Skip to content
 

Bell Labs

Sining Chen told me they’re hiring in the statistics group at Bell Labs. I’ll do my bit for economic stimulus by announcing this job (see below).

I love Bell Labs. I worked there for three summers, in a physics lab in 1985-86 under the supervision of Loren Pfeiffer, and by myself in the statistics group in 1990.

I learned a lot working for Loren. He was a really smart and driven guy. His lab was a small set of rooms—in Bell Labs, everything’s in a small room, as they value the positive externality of close physical proximity of different labs, which you get by making each lab compact—and it was Loren, his assistant (a guy named Ken West who kept everything running in the lab), and three summer students: me, Gowton Achaibar, and a girl whose name I’ve forgotten. Gowtan and I had a lot of fun chatting in the lab. One day I made a silly comment about Gowton’s accent—he was from Guyana and pronounced “three” as “tree”—and then I apologized and said: Hey, here I am making fun of your accent, but really it’s impressive you speak English so well. After all, I can’t speak . . . um, what’s your first language? To which Gowton replied: English. I speak only English.

As I was saying, Loren was serious about his work. He was in his mid-40s and was married with a little baby, but he came in to the lab every day and worked the whole time. I remember one day we were leaving and I turned off the lights. He went back in and turned all the lights back on. I think he objected on principle to energy conservation. He also had the idea that the U.S. should mint one-dollar silver coins with less than $1 worth of silver in them, thus making a profit on the difference. He and his family lived in the suburbs and he said he hadn’t been to NYC in a few years. He wasn’t interested in literature and said that when he flew to conferences he took some light physics reading for the plane. He also had a story about when he spoke on his research at a conference and some Japanese guys listened to his talk and then went off and ripped off his ideas.

The second summer I worked at the Labs, I was making a huge amount of progress near the end, but then I had to stop the project in the middle because I’d planned a trip to Europe. He told me I shouldn’t do it, that Europe is no big deal but I’d regret not finishing the project. The funny thing is . . . he was right! I did not enjoy my trip, and I would’ve got immense satisfaction from getting the work done. (I worked long hours during the last week gathering data, running computer programs, and writing things up—we could actually write papers longhand and give then to the Typing Pool to get them typed up!—and we did get some publications out of it. But the one idea that was truly mine—a demonstration of the superheating of defect-free silicon—never quite got into publishable shape, and I don’t think Loren or the others ever followed up on it.) Another one of Loren’s sayings was, You’ll have plenty of time to relax when you’re dead.

Bell Labs was just an amazing place and we all felt privileged to work there. When I worked there in 85 and 86, I stayed in some dorm housing they arranged at a nearby college campus. One of my roommates in 85 was a fun guy named Bob Miller who had a thick Pittsburgh accent and a Pierce Brosnan-like attractiveness to women. Just about every weekend we drove to the mall or the beach, or both. I have no idea what’s up with him—it’s not so easy to track down someone named Bob Miller! We had two other roommates who were a couple years older: Guy Marong from Luxembourg and a guy whose name I forget from the Bronx who had a tough life—during the year he went to school at the College of Staten Island, which he said involved an exhausting subway/ferry/bus commute. This guy was a big eater—for dinner he would fry up three big burgers on the stove—and he had some great stories. As a teenager, he told us, he’d worked