skip to main content
Home  /  Interviews  /  Peter Szolovits

Peter Szolovits

Peter Szolovits

Professor of Computer Science and Engineering at MIT

By David Zierler, Director of the Caltech Heritage Project

May 31, 2022

DAVID ZIERLER: This is David Zierler, Director of the Caltech Heritage Project. It's Tuesday, May 31st, 2022. I am delighted to be here with Professor Peter Szolovits. Pete, it's great to be with you. Thanks so much for joining me.

PETER SZOLOVITS: Nice to hear from you, nice to join you.

ZIERLER: To start, would you please tell me your title and institutional affiliation?

SZOLOVITS: I am a Professor of Computer Science and Engineering at MIT. I have been here for almost 48 years, so this is my long-term home.

ZIERLER: Pete, just a few questions for context before we go back to your undergraduate experience at Caltech. Where did you do your graduate work?

SZOLOVITS: Also at Caltech.

ZIERLER: What have been some of the major areas of research that you have pursued in your career?

SZOLOVITS: I have been interested in artificial intelligence, and in particular, in how one represents knowledge of the real world in a computer, and how one reasons with it. Very early in my career, I got interested in healthcare as an application area for these ideas, and so since 1974, I have been working with doctors on trying to understand mainly what would be an appropriate basis for decision support systems. The idea is that you could build a system that provides a second opinion for a doctor who is facing some complicated medical case. Of course, in the early days of my work, we didn't have a lot of data to go on, and so most of the design of these systems was based on trying to elucidate human expertise. I read a lot of medical books, and a lot of journal articles, and interviewed a lot of doctors, and tried to formalize what knowledge they had in their heads. Then about halfway through these nearly 50 years, large datasets started to become available, and so now it is more common for people to try to use machine learning methods on these large datasets. One of my technical interests is how do you combine a priori knowledge with what you learn from these large datasets, because the datasets often have built-in biases, and they are incomplete in various ways. Medicine has been around for thousands of years, and people have learned a few fundamental things. Like until 1680 or so, hadn't been published that the heart pumps blood. That seems like a useful thing to know, and actually remarkably difficult to learn from, say, intensive care data recordings.

ZIERLER: Beyond datasets, what have been some of the key technological advances both on the hardware and software side that have allowed you to pursue new research endeavors?

SZOLOVITS: Nowadays, the standard way of approaching machine learning problems is with deep neural network methods. I remember, about maybe five or six years ago, Geoff Hinton, who was co-winner of the Turing Award in 2018, gave a talk at MIT, and he was saying that he had proposed some of these ideas, in around 1980, but we were six orders of magnitude short in computing power to be able to make them practical. He said, "Now we've got ‘em." I think starting about a decade ago, computers became powerful enough that ideas that weren't that novel in terms of the fundamental idea all of a sudden became applicable to large datasets and actually get some really interesting results. The other thing that happened is in knowledge representation, which as I said, was one of my interests. The paradigm for that was symbolic logic, and so you would make statements in logic about characteristics of individual entities or about relationships among sets of entities, and these were typically extracted from the knowledge of a human being or some formalized knowledge in a textbook or article. Again, about a dozen years ago, people invented this idea that you could use co-occurrence of words or concepts in text as a basis for building language representations. If I try to understand English language, I can create a very large database that says which words occur just in front of or just behind, or two words in front of, or two words behind, or …, some other word. From that data, you can create what it called a vector space embedding. You assign to each word a very high dimensional vector representation of that word that is derived from the probabilistic relationships to all the other words around it. That turns out to be a remarkably effective representation of the word. When you then try to do something traditional like part-of-speech tagging or word sense disambiguation or something similar, those numerical embeddings give you a basis for doing these deep neural network computations that actually do a better job of solving those problems than the symbolic methods that we had been using earlier. The symbolic methods often have "sharp" boundaries of applicability, where their performance drops off as you move away from the examples people had in mind when they developed them. By contrast, the newer methods, based on probabilistic data and using continuous representations, seem to lose accuracy much more gradually.

Then about five years ago, people developed something called the transformer, which is a particular variety of this vector space embedding idea, that requires even more computation. Fortunately, computers keep getting faster, but I saw an estimate that one recent giant language model cost ten million dollars to train. The original idea I mentioned from a dozen years ago gave you a single representation for a single word or token, whereas the transformer idea gives you a context-dependent representation, which means that if I use a word like "bank" and it can mean either riverbank or a financial institution, you actually wind up getting different representations for that word depending on the other words around it. If the other words around it are about money and deposits and interest and so on, you're going to get the financial bank, and if it's around erosion and boat launches and stuff, then you're going to get the other one.

