, judging what you say, with the reluctance of people involved in LIGO, the actual reluctance to give data to the theorists, because of course the theorists are going to end up being the ones doing the data analysis. It's going to be a problem if they've never seen any data at all. Yeah, well, and it cuts both ways. That actually, that's another place where there's this kind of tension that pulls in both ways. On the one hand, the point that you just make is that if they've never seen any data at all, they're never going to get educated. On the other hand, you also have this problem is, you know,

2:07:30 geez, we give them this data, and then how do we keep them going out and spouting off nonsense and attributing it to us and having it reflect back on us? Because anything that is said about the 40-meter data or any kind of data analysis is going to reflect on the lab and the project. They need to exercise considerable control over who sees the data, and if you see the data, you have to agree to certain rules about what you can say and who you can say it to, and approvals that you have to get for writing papers or giving talks in which you even allude to anything that you've done. and that is in part because of this very issue that they recognize that there's certain reality checks that are missing in this community of people and also from experience in seeing some other talks that have been given by theorists who have had access to data where the instrument no longer exists or the experimentalists who took the data weren't involved involved in the apparatus. In particular, some of the talks that have been given about the 100-hour run, Cardiff, or excuse me, Glasgow-Garking 100-hour run, as analyzed by Cardiff, were very embarrassing to those experimental groups. And that's made people here very sensitive to the issues as well. The experimental groups weren't happy with... No, they were not happy. They were not happy. But that's yet another thing where they didn't have a lot of control because they gave the data to one of the proverbial 800-pound gorillas. And he can sit wherever he wants. Right. The...

2:10:00 what are the I guess I'm trying to think what are the problems obviously Cardiff wasn't saying that they were seeing no but there were discussions about well for example there were simple things technical issues, you know, the squared modulus of a Fourier, of a discrete Fourier transform is not a good estimate of a power spectrum. That wasn't realized. But that was, you know, kind of the first thing that a number of people in the audience where I first saw this work presented, experimentalists, looked at some of the figures that they were showing for power spectrum and said, well, geez, that's ringing next to all those peaks. How did you window the data? And, you know, they kind of blinked their eyes and said, window? What's a window? And so, you know, there are things like that that, of course, throw the whole thing. And then there were discussions about, you know, interesting artifacts that they found in the data, never claiming that they were talking about events, never doing anything like that. But they would talk about interesting artifacts that the experimentalists were just kind of saying that, well, you know, geez, that's just because this was a prototype instrument. That's not something that you should be focusing any attention on. Why didn't you, you know, ask us before about that? I can tell you exactly what that is, type of thing. And so there was You know, this kind of thing culminated in the recent gravitational wave data analysis workshop at Arce in November, where at one point the most absurd conversation in the audience took place in the question and answer session And Bernie Schutz tried to lecture Ray Weiss on the fact that, well, you know, there are motors in laboratories, and that disturbs the line frequency and can show up in spectrum. And, I mean, I thought Ray was going to have another heart attack. I mean, you know, of all the people to be lecturing Ray on this thing is Bernie.

2:12:30 but it's that kind of thing where the theorist thinks that they understand something that they don't or hits upon something where, you know, Ray's attitude was that you shouldn't be focusing your attention on trying to find ways to filter out line frequency noise from the spectrum, from the gravity wave output of the detector. Because if that noise is there, then it's because we've done something wrong in building the detector, and it's not operating well. And so if it's there, you know that something's wrong with the instrument. You don't filter it out and then look for gravity waves. And it was that disconnect that was taking place. there. And that was a reflection then of what was also being seen in some of the other earlier discussions. So there's still a cultural conflict because theorists who are doing the data analysis, no matter how sophisticated they are, theorists are still relatively naive when they're approaching something. Yeah, never built an instrument, never operated an instrument, don't have any kind of feeling for when it's working or theorists have, by and large, not focused any attention at all on data diagnostics, on looking at the data from the point of view not of looking for gravity waves, but trying to diagnose the instrument, or looking at anything but the gravity wave channel. I mean, of the volume of data that LIGO is going to produce, you know, a percent of that is going to be the gravity wave channel. there is a huge amount of other instrument monitors that dominates the data flow and are basically at some level you have to look at to determine the health and status of the instrument. And before you look at the gravity wave channel, those have all got to be sorted out and understood. it. But theorists have been focusing on the gravity wave channel as if it is the one and

2:15:00 the only thing about the instrument. And that's, it's kind of a tail wagging the dog. So a great part of the problem that's being faced is that you're liable to end up with this disconnect, as you say, because the model at present might seem to be that the experimentalists are going to be looking at the 99% of the input coming out and the theorists are going to be left looking at this. The 1% and if the two groups aren't talking to each other then the theorists are going to be claiming they're going to be seeing phantoms or seizing upon things that and are just not relevant, they shouldn't be spending their time with. That's interesting. It'll be interesting to see how this will be done. Yeah, I think the community has got some really... very rough seas ahead of it. Yeah. Everybody's going to get to know each other very, very well. And that's a good thing. Sure. It's a good thing. It's going to be interesting getting there, though. Yeah. It's going to be very interesting to observe it. Okay. Well, thanks very much. Certainly. My pleasure. My pleasure. My pleasure.