Notice: We are in the process of migrating Oral History Interview metadata to this new version of our website.
During this migration, the following fields associated with interviews may be incomplete: Institutions, Additional Persons, and Subjects. Our Browse Subjects feature is also affected by this migration.
Please contact [email protected] with any feedback.
Credit: Simons Foundation
This transcript may not be quoted, reproduced or redistributed in whole or in part by any means except with the written permission of the American Institute of Physics.
This transcript is based on a tape-recorded interview deposited at the Center for History of Physics of the American Institute of Physics. The AIP's interviews have generally been transcribed from tape, edited by the interviewer for clarity, and then further edited by the interviewee. If this interview is important to you, you should consult earlier versions of the transcript or listen to the original tape. For many interviews, the AIP retains substantial files with further information about the interviewee and the interview itself. Please contact us for information about accessing these materials.
Please bear in mind that: 1) This material is a transcript of the spoken word rather than a literary product; 2) An interview must be read with the awareness that different people's memories about an event will often differ, and that memories can change with time for many reasons including subsequent experiences, interactions with others, and one's feelings about an event. Disclaimer: This transcript was scanned from a typescript, introducing occasional spelling errors. The original typescript is available.
In footnotes or endnotes please cite AIP interviews like this:
Interview of David Spergel by David Zierler on November 17, 18, 2020,Niels Bohr Library & Archives, American Institute of Physics,College Park, MD USA,www.aip.org/history-programs/niels-bohr-library/oral-histories/45288
For multiple citations, "AIP" is the preferred abbreviation for the location.
In this interview, David Zierler, Oral Historian for AIP, interviews David Spergel, Director of the Center for Computational Astrophysics at the Flatiron Institute, and Charles Young Professor of Astronomy on the Class of 1897 Foundation, Emeritus, at Princeton. Spergel describes his transition to the Flatiron Institute and he shares that he will become the president of the Simons Foundation in summer of 2021. He explains his initial connection to Jim Simons and how the Institute differs from a traditional academic environment. Spergel describes New York City as a burgeoning center for machine learning both in academic and industrial research and he conveys his long term interest in determining the future value of machine learning to multiscale physics. He recounts his childhood on Long Island and what it was like to have a physicist for a father, and he explains his undergraduate experience in the physics program at Princeton, where Jim Peebles was a formative influence. Spergel describes his graduate work at Harvard where he worked with Bill Press on the solar neutrino problem and James Binney on orbital dynamics, and where he learned about superconducting cosmic strings. He discusses his postdoctoral appointment at the Institute for Advanced Study, where he became interested in galactic orbits and where he realized the value of the data coming out of COBE. Spergel describes his subsequent appointment to the faculty at Princeton and the promise of string theory at this time. He describes the notion of a multiverse as a non-scientific tautology and he explains why his favorite paper is Wigner’s take on the “unreasonable effectiveness of mathematics.” Spergel describes the origins of WMAP, the turning point this collaboration offered his career trajectory, and how the project allowed for pathbreaking avenues to measure the properties of the universe. He surveys the various ways that inflation, expansion, and acceleration of the universe fit with WMAP, and he explains how this collaboration is driving the next generation of experiments, and in particular, the impact in advances in detector technology. Spergel describes his involvement in the Roman Space Telescope Project and the budgeting challenges it has experienced during the Trump administration. He discusses his advocacy work in Congress on behalf of NASA. Spergel surveys his career as a teacher and a graduate mentor, and he describes how the culture of inclusivity at Princeton has improved over the years. At the end of the interview, Spergel shares his plans for the future of the Simons Foundation, he explains how he will attempt to remain close to the science, how he will use his new position to continue to promote diversity in STEM and to support cutting-edge research across a broad array of scientific endeavors.
This is David Zierler, oral historian for the American Institute of Physics. It is November 17th, 2020. I'm so happy to be here with Professor David N. Spergel. David, it's good to see you. Thank you so much for joining me today.
OK. So to start, would you please tell me your title or titles and institutional affiliations, and I pluralized both because I know you have many.
I am the Director of the Center for Computational Astrophysics at the Flatiron Institute, a new institute funded by the Simons Foundation. To give the full title, I am the Charles Young Professor of Astronomy on the Class of 1897 Foundation, Emeritus, at Princeton.
[laugh] David, let's unpack both of those a little bit. First of all, who was Charles Young and what is your connection to that named chair?
Charles Young was professor of astronomy at Princeton back in 1877. He had moved to Princeton from Dartmouth. I'd have to look up the history to be certain of it. The chair was set up back in 1927 when Henry Norris Russell, who was leading Princeton astrophysics, was offered a professorship at Yale. He was a Princeton alum, and his classmates endowed a research chair for him at their 30th reunion. It was one of the first research chairs endowed at Princeton, and it was a very nice one as it came with such substantial research funds. Henry Norris Russell held it for many years. After him, Lyman Spitzer held the chair. After Lyman Spitzer, Jerry Ostriker. After Jerry Ostriker, Scott Tremaine, and then I held it for a bit more than a decade.
And when did you go emeritus?
I went emeritus just over a year ago. I joined the Princeton faculty back in 1987 and was there until I moved to New York. When I first moved to New York to set up this new institute, I kept my Princeton position. I had a joint appointment for three years. I went back and forth. I felt that I sort of didn't want to hold two really good positions, so decided to take emeritus status.
It really wasn't fair to the field to clog the system up. Princeton had been very good to me. I felt that it wasn't fair to Princeton to be part-time in that position. I also felt by taking emeritus status and giving up tenure, I was sending a message to the Simons Foundation and to the staff at the CCA. I chose emeritus status at a very young age I was 59 when I become an emeritus faculty member.
I think that's actually worked out well. Princeton recently hired Eliot Quataert from Berkeley and that's a terrific move. I had been part of a series of really superb theoretical astrophysicists who held that chair. It was an honor to be on that list. I think Eliot is a great successor to that, so I was pleased how that worked.
New York has been really an exciting opportunity. Basically, Jim Simons turned to me and said, "We want you to set up a center for computational astrophysics," gave me the funds to operate it, the resources to hire about 60 people, and the flexibility to do it in a way that I thought was most effective. I got to help design what is a beautiful space that I'm in right now, which is this Flatiron Institute, and got to very quickly build up something that I think has become an important center. We did this by starting out with a lot of joint hires: David Hogg from NYU, Greg Bryan from Columbia and Rachel Somerville from Rutgers. We then did a series of new hires,: Will Farr and Phil Armitage with Stony Brook and Yuri Levin with Columbia. Will moved from Britain, Phil from Colorado. Yuri Levin moved from Australia to Columbia for a joint position. We then hired a whole series of joint junior faculty to new positions. So we've, hired two at Columbia, Colin Hill and Melissa Ness, Blakesley Burkhart with Rutgers, Ruth Angus with the Museum of Natural History, Ken Van Tilburg with NYU and Chiara Mingarelli and Daniel Angles-Alcazar with U Conn. We’re just about to pursue two new joint hires, one with CUNY and one with Cooper Union. We’re also hired about a dozen people full-time at the junior faculty level. After two years, Rachel Somerville moved to full-time, and Shirley Ho moved here from a tenured position in Carnegie Mellon and a staff position at LBL. In addition, we’ve got about 25 postdocs, five visiting graduate students who are here for five months, and a number of sabbatical visitors.
David, when did you connect with Jim Simons? How did that come together for you?
I had met Jim by being on an advisory board for him for math and physical sciences. I knew him a bit but not very well. He decided to set up this institute about seven years ago. He staged a retreat to Buttermilk Falls with a group of leading scientists looking for ideas on new directions for the foundation: he was expanding the Simons Foundation and wanted to think about what to do next. Ingrid Daubechies, the applied mathematician from Duke, at the meeting suggested, why don't you set up an institute focused on computational sciences? Ingrid may have wanted that institute at Duke. But Jim liked the idea and thought it would be interesting to set one up here in New York. He started by hiring Leslie Greengard to run a program initially called SCDA (Simons Center for Data Analysis) that grew into our computational biology institute, and then started looking around for other areas. I got a call from Andy Millis, who was, I think, associate director for physics at the Simons Foundation at the time, saying that they're interested in computational astrophysics, would I be willing to run a half-day workshop on what are the opportunities in computational astrophysics? I had no idea at the time that this was something that was going to lead to a position here. I just thought this was an opportunity to kind of promote the field and convince the Simons Foundation to make an investment in this area. I put together a half-day program, invited in Eliot, Lars Hernquist from Harvard, Rachel Somerville, Shirley Ho, David Hogg, and Greg Bryan. A number of the people who were there that day become involved in the center. We did a half-day program, at the end of which Jim asked me to come to his office and offered me this job on the spot, and sort of outlined what would be the resources available. I then took about a month trying to figure out whether I wanted to take this job. At the time, I was exploring another opportunity that I was on the short list for. I had been at Princeton for 30 years and had been department chair for 10. I was ready for the next challenge. This was an interesting challenge, and so I took this up.
Now, of course, the Simons Foundation precedes the Flatiron Institute, but does the Center for Computational Astrophysics precede your tenure or were you the founding director for that?
I was the founding director. Will this transcript be published after December 16th?
No. It won't be published before December 16th?
OK. Then, I can tell you that I'm going to then become the president of the Simons Foundation. That will become public knowledge December 16th.
So I'll actually stay here, and Jim Simons will announce his retirement and I'll start running the foundation here.
David, for better or worse, given your experience and those of many of your colleagues, to what extent is the business model of the Flatiron Institute sort of—I don't know if poaching is the right word, but making offers that are just extremely too attractive for people to retain solely academic appointments?
We actually haven't moved that many people to full-time positions. What I've tried to do is to make New York a real center for computational astrophysics. So I've have bribed universities to make hires.
The model that we've used for a lot of these positions is we've picked up half the cost of the position for the first three to five years and picked up the startup.
This makes it much easier for a university to add a new line, often commit to new hires in advance of retirements. I like to joke that the position is almost free for five years and then there's a new dean so that's the next dean's problem.
It's been very attractive to the surrounding universities. We've added nearly a dozen positions in the region in computational astrophysics. In the pre-COVID days, we were running about 50 workshops a year: many of them would attract a lot of the locals. We would run weekly meetings where people would be coming from Yale and Princeton and Stony Brook, and certainly from closer institutions [NYU, CUNY, Rutgers, Columbia]. We had a lot of people coming through from Europe: we were able to really take advantage of New York's location as a travel hub. I think that we have succeeded in the four years to really create an intellectual center where lots and lots of people would interact. I'm sure when COVID is done this will all come back. I felt that that was one of the kind of valuable ways we can use these resources. We've also tried to create new types of opportunities with the permanent positions: a number of the people we've hired are doing things that are really valuable in terms of developing codes and supporting those codes for the scientific community. Many of them worked on code development, something that are not directly valued in the universities, but important to the field
Some of this work is already supported in the national labs and places like Space Telescope Science Institute, but the focus there is often service for particular telescope projects or DOE mission-driven for the labs. We have the flexibility to support a number of people who are important to astrophysics and don’t fit into the universities or the national labs. I think that's been primarily how we've been using those resources.
David, it sounds like your collaborations and affiliations with universities confers all of the benefits of having those institutional connections with none of the constraints of actually being either physically or administratively located within a university setting. So I wonder if you might explain how that translates to what kinds of things the Flatiron Institute, and specifically the Center for Computational Astrophysics, has been able to do as a result of this very unique place that it exists in the larger ecosystem?
A bunch of different things. One of the programs we set up is a visiting student program where students come for five months here. We provide housing and full support, and they work on a project with some of the research staff here. And this is a model where kind of—and we've actually expanded this through the Flatiron Institute to some of the other centers (Computational Biology, Computational Neuroscience and Computational Quantum Materials). \Students doing computational work at their department and are often the only computational student, can come here and be part of a much larger community. Typically, most departments will have only one faculty member who thinks of themselves as a computational astrophysicist (or computational biologist). There's 60 people here on the staff. When you consider all the graduate students working with people here, there is more like 100 people coming here each week. I have four students from Princeton working with me who are coming here once a week, and the same is true of many of the people here. They've got students in surrounding places. So you have a community of more than a hundred people doing computational astrophysics that you're part of, and that's, I think, been very valuable for people. We set up a program this summer that worked very well, partnering with the National Society of Black Physicists. I felt that with COVID shutting everything down over the summer, it was clear to me in April that many people would lose their summer internships and not have opportunities. And I think for many people the chance to work before going to graduate school is just so important to have the experience of research.
The people who have been hurt by COVID the most are the people with the fewest resources.
I think you see that both in terms of the deaths from COVID among the African American and Hispanic communities, but I think it's also the people who are most hurt by school closures and cancellations of summer programs. So we partnered with NSBP. We had 20 students this summer working remotely with our research staff and they had a great time, and papers and projects and came out and everything. And one of the nice things about being funded by the Simons Foundation and basically—I have a budget that I control pretty much as I see fit for supporting research that I can move quickly when there's opportunities like that. It’s also the way we can move quickly when there's something interesting scientifically. When neutron star collisions happen, and we want to organize a workshop two weeks later, we can just do it. And we don't have to apply for some grants to do it, we just do it. So that's really been fun to have that ability to make things happen.
This is a question I posed to Geoffrey West at the Santa Fe Institute, and I think it's quite relevant for you, as well. I've talked to so many deans of science and department chairs about the way that COVID is making everyone rethink really fundamental things in higher education. And people are legitimately concerned about—even at institutions like MIT and Caltech, they're really concerned about what's the value of an education if it's held online. And so I want to ask you, in what ways does COVID perhaps offer opportunities for the Flatiron Institute in its education mission, in terms of just rethinking the way that science is taught and the physicality of where science is taught? Has COVID offered any opportunities for the Flatiron Institute to offer new models in physics education and science education generally?