In the medical world, that has been an important advance, because the same abbreviations, for example, gets used for all kinds of different concepts. There was an interesting paper from people at the National Library of Medicine who created PubMed, where they found a two-letter abbreviation that appears in many, many abstracts of PubMed articles, but with 45 different meanings in different articles. One funny one we found in some of our own work is where we analyze a large dataset of intensive care data. Nurses use abbreviations all the time in their notes, and so you read about "BS" in the note. Of course BS in colloquial English stands for bullshit, but that's not what they have in mind. [laughs] They're talking about bowel sounds or breath sounds or blood sugar or bedside or bariatric surgery or …. Again, there are many, many different meanings of that two-letter abbreviation in addition to the colloquial word.

ZIERLER: To give a sense of the beneficiaries of your research, is the funding that supports what you do coming more from the health science side, or the computer science side?

SZOLOVITS: It has varied a little bit over time. I'd say up until around 2000, it was coming mainly from the parts of NIH that thought that computing was important. For example, I was, I think, for a number of years the largest funded project from the National Library of Medicine, which is one of the NIH institutes. That's because their conception of how to make progress is very much tied into computer use. They run PubMed and the NCBI genetic and clinical data repository, for example. They're very interested in automatic classification of documents, medical reasoning, retrieval of documents, matching of gene sequences, and so on, so they were my primary source of funding for a long time.

In about the last 20 years, what I found along with many of my computer science colleagues at MIT, is that the growth in expenditures in the federal government has not kept up with the growth in demand by researchers. I would say when I started in the 1970s, 98%—I just made up that number, but it was almost 100%—of all the funding for Project MAC, which became the Lab for Computer Science and the AI Lab at MIT, where those two later joined to be what is today CSAIL—Computer Science and Artificial Intelligence Lab. Almost all of the funding came from either ARPA (the Advanced Research Projects Agency of the DoD—before it got the "D" put in by the Republicans after they took control of Congress in 1994), NSF, and the Department of Energy. A little bit from NIH. The other 2% roughly came from gifts from grateful alumni and occasionally some joint agreement with companies. Today, we're probably getting over half our funding from those companies. They have become much more interested. It's not just an intellectual property play. They really seem genuinely interested in supporting fundamental research, because they can't afford to do that themselves. It's too expensive, but they figure by throwing smaller amounts of money at universities, and by involving their engineers in joint research projects, they can learn how to do this and how to apply some of those results to problems that they're actually interested in.

ZIERLER: Before we go back to your undergraduate days at Caltech, just a forward-looking question: With all of the excitement around quantum computers and existential questions about what they might be good for, do you see promising avenues, should we attain a scalable quantum computer, for your field?

SZOLOVITS: I have no idea, is the real answer. Scott Aaronson, who is a great technical wizard and philosopher of this area, is now a professor at UT Austin, but he was at MIT for a number of years, and we used to discuss this all the time. The problem is the theoretical results aren't very convincing yet. The original idea was that because quantum states can represent an exponential number of non-quantum states, that all of a sudden we would get an exponential amount of computation. I was always skeptical of that. I was a physics major as an undergrad, and one of the things you learn in physics is that nature is not kind [laughs] to that kind of idea. If I try to represent 100 billion states in some small collection of atoms, noise is going to kill me at some point. I haven't seen a good theoretical analysis of exactly what the noise thresholds have to be. It has become more of an engineering problem that I know Caltech is very much involved in, and IBM is very much involved in, and I have colleagues at MIT who are very much involved in it. It has become kind of an engineering discipline of how do you create these large number of cubits that interact with each other but that don't interact with their environment in a destructive way. There's a lot of effort on self-correcting codes and things like that, which seem to me less fundamental and more engineering problems.

I am an engineer, so I'm not scoffing at that. It's just that since we don't have good theoretical results on exactly what these computers are or are not able to do, it winds up becoming discussions of the engineering problems of making larger numbers of cubits that work together. Then a lot of it is sort of empirical argumentation that says, "I can solve this particular problem super-fast with a quantum computer." There was a result in 2019, where Google announced that it had achieved quantum dominance, i.e., showing that a quantum computer could solve a specific difficult problem many orders of magnitude faster than it could be done on classical computers. Then IBM came along and said, "Well, we can actually achieve most of that result with classical computing." That's not a very satisfying situation. More claims from other groups have come since then, though most seem to be about sampling of quantum mechanical states, which seems like a particularly apt use for a quantum computer. Whether quantum computing can efficiently solve general exponentially hard computing problems is, I think, still unknown.

ZIERLER: Maybe another way of getting to the question—beyond the obvious limitations of quantum computation, in your research, are there any obvious walls that you hit working in a classical computing world?

SZOLOVITS: Sure. Let me tell this to you in the form of an anecdote. When I joined the faculty at MIT, about 45 of my close colleagues and I shared one PDP-10 computer that had a microsecond cycle time—it could execute about a fraction of a million operations per second, since most took several cycles. Its memory was a vast (for its time) roughly five megabytes, compared to 64 gigabytes in my latest laptop computer. I spent enormous amounts of time trying to be very clever to optimize algorithms that would run fast and in very limited memory on such a slow computer. Now when I look back on it, I realize that I can just beat a lot of those problems to death with my 5-gigahertz-cycle-time computer, because it runs so many orders of magnitude faster that I don't have to be as efficient.

But, there are a lot of problems in the work that I do where in principle, it would be possible to try all possible interpretations of some set of data, but there's an exponential number of those, and we don't have an exponential amount of computer power. If quantum computers actually gave us an exponential amount of computing power, then I could once again stop being clever and just beat it to death with the exponential amount of power.

ZIERLER: Now let's go back to when you were at Caltech as an undergraduate. Generationally, I would have guessed you were a physics major because that was the major to be at Caltech for your generation. What years were you there as an undergraduate? When did you start?

SZOLOVITS: I started in 1966 and I got my bachelor's degree in 1970.

ZIERLER: To set the stage for the basis of our discussion today—your involvement in questions about admitting women to the undergraduate student body—in the late 1960s, were you generally aware of the women's rights movement beyond Pasadena? Was there a larger social context that fostered these discussions among students and faculty at Caltech?

SZOLOVITS: Yes, certainly. We all watched Walter Cronkite on the news. It was a time of ferment. There were attempts to lower the voting age, to lower the drinking age. We protested against the Vietnam War. I was very afraid of getting drafted. Some of my classmates did in fact get drafted and wound up serving in Vietnam. I wound up as a graduate student joining Air Force ROTC as a kind of draft dodge, and then by the time I got my PhD, they were getting rid of officers, and so I wound up only serving on active duty for three months. Then I was in the Reserves for ten years.

ZIERLER: Before we get to some of the institutional mechanisms that made this transition possible, colloquially, among students, in conversation with faculty, what were people talking about that precipitated the decision to admit women as undergraduates?

SZOLOVITS: I don't know what everybody was talking about. I think part of that ferment for equality certainly suggested that there's something weird about a school that wouldn't admit women as undergraduates. Caltech always admitted women as graduate students, but in very small numbers. I got involved in student government, much to my mother's horror. I can't remember exactly what position I held. By the time I was a senior, I was VP of ASCIT and head of the Board of Control, but I think before that, I had been secretary of the Board of Control, and so I had access to the archives of all the student government minutes.

ZIERLER: Let's just translate the nomenclature here. What does ASCIT stand for?

SZOLOVITS: The Associated Students of the California Institute of Technology. The student government.

ZIERLER: And Board of Control?

SZOLOVITS: Board of Control is a funny term. At Caltech in 1931, when the old dorms were built, the students petitioned the faculty to institute an honor system. The faculty agreed. The Board of Control is essentially the trial court for people who violate the honor system. It's a very interesting arrangement, because first of all, fortunately, at least in my day, there were very few people who actually violated the honor system, and so it created a wonderful atmosphere at Caltech where we had un-proctored, limited time, closed book, take home exams, where your professor would say, "Here's the exam. Take no more than two hours finishing it. Don't consult anybody else or any notes and turn it in by 5pm tomorrow." I don't know the percentage, but the vast majority, almost unanimously, people obeyed those rules. The reason it worked is because it worked. I remember when I was a freshman, there was a camp that we were invited to at the beginning or just before school started, where the upperclassmen would indoctrinate the freshman and say, "Look, this is how we function at Caltech. It works. It's wonderful. Don't fuck it up." [laughs]

ZIERLER: [laughs]

SZOLOVITS: It kept working. Every once in a while, when I'm back on campus, I talk to some of the undergraduates, and as far as I can tell, it's still working. It has gotten more bureaucratic because of various national or federal laws about the rights of students accused of various things. We really were kind of a star chamber. You had no right of representation, and so on. But as I say, very few cases. We were meting out punishments anywhere from, "Don't do it again" to expulsion. Officially, of course, we were a student board, so we were not empowered to expel people, but we would recommend to the dean of students. In my experience, the dean never rejected our recommendation. It was a great system.

Anyway, in the course of some of my duties in the student government, I remember reading through the archives of the minutes of the student government. I remember being surprised that there had been women undergraduates at Caltech until about 1952 or 1953, at which point the faculty in its infinite wisdom decided that all the women—there were few women—but all the women that had graduated from Caltech wound up becoming wives and mothers and dropped out of doing science and engineering, and therefore it wasn't worth [laughs] training them, so they changed the rules and they said, "No more women."

ZIERLER: In the late 1960s, from where you sat, what aspects of this decision to readmit women were driven top-down—in other words from administration and faculty—and what aspects were really coming from the students themselves?

SZOLOVITS: I think the students agitated before the faculty or before the administration, but I think we didn't find a huge amount of resistance. I think that once the case was put to people, they more or less agreed with us, and it happened relatively quickly. There were other things that the students also urged. For example, before the decision to admit women, we were arguing that Caltech ought to join up with a women's college in some sort of partnership.