There were some things we were doing already that put us in a good situation for continuing. So one of the things that was nice about building a building four years ago—or redoing a building—is that it was clear already that things like Zoom would be very valuable. So all of our classrooms are equipped with microphones in the ceilings and multiple cameras so every room we have is set up so that it could be used in the hybrid form. And so pre-COVID we were already running workshops where we had lots of participants in the room but also lots of people remote. And I think we will go back to that with even more remote participation. We were already recording all of our talks and we run summer schools where we have advanced topic lectures, and all that's posted to Zoom. And I would say this isn't a COVID change, we had done this before COVID, and I think we will just continue to do more of it. On the other hand, I have been suspicious of on-line education. Now, I taught an online class, I have an online course called Imagining other Earths. That was a freshman seminar that 40,000 students have taken. It's been available for eight years. So I certainly see a value in online education, but I do not see it in any way as a replacement for in-person education and the experience, perhaps most importantly, of working together with other students. In a way, the lectures could be online, but you need to be gathered in a room with others to do problem sets. That's how I learned physics as an undergrad. Some of the people I did problem sets with are still close friends. And I think that's very important for students, and I think they've really lost out this year at places like Princeton that went remote. I have a lot of respect for the places like Duke and Cornell that have made the effort to stay open.
And, David, just so I understand, the educational offerings of the Flatiron Institute, from every level from undergraduate to postdoc, do you see them as complementary to traditional academic environments or are any of those offerings understood to be alternatives to traditional academic environments?
So the model—we do not give out PhDs, we do not give out degrees. But, on the other hand, if you're a student and want to learn about machine learning and astrophysics, and that course is not really going to be open at your university, or you want to get really more deeply into computational fluid dynamics, we have advanced lectures and schools.
Advanced courses for specialists that we provide, often things a typical graduate program will not be able to offer.
And this is true even at the postdoc level, people are not being postdoctoral researchers at the Flatiron Institute, are they?
Oh, no, people are being postdocs.
Oh, they are? OK.
Yeah. We are competing with universities in many ways for postdocs, so we will compete with Harvard or Berkeley for a postdoc.
Right. So four years is a relatively—it's barely one academic generation. How have the postdocs fared counterfactually, to where some of these postdocs might have gone otherwise?
Well, our postdocs have done very well. We very quickly got excellent people and they went to good places. The counterfactual is very hard to know because the people we made offers to, had they gone to Berkeley, they would've done well there, too. Now, one of the things we've done again regionally is we've made a whole bunch of joint postdoc offers, too. So a lot of our postdocs are spending two years at Columbia, two years here, and going back and forth. I've used that to also lengthen the postdocs, so some of them are four or five years. I would like to see a world in which people did not have to move around so much as postdocs. I think that that's been—it's difficult on peoples' lives. I think it's something that—it hurts everyone. It tends to hurt female postdocs more than males because spouses or partners are less likely to move. So the ability to be able to offer for these longer-term things and having the flexibility to do these joint positions I think has been working well for the people involved, so that's been good. And with the postdocs—one of the things I do this afternoon is I will chair the meeting to look at the candidates for the next set of hires—is we're hiring them as a cohort rather than hiring them stove-piped to particular projects attached with individual grants. And I think that tends to give a more diverse group of postdocs in multidimensions, certainly more diverse in terms of gender, and we at least have a modest number of people from under-represented groups, but also diverse in terms of we have a bunch of people whose backgrounds are more in particle physics or computer science. And this, I think, creates a place where we can learn from them and they can learn from us. So being able to view things as a cohort is sort of an opportunity to structure things differently.
And it certainly sounds like there are multidisciplinary opportunities that are more easily transversed than they would be in traditional academic departments?
Yeah. Now, one of the areas where I think we—and, again, it's four years, so I think we've had an impact but it's very early. I invented this place from scratch and made up all these things from scratch, so I'm seeing how things go as they come along. One of the areas where we have made a big investment in machine learning and its applications. All of our centers here have people working in machine learning. Shirley Ho, who is running our group here, has built up a pretty big group of people. She has a crazy number of students working with her—20 graduate students all around the country working with researchers here on machine learning applications. David Hogg is also training a lot of students and postdocs in statistics and machine learning. There's a lot of interest among the students in applying machine learning to problems in physics and astrophysics, and not that many people trained in it. We’ve been working pretty closely with people at Center for Data Science at NYU and some of the folks like Bin Yu at Berkeley. We've been learning a lot from the machine learning community and we've become a place where people can learn statistics and machine learning and apply it to astrophysics. And I sort of see that as one of our opportunities is to move into new areas like that.
And either for financial need or for building those institutional relationships, is there ever any need to look for funding sources beyond the Simons Foundation, like the NSF or DOE or DOD?
Because of our tax status we cannot accept funding from them.
Now, that's here. Now, remember, lots of people have joint positions.
So I actually hold grants at Princeton. Many of my colleagues here with joint positions are holding grants at Columbia or NYU or Rutgers. The model we had is that—again, it's four years old, so, like, the model is still being invented—is that the graduate students are sitting at their home universities. The grants supporting the graduate students who are sitting at the universities. The postdocs are being supported entirely here.
The graduate students typically spend three, four days a week at their home university. They come here one or two days a week. One of the delicate balances with the surrounding universities is to make sure they feel that we are complementing and enhancing what they're doing; we're not stealing their students and faculty.
Yeah. And I can think of a million examples, but one that just comes to mind is DARPA's interest in quantum computing must have a lot of overlap with the Flatiron Institute's interest in quantum computing.
Some, but we're interested more in quantum materials, so Antoine Georges' and Andy Millis' focus is on solving many electron problems in quantum materials which is different from the quantum computing. Our interest in machine learning is more resonant with a lot of the federal funds going to machine learning.
So one of the things we're doing—and I think there will be opportunities to do even more of this—is we've been partnering with NYU, Columbia and Princeton on proposals for centers for machine learning applications in the sciences. This is something where we're also taking advantage of the fact that New York is one of the centers for machine learning also on the industry side. Within a mile radius of where I am now, we've got Facebook's research center with Yann LeCun, and Samsung's and Microsoft and Google's big research centers. Pre-COVID, we were running—and this is Flatiron wide, but led out of astrophysics—weekly machine learning seminars. These attracted a lot of the industry researchers to. One of our abilities is the power to host meetings, where it's easier for someone from Facebook to meet someone from Google here than at either building.
It's even easier for NYU and Columbia to meet here.
Sure. It's a safe space.
Exactly. One of the things we've been trying to do is encourage those conversations. I am convinced that machine learning will be useful for physics. There's lots of little things that you get to do better. What's not clear to me yet is whether it will be transformative.
I’m working on two big applications of machine learning to physics with various students and postdocs. One is a big program of forward-modeling the universe: start from initial conditions and run them forward to match our observation of the large-scale structure of the universe. Our goal is to infer the initial conditions of the universe.
Now, to do that fast enough so that you can recover the 1011 values that you need for the initial conditions, you can't run an N-body and hydrosimulation that does galaxy information that many times. We are developing neural nets that learn from the hydrosimulations. We’ve run some of the largest set of N-body simulations and some of the largest sets of hydrosimulations here. We have a lot of computing here. We’re developing machine learning techniques to learn how to forward model from the simulations. This work raises interesting kinds of computer science questions about including symmetries in the neural networks. My hope is over the next five years we'll have the ability to do this forward modelling. We're now working on the inference problem with, if we could forward model, what's the best way to do the inference problem?
In determining the threshold for what counts as a transformative advance in physics, in this model, what are the foundational questions that you're after? Are these things like discovering dark energy, dark matter, that kind of thing?
That's one set of things, and we can make progress on that. The other—and this is the second strand in my current machine learning work —is can we do multiscale physics with machine learning? So the way I think about this is, let's say I have some small box that I can simulate and evolve forward in time. I want to coarse grid the small box it so that I can have a set of rules describe what's happening in the small box, embed it in a bigger box, and evolve it forward and get statistically the same kind of result in the small box. So I can take a problem—and one of the things we've been looking at for this is turbulence and magnetohydrodynamical turbulence. So turbulence is, I don't know, a 150-year old problem. We know the underlying physics, but we can't simulate the dynamical range needed for the laboratory, and we'll probably never be able to simulate the dynamical range needed for astrophysics with just direct numerical simulation. So we always have to come up with subgrid models. And the way that I think about a subgrid model is I have some—say I could do all the physics right and I had some box with a billion numbers that I want to evolve forward and capture in cloud formation what's happening on the droplet scale, and turbulence what's happening on small scales, can I learn the laws of physics I need to have a more accurate subgrid model, to do something better than approximating the small scale physics with an effective viscosity? And we're working on a couple different approaches of doing that using machine learning. I think of machine learning as being an effective way to get a low-dimensional representation of a function in a high-dimensional space. The classic machine learning problems I think about is I give you lots of images of cats and dogs. Cats and dogs are pictures, and let's say with a megapixel camera so they represent points in a million-dimensional space and intensity as each picture, and I want to draw a surface in that million-dimensional space that separates cats from dogs. So I have a low dimensional representation of the key distinctions in that high-dimensional space. So what I'd like to do is have a low-dimensional representation of, say, something like turbulence where I describe it by more than a single number, an effective viscosity. I'd like to have a representation—I don't know, 10 numbers, 100 numbers. Can we use machine learning to do that? If we can, that to me would be transformative: there's a whole class of problems in physics that we can’t solve now. multiscale physics problems. These might now become doable. So that would be a fundamental progress not in understanding the basic laws but our ability to solve and make predictions in complex situations.
So for one obvious really big example, in just your understanding of what gap machine learning might fill, how might machine learning get us closer to understanding how gravity works with the standard model?
I'm not sure. So I see it more—the two strands we've been looking at are, can we use it to make more accurate predictions from observations? Can we get at things like neutrino mass? Which would be nice to know—I don't know at the end of the day how fundamental it will be—but by getting more value out of the observations we take. And the other is, can we address these multiscale problems?
Well, David, let's do some oral history. We've been talking—
—in this amazing discussion about all of your present—it's so exciting to hear that, but let's take it all the way back to the beginning. I'd like to start first with your parents. Tell me a little bit about them.
My father was a physics professor, so I'm in the family business. My father grew up on the lower east side of New York. His parents were immigrants from Eastern Europe who came over around 1910. My grandfather actually before the first world war he came from what is Galicia was then part of the Austria Hungarian Empire. My grandmother on that side came from the region around Warsaw in Poland. My grandparents ran a grocery store. My father went to college. He went to RPI and then Rochester where he worked with Bob Marshak.
And then, the story my dad told was Bob Marshak told him that he would make sure that my dad never get an academic job. After graduation in 1965, my father went to work for Grumman Aerospace.
In the '60s, City University was expanding,. He moved to be one of the founding faculty at York College in Jamaica, which serves mostly first-generation kids growing up in Queens. He was a physics professor there and eventually became the chair of natural sciences at York College. He retired a number of years ago. My mother grew up in Queens. She went to Cornell School of Home Economics [Now, Human Ecology]. She then would go on to teach high school for many years: high school home economics. She got tired of that and went to law school in her late 40s. She practiced for a few years and then taught law classes at York. I grew up in Huntington, Long Island: a comfortable middle-class upbringing with academic parents.
Your parents are Jewish?
My parents are Jewish.
Did you grow up Jewishly connected at all?
I did. I was bar mitzvahed. My father was the vice-president of education at the conservative synagogue, so I went not only to Hebrew school but Hebrew high school. I can read Torah, I know the tropes, I know a little bit of Talmud. I went to public schools, but am modestly knowledgeable—not deeply knowledgeable, but modestly knowledgeable about Judaism.
Did your father involve you in his career at all? In other words, did you know what it meant to be a physics professor even when you were a kid?
Little bit. Not deeply. And I think in some ways one of the things that was helpful for me—well, certainly, I was exposed to it—but I also had a sense that it would be a career that I would enjoy. And while I certainly ended up going on a track that took me to sort of R1 top university path, I knew that I would be comfortable teaching physics at a mostly teaching institution. That to me was like a good thing to do and would be a fulfilling life. 'Cause I saw my dad was happy doing what he did. And I think that, looking at a lot of the students and postdocs who've worked with me, a lot of them feel real pressure to—if you're not at MIT or Berkeley, you're somehow just not a success. And I just think that was really good to have that background.
When did you start to get interested in science? Was it early on?
Pretty early on. I was interested in math and science. I was on the high school math team and did well in the competitions. I participated in what was then called the Westinghouse Science Talent Search.
And you probably benefited from excellent science and math education in high school?
I went to a good suburban high school, and I had some good teachers. It wasn't off scale great. It wasn't like Stuyvesant, it wasn't the very best, but it was a good education. I also read a lot
And when you were thinking about undergraduate programs to apply to, were you specifically thinking about physics programs?
I was thinking about physics programs. I was also interested potentially in going into law. When I applied to colleges, I was debating between physics and law. Back in those days, you could apply early action to multiple places, so I applied to MIT, Yale, Princeton and Harvard.
What was your batting average?
Three out of four. Everywhere but Harvard. And I was on the waiting list at Harvard.
I wonder if your father, just being in the field, if he had useful input on the different kinds of physics programs that you were considering, because these are schools with very different approaches to physics.
He gave me good advice along the way, but that particular decision less so. That particular decision, that was actually more driven by the non-physics aspects of it. Yale physics at the time—may even still be true—I felt that Yale didn't treat physics like its most important department; it prioritized humanities. So it came down to Princeton and MIT for me, and then I decided I wanted to be in the liberal arts environment.
Where you could focus on physics but have a broader education?
Did you do a good job of taking advantage of that?
Yeah, I did. I'd usually take one extra course a semester beyond the standard, so five classes a semester. I studied art and music and a lot of history and political science classes.
Who were—this is not a question I would normally ask because undergraduates, their worldview is rather narrow, but given the influence of your father, I wonder if you were alive to the concept of trendiness in physics or what was considered at the time the really important research that was going on or the luminaries in the faculty? On that basis, what was really exciting during your time at Princeton as an undergraduate?
I was very lucky as an undergraduate in terms of who I was exposed to. I took, as a sophomore, quantum mechanics from Jim Peebles and really liked Jim as a professor. He was a very good teacher.
He taught quantum mechanics when I was a sophomore and stat math and thermodynamics when I was a junior. I ended up taking four out of my eight core physics classes from Jim Peebles.
Basically anything he taught you wanted to take?