ZIERLER: Like a Harvard-Radcliffe kind of thing?

SZOLOVITS: A Harvard-Radcliffe kind of thing. There were already I think two programs in the 1960s. There was a five-year program with Occidental College that allowed you to get I think a dual bachelor's degree. It was a BA from Occi and a BS from Caltech. I've forgotten exactly what the rules were, but it was a five-year program instead of a four-year program. There was something comparable I think with Pomona College as well. We were arguing that Caltech ought to make some formal arrangement to merge with one of these colleges, or one of the women's colleges in the Claremont complex. Like I think Scripps, I believe, is still all women. Pitzer used to be all women at that time. CMC used to be all men, so they already had the same issue. That didn't go anywhere. In fact, there was a discussion about Immaculate Heart College, a college run by a bunch of nuns at the place where the American Film Institute now is on Western and Los Feliz in Hollywood. That college I think was about to be trashed when they sold the campus. We also argued that maybe we should incorporate that college. Now, I have to say, part of the motivation, at least on the students' part, was not just high-minded notions of equality and equity.

ZIERLER: I was going to ask, yeah.

SZOLOVITS: Dates were hard to come by [laughs] at an all boys' school. Many of us were either commuting out to Pomona or finding other places to find dates.

ZIERLER: I wonder if, today, when we talk specifically in the sciences about diversity and inclusivity, one of the things that always rings true is that for science and engineering to progress, you really do need different perspectives. That's not just sloganeering; it's the real deal. We really do need that diversity. Among the highfalutin' reasons to admit women, and thinking about the larger social context, did the undergraduate student men appreciate that women really could be Caltech material in science and engineering? That they could be every bit as strong and robust in their studies as the men?

SZOLOVITS: I suspect that there was a diversity of opinion among the students. People came from all over the place. I'm an immigrant to the U.S., but in my time in America from the time I was ten years old, I grew up in Los Angeles. L.A. is a pretty liberal, tolerant kind of neighborhood, so to me, that seemed like a totally obvious fact. I went to a high school where the woman who beat me out for valedictorian [laughs] was a woman, and so it did occur to me that there were smart women who could do very well in the sciences [laughs] and engineering.

ZIERLER: When these conversations started to gel among the undergraduates, what were the infrastructures available to make decisions, make recommendations, and agitate, to the faculty and administration?

SZOLOVITS: Part of it is just that Caltech is a very small place, and so I got to know a lot of my professors personally, and we would just talk. They would come by the dorms for lunch and dinner and sit around and chat with students, or I would grab somebody after class and we'd shoot the bull about some interesting topic. That was I think a wonderful situation that meant that we could have some influence, at least in the thought process. The other thing that happened is—do you know about Joe Rhodes?

ZIERLER: I've heard the name, yeah.

SZOLOVITS: Joe was student body president for two years during my sophomore and junior years. He was a year ahead of me in college. He passed away, unfortunately, about a decade ago, but his ambition was to be a politician. He saw himself as the next Mayor Lindsay of New York City, except he was from Philadelphia. I think he in fact wound up as a Pennsylvania state legislator and then served in the executive office of Pennsylvania as secretary of something-or-other. Joe was an unbelievable dynamo. He was student body president I think for two years, elected as a sophomore, just because of his incredible personality. Oh, he was also Black, which is an important feature—well, he was like Tiger Woods. I think his mother was Thai or some East Asian variety and his father was an African American man. Joe I think had a strategic vision that probably not many of us in the student body could articulate or could formulate as clearly. His vision was to argue that students ought to be involved in running Caltech. I remember the argument that I formulated in favor of this: that professors who don't make tenure are only there for seven years. Students like me who are there for eight years have roughly the same length of tenure as tenure-failed faculty, and so why should I not have an opportunity to make my views known about what Caltech ought to be like, during the time that I was at Caltech?

ZIERLER: Interesting.

SZOLOVITS: I think arguments of that sort held some sway, and by either 1967 or 1968, the faculty actually added student members to all of the faculty committees except the Academic Freedom and Tenure Committee.

ZIERLER: Because that was confidential material?

SZOLOVITS: Because there was this history of the Board of Trustees being forced to resign by Lee DuBridge back in the 1950s when they were trying to fire Linus Pauling, and so this was sort of very high-stakes inside pool kind of stuff, and the faculty decided that that was really their purview, and the students had no voice in that. But every other committee, I believe, had a non-token representation of students on it. The majority of members were still faculty, but I think, for example, I was on the Admissions Committee for three years, and I think there were maybe five of us out of like 15 or something who were students on the committee.

ZIERLER: Is that to say that the most important discussions about this topic happened between students and faculty, and not students talking to other students and then presenting these conversations to faculty?

SZOLOVITS: I think to first approximation, yes. It's interesting, because I remember at that time, for example, students were tremendously at odds with the faculty and administration at Berkeley and at Columbia. There were riots and people taking over the president's office, and stuff like that, at these other schools. That never seemed like the Caltech style. [laughs] I think partly it's because the faculty were more responsive to our desire to get involved in those conversations, and I think on the whole, we felt like they took us seriously.

ZIERLER: Negatively defined first—and if you don't want to name names, that's certainly fine—but were there faculty who were against the idea, who said, "That's not what Caltech is, we should not admit women"? Were faculty talking openly like that at all?

SZOLOVITS: Mercifully, I don't remember specifics. I know that there definitely were faculty who thought that all of this stuff was a bad idea, that the faculty should run the place and students—I think somebody even wrote something in the student newspaper that said—what was it?—it's like we're a hotel, and the faculty are the staff, and the students are the guests, and the guests don't tell the hotel how to run itself.

ZIERLER: [laughs] If you can't remember specific individuals, there was a distinct—at least minority? Was it a minority of faculty who were against this decision?

SZOLOVITS: I assume it's a minority because the decision [laughs] went the other direction.

ZIERLER: Yeah, but the decision ultimately came to the president and the board. It's not a democratically situated kind of system.

SZOLOVITS: But I think the faculty had to be strongly in favor of it. I met a lot of our trustees at the time. There were two Democrats on the Board of Trustees at that time, Tom Watson Jr., head of IBM, and Dreyfuss, who was a Caltech Mech E grad, who became a very famous designer. Typically they were the favorite trustees of the students, because you could talk to them about progressive issues, and they would be willing to have those conversations.

ZIERLER: This is probably Henry Dreyfuss you're talking about?

SZOLOVITS: Henry Dreyfuss, yes. He was a really cool guy. Although, he and his wife committed suicide, which is not a nice way to exit. I think she became ill, and they decided to kill themselves instead of going through whatever was in store for her.

ZIERLER: Maybe you have a better, happier memory of the faculty who were real leaders in this. Do you have anybody in mind that was really out in front on this issue among the faculty in admitting women?

SZOLOVITS: I don't have an encyclopedic catalogue of who they were. I remember people like Harry Gray, Jerry Pine, Norman Davidson.

ZIERLER: What about Feynman, given his stature at Caltech?

SZOLOVITS: I don't remember what Feynman's personal opinions were on this. Normally, he was at some extremely high level of intellect that didn't bother with such trivia. Although I do remember that he came to the defense of a topless bar in Altadena where they were on trial for whatever it was, violating some city ordinance against topless bars, and Feynman took the stand and said that he had done some of his best physics work at that bar [laughs] inspired by the topless dancers. [laughs]

ZIERLER: I knew there was a reason I wanted to ask about Feynman. I just didn't know it was that! [laughs]

SZOLOVITS: [laughs] He was a very interesting character. I was really sorry—he gave the Feynman Lectures in I think 1962 and 1963, so I just missed them by a few years. But I studied physics from the Feynman Lectures, and they were taught by Ralph Leighton and Robbie Vogt and Barry Barish who had all been involved in—I think Leighton and Vogt are actually coauthors on the Feynman Lectures, because they turned Feynman's notes and lectures into the textbook. Barish, I think, had been involved in teaching the class. Of course, he became famous for winning the Nobel Prize for the LIGO experiments.

ZIERLER: What was the paper trail like? Do you recall a white paper, any documentation that was presented, either from students to faculty, or a joint paper that was presented to administration and trustees? What did that look like, or was it all verbal?

SZOLOVITS: My recollection is that it was mostly verbal. I think if you look at the student newspaper of that era, there may have been editorials or op-ed pieces or something that are relevant. If the Associated Students has maintained their archive of minutes, you might find that instructive, if they're willing to let you see them, which I suspect they might be.

ZIERLER: What stands out in your memory as some of the dramatic moments leading to the decision itself? How did that play out?

SZOLOVITS: To me, the first dramatic moment was when the faculty agreed to put students on faculty committees, because that gave us as students a platform from which to argue our case for everything else. It wasn't that admitting women was the main goal of the students. We also wanted the war to end, for example. When Lee DuBridge went off to become Nixon's presidential science advisor, we wanted a chance to influence who the next president of Caltech would be. In fact, I was on a committee that interviewed the final three or four candidates. We recommended for a different person than Harold Brown, but Brown became the choice of the trustees.

ZIERLER: What were the motivations in supporting the other person?

SZOLOVITS: I think we were worried that Brown was at that time very much associated with Edward Teller and Lawrence Livermore. He had been Teller's deputy chief of Livermore and then he became one of McNamara's whiz kids, so he was associated with the war in Vietnam and with the military policy. In fact, after he became president, I was among the group of people that went to him and said, "Look, we know from private discussions with you that you're not a big fan of the Vietnam War, but why don't you speak out against it?" He said, paraphrasing, "Well, because I still have some influence in Washington, and I don't want to give that up. If I speak out publicly, then that goes away."