Exactly. During my sophomore year, Jim gave a lecture on the current state of cosmology as the last lecture of the quantum mechanics course. I really regret that I can not find the notes for this class. Jim and I discussed this, and he doesn’t have the notes either. I also asked several of my classmates— no one seems to have the notes. He gave a lecture on problems in cosmology. He talked about the flatness problem. He talked about dark matter. He talked about his understanding of cosmology in 1980. I was at the Nobel Prize ceremony when he got the Nobel Prize and I've known him well I really regret I don't have those notes and he didn't have those notes: it would have been interesting to capture his view at that moment. As a Princeton student, I knew about dark matter and things like that very early and understood that it was an important problem. I did my first junior paper with James Binney, who was a visiting assistant professor at the time, in galactic dynamics, and that was wonderful. My fall semester junior paper turned into a published paper. During the spring of my junior year I worked with Jill Knapp and developed a proposal to observe at the Very Large Array. As a senior, I observed at the VLA and at Green Bank. I had real significant radio astronomy experience. It snowed the entire time that I was at Green Bank so that didn’t lead to any science. The VLA observations lead to my second published paper: a mixture of observations and modelling.
What was Knapp working on when you connected?
She was working on asymptotic giant branch stars.
My senior thesis looked at the transition of asymptotic giant branch stars to protoplanetary nebulae. We focused on how at the end of the mass loss phase the star ionized the gas around it. In my junior paper, I made estimates showing that the radio emissions should be detectable. I detected it in the observing run at the VLA in the Fall of my senior year and then spent the spring modelling the data. I was observing at the VLA when it was about half built and pretty quickly learned the software needed to do the mapping.
David, I love to ask this question because it means different things at different places at different times. If you can isolate in your memory where these terms were and how they were used at Princeton when you were an undergraduate, both substantively, scientifically, and administratively, where is cosmology, where is astrophysics, and where is astronomy?
Princeton had two department then and it still has two departments: a physics department and an astrophysics department. The astrophysics department was always astrophysics, it was never really astronomy. The focus on theoretical astrophysics goes back to Henry Norris Russell applying ideas from physics to understand spectra, and Spitzer studying plasma physics. Theory was always the focus at Princeton, perhaps because it was not close to a mountain. The department was called the department of astrophysical sciences. I think it was originally department of astronomy. I think Lyman Spitzer made it the department of astrophysical sciences in the '50s when he set up the Princeton Plasma Physics Lab and wanted that to be part of it. The physics department didn't want plasma physics. I believe that there was a third branch back in the 1950s. The physics department didn't really want atomic physics.
But Spitzer wanted to be able to measure things like oscillator strengths to interpret spectra. So the astrophysics department absorbed all those things into astrophysical sciences. Subsequently, in physics decided that atomic physics and cosmology were interesting areas.
Would anybody have called themselves a cosmologist at that point?
I'm pretty sure Jim Peebles would have called himself a cosmologist.
When I was an undergraduate, the senior people in cosmology were Jim Peebles and Jerry Ostriker, who was chair of the astrophysics department. Ostriker and Peebles had written papers back in '74 already on dark matter. I'm an undergraduate from '79 to '82.
I think in writing those papers, Jim would've seen himself as the cosmologist, Jerry as an astrophysicist.
When I was an undergraduate, I got to know Ethan Vishniac, who had been a graduate student with Bill Press and was then a postdoc with Jerry Ostriker pretty well.
So you connected with Bill Press at Princeton, not first at Harvard?
No. I connected with Ethan. Ethan influenced my thinking about where to go to graduate school. In my graduate school application said that I wanted to do cosmology. I was interested in the growth of perturbations, how galaxies emerged.
This was around the time Bardeen was doing his work on gauge-invariant formalism, and I liked general relativity, I was intrigued by all that, so that was my plan, to go work on that. So I knew I wanted to do cosmology at that point.
Where was general relativity at Princeton when you were an undergraduate? What kind of exposure did you have to GR?
I took a course in general relativity with Rich Gott in my sophomore year, which I really enjoyed. Sitting in my office, I can see on my shelf one of the books that convinced me to become a physicist, Misner, Thorne, and Wheeler’s Gravitation. The book had track one (which was for undergraduates) and track two.
We learned track one as undergrads. Princeton finished its classes in December, finals didn't start 'til mid-January. You had sort of four weeks between the end of classes and the start of finals. I actually read most of Misner, Thorne, and Wheeler during that month in my sophomore year. I was home with my family. I would read the book when I was home in Long Island When we went skiing and I would read at night and ski during the day. I would spend two to three hours each day reading Misner, Thorne, and Wheeler. I just thought that was so beautiful.
And given your interest in astronomy, did you ever consider for graduate school a nontheoretical avenue of pursuit?
No. I wanted to be a theorist. I did enjoy doing the observations, but loved theory. For graduate school, the debate for me was between Harvard and Caltech. Princeton did not, and still does not admit its own undergraduates. I think it's a good policy—There was a view that you should go somewhere else. I debated between working with Roger Blandford at Caltech and working with Bill Press at Harvard. There were probably three factors that led me to go to Harvard. One was, my girlfriend my senior year had applied for and gotten a Rhodes scholarship to go to Oxford. I had written to James Binney about spending a year at Oxford and decided that would be worth doing. By the time all that happened, that relationship ended.
But I was still interested in going to Oxford and Bill Press was supportive of my taking a year off. In fact, we would manage to count it as a Harvard enrollment year on leave to Oxford. Roger Blandford was not supportive of that at the time. The other was when I went to visit Caltech, I asked the graduate students, "How do you meet women here?"
They told me, "You go to Pasadena Community College. You sign up for a sports class. You pretend that you're not a Caltech student and you can meet them. On the fifth date you confess you're at Caltech."
I went to Harvard to visit, and there was a steady flow of undergrads from Princeton to Harvard, so I knew a lot of people there already, and it was clear that I would have a social life.
Or at Caltech you could just be Bill Press and be the kid of a Caltech professor and marry another kid of a Caltech professor.
That was a possibility.
Also, I just hit it off with him [Press]. We got along well.
What was Bill working on when you first connected with him? What was his project at the time?
His big project was his book, Numerical Recipes. He was working on writing that. Before Harvard, I went to Oxford for a year, I worked with James Binney, we write a couple more papers together. I learned a lot of dynamics. I went to some of my first conferences. When I show up at Harvard as a graduate student, I am pretty far along. I had taken my AP tests in high school. I didn't have an AP physics in high school, but I just taught it to myself. And, of course, there it was helpful, if I had a question, I asked my dad, but I mostly just self-taught. So at Princeton I took, as an undergraduate, most of the required graduate classes my senior year. I was mostly taking graduate classes. So I took a class with Lyman Spitzer. I think I am the youngest—I know I am the youngest person to have taken a class with Lyman Spitzer.
[laugh] And what was the department at Harvard that you were part of? It wasn't the physics department?
I was in astronomy.
I was in astronomy. I did sit in on the field theory classes at Harvard physics, so I was in class with Lisa Randall and the people like that. And actually at Oxford I got to know Brian Greene who was there with me. We worked together some. It was my first year in graduate school. So when I got to Harvard for graduate school, I had taken most of the graduate classes at Princeton, so I took the exams for the Harvard graduate classes when I arrived in September. This meant that I could skip them and graduate more quickly. I would go on to take my qualifying exam at the end of my first year. I only spent two and a quarter years at Harvard as a graduate student. I got my PhD very quickly. When I showed up in Fall 1983, Bill Press suggested to me to work on the solar neutrino problem and to work on an idea he had of having dark matter particles in the star serve as a way of transporting energy efficiently.
This was a dynamics problem. I had to compute the behavior of orbits of dark matter particles in the regime in which they had occasional collisions. It turned out to be a problem that I was very well suited for as I had spent the year with Binney learning about orbital dynamics.
It was just an orbital dynamics problem now with scattering, and I was able to develop a kinetic theory for it. This work led to our first paper on transport and looking at the dark matter properties. We realized the dark matter would annihilate, and Katie Freese—that was now my second year there—Katie Freese had shown up as a postdoc to work with Bill. Katie and I worked together pretty closely on that, and we wrote a paper with Katie and Lawrence Krauss on dark matter annihilation.
That was a big deal, that collaboration?
That was a big—yeah, back then. When things later came out about Lawrence Krauss a few years ago as part of the Me Too movement, Katie and I emailed each other and said, we were ahead of the curve. We knew he was a problematic guy 30 years ago.
But I won't go into more detail on that experience, but—
But 30 years ago is a very long time ago in terms of what's acceptable and what's not, or what passes and what's not.
Well, he was very aggressive. I was a second-year graduate student, and he was pretty aggressive with intellectual property.
Bill Press protected his students, so I was ok. Soon after we finished our paper, Katie had a visitor, Andrzej Drukier, who was working on dark matter experiments. Katie brought the three of us together, we started talking. He talked about what one could do with these experiments. I started thinking about what the dark matter experiments would do. I realized that there would be this modulation effect; that, as the earth moves around the sun, its velocity relative to the dark matter halo would change with time. That would produce a modulation in the dark matter signal, that there'd be a higher rate in June when you're moving faster relative to the dark matter halo and the earth's motion is parallel to the sun's motion around the galaxy. In January it's antiparallel so there's a lower rate. And one of the ways you would know there's dark matter there is seeing this annual modulation. Now, the Italian DAMA experiment has claimed to see this signal It's certainly a statistically significant signal. However, they still haven't convinced most of the community that this is an astrophysical signal as opposed to background, but this was the distinctive signal, and we published that paper. And then, the final part of the thesis work was a collaboration with Andrzej, Graciela Gelmini and Frank Avignone. Frank Avignone and his group had a germanium experiment that was measuring germanium double beta decay. Andrzej recognized that this was a good experiment, it was very sensitive, it might be one we could constrain dark matter with. We did the first analysis actually of dark matter constraint from an experiment and we had published what would be the first dark matter exclusion plots as functions of dark matter mass and cross-section, these are now standard. We were actually able to rule out heavy fourth-generation neutrinos as a dark matter candidate. This paper was the fourth piece of the—or fifth piece as the thesis work. At the time, I also was working with George Field and Alex Vilenkin on cosmic strings. We met at the weekly early universe meeting that Alan Guth, Alex Vilenkin and Lawrence Krauss organized. This meeting brought together people from Harvard, MIT, and Tufts. I learned about the ideas of inflation and cosmic strings and things like that from the meeting I would write a paper with Alex Vilenkin and George Field on superconducting cosmic strings, an idea that Ed Witten had developed and looked into those. So I sort of learned a lot of stuff through those gathering.
David, I'm curious about your dynamic with Bill as your graduate advisor, part of the thing that's so fun about him is that he's perennially young. He's always young no matter where he is, right? Given how young he was looking back, did that affect your dynamic? Did you work with him more in a collaborative way than you might have with the more senior person?
Bill was always dynamic, but he—I did not think of Bill as young.
Even though he must've been, what, like, 32, 33?
He was, like, 33, but I was, like, 21.
Yeah. You're both young.
So we were both young, right. So I remember—every now and then I do this calculation where it's like, OK, I'm now—Bill back then seemed not quite old, but he wasn't young for me. Bill seemed middle-aged at 33. And I'm now 43, or I'm now 50, or I'm now 59, and, like, Bill, already—I now look at people who are 33 as so young, but, yeah. No, Bill was department chair, and I hadn't appreciated Bill's youthfulness 'til I got much older.
He was my advisor. He was older. The same was true for James Binney, who I had worked with earlier. I did my first paper with James. I was 19. James was 30. I thought of him as much older than me—the dynamics group was James and Martin Schwarzschild. Martin was then 65—maybe 67. To me, they were just older.
How closely connected was your research with what Bill was doing at the time, just in terms of as a symbol of your intellectual partnership?
Bill was, I think, spending probably most of his time on his Numerical Recipes book, but the thing we did together, we were collaborators. I would say the first two papers, this Press and Spergel, Spergel and Press paper really followed a model that I have often used as an advisor—Bill gave me the idea, I developed it, I would meet with Bill pretty regularly, present what I had done, he would encourage me, make suggestions. It was a mentor-mentee model. The next papers were with Katie Freese, and that was as collaborators. Katie was a couple years older than me, but she and I pretty quickly collaborated as equals, bouncing ideas off of each other. Bill was actually not a co-author on some of those papers, but we would be telling him what we're going. He was a good mentor for me in that, when I made that transition to that kind of work, he encouraged that. That was the next step to my intellectual development. Later, when I was a postdoc at the Institute and also as an assistant professor, Bill and I worked together on a couple of projects. At that point, we were collaborators.
Who else was on your committee, your thesis committee?
It was Bill Press, George Field, and it was supposed to be Alan Guth for the exam, but Alan couldn't make it, so Lawrence Hall was on my committee. John Huchra was the fourth member of the committee.
Ah. Alan would've been an outside reader from MIT?
Did Harvard generally have outside readers for astronomy theses?
No, but Alan was sort of outside but not too far outside. He was spending a day a week at the Center for Astrophysics.
One of the things that Irwin Shapiro, who was director of the CFA, recognized was that the value of science at this new intersection—it was an exciting time in this area between particle physics, and astrophysics and cosmology was an area with a lot of possibility. Irwin had Lawrence Krauss come up a day a week from physics, had Alan Guth there a day a week. I think he was paying Alan as a consultant—I think. Not 100% certain but I'm pretty sure of that. So he was around.
After you defended, what were the most compelling postgraduate opportunities for you? What were the things that you were considering?
Before I defended, John Bahcall invited me to come to the Institute for Advanced Study for a week and offered me a long-term membership at the IAS,. It was a five-year postdoc and it was very clear that that was a great opportunity. I didn't actually apply for any jobs.
They reached out to you and it was a done deal?
Yes, I didn’t apply. They offered me the job.
And the idea was the Institute would've been head and shoulders above anything else that you might've reached out for so why bother?
Yeah, yeah. That was such a great opportunity.
How much of the attraction was working with John Bahcall and how much of it was just being in that intellectual environment for you?
A bit of both.
And also, I knew I would return to interact with astrophysics at Princeton.
So you saw that as part of the deal, that you could have that opportunity to be part of the broader physics and astrophysics community at Princeton proper?
Yeah. When I was a graduate student, Jerry Ostriker came up to visit Harvard for a week and told me that they would do a faculty search at Princeton and I would likely be would one of the top candidates.
I knew that if I were to go to the IAS, I would be on track for this faculty search, too.
So you did see the postdoc as sort of a soft landing to a faculty position at Princeton?
That's right. I actually defended my thesis in December 1985. I stayed at Harvard 'til May and then took off for about two-and-a-half months to travel around the world. I got one of these round-the-world tickets and went with a friend and went all over. Then went to the Institute. They did the faculty search in the fall of my first year as a postdoc. I was offered the faculty job—I don't remember the date, but winter or spring, sometime that first year. I negotiated to start my faculty job but be on leave the first year. I just stayed at the institute for two years. My second year at the institute I was on leave from the university and then moved over as an assistant professor.