ZIERLER: What did that comment strike you as? Was that reasonable? Was that cowardly? How did you feel about that?

SZOLOVITS: I thought it was actually reasonable. Brown was pretty much a straight shooter. I remember when we were interviewing him, we asked him—because DuBridge had been president of Caltech for 27 years or something, and the students thought that that was too long. He was a very dyed-in-the-wool, old style president, and so we thought it was important to have a little bit more frequent turnover. I remember we asked him how long would he serve as president of Caltech if he was selected. His answer was prescient. He said, "Until the next Democratic administration is elected in Washington."

ZIERLER: [laughs] Indeed!

SZOLOVITS: And sure enough, the next Democratic administration was Jimmy Carter, and Brown left to become Secretary of Defense.

ZIERLER: And he preserved that influence.


ZIERLER: From that comment or from other interactions, what sense did you get about Brown's politics regarding women and their admittance to Caltech?

SZOLOVITS: I don't remember whether he was wildly enthusiastic, but he certainly was not against it, and I thought he was in favor.

ZIERLER: To the extent that Caltech looks at peer institutions—what's going on at Stanford, Princeton, Chicago, places like that—was the sense among undergraduates that on the gender issue, Caltech was behind the times?

SZOLOVITS: Absolutely, yes. Caltech often looks to MIT for ideas about things to do, and MIT always admitted undergraduate women. MIT didn't have a faculty meeting in the 1950s to say, "Let's stop admitting women because it's worthless to educate them." For example, when I started at Caltech, the SURF enterprise, the Summer Undergraduate Research Fellowship, started, and it was acknowledged to be a complete copy of the UROP program at MIT, which worked very well. Now, as an MIT faculty member, it still works very well at MIT. I thought that was a good theft. Places like Stanford and the big universities were seen as sufficiently different that I'm not sure that they served the same kind of guiding role. Stanford has a law school and a medical school and a Department of Lower Yoruba and things like that, that just are not what Caltech is about.

ZIERLER: To the extent that you had visuals on what was happening between the trustees and the administration, what was the mechanism? Was it simply an up or down vote? How did they come to the decision and then how did they announce it?

SZOLOVITS: I don't know. I was able to attend a few trustees' meetings when I was in the student government because they invited us to those. But I suspect that a lot of those discussions happened before the trustees' meeting and not at the trustees' meeting. Typically, the administration would ask for support from the trustees but not engage in a long discussion about, "Which way should we go?" For one thing, I think the administration felt that they knew better than the trustees what ought to happen, and so they saw their role as convincing the trustees to do what they wanted to do.

ZIERLER: Do you have a sense of the timing, when actually this decision was made, when it was communicated to the students?

SZOLOVITS: I think it was made in 1969. I don't remember what month. I know that by the fall when Caltech announced its admissions season—you could apply for early admission, early decision, by maybe September or October, and by then, the policy was clearly in place, because they had not welcomed applications from women in previous years and they did that year.

ZIERLER: In 1969, you were a junior?

SZOLOVITS: I was a junior up to June of 1969, and then I became a senior in September.

ZIERLER: You were there for the first women admittance to Caltech, that first year?

SZOLOVITS: I admitted one of the first seven women to Caltech.

ZIERLER: Oh, that's great! Now, were you on the Admissions Committee? Did have a sense of—was there a flood of applicants as a result, or what did the numbers look like?

SZOLOVITS: The numbers were pathetic. The Caltech class has grown a little bit since then, but in those days, we admitted about 230 students to get a class that was pretty close to that. Most people who were admitted actually came. Or no, maybe we admitted more, I can't remember, but we were shooting for about 220, 225, as the class size. I think we admitted seven women. [laughs]

ZIERLER: Now, how much of that was bad PR? Did Caltech not do a good job conveying that women were now welcome to apply?

SZOLOVITS: Think about it. If you were a senior in high school, and you had an opportunity to go be one of a very small number of women at this previously all-male school, how attractive would that be to you?

ZIERLER: Fair point. That must make the original seven quite special indeed.