How much exposure as an undergraduate did you have to the Institute? I mean, was this a mythical faraway place for you?
It was a mythical faraway place.
Yeah. Were there even opportunities for undergraduates to be invited? Would they even be welcome at the institute for seminars, even precocious undergraduates such as yourself?
Nope. No. I would try to change that in later years, never quite successfully. There was this Tuesday lunch, which was famous, where people would come—and this is before the archives, so it was a time at which a lot of new results were shown there—and people would sit around this U-shaped table. And when I would first come there as a postdoc, John Bahcall, Lyman Spitzer, Martin Schwarzschild, Jerry Ostriker would sit at the head of the table, and the rest of us would sit below. It was really pretty hierarchical.
And what was John Bahcall working on when you connected with him?
A solar neutrino problem.
And also some stuff in galactic dynamics.
Did you see this in terms of Bill Press's interests and motivations for you career as this was very sensible in terms of where he wanted you to go and who he wanted you to work with?
Yeah. He encouraged me to work with John, both because of John's role as a mentor and John was very influential in being able to place his people in top positions. If you look at who went through the institute in sort of that 20-year period, John's first postdoc is Martin Rees, and then it's Roger Blandford, it's Lars Hernquist. You look through the list of people, say, who are now in the National Academy and who are theoretical astrophysicists, and a substantial fraction of them were postdocs with John. So it was clearly a good opportunity to go there as a postdoc.
Beyond the sort of cartoon version of the Institute as a place where geniuses just gathered for tea and discussed big ideas, in what way in the day-to-day was being in that intellectual environment specifically useful to your career at that point? Why was it good to be there for you?
There were some smart people there. In many ways, one of the most productive things that came out for me of that year or two years was Leo Blitz was on sabbatical there. Leo Blitz was a radio astronomer at the University of Maryland. He was doing a lot of work in galactic dynamics, galactic structure with molecular cloud observations. I was interested in dark matter and I had gotten interested in dark matter experiments. And I knew for dark matter experiments we needed to know the dark matter orbits. And I knew if the dark matter was on radial orbits it would be different than the dark matter being isotropic in terms of its prediction for experiment. I also knew as a dynamicist I would make the galactic halo turn into a bar if it was on radial orbits, so I wanted to constrain it. So I started talking with Leo about this and this grew into a collaboration where we wrote a couple of papers, one showing that the dark matter halo had to be relatively spherical because of the way the disc rotated. We used the HI observations to trace the distribution of gas and showed that the gas was on close to circular orbits: the halo had to be relatively round so the dark matter orbits couldn't be too radial. And that got me interested in galactic dynamics and the idea of a bar. Leo and I then started looking at what could we tell about a bar in the Milky Way. I realized that there would be this perspective effect that if our galaxy was barred, the near side of the bar would appear brighter than the far side and would seem more extended, and we just sort of worked through what it would do. Leo realized that there was this infrared data that was just published from a balloon flight from a Japanese group that we could look at. We were able to show evidence for that from that flight. COBE was launched at this time. The COBE data was not yet public, but the COBE team would then reanalyze the results, and we would subsequently analyze the COBE data. This analysis gave the first photometric evidence that our galaxy was barred.
There was earlier work by de Vaucouleurs based on the gas motions in the inner galaxy that suggested our galaxy was a barred galaxy, but I think our work played a pretty significant role in shifting the paradigm of how our galaxy is viewed. We convinced many people we lived in a barred galaxy. I started this work as a postdoc. By the time, we published the papers, I was already a second or maybe third year assistant professor. One of the big benefits for me of being at the IAS was the opportunity to work with Leo and learn from him there. He was another important mentor for me.
David, where was string theory at the Institute when you were there? Was this a big deal at that time?
It was a big deal at the time.
Who was talking about it, what were some of the exciting developments around string theory, and did you see any relevance of this work to your research?
So, yeah, string theory was already a big deal when I was at Harvard. It continued. Ed Witten was the guru of string theory and was around. The string quartet papers with Gross and Harvey, those were all happening. String theory seemed very promising, seemed to be a path to a predictive theory of everything. I was very interested in seeing what ways could we say something interesting about string theory from cosmology.
Among the things we looked at when I was a graduate student at Harvard was what happened if you had E8 x E8 theory. It implied the existence of a whole another sector that looked like ours. We asked what would that mean in terms of inflation? So we thought through some of that. Ed Witten wrote this paper on superconducting cosmic strings and one of the things I learned a lot from at the institute was thinking through what that would mean physically. I worked on superconducting strings with Jeremy Goodman, who was another long-term member at the Institute who had moved with me—the two of us would be hired together to become junior faculty at Princeton. He and I worked with Tsvi Piran, who was visiting faculty at the Institute, and wrote a paper where we worked through the electrodynamics of superconducting cosmic string. We basically took Jackson’s book on classical electrodynamics, which treats a charged point particle moving on a world line, and worked out what that would do for a superconducting cosmic string moving on a world sheet. So that was fun. And then I started working with Bill Press actually on the electrodynamics of cosmic strings: could you make a dynamo out of a superconducting cosmic string?
David, just to fast forward to the present on string theory, of course there are two camps, and the range on either side is there are people who have lost patience altogether that string theory is going to yield anything fundamental, and then the other side, there are people who think that lots of exciting things are happening and will continue to happen in string theory. I don't have to name names. You can probably think of who represents each side of the spectrum very well for yourself. But I'm curious, to fast-forward to now and looking to the future, where do you see yourself on that spectrum?
Closer to the optimist about string theory in that I see—string theory has certainly taught us a lot of interesting mathematics. It's turned out to be a useful way of connecting different areas of math and we certainly learned some things from there. The AdS/CFT correspondence has proved to be a powerful way of understanding complex theory. String theory has not lived up to our hopes as being this clear path towards a theory of quantum gravity, but there's sort of no guarantee that we will figure out quantum gravity in the next 50 years. So it may turn out to be the right direction, but we won't know for a long time. On the other hand, I do think we put too many eggs in one basket. I do think it was a mistake to encourage so many of the brightest minds in physics in the past three decades to go into string theory. Maybe if we had half the number of people go into string theory we would've made as much progress and had people explore other ideas. But as a cosmologist thinking about the early universe, I think that string-inspired theories are useful things to think about. On the other hand, I don't think we know that string theory is the extension of standard modeling that will work in the end.
I wonder, David, if we can go back to an earlier part in our conversation. You know, historically one of the criticisms, of course, of string theory is that ultimately it dead ends into untestable propositions with things like multiverses.
Where might machine learning bridge that gap? Where might machine learning offer modes of testability that currently don't exist in defense of string theory?
I don't see it. Machine learning has offered us ways to explore high-dimensional spaces effectively. If string theory had calculations that needed to be done that you couldn't do because of the dimensionality of a problem, machine learning might be able to help. The problem for string theory with multiverses is not that they can't do the calculation, it is the answer depends on unknown priors.
I'm very Popperian in my approach to problems. If things aren't falsifiable, they're not interesting, and the multiverse to me isn't interesting. I had the good fortune on working cosmologic microwave background at the time at which—we know stuff now that we didn't know in 1980, and that's great. I don't see the multiverse advancing. I don't see ideas being ruled out. I have worked on ideas like the formation of structure with cosmic strings and the formation of structure with textures. While speculative, those ideas to me were interesting because they could be wrong. The multiverse is not a testable set of ideas, so it's just not interesting to me.
But that's separate from saying, if push comes to shove, do multiverses exist or not? That's a separate statement.
That's separate, but that means it's a statement we'll never really know the answer to.
So there are things that could be true, but we can't disprove or prove them. That, to me, is—
So you're saying as a physicist and not a metaphysicist, it's just simply not interesting to you?
Yeah. It's just I don't think it's likely to advance. I think of myself as a pretty practical physicist. I want to have theories that have implications for observations. Whether or not they're observations or experiment that we can do today. We don't have to be able to do that experiment or observations today, but that actually makes eventually testable predictions. I mean, that's why to me things like strings or textures in the early universe were interesting. That's why right now I have a student thinking about neutrino-neutrino interactions and how they might affect the microwave background, because you can test and disprove it and learn something. I think string theory is not untestable. There are testable ideas that come out of string theory: for example, the existence of lots of light axions, and Compatification leads to many dilatons. There are other potentially observational consequences. String theory often produces things that look like Chern-Simons terms in the Lagrangian when you compactify the Calabi-Yau manifold So that, then, has interesting potential physical implications. For me, this makes the string-inspired models, makes it worth understanding. String theory has certainly revealed a very rich mathematical structure, and people have learned a lot exploring in this mathematical structure. It's not something I've done personally, but I think there's a lot of great work there. The multiverse stuff, there's just not—it's not very deep. The anthropic principal is not a rigorous concept. If you talk to a philosopher who's really thought more rigorously about the philosophy of it, they often discount the work. To say that the universe exists because we're here, it’s a tautology. To me it's a bit like someone asking you, like, why are you wearing a red tie today? And my answer is, you're only asking that question because I'm wearing a red tie.
It's a tautological answer.
It doesn't guide you to the next step. So to go back to that 1980 lecture by Jim Peebles that I went to as a sophomore, one of the things he talked about was why the universe is so close to flat. He talked about the fact that it required this fine tuning, that back at Big Bang nucleosynthesis, omega was different from one by a part in 1024and one of the ideas put forward for that was anthropic: if it were any larger, the universe would have collapsed. If it were any smaller, the universe would be empty. In the time between when I attended that lecture and when I went to graduate school, Andrei Linde, Guth, Sato, Steinhardt and others developed inflationary theory. It offered a dynamical solution to that problem. Whether inflation is the ultimate final theory or not we don't know yet, but it's been very successful in fitting a host of observations.
And when you say, "dynamical solution," what do you mean? What does "dynamical" mean in this context?
The dynamics of the scalar field drives the expansion of the universe. When the field is in the false vacuum state, the universe is dark energy-dominated. This dark energy drives inflation and makes omega equals one becomes an attractor rather than a repeller. This is physical explanation of why, for a wide set of initial conditions, the universe ends up flat. So I find that really attractive to have field dynamics explain the initial conditions. The anthropic solution to a problem I view as surrender. It's basically saying we will never figure out a dynamical system solution; we will not be able to understand this in terms of fundamental physics. The anthropic principle says a very set of initial conditions are need for sentient life, so we must have had those very peculiar initial conditions. Even if that turns out to be the correct answer, we’ve learned nothing. We've certainly not worked on the problem of the origin of the universe enough to give up. So if other people want to surrender, that's fine.
David, while we're on the topic of your lack of charity towards tautology as it relates to unknowable questions in physics, I wonder what your response might be to those who ask, and particularly might have an agenda behind the question, how was the universe created out of nothing by itself? Does physics offer—does cosmology offer a credible answer to the idea that something can be created out of nothing, or do you simply reject that also as a tautology?
No. It's an interesting question, and it's one that we may not—without a theory of quantum gravity, we may not be ready to answer that question yet, but it's an interesting one. I'd say the hint we have for that is that the total energy of the universe is very close to zero. This is something we now know observationally from measuring the geometry of the universe and seeing that it is close to flat. A flat universe implies that the total energy is zero, and we know that we can make quantum fluctuations of zero energy very easily. So I think that's a hint of what the answer might be. So without a theory of initial conditions, without a theory of quantum gravity, I don't think we can give a really solid answer. I do think that we have a hint that the quantum creation of the universe is at least worth thinking about.
And with or without machine learning, are you generally bullish on the idea of a grand unified theory?
I don't actually see machine learning as vital to the project of unifying the forces. It has worked really well up to now. For me, the most wonderful paper in physics is Wigner's paper on the unreasonable effectiveness of mathematics—
—which I'm constantly giving to students to read. One piece of the unreasonable effectiveness of mathematics is that symmetries have turned out to be so fundamental for our understanding. The success of the Standard Model, the role of the Higgs, we've gone very far. Electroweak unification has worked, so why not grand unification? I am very optimistic that we will achieve at some point grand unification. I see machine learning—it's a tool for doing calculations and doing statistics in high-dimensional spaces but it's just a mathematical tool. That's all it really it is.
And it's dependent on advances and observation to yield data for the machine learning to process?
That's right. The machine learning is limited by the training set. The computer doesn’t learn about new regions of parameter space.
David, when you joined the faculty proper at Princeton, in what ways did that change or influence your research agenda or not? Were you looking to carry over the same questions from the Institute to the faculty?
I continued in much the same direction. I was working on projects ranging from the more observational, studying galactic structure with Leo Blitz, to thinking about defects. For the next step in my research, an important influence was that soon after I started as a faculty member, Neil Turok moved to Princeton as an assistant professor in the physics department. Now, at the time, both Neil and I lived in New York City as both of our spouses were working in New York. My wife was first a medical student, then an intern, then resident and finally a fellow at Mount Sinai and then Columbia. I commuted from New York to Princeton for six years, and Neil and I would take the train together. Neil had started thinking about textures, so I would work with him on textures. We found analytical solutions to texture equations and thought about their astrophysical implications. Neil came at this with a deeper understanding of fundamental physics than I had, and I came at it, I think, with a deeper understanding of astrophysics and computational methods. We wrote a couple of papers and had a productive collaboration. We were thinking about the predictions of the texture model for the microwave background and large-scale. When the COBE results came out, I was at the AAS meeting. I gave one of my first national press interviews about it. I was quoted by Mike Lemonick in , Newsweek saying “we are dead”. The level of fluctuations seen by COBE was too low for defects to seed the observed large scale structure formation. While it would take my colleagues who worked on defects another two or three years to be convinced of the death of the model, it was apparent to me already. I stopped working on defects. I had just gotten tenure at Princeton. I was a little depressed. I had spent about five years working on the texture mode and wondered what to do next. I decided with COBE out, it really changes cosmology; we should have a workshop and bring together the leaders in the field. I went and talked with Dave Wilkinson, Jim Peebles and Lyman Page about hosting a meeting in Princeton. Lyman I have known since graduate school.
He had moved to Princeton. He was a graduate student at MIT. When I was at Harvard, we had a close mutual friend, Lynne Deutsch, who was my officemate in graduate school and would become a Professor at Boston University. Sadly, Lynne passed away quite young.