SZOLOVITS: Yes. The woman that I was a particular champion of, I remember—the way the Admissions Committee worked in those days is that everybody had a region where they were responsible for the applicants from that region. Caltech has I think still many more applicants from California than it does from Iowa. Mostly, the people who traveled to the Midwest or the East Coast were faculty. The students were mostly assigned to regions in California. I shared a region with Dick Dean (math professor) and Jerry Pine (physics) in different years, in I think the San Fernando Valley and parts of Santa Monica or something. I had this district and I remember going to the school where there was a young woman who basically ran the school. She was in their student government and she was active in clubs. Every teacher said, "She's the most brilliant generator of interesting ideas" and so on. One nice thing about Caltech Admissions is that we got to interview the teachers as well as the students, and so you come away with a much clearer idea of what a student is like. MIT has these alumni interviews which I think are relatively worthless, because they're not calibrated. So, that was good, and I came back and I was really enthusiastic about this woman, but it turned out she didn't have all 800s on her SATs. The way the Admissions Committee worked is you went round robin. When it came to be your turn, you would raise your top candidate, and you would propose them. Then if somebody around the table said, "Well, I have somebody who's better than that," then that person might get deferred. I raised this woman at some round in the admissions process, and I think it was actually Harry Gray who said, "Oh, I have a better candidate." Harry had a reputation as having a good eye for beauty. This girl happened to be very pretty, in addition to her talents. I remember walking over to Harry and showing him her picture [laughs], and he acquiesced to admitting her. I can't remember her name anymore, but I looked her up at one point a couple of decades ago, and in fact she graduated, she got a PhD in biology at Berkeley or someplace and she became a college professor. So, it worked out well. I felt gratified. She was one of seven.

ZIERLER: Did you have the pleasure of having any of those seven in the classes you took as a senior? Or they must have all been freshman, so I guess not?

SZOLOVITS: No, they were admitted just as I was graduating. We admitted them in the Spring of 1970, which meant that they became freshmen in the Fall of 1970. By then, I was a graduate student.

ZIERLER: What about accommodations? Was there a new dorm that was built? Were there bathrooms that were added? What was done immediately, if anything?

SZOLOVITS: Immediately, they took one of the alleys that runs between Blacker and Dabney and converted it into a women's corridor, women's alley. Those bathrooms became women's bathrooms. I don't know if they took the urinals out [laughs] or what they did with them. The women were there, together, in one alley. Otherwise they were integrated with the houses.

ZIERLER: As you said, the numbers were pathetic. Did you have a sense, at least institutionally at Caltech, that female admitted students would always be tokens, or was there a momentum even from the beginning that at some point this will build to something approaching equity between the genders?

SZOLOVITS: I think very much the latter. I think people understood that this was just the crack in the door, and that it was going to change over time. What's the percentage of women in the Caltech undergraduate classes now?

ZIERLER: Oh, I don't have a number offhand, but it's much better, way better than that from the beginning, obviously.

SZOLOVITS: When I came to MIT, if I remember right, the fraction of undergraduate women was about a quarter. Now, it's about half, so we've reached parity.

ZIERLER: I just Googled it—54% male, 46% female, according to U.S. News from a couple years ago.

SZOLOVITS: Okay, so you're comparable to MIT. I think that's very much what we were expecting.

ZIERLER: Of course here it's still science and engineering where generally women are still underrepresented.

SZOLOVITS: Yeah. At MIT, I heard a talk by one of our administrators a couple of years ago about this. Women are generally underrepresented in engineering and overrepresented in the sciences. I think both math and biology are significant majority women, and my department, for example, which is electrical engineering and computer science, is still I think on a good day about 30% women.

ZIERLER: Looking back, Pete, what gives you the most satisfaction in your involvement in this issue?

SZOLOVITS: That we succeeded. [laughs] For a number of different reasons. One thing is that when you invest a lot of effort in something, having it pay off is just glorious. The other, of course, is that I still believe that our goals were the right goals.

ZIERLER: How much credit does Brown get in all of this, do you think?

SZOLOVITS: I don't know. Some of what happened happened before Brown. For example, Brown became president in 1969, I think, and I think this was already well underway.

ZIERLER: He inherited something that already had momentum, you're saying?

SZOLOVITS: I think that's right.

ZIERLER: It's more like he would have had to go out of his way to block that momentum?


ZIERLER: In terms of leadership prior, what about DuBridge? Was anything happening that DuBridge deserves credit for?

SZOLOVITS: Probably. The one thing that I've always held DuBridge in terrific respect had to do with the Pauling incident that I told you about. Basically, DuBridge organized the faculty to say that if the trustees fired Pauling, the entire faculty would quit.

ZIERLER: That is amazing.

SZOLOVITS: [laughs] Pretty good!

ZIERLER: Have you been an active alumni in the years since? Have you stayed in touch with Caltech?

SZOLOVITS: A little bit. I send you guys money, and my late wife and I, when we drew up our legacy plans, decided to leave a bunch of money to Caltech. [laughs] Unfortunately, as a professor, and she was a do-good lawyer, neither of us made the kinds of oodles of money that big donors make. But we left or we will leave something mainly to support undergraduate research which I think was really a critical component of my education.

ZIERLER: The basis of the question, of course, was if you've been able to derive any pleasure over the years in looking at how women have done at Caltech, based on part on what you were able to accomplish?

SZOLOVITS: Yeah. Not in detail. Before the pandemic, I did get to Caltech fairly often. My son lives in Los Angeles, and one of my brothers-in-law lives on San Pasqual, right near the campus, and so usually when I'm in Southern California, I crash in his guest room, and Caltech is one of the places I tend to go pretty often. But not in any real formal sense. I think now your development office likes to stay in touch, because they're licking their chops waiting for me to die [laughs].

ZIERLER: [laughs] Pete, just to bring it full circle to earlier in our conversation, what were the circumstances of your decision to stay at Caltech for graduate school?

SZOLOVITS: The draft.

ZIERLER: As opposed to going somewhere else, I meant.

SZOLOVITS: Interestingly, I had actually decided to go to Harvard for grad school, in computer science. I was admitted, and then my draft board sent me a nice letter. You don't remember, but if you read your history, this was the era of the draft lottery. I got number 70, and my draft board assured me that they were going to draft up to somewhere in the low 200s, and so I was screwed. What they said is, "If you start an academic year in your graduate program, we will let you defer reporting for duty until the end of the academic year." I figured I could get one year out of that if they didn't draft me over the summer in 1970, which I didn't think they were going to do. But, that wasn't very satisfying, because I wanted a PhD and I wasn't going to do that in a year.

I actually went to Harold Brown and I said, "Dr. Brown, I have this predicament." [laughs] He was very sympathetic, and he was of course very informed. He suggested the course of action that I took. He said, "Well, join ROTC. That automatically gets you a two-year deferment. After two years, you will be halfway through a PhD program, so then when you get your commission, you write to whatever service and say, ‘I'm halfway to my PhD. Could I please have another two-year deferment in order to finish my PhD, and then I will be much more useful to you as a PhD-level researcher.'" The other advice he gave me was, "Among the different services, join Air Force ROTC." I said, "Why is that?" He had been secretary of the Air Force so he probably had some prejudice in favor of the Air Force, but he said, "In the last couple of rounds of RIFs, reductions in force, the Navy got hit, and then the Army got hit, so it's the Air Force's turn next. What you should think about is that four years from now, the Vietnam War will probably be over, and the military will be getting rid of people, and so they might get rid of you, in which case you wouldn't have to do your four-year active duty service."

ZIERLER: Now, that is some strategic advice right there.

SZOLOVITS: And that's precisely what happened. The only problem I ran into is that Harvard in a fit of moral clarity had gotten rid of their ROTC unit, and so I couldn't join ROTC at Harvard. Brown actually was very kind. He called up Jerry Wiesner, who was president of MIT at the time, and he said, "This guy wants to go to his PhD at Harvard. Could he join ROTC at MIT?" Wiesner said, "Hell, no!" [laughs]

ZIERLER: So you stayed.

SZOLOVITS: I had also been accepted by my department. I couldn't have stayed at Caltech as a physics major, but I had decided to switch to computer science anyway. Computer science, or what at that time was called information science, had offered me admission as a graduate student, so I stayed there, joined Air Force ROTC, got my commission two years later, got my educational delay, and another year after that, I got this nice letter from the Air Force saying, "You're redundant and useless to us. Please go on active duty now for three months, and then we'll forgive you for your four-year active duty obligation."

ZIERLER: That's great. Pete, last question to bring it back to the main topic of our conversation. At a later stage in life, and because Caltech had long admitted female graduate students, in what ways did your graduate experience deepen your appreciation for what Caltech had decided thanks in part to people like you?

SZOLOVITS: At the time I was a graduate student, my entire department was tiny. We had four permanent faculty and one visiting professor, and just a dozen graduate students. I think none of them were women, and so I had no particular experience with women graduate students at that time. When I came to MIT, my initial group of graduate students were all men, because we didn't have that many women graduate students. I was pioneering in a field that was not very popular at the time, sort of medical AI, and there happened not to be any women who were interested in that. I did serve on one woman's PhD committee who was a natural language processing person, which is something I knew very little about. That was my only connection with a woman graduate student.

Then by about the early to mid 1980s, about ten years into my career, I did attract three women graduate students who worked with me. Two of them got PhDs and went on to very successful careers. One of them became chief technology officer of Samsung Corporation. Then when she retired from that, she's now a professor at McGill University in Montreal. The other one became a research leader at MITRE Corporation, and then her husband moved to Toronto, and so she got involved in the Canadian national healthcare system as a consultant and has been doing that for a number of years. So, they worked out well. The third one stopped with her master's degree, and I think she is now working as a researcher at the University of Washington. So, they all did fine, and ever since, I've always been on the lookout for bright women who are interested in my area. Last year, I finally had one of my women doctoral graduates who became an MIT professor.

ZIERLER: Oh, wow.

SZOLOVITS: Which is kind of nice.

ZIERLER: Now that's full circle. That's great.


ZIERLER: Pete, on that note, it has been great spending this time with you. I'm so glad we were able to capture this history. Thank you so much.