I organized this meeting. George Smoot and Chuck Bennett and people from COBE came. And at the meeting, I remember Rich Gott, a colleague in the astronomy department, got up and said, "COBE proves the universe is flat." I thought, how do we know that?
This motivated me to start thinking about negatively curved universes. Marc Kamionkowski had just arrived at the Institute for Advanced Study as a postdoc. Soon after he started at the IAS, he walked into my office and said, "Let's work on something together." I suggested we think about the predictions for negatively curved universes. We worked through the theory and realized that in a negatively curved universe the Doppler peak would shift, acoustic peak would shift to smaller scales. The next set of microwave background experiments could measure the geometry of the universe. We then wrote a follow-up paper with Naoshi Sugiyama that looked at this in more detail. We modified his code to be able to make microwave background predictions for a negatively curved universes. I then giving talks about how we could measure the geometry of the universe with microwave background measurements. Lyman was excited by this, and Lyman and Chuck talked about it and they decided to bring me on as part of the proposal to NASA that became WMAP as the house theorist. (At the time, Dave Wilkinson was still alive and an active part of the project. We called it MAP [Microwave Anisotropy Probe]). For me, joining the MAP team was a big transformation in how I did science. Up to then I was working as a theorist on lots of little projects moving around from area to area.
David, I just want to interject at this moment and return to the concept of themes, because in your career, this sounds like it's a real narrative turning point in terms of tenure, in terms of you are now latching onto a formative project that is really—it's going to make the name for yourself long term, right?
So to return to the concepts and themes surrounding astronomy and cosmology and theoretical astrophysics, given the fact that these terms are fluid administratively, scientifically, substantively, and also in terms of your own—the niche that you felt like you filled within the faculty or within the collaboration, how were these terms changing in your mind in terms of your approach to the research broadly conceived?
At the time, I thought of myself as a particle astrophysicist. At one point, I used to claim that I had the office in the astrophysics department closest to physics—
—as I was the person in the department with closest ties with physics. I was also in this position where most people who thought of themselves as particle astrophysicists came from physics.
Came from particle physics?
Came from particle physics. But I felt that I was on this trajectory where, for some—I don't know, accident or intentional—I was always in an astrophysics department. I was in the astrophysics department as an undergrad, as a graduate student of Harvard, I was the department of astronomy. At Princeton, I returned to the astrophysics department.
Right. But just so I understand, you bypassed that big narrative of, like, particle physicists post-SSC who were looking for new frontiers and that was their entrée to astrophysics.
You bypassed all of that.
That's right. Well, these were my collaborators and people I was learning from, but I felt—I always had a foot in the astronomy side and a foot in that community, so I spoke to both. But I would feel sometimes that working in an interdisciplinary area, there were times I felt I didn't fit in in either community. My astronomy colleagues saw my thinking about cosmic strings or string theory as working on made-up physics. I know this was true in my tenure letters and my external reviews.
Only two years ago, Sandy Faber, a very distinguished astrophysicist, asked me, “Why did you work on textures?” It is make-believe physics. I thought of it as exploring the physics we don't know yet. Astrophysicists can be intellectually conservative. John Bahcall was one of the people uncomfortable with particle astrophysics. Although, John was the person who I would argue did the most important work in particle astrophysics.
The solar neutrino problem turned out to be the particle physics we learned through astrophysics. But I think he would've been very offended to be called a particle astrophysicist.
Actually, I know he did not like being called a particle astrophysicist. But I point out to him that he was at the forefront of it. Many of my colleagues in astrophysics were nervous about the interactions with particle physics.
On the other hand, some of my colleagues in particle astrophysics would've viewed astrophysics departments as kind of backward and fuddy-duddy.
They felt they didn't have anything to learn from them. But I think, of course, there's an enormous amount to learn from traditional astronomers. So on a good day I felt that I had the advantage of going back and forth between those two worlds.
On a bad day I wondered where I fit in.
David, I want to ask about your early impressions of WMAP in the context of – at what point you realized that you were part of something fundamental? So just for an example, to set the stage, one of the things that's so fun about LIGO is that it was almost nothing for so long until very quickly it was everything. So that narrative of LIGO is it has that dramatic—it almost could've been a dud for really decades until the detection, right? At what point—now we know what WMAP has accomplished and what it will continue to accomplish. Did you appreciate even at the beginning that it was set up for greatness, that it was conceived to do fundamental things?
Yeah, yeah. No, it was pretty clear—so I remember going to this first meeting where we started to design the experiment, and I already knew that we could measure the geometry of the universe. That was already clear to me before I walked into our first meeting. As soon as I saw the sensitivities that my experimental colleagues felt they'd achieve, I did a very rough calculation of how much we could learn about the basic properties of the universe. This became central to our proposal where I showed how accurately we could measure cosmological parameters. The calculation I did in the proposal arguably was actually a little bit of an oversell. For each parameter I assumed we knew every other parameter perfectly, so I just looked at one at a time. The simplification let me do the calculation more quickly. Soon after we started looking into MAP’s capabilities, I gave the Warner Prize lecture at the AAS meeting in January 1995 in Tucson where I talked about measuring cosmological parameters from the CMB: this was a new idea at the time. Marc Kamionkowski had heard the result and was excited by it and wanted to do a more detailed study. Marc had just started as a faculty member at Columbia. Marc brought in Arthur Kosowsky, who was a new postdoc at Princeton and was another former student of Mike Turner and Rocky Kolb's from Chicago, and Jerry Yungman. In 1997, the four of us did the first Fisher matrix calculations of how well MAP could determine the basic cosmological parameters. MAP launched in 2000, our first paper comes out in 2003, In our 1994 proposal, we had shown that if the experiment worked the way we hoped, that we could measure the basic parameters of the universe's age, of composition, and so on very accurately. It was clear to me at that point that the most interesting possibility would be if we could rule out Lambda-CDM rather than fit it to the data. Already in '94, before the supernova measurements, it was already the preferred model, but certainly not established the way it is today.
In the Warner Prize lecture, I got up and said, "If we build this experiment we will measure the fundamental properties of the universe. And we'll have one of two possibilities, either we'll rule out what's now the most popular model or, if it's successful and if its adiabatic fluctuations, we will measure these basic properties to high precision." So I was pretty convinced that we would do that early on. Now, I don't think I convinced most other people. People thought it wasn’t possible. I remember talking to Mike Bolte after the results came out a decade later and he said, "You were right that that would work this way." But experiments often promise more than they deliver, and I think that was probably most people's, say, interpretation of my talk. I went around the country d in the mid-'90s, gave a lot of colloquia where I talked about all the things we would learn. I suspect most people felt I was still this fairly young guy—I was 35—was a little suspect. I'd done all this stuff on strings and textures.
That's right. That's right. How did the collaboration work in terms of the different personalities and sensibilities involved? Who added what and in what ways?
It worked really well. So what worked well was a bunch of them had worked together on COBE. I was on sabbatical visiting Leo Blitz in '94 when the proposal happened, so I got to know Gary Hinshaw and Chuck Bennett very well. Our families would do things together. I developed a very good relationship with the Goddard people. Lyman and I already knew each other very well. It was a really good team. In a way, we were helped by the fact that the team was too small for the task we had so we all had to work really hard. There was no time for politics. There was no fighting over turf because so much had to get done and there were so few of us. Lyman really led the optical design and a lot of the engineering design. Norm Jarosik played a big role in detector development. Gary Hinshaw became a really good leader and managed the software. Gary and I knew each other as graduate students at Harvard and we worked together quite well. I put together a very good group of students and postdocs around me who developed the software for analysis and interpretation. Chuck Bennett was a terrific project leader. He interfaced with NASA, he managed the engineers, he had a vision of how to get the most out of the scientists on the team. It was a really well-run project.
What were some of the major institutional funders that were necessary for WMAP's success?
We competed in the Explorer program, so we sent a proposal in and it was one of 40 that were sent in and it was eventually selected. We were pretty much fully funded by NASA. NASA was the major and the dominant source of funds. We had some additional support from Princeton. Princeton gave Lyman some leave, and I had some resources to hire some students and postdocs, so Princeton certainly helped. But it was almost all NASA funding.
In what ways did you contribute to the nuts and bolts aspect of WMAP, just in terms of conceiving the actual instrumentation?
The instrument design really came from Lyman and Chuck and others. They made the key choices for the detectors and the differential measurements. I developed the science requirements. I understood that to be able to measure the cosmological parameters accurately we needed to know the beam very well, so I showed how we would have to measure that to a high precision. The work pushed the team to develop ways to calibrate the beam. I showed that polarization measurements could be useful for measuring the optical depth, and starting to search for gravitational waves. I made the science case for measuring polarization to test the nature of the fluctuations. The experiment was “sold” in the proposal as a temperature experiment. That's what we promised NASA. However, we tried to make sure we could also achieve interesting polarization results. I helped make the case to achieve that sensitivity.
David, last question for today's session and then, of course, we can pick back up on this tomorrow. I want to return to the idea of terms as they relate to WMAP. So when we're understanding the Standard Model of cosmology, we have inflation, we have expansion, and then shortly after with Saul Perlmutter we have acceleration, right?
What are these three terms and how does WMAP enhance our understanding of one or all of them?
Let's come back to that, particularly I would say the history of acceleration. Saul and Adam played an important role in establishing it. We're doing this proposal, writing about this before their results come out. Already, it was the standard theory. We should probably talk—let's talk more about that this tomorrow. WMAP—and this is a longer conversation and have an 11:00 call—telling us both about the fluctuations that trace initial conditions that came out of inflation and how these fluctuations evolve to become galaxies. And how they evolved depends on the composition and geometry of the universe. When we proposed this and when we looked at this, it was getting at both pieces of this.
And on that note, I'll let you go. We'll pick right back up tomorrow.
Terrific. So I'll see you tomorrow.
OK. Thanks so much, David.
OK. This is David Zierler, oral historian for the American Institute of Physics. It is November 18th, 2020. I'm so happy to be back with Professor David Spergel. David, thank you so much for joining me today.
OK. So I want to build out on my question from yesterday where the last thing we touched on were these terms "inflation," "expansion," and "acceleration" of the universe. So let's just go chronologically in understanding the shoulders upon which WMAP was standing. Where was Alan Guth and inflation at the point of conception for WMAP?
The inflation papers came out back in sort of 1982 to 1985,. By the time that WMAP was conceived the basic theoretical framework was pretty clearly established. The predictions of inflation were well established,: nearly scale-invariant spectrum with a spectral index slightly less than 1, and Gaussian and random phase. There was weak evidence for these predications. The COBE data was consistent with Gaussian fluctuations but really wasn't that sensitive to non-Gaussianities. The balloon and ground-based experiments were consistent with Lambda-CDM but not really able to measure with precision yet the spectrum, so that it was an idea that was theoretically attractive. It was something that people took seriously. Back in 1994 or 1998, I don’t think that many people understood that the upcoming CMB experiments, such as WMAP and Planck would measure the spectral index, constrain the tensor amplitude and the amount of non-Gaussianity with precision. On the other hand, the evidence for the expanding universe y goes back 70 years or 80 years before the mission. The data had been built up over time and basic paradigm of an expanding universe was pretty well established. There was growing evidence for the accelerating universe through the '80s and '90s. We had a lot of evidence suggesting that the total density in matter was less than the critical density. Dynamics, measurements of large-scale structure and cluster counts implied that the matter density was about 20-30% of the critical density. Inflation predicted a flat universe. There were papers by Peebles back in '83 and by Ostriker and [Paul] Steinhardt and Michael Turner and others suggesting our universe is vacuum energy dominated universe, a Lambda-CDM cosmology. By the early 1990s, the predictions of a Lambda-CDM cosmology were well understood.
I remember back in 1994, I helped organize a conference in Princeton—
We had a debate between advocates of the different cosmologies. We could not get anyone to defend a matter dominated CDM cosmology, standard CDM. To fill the slot in the debate, I volunteered represented it in the debate. I did not think that the evidence for it was compelling, but to complete the conference I did it. I presented some new data that Julianne Dalcanton had gotten in her thesis showing that there were larger numbers of low-surface brightness galaxies than people appreciated. I made the case that there could be enormous amounts of mass associated with these low-surface brightness galaxies and the total density of the universe in matter could equal the critical density. I won the debate and standard CDM got the most votes. . This victory was more of a demonstration of my high school debate skills than a convincing scientific argument.
The debate was somewhat serious, but also somewhat in jest. It was the last session of a meeting. The supernova data showing acceleration was very important in some ways. While there was already growing evidence for vacuum energy, the supernova data and it was the “tipping point”. It convinced people outside cosmology to take dark energy seriously. In that sense, it was similar to Vera Rubin’s work on dark matter. Prior to her work, people who were sort experts in the subfield thought there was a case for dark matter. Her work convinced the broader scientific community.
Rubin's work really convinced the broader scientific community, and in the same way Perlmutter’s and Riess’s work convinced the community that this was likely true. People were leaning towards it. Data from some of the ground-based experiments, the Miller, Page, and Devlin work with TOCO and then the work with BOOMERanG with the CMB and MAXIMA, TOCO actually first, showed pretty clear evidence for an acoustic peak at about the degree scale. This suggests that the universe is flat. Now, there was some ambiguity at the time because there could be contributions from things like isocurvature fluctuations. These pre-WMAP experiments had already seen the first acoustic peak. WMAP measured the first peaks with precision. However, the basic picture of the properties of the fluctuations was already there, not solidly in place, but it was the preferred model.
Mm-hmm. To what extent did WMAP—not only did it demonstrate how much of the universe is comprised of dark energy, but to what extent did it further our understanding, our ongoing understanding, or what dark energy actually is?
So let me just step back a bit and try to address all three questions with WMAP. When we look at the microwave background, we learn both about the initial conditions in the early universe that seed the fluctuations and how the fluctuations evolve. We learn both about the early universe at very high redshift and the physics of the universe (and hence its properties) 300,000 years after the Big Bang. WMAP showed that the fluctuations in temperature and polarization are correlated on very large scales: the detection of these adiabatic super-horizon fluctuations is one of WMAP’s most important results. If there was no inflation, then WMAP’s detection implied that these fluctuations were already in place before the universe started expanding. Inflation could explain these large-scale fluctuations: since inflation predicted the universe expanded superluminally. The temperature polarization correlations are strong evidence that either there was something special in the initial conditions or the universe went through a bounce, or you had something like inflation. You were really driven to something beyond what was the standard Friedmann-Robertson-Walker expanding universe model. I think that was one of the important things we learned. We were able to measure the distance to the surface of last scatter by seeing the acoustic peak, and that gives us a constraint on the—basically measures the interval of the Hubble parameter from here now back to redshift of 1100. And that gives us a lot of constraints on what the dark energy could be.
It required the dark energy to have an equation of state close to that of vacuum energy. Another important thing that the WMAP measurements told us is that it requires that there be dark matter back at that time and that dark matter was non-baryonic.
In order to see the acoustic peaks the way we do, and particularly to see the third peak, you're seeing—so the physics of the microwave background fluctuations, or you've got this baryon photon fluid, where the electrons and protons are coupled by their electromagnetism and the electron and photons by collisions. The protons-electrons and photons form a single fluid that responds to the fluctuations by producing soundwaves. The soundwaves propagate out from dense regions and the cold dark matter stays in place. The deeper gravitational potential well associated with the dark matter produces a cold spot surrounded by a hot ring generated by the outflowing sound waves in the proton-electron-photon fluid. Without cold dark matter, you won’t get this distinctive pattern.
We see in the microwave sky the signature of non-baryonic dark matter. WMAP provided firm evidence that dark matter is something that doesn't interact with photons, electrons or protons. It's something new. Something beyond what's in the Standard Model.
How useful by order of elimination, in terms of understanding dark matter, if we can clear away what it's not, what does that leave us with?
I think that's how we have been making some progress, at least.
By figuring out what it's not, we move to figure out what it is. WMAP and the CMB observations show it's something beyond what we seen in any terrestrial experiment up to now. It eliminates many possibilities: very low mass stars or planets or black holes that formed after redshift of 1100. This dark component was there back at redshift of 1100. Since the universe was quite simple then, it's sort of cleaner to understand.
Mm-hmm. One of the great success, obviously, of WMAP is determining with great precision the age of the universe. I wonder if you can talk a little bit about the powers to which we can extrapolate in considering cosmic eschatology. In other words, what does this data tell us about how long it will take before the universe to end?
We know that the geometry of the universe is close to flat, so then its future is governed by the properties of the dark energy. If the dark energy is a vacuum energy, the universe will expand forever in accelerating pace. It will get emptier and emptier. It will be a universe that ends in ice.
If we look back to the very early history of the universe, the universe went through a dark-energy driven accelerating phase that we call inflation. It's what generated the fluctuations. The inflation-driven accelerating phase ended with reheating, and the universe became dominated by radiation. We could be going through a similar temporary phase of dark energy driven expansion. The universe could be accelerating today because we have a scalar field trapped in a false vacuum state. It could either tunnel out of a minimum or roll down a hill and exit the false vacuum state. If this happens in our future, then the universe could become dominated by matter and radiation again. Our current dark energy-driven expansion could all be a temporary phase.
However, given that the observed equation state of dark energy is very close w = -1 (vacuum energy domination), we can expect to be in the accelerating expansion phase for tens of billions of more years.
The most dramatic ending would be if the dark energy per unit volume grows with time. This possibility violates the strong energy condition: the vacuum energy just keeps growing and growing. In that case, the universe will be torn apart in a big rip. But even that ending is many billions of years in the future.
And when we're talking about tens of billions of years, it's possible then to anthropomorphize the universe. We might very well be in our teenage years; we might have a lot more ahead of us than we do behind us?
I think that's likely true, then.
David, obviously, the outside world got to understand the enormous success of WMAP through all of the incredible awards that were bestowed on the primary researchers and the collaboration as a whole, but from the inside, when did you and your team really know how special this project was, when did that happen?
You know, it didn't—the lead up to the launch, we had a sense that if things worked well that we would be able to make tremendous progress. We suspect that this could be a really big advance. We had a reasonable sense of that. When the data arrived—we were really focused on what was in front of us and didn’t really have time to think about its potential impact. I'd never worked harder in my career than I did in the two years between launch and when we put out our first paper. My wife and I just had an infant at the time--- it was a very busy time. All of us on the WMAP team were focused on trying to understand the instrument, trying to understand the data, and trying to understand any potential anomalies in the data. As we worked on this, we knew that all the data would go public when we released our papers, and that lots of people would pore through our analysis. They would all try to figure out where we went wrong.
We knew that there would be a lot of people sort of trying to—this is intellectually healthy, that they're trying to, in a sense, make their name by figuring out what we did wrong. We knew we had to be very careful. We knew that everything that we published was going to be checked thoroughly. We got the first papers out within six months of completing the first full survey. We were both trying to work really quickly and really carefully. This required working really hard. We had a really terrific team of people working on the analysis. The students and postdocs I worked with on the first WMAP analysis are now all very successful: Eiichiro Komatsu is now Max Planck director, Hiranya Peiris is now director of the Oskar Klein Institute in Sweden, Licia Verde is the leading cosmologist in Spain. They all went on to great success, and they were just a really great group. We were working closely with Lyman Page and his group at Princeton and Gary Hinshaw and Chuck Bennett and their team at Goddard. We were going back and forth to Maryland every couple of weeks and on the phone regularly. It was an exciting and very intense time.
Now that we're 10 years out, I'm curious, is there anything that we're still learning from the data? Has everything been pored over at this point?
I think everything's been extracted from the WMAP data. Planck basically reproduced the results of WMAP and has higher sensitivity, so most people now rely on the Planck data. We recently put out some papers using our experiment in Chile, the Atacama Cosmology Telescope, where we combined WMAP and ACT and got precision comparable to Planck. In this analysis, we wanted to use only WMAP on large scales so that we could have constraints independent of Planck. WMAP and Planck would agree remarkably well.
If you look at the WMAP and Planck maps, they're seeing the same pattern of fluctuations in the sky. The analyses of the data yield consistent cosmologies. Planck confirmed all the measurements we made with WMAP. They did it with higher resolution and sensitivities, so went beyond our measurements. We are now improving on Planck (and WMAP) on small scales with ACT. The ACT telescope has now covered 20,000 square degrees, so half the sky with five times the angular resolution of Planck. Just as Planck was more than twice the angular resolution of WMAP, ACT is now five times Planck’s resolution. We have—depends where you look on the sky—a factor of two to four more sensitive maps. We are now comparing our maps to Planck in the same way Planck compared to WMAP. The CMB measurements have been a tremendous experimental accomplishment: WMAP, Planck, ACT, and the South Pole Telescope are all seeing the same pattern of fluctuations. The maps agree in detail. They're using different technologies at different sites, and different telescopes. , The very sensitive measurements looking at fluctuations at a millionth of a degree and are showing a consistent picture. For me, this really strengthens the case for the cosmological conclusions that we draw from the data we have.
It's only 10 years, of course, since the end of the collaboration with NASA, but I wonder in those 10 years how much you've thought about how valuable private space missions like Blue Origin or SpaceX, where might they factor into similar collaborations in the future?
Well, I see them more as suppliers rather than driving the collaborations.
When you launch a scientific satellite, in some ways it doesn't really matter who builds the rocket that gets it into space. I'm now working on the Roman Space Telescope: whether the Roman Satellite launches on a rocket built at Marshall Space Flight Center or a rocket built by SpaceX, or a rocket built by Blue Origins, doesn’t’ matter to me. They all get us to space. The advantage of SpaceX and Blue Origins is these new space companies have shaken up the industry. They've made it cheaper to get to space. Before these “New Space” companies started to compete with traditional suppliers that had formed the United Space Alliance, the cost of rockets were going up and up. If it costs $300-million dollars to get to orbit, you're not going to build a satellite that costs less than $300-, $400-million dollars.
These high launch costs pushes NASA to build only really big expensive satellites. WMAP was a fairly inexpensive experiment by these standards. It cost about $125 million, and it launched on a Delta II rocket. They stopped making Delta II’s. There was not a demand for it. Because SpaceX is now flying Falcon 9s, we now have an inexpensive launch vehicle available again,. We can build small missions again. I think it's very valuable for NASA to have a mix of small, medium, and large missions.
The large missions can take decades to build. The Roman Space Telescope starts with the SNAP mission proposal that Saul Perlmutter develops in the 1990s and it's going to launch in 2025. With WMAP, we start working on it in '94, put the proposal in in '96, so it was accepted in '97. It launched in 2000.
And now that we're 20 years out from the launch, what are some of the technological or theoretical advances over the past 20 years that, let's say you were doing WMAP today, how would you do it differently as a result of those advancements?
The detector technology has advanced tremendously. At WMAP’s highest frequency channel where we had the largest number of detectors, we had four detectors. We knew each detector. We knew each of its flukes and properties. Now we're looking at data from ground-based telescopes that have tens of thousands of detectors. We’re thinking about building telescopes that will hundreds of thousands.
Is that to say that there's a Moore's Law of detectors, that it is just going to keep getting better?
Yes, there is a Moore's Law of detectors. And my colleague, Suzanne Staggs, has a very nice plot of detector sensitivity from Penzias and Wilson’s seminal experiment in 1965 to today’s detectors. It has followed a version of Moore's Law with sensitivities doubling time every three years. It's a slower doubling time than Moore's Law for computation. The experiments just keep getting better. The increase in sensitivity of the experiment is due to having more detectors in the camera and due to having each detector be more sensitive than the previous generation.
If we were to do things today, we would certainly have more detectors. We would have more detectors and high frequency channels. Planck had higher frequencies and more detectors. In our original proposal we said we would explore having a 120 gigahertz channel. When we went to assemble WMAP, the detectors at 120 weren't ready. We were lucky to get the 90 gigahertz into the experiment; they were very important for our measurements. We were using what was the best technology available a couple years before launch. Technology has advanced, so we would certainly have a much more sensitive experiment. We optimized the experiment for control of systematics for our temperature measurements, not for control of systemics of polarization measurements. Now people who are designing polarization experiments are putting in half-wave plates or other ways of modulating the polarization measurements. I would've put in polarizers at different orientations and included lots of detectors. I'd do what we're doing at ACT. I'd put in the polarizers at a whole bunch of different angles so that, as we scan, we'd be scanning polarization at different angles. That would've led to better control of systematics. When we put the polarizers into WMAP, we had to choose their orientation and didn’t make the optimal choice. We could have made a choice that would've given us 20, 30% better sensitivity on large angular scales just with the detectors we had, if we understood better how polarization map making would work. That's something—three years after launch, we knew enough to know to do that differently. I don’t think that we wouldn't change the core design of WMAP: a differential experiment to have control of systematics. I think the choices we made with WMAP were good choices. Planck was great, but in the end, we can do most of the high-resolution work on the ground.
So the choice to build something smaller that just got the large angles from space I think worked well.
Smaller and less expensive? There's more flexibility there?
Small means cheaper. You tell me how much a satellite weighs and I can tell you how much it costs.
Right. David, back on planet Earth, I want to ask about department service. I want to ask specifically—there's a range in terms of responses to when it's your turn to become department chair. On the one hand, many people say, well, it's just something I have to do, and I've been tapped and so I'll do it. And on the other side of the spectrum, it's a real opportunity for me to institute some change that I see needed. Where did you see yourself on that spectrum and how would you describe your tenure as department chair?
For me, it was an opportunity to build and strengthen the department. I was associate chair for, ten years before I was department chair.
I was involved in department leadership for, like, 20 years. I helped sort of shape the department. I felt the biggest challenge was generational transition. There was a generation of really great scientists who were 20 years older than me: Jerry Ostriker, Jim Gunn, Bohdan Pacyznksi, Neta Bahcall, all of whom were in the National Academy. Many had won major prizes and had played big roles in shaping the field. They were all going to eventually retire and bringing in the next generation of faculty was important. Jill Knapp, who was director of graduate studies for decades, really helped create, a very supportive environment for our graduate students and undergraduates. She was another key faculty member from that generation. I felt it was important to maintain and strengthen that supportive culture. When I became chair, the department was deeply involved in the Sloan Survey, and I thought it was important to identify the next big project. These were the challenges that I saw as department chair.
Was your part in the Roman Space Telescope, did that come directly after WMAP? Was that sort of the next obvious project to work on?
Yes and no. I got into that in some ways in a funny way.
Meaning it was already underway by the time you joined?
Well, I'll tell you the story. So my first involvement was as part of the decadal survey: I chaired the cosmology panel in 2010. After WMAP I felt it was my job to be willing to serve on national committees and panels. We identified the top questions for cosmology including “what is dark energy” and “what is dark matter?”. We made the case for building projects like LSST and what would eventually be WFIRST. After the decadal survey was done, I then became chair of the committee on astronomy and astrophysics.
I had already been working on designing coronagraphs. In fact the delay in WMAP of about six months meant I had six months without teaching that I wanted to do something new with back in '94 or '95—no, no, it was later—'99 and 2000. I came up with a new design for a coronagraph, so I knew something about optics.
John Grunsfeld, who was then head of the Science Mission Directorate, called me and said, "NASA has just gotten this 2.4-meter telescope from the National Reconnaissance Organization. We have to decide if we want to keep it. This has to be kept quiet. Can you do a quick study of whether this new telescope could be repurposed as WFIRST? (WFIRST was planned as a 1.6m telescope) here's what we can tell you about the telescope." I then started to work with Jeremy Kasdin, who was my colleague of many years in mechanical and aerospace engineering. Jeremy had become an expert on coronagraphs. Eventually started also working with Alan Dressler, and Matt Mountain at Space Telescope. We quietly worked together and did this little study and made the case that we should do this. And then, in June 2012. I remember the date because I was at my college 30th reunion, I took Joel Achenbach, who was the Washington Post science reporter, aside. I leaked to him the news of this telescope and said all of this would come out in a month. When it was ready he should run with the story. NASA tried to quietly announce it by just mentioning it in public at an advisory committee meeting. Because of Joel, this “quiet announcement” ended up on the front page of the Washington Post. This created the excitement in the astronomy community that gave us momentum to start to build support for the mission. When that happened, the head of NASA did not want this mission built. He was afraid it would be too expensive and detract from other priorities the agency had. So he would not allow anyone from NASA to study it, so that summer—this was always one of the nice things about being department chair, I had some funds, I could organize a conference with internal funds—I organized a conference in Princeton in August of 2012 on the mission and brought together people from industry and from a lot of different science communities that might use it. We then put a conference proceeding that we posted on the archive that laid out the science case for the mission (https://arxiv.org/abs/1210.7809). With the support of some colleagues within NASA, we then were able to convince NASA headquarters to start a study of the science yield from this repurposed telescope. Neil Gehrels and I led that study. I had never been involved in an optical telescope or infrared telescope before, but I was certainly interested in the science. I was interested both in the exoplanets and in the dark energy science, and I thought an all-sky infrared survey would be valuable. I found myself in the position of helping to develop the mission. I sort led the first two papers developing the case. I worked to build both community support and political support. I found myself going to Washington a couple times a year walking the halls of Congress making the case for the mission.
What did you learn about advocating for these kinds of projects at the federal level? What did you learn about who the most important players were and what the most compelling arguments to make were?
I learned the most important players were the staff: the congressional staff and the people at Office of Management and Budget and OSTP. I found that the most compelling arguments was the science. We had a good science story to tell and people get excited by the science. With Congress, it was very important to have people from members’ districts come in and talk. At one point, I was organizing a group of 100 people from all over the country to go to Washington. I would email and call. I would use in my prestige, my connections, and visibility in the community to twist people's arms. I have lots of former students. I've got something like 100 former graduate students, postdocs, undergrads. I'd call people and I'd say, "Come to Washington." For example, Matt O’Dowd from Lehman College in the Bronx came and talked with José Serrano's staff. José Serrano was chair of the appropriations committee. He was so excited that people in his district in the Bronx were interested in this NASA project. He was pleased that people from his district were coming and talking to him about quasars. I would go look and see who is on the Appropriations Committee. I would then look at what universities were in those districts, then beg people to come to Washington and help present the case. It worked.
I had also done a lot of service work for the National Academy. After being chair of the committee on astronomy and astrophysics, I chaired the Space Study Board for three years and was on the NASA advisory council. I could walk the halls of Congress wearing those hats and was able to say that this mission was the number one priority of the decadal survey. I wasn't working for my own personal funding—I was advocating for the field. I remember once meeting with some members of the House Science Committee. This is when it was controlled by the Republicans, who generally had been very supportive of basic science. However, there was this one staffer who was very aggressively questions Jeremy Kasdin and me: "Why are you here? What's in it for Princeton? How much money do you get out of it?" I tried to explain that I will get a little bit out of the funding: I'll get a couple months summer salary. I'll get support for a graduate student. It'll probably be worth $150,000-a-year and that is all that Princeton will get directly from the funding. It's not nothing that I get from this, but that's not why I'm here. I'm here because I'm excited about the science and I'm doing this for the field. WFIRST will launch in '25 and '26 and produce most of its result in the 2030s. I'm not going to be the one leading the WFIRST analysis. I'm not going to be playing the role in 2033 that I played in 2003 where I led the WMAP analysis. It's just a different stage of my career. You do it because you want that next generation for the opportunity to push the science forward.
Speaking of stages in life, are you surprised at the gestational development of Roman Telescope as compared with WMAP? Does it strike you as particularly long given what it's set up to do?
I always felt it was shorter. I'm an optimist. But I knew—the gestational period is just longer for these bigger projects.
So it's just a function of the size of the collaboration that's making things—
Since we're in the middle of it, what are some bumps that have come up along the way that you may or may not have been anticipating?
President Trump canceling it three times.
Setting the budget to zero.
I hadn't anticipated Donald Trump.
Yeah. So in what ways have you and your partners kept it on life support given these existential threats?
It would get zeroed. I learned all sorts of things. I learned—the day the budget comes out and it's zero, as soon as it was there, I had tweets ready to go. I tweeted out—I had built up, because of my experiences with BICEP where I was the person who pointed out the problems with that project—a lot of science reporters follow me on Twitter. I had talked about the problems with the gravitational wave claim on Twitter. When I tweet out comments on the NASA budget, it gets picked up by the newspapers. Journalists want a rapid response so they can write the story quickly. I try to tweet quotable things that can go straight into the newspaper. I learned that the first response gets quoted. The first response to a budget cancellation helps shape the conversation. Then you have to explain to the Europeans and the Japanese who are partnering with you that the President zeroing a project doesn't actually mean a cancellation. They say in Washington, the President proposes and Congress disposes. Even when the Republicans controlled the House and the Senate, there were people in the Republican party who were very supportive of the basic science as one of its core roles of government. I had to explain to the Europeans and Canadians that we had a good chance to overturn the cancellation and they shouldn't abandon us yet. What we had to worry about was our foreign partners dropping out when they saw the lack of support in the President's budget. We worked together with our industry partners to save the mission from cancellation We have calls every month at certain times with people from the main suppliers, we coordinated political strategy with people of Space Telescope and JPL. It is important to have a consistent message. It's important when you go in and talk to Congress that they're hearing the same thing from everyone who talks to them about the state of the budget and the state of the project.
David, given the fact that you emphasized the importance of the collaborations in Europe and Japan, I wonder if you can talk a little bit about one notable country that you left off that list, which, of course, is China, which today is supporting basic science and cosmological research endeavors at a rate that probably dwarfs the United States in the post-Sputnik era. So I want to ask specifically, to the extent that you've operated in a political realm of advocating for greater funding in the United States for these missions, where is the tinge of nationalism in terms of sounding the alarm that China is really going to leave us in the dust in the 21st century? How do you feel about something like that?
All right. So I took the approach of emphasizing different parts of that story depending on my audience.
When I was talking on the Hill to certain staffers, I would emphasize that the Chinese are making a big investment; they're going to launch their 2-meter telescope off their space station, the Chinese Space Station Optical Survey Telescope, to do the same big core science that we hope to do with Roman. I would say that dark energy was something that was discovered by US scientists and we want to maintain leadership in this basic research area. I would emphasize that China was doing a very similar mission: they were repurposing some of their spy telescope technology just like we are with Roman. Were we going to abandon US leadership? Were we going to make the investment or not? I had my nationalistic story that I told when it was needed. On the other hand, when I sort of step back as a scientist, I think it's really good that China is making those investments and—
Because it's just good for science and somebody should be doing it?
It's good for science. I'm cochair of the International Astronomical Union's Committee on International Collaboration in Space. At the meeting at which the Chinese talked about their mission plans, I pointed out to them they had to start thinking about the effects of the free-oxygen radicals in the atmosphere on the UV coatings on their telescope. I think that comment got me on their advisory board, so I actually have been to China a number of times to give them some advice on their project. I was speaking at the World Laureates Forum in Shanghai earlier this month— I spoke to both the political people and the scientists, above a vision that I've been pushing of a common world astronomical all sky map. In 2030, the Roman Telescope will survey the sky in the infrared, Euclid will have surveyed it in the optical, the Rubin telescope will be surveying the sky at multiple bands in the optical in the optical, and the Chinese will have surveyed the sky in the ultraviolent with their telescope.
Which is complementary research, you're saying?
Which is complementary. We want to be able to combine in a single data set all these multiwavelength images of the sky,. This rich data set will tell us a tremendous amount about everything from exoplanets and stars to cosmology. We want to make sure that we have the data in a format that is usable by everyone. Right now there are a lot of constraints on NASA collaborating with China, driven by a combination of really legitimate security concerns, their technologies that they don't want stolen, and some jingoism. It's sort of everything—there's legitimate and illegitimate reasons behind this.
Yeah. David, on that note, let me ask you then, given your current role at the Flatiron Institute and beyond, given that there are built-in flexibilities that you have that you wouldn't have as a PI with NASA or even at an American university, could you see, in light of the fact that the nexus of so much scientific research and funding is going to be in China, for better or worse, for the 21st century, can you see the Flatiron Institute setting up a campus in Beijing? I mean, is that in the realm of possibility?
No. On the other hand, I can see, and I've actually offered to host the meetings to bring together scientists from China, the US, and Europe to collaborate together more effectively.
I don't think we need to be funding Chinese science. The Chinese are doing a fine job.
I'm not thinking about funding Chinese science, I'm talking about simply having a presence in the place where all of this is happening.
I think we'll have a presence by having—I already hired the best young computational astrophysicists away from China.
We will continue to do this. I think we want to continue to enable outstanding Chinese scientists to come to the US, spend time, interact with scientists here. I think that the open and free exchange of ideas with our Soviet colleagues in the 1950s, '60s, and '70s played a role in undermining the communist system and promoting freedom and democracy.
Not that the world—Russia is far from perfect, but we actually did end the Cold War. With China, the free exchange of ideas is good for everyone. On the other hand, I do not see the focus of Simons Foundation funds as going to supporting work in China. I think it will be much more likely focused in the United States.
David, I don't want to ask to talk about all of the awards that you've received. That would be way too much of a burden. But I do want to ask if there's one that stands out for you as being most personally meaningful?
I got the MacArthur Fellowship early in my career, it enabled me to set aside the money. It all went into the college funds for my kids and basically paid for four years of Princeton for one, four years at Chicago for another, and is paying for four years of Princeton for the third. It removed all that worry, which was great. I think that was the most impactful. The Shaw Prize and the Breakthrough Prize had the best parties.
Those were fun.
Yeah, yeah. How do you see the Breakthrough Prize in terms of an opportunity for a very 21st century mindset to emphasize the fustiness of the Nobel Prize? How do you see it in those terms? In other words, you get the Nobel Prize, but the fact that it's by definition limited to three people, which, of course, is antithetical to the very concept of big science and collaboration—
—as they exist today.
The Breakthrough Prize is more flexible, which is great. I think that flexibility is useful. I like the focus on relatively recent science, what the Nobel Prize has drifted away from. How the Breakthrough Prize will evolve with time will be interesting to study over time for some future historian. The Breakthrough Prize is shifting in the same direction as the Nobel.
They just gave a prize to Steven Weinberg, right? I'm on the selection committee. I would like the Breakthrough Prize to make more use of their flexibility to award people who would not get the Nobel—I think targeting big collaborations has been great. Yeah. Major contributions to the advance of physics is no longer driven by Einstein working alone on his theories. It would be good to use the prize to recognize how science has changed more.
David, what were your reactions to Eric Adelberger and his team winning the Breakthrough Prize for negative research, essentially disproving the Fifth Force?
Oh, I thought that was great.
I thought that was great. I was a big supporter of that. Another area where I'd love to see a prize given is for the dark matter experiments. There's been tremendous experimental work done and they haven't seen the dark matter. You have to ask, what's the role of a prize? I think that one of the roles of a prize is to send a message to the physics community: This is something we value. By awarding work like Adelberger's group, we recognize that a long a careful program of looking for new physics, even if it doesn't turn up a dramatic result, is valuable.
What does that say about science and society that Ephraim Fischbach did not get the Breakthrough Prize for introducing the research that got it for disproving the Fifth Force? How do we understand how these things work?
There were a bunch of things that drove those experiments. Fischbach's work was one, but I think a lot of it was for also the cleverness of that experimental approach that Adelberger developed and his group. Another thing these prizes do is also encourage young people to go in a certain direction. I would like to see over time a more diverse group of people get the prize. The prize sends a message about who does science.
Which is to say that there are lots of women and Black scientists who are doing incredible work right now?
Yeah. And I think with these prizes it's too easy to get obsessed with, well, is X better than Y?
What metric are we using?
I think it's more important to think about what the impact of the prize is. What is the message we send?
David, one aspect we haven't yet talked about in your career is your work as a teacher to undergraduates and your work as a mentor to graduate students. So first, on the undergraduate side, over the course of your teaching career, what were the most fun and meaningful courses for you to teach undergraduates?
Oh, I always like moving around and teaching different things, so I tried to move through the curriculum. I actually liked teaching the freshman seminars when they're just starting, you know, in small group and getting to know them. And I also liked teaching the upper-level undergraduate courses because it is the first time that you really got them deep into the subject and you got them to think like physicists I enjoyed teaching the upper-level cosmology class. Probably my favorite class is the upper-level cosmology class.
Given the fact that you were an undergraduate at Princeton and you taught Princeton undergraduates yourself for 30 years, what might be a common myth about a Princeton undergraduate that you might be well positioned to dispel?
Well, I think most of the myths are probably right. They tend to be bright and some of them tend to be privileged. There's some who are aren't privileged, but most of the physic students are very strong.
Have you converted any undergraduates into graduate mentees?
No. We always encourage our undergraduate students to go somewhere else. So I've stayed in touch with—I actually just wrote a recommendation letter yesterday for one of my undergraduates applying for postdocs, so I continue to support and mentor them. I worked with Dan Eisenstein who's in the national academy now, John Kovak who led the BICEP experiment, and Cullen Blake, Eric Gawiser, and a whole bunch of people who won Hubble Fellowships. So I have a lot of successful undergraduates, but I always encourage them to go to graduate school somewhere else.
David, I'm curious, given your commitment to teaching, as a theorist, for experimentalists, people that work in observation in your field, it's easy to involve undergraduates in the lab work and instrument building and things like that. What opportunities as a theorist do you have to engage students, not just in a way that's nice for their experience, but is actually useful for the work that you do?
Well, the work I was doing in galactic dynamics and galactic structure tends to be more accessible to undergrads.
So I engage them in that. I engage them sometimes in some analysis work. Nick Hand did an excellent senior thesis, he made the first detections of the kinematic Sunyaev-Zel’dovich effect: the large-scale motions of galaxies relative to the cosmic microwave background. Charlie Steinhardt did some interesting work on the time variable fine structure constant. Eisenstein worked on galaxy formation with textures. For the good undergraduates, I could give them more speculative projects than I give to a first-year graduate student. For a first-year graduate student there's an expectation that research will definitely lead to a paper. With an undergrad, if they learn a lot, it is a success. —If a project works, that's great, but if it doesn't work it's OK. I would sometimes give higher-risk projects to undergrads.
And given your long tenure at Princeton and the fact that you started probably as an undergraduate not too long after they infamously installed a women's bathroom in the department—to give an idea of the fact that women were just simply not part of the equation for so long—in what ways, in particular in light of your leadership position there as associate chair and department chair, what strides did Princeton make in terms of celebrating diversity and inclusivity in STEM? And to fast-forward to today, what work remains to be done?
Where we made, I think, a lot of progress was on gender diversity so that the—towards the end of my tenure as department chair, we had often had years in which the majority of our undergraduate majors were female. I hired Jenny Greene, Jo Dunkley, and Eve Ostriker to our faculty. We have three tenured women: two theorists and one observer, who joined the department as faculty. We had a lot of outstanding female graduate students and postdocs. o I think we've made a lot of progress there. I think in New York we've made a lot of progress in diversifying computational astrophysics, about 40% of our staff is female. When I started there were women in the field, but sometimes you would go to a seminar and there would be only one woman in the room.
I think we've made good progress increasing the number of women in the field. I think we have a long way to go in increasing the number of underrepresented minorities. Now, I've been able to be the mentor for a number of outstanding African American scientists. I actually got an award for my mentorship from the National Society of Black Physicists this year, which was nice.
Oh, wow! Congratulations. That's awesome!
Yeah. I brought Neil Tyson to Princeton as a postdoc, so we worked together when he was a graduate student, and I was his postdoctoral mentor.
Ah. On that note, another question I don't want to burden you with in detail, I certainly don't want to ask you who your most successful graduate students have been because you have so many of them, right?
So, instead, I want to ask—instead of asking to name names, if you can reflect on what you've learned in terms of common characteristics that your most successful graduate students have exhibited during your time working with them, what are those sort of personality traits or scientific talents where you were able to tell, after all of your years mentoring, this person is really going places?
So it's a combination of things. You need some technical ability. If you don't have that, it's just not gonna happen. Perseverance, you know, that you keep at something, but also a certain flexibility. If you keep at it long enough and it doesn't work, you try a new approach. There's another element of risk-taking for the ones who I think have been most impactful, a willingness to go out on thin ice a little bit, a willingness to go into new areas. Julianne Dalcanton did her thesis with me and she worked on low surface brightness galaxies using these new detectors. She worked with Tony Tyson and me, and I'm not a detector guy. I know something about them but wasn’t in any way expert. But she was kind of willing to go out and to address the science problems she wanted to address. That was a thing that she needed, so she did it. Eiichiro Komatsu was really willing to go try new things, move into new areas, and he's now Max Planck director. I think those are some characteristics I see. I look at students, I actually think the group of students I have working with me right now are as strong as any I've had. I think about this when I think about next steps in my career. I was out at Princeton on Monday. Do I want to stop taking the train to Princeton in the midst of the pandemic to see my students? No, I think it's really important to see them and they're really great, so I go.
Yeah, yeah. And do you see your work as a mentor to postdocs at the Flatiron Institute administratively or substantively, is it really any different than being a postdoc mentor at Princeton?
Not a lot. It's a similar situation, actually, where I'm the primary mentor for a small group at Flatiron, but I'm the secondary mentor for many people. As department chair, I felt responsible for all the students and postdocs, and here as director I feel responsible for all of the people at the Center—I try to let them all know that they can come to me for advice and guidance.
David, we frontloaded our talk with a very extensive conversation about your current work, so for the last part of our talk I want to ask you specifically one area we didn't talk about, which is the fact that you are now moving into this leadership role at the Simons Foundation. How are you going to ensure that you don't move too far away from the science? What are some steps you're taking to ensure that you're going to remain as involved as you want to be in that regard?
I'm setting up a research group that will be directly attached to me that will involve a number of the people working on the Simons Observatory, so I'll continue my involvement with the CMB experiment. I'm going to hire someone who is a postdoc here who's been working with me at Princeton to come back in here to be here in New York. He will work on applying machine learning to cosmological simulations. I have a couple of staff members working full-time with me. They’ll help me continue to work with graduate students and postdocs. I think I will end up working in a mode where I with people through weekly meetings—already I tend to meet with my graduate students once a week and they give me a report on what they've done. What I want to be able to continue to have is people for them to go talk to between the weekly meetings—if they're meeting with me on Tuesday and they need someone to really help them with their code on Thursday, that that person's around.
When I worked with Jean Quashnock, who was my first graduate student, Jean and I sat in front of the screen and I went through his Fortran code line by line to help him debug. In retrospect, that was not a good way to be a mentor. I was too involved, too much in the weeds, and I should've given him more independence and let him take a little longer to debug the code. He would've figured that out on his own. He didn't need me to do that.
My kids were really amused when there was a cartoon by Dilbert about the pointy-headed boss telling someone to do work in a language he couldn't read.
But I'm now that guy, right? I'm doing a lot of work using machine learning but don’t do any coding. With WMAP, I wrote a lot of the software. I built a Beowulf cluster to run the map-making codes. Back, then I wrote a lot of WMAP code. On the other hand, I have hardly done any coding since becoming the Director of a computational astrophysics institute.
I tend to write equations on the board and sketch out what I want to do next and talk with people. I like working in that mode now, and that's something I can do, I think, while being President of the foundation Jim Simons says that it takes about 50% of his time to be in meetings for the Foundation.
I'm hoping to spend the other 50% of my time continuing my research program. I've scaled back a bunch of things I've been doing with the National Academy. I said I would scale back some of the stuff I'm doing with Roman. I'm going to step down from some of the committees I'm on so that I want to make sure that the time I have has some time for research. I will continue to do some political things. The next call I have after this is with Senator Schumer's office to work on a bill helping them restructure the NSF.
So you're not walking away from science policy too quickly?
No, no. This is very clearly with the Presidential transition a moment of opportunity.
Right, right, right.
But I enjoy doing that with 5 to 10% of my time. I was asked if I wanted to go get a job in this next administration. I was actually asked if I wanted to serve in the Trump administration. One of my relatives is a good friend of Jared Kushner's, and I'm the smartest cousin in the family, as it were.
He asked me if I would be interested in being a science advisor to Trump. Now, this didn't come from Trump but from my cousin. I actually thought about that and said, you know, it's important for them to get the best advice they can regardless.
Regardless of it's listened to, you mean?
Listened to. Just basically get the best advice regardless of whether or not I'm not very happy with the government putting children in cages. I would be willing to do it. But then they asked if I had given any campaign contributions. I had maxed out my donation for Hillary Clinton and the like. And then he said, “That's information is all publicly available. It's not even worth putting your name in “
And that was even before you knew what Trump would do to the budget for NASA? [laugh]
David, for my last question, I want to subdivide it into two broad areas. So just to put a historical marker on this, it's the end of 2020. Our conversation ultimately will become a capsule in the archives of the Niels Bohr Library, and everybody should know that the whole world is looking forward to getting out of 2020 as soon as possible, right? So on that basis, there's two aspects of your career looking forward, let's say the next 10 years, to counter this pessimism for all kinds of reasons to be optimistic. So first I want to ask, if we could do sort of a decadal survey fast-forward, for the Simons Foundation, for the Flatiron Institute, circa 2030, best case scenario, what does that look like for you?
OK. So the Simons Foundation continues to support excellent science. We're playing an increasing role in diversifying the field by funding faculty positions at places like City University and the Historically Black Colleges, and increasing the gender diversity fields like physics and mathematics. We’re doing some things that the federal government can't do. We're playing an important role in connecting China to the world scientifically and retaining excellent scientists. The Flatiron Institute helps establish computational science as an equal of theory and experiment. People realize that for understanding a whole host of problems, simulation is essential. We play a significant role in advancing the role of machine learning in physics, and I see that as a two-way street, both using physical models to inform what's going on in machine learning and the role of symmetries, and using machine learning to address a class of problems we haven't been able to address before. Flatiron will have had a significant fraction of the people who are computational scientists pass through it as either graduate students or postdocs.
It's going to be a must-stop like CERN or SLAC for particle physics?
That's the hope. And that we both—and that gives us the ability to train excellent people but also create the right culture for the field, and that includes things like stressing the importance of reproducibility, open codes, and building kind of an open and inclusive culture. The Simons Foundation has other pieces where it would be exciting to see progress on. They're the leading private—we, we are the leading private founder of research in autism. There's a real possibility of doing translational work there. I think we are already learning that autism is not simple--- there is not a single cause of it. It's a label that identifies probably multiple different potentially treatable problems and that we make progress in treating autism. I think that would be cool. I like to fantasize what will happen in the next 10 years, and that's just some of the things we're working on.
What about coronavirus research, for example?
It's done. Coronavirus research was an important area of progress in 2020. The mRNA vaccine—this is something that I think is a great scientific success, one that I think we played an indirect role in, actually, through our support of the archive.
We're the leading supporter of the xxx.arxiv.org. This helped inspire the bioRxiv which really helped with dissemination of information about COVID. Today Pfizer announced 95% effectiveness for RNA virus, and that it seems to be safe and effective. I think one of the things Simons Foundation will help with is remote learning—one of the things we learned in 2020 that we want to apply in this decade from remote education to remote work, what are the things—there are bad things about that, but what are the good things we want to bring forward? I'm actually, from my readings, and this is outside my expertise, but I've, like everyone, been fascinated by the coronavirus, these RNA vaccines are a whole new way of developing vaccines.
Pfizer and Moderna's success in this suggests that this could be applied to other areas successfully.
Let me revise the question. I consciously said coronavirus research and not COVID research because, even if COVID is going to be solved hopefully by mid-2021, there will be the next coronavirus around the corner. So on that basis, in what ways do you see long-term—the Flatiron—the work that's being done in machine learning, to sort of not be caught flat-footed for the next coronavirus pandemic?
Yeah. So the systems biology people, we're working on basically developing pharmaceuticals that might be good fits to the coronavirus.
Like a universal approach, essentially?
It's a more universal approach. There are interesting potential machine learning applications in drug development and materials development. For the computational materials people to be able to y make designer materials, some of which might be bioactive materials, some of them will be used for energy purposes. We might be able to use neural nets to explore high-dimensional spaces for solving the many electron problems and finding new materials from superconductors to biomaterials in the next decade. We’re well positioned to be a place where that happens. We can be making significant contributions to fields like systems biology. These are areas where either I hope we'll do them internally or we'll serve as a place that facilitates discussions and conferences, or we support research through our collaborations program. This program supports collaborations at the couple-million-a-year level. Right now one of the collaborations that we're funding is exploring optimized stellarator design. They are using machine learning techniques there, but their team also has mathematicians using some of the deeper symmetries of the magnetic fields to advance their design. We are discussing working with DOE to build a protype stellarator. This is an area where I think foundations can really contribute. DOE was spending 99% of its fusion experiment work funding on Tokamak designs, one particular fusion design, and almost nothing on stellarators. We more than doubled the expenditure on stellarators. If that turns out to be the right bet, that'd be awesome. Foundations can and should do higher risk projects. To go back to your question on SpaceX, one of the differences that I see between SpaceX and NASA is that SpaceX is allowed to fail. If you look at SpaceX—well, to me, one of the remarkable things they can do, and they just did this with the manned vehicle, is they can land the first stage back on the ground. It flies back down to ground, they land it on the ocean, they take the first stage back, they fix it, they refuel it, they reuse it. And the first, I don't know, five times they tried to land the first stage it failed. It broke, it sank, all sorts of things went wrong. But they learned from every one of them.
But, David, to be fair, wasn't NASA allowed to fail in 1961?
They were allowed to fail in 1961, which is why NASA made lots of progress in the early '60s. But NASA today is not allowed to fail.
NASA is very proud of their 90-something-percent success rate on missions, and it's one of the things I see drive mission cost is they are very careful about integration and test—no NASA manager will get promoted because something failed. As a result, they're not willing to accept high-risk missions. The peer review process for grants tends to reject high-risk proposals. I have found in my own grant proposals, when I proposed to the work that led to the case for WMAP, I couldn't get funding. Once we wrote the first papers and people saw what we were doing, that was useful. I had an easy time getting funding when people had already seen. It's just new ideas are just hard to get into the system when too many people have to sign off on it. We are in the position where we can be a bit quirkier.
You're not too big to fail is what you're saying?
Yes. That's right. If there is something that I'd like to push the Foundation to do more, it is to fail. Over a decade ago, I was chair of the board for the Institute of Theoretical Physics. ITP (now, KITP) which runs these programs every few months, and they have four or five a year. We were reviewing their program over the last five years. I said, OK, out of those 20 programs, which one was the biggest failure, and which one was the biggest success? And they were uncomfortable answering me.
Maybe they didn't even have the metrics to really give an answer about what constitutes success and failure to begin with?
Yeah. And you want to have the metric. What do we think for this effort, what's a success and how would we measure it? And what's a failure? It’s not bad to have some failures. If you're not failing—I mean, I encourage students this way. It's like not everything you try should work. You should be willing to try things and give them up. You don't want to stick in a failed direction for years, but sometimes you want to try something, and this goes actually back to your question on undergraduates.
One of the nice things I felt I could do with undergraduates is give them things that might fail.
Right. The typical Princeton undergraduate is not used to failing. That's an alien concept.
David, for the last part of my question, emphasizing optimism for the next 10 years, so in the world of observation and astrophysics, given what the Roman Telescope ideally will do by 2030 and all of the other super-exciting projects that are in train right now or are about to be, best case scenario, what will we know about the universe in 2030 that we don't know today?
We'll know that dark energy is dynamical, that's it's not a cosmological constant.
From that we'll have a sense of—we'll know about the universe's future. We have lots of different ways of going after dark matter. One of them shows us what dark matter is. Yeah. This is in the optimistic scenario. We detect gravitational waves from the early universe. Maybe those gravitational ways are mildly non-Gaussian, so we get a hint of the physics of the very early universe. Those are things—I'd be surprised if we don't know the neutrino mass. That's something I'm almost willing to bet we will have measured just given the capabilities of the experiments.
That's one I'm willing to guarantee, by 2030 we'll have the neutrino mass measured. So those are big questions that will make progress on cosmology.
What about dark matter, what are you most focused on with dark matter?
Oh, I think dark matter is on that list. We could either detect some observational signature of dark matter or some underground experiment will just show dark matter particles. Sensitivities improve and you just don't know when we're going to get lucky. I think if I get to be optimistic, I'll say that one of those experiments get lucky, they get to the right sensitivities that we see something. We detect signs of alien life. I don't know if that will be lucky or not.
Maybe m the Mars missions have sample return, and we find signs of early life on Mars. That's something that will plausibly happen there.
Maybe machine learning turns out to be a powerful tool that actually lets us do things like everything from designing new vaccines to designing new materials to understanding ways of describing multiscale physics and things like turbulence and weather. All plausible things to happen in the next decade.
Well, David, it's been absolutely phenomenal spending this time with you. Thank you so much for talking with me and for sharing all of your insights and recollections over the course of your career. I really appreciate it.
Oh, it's a pleasure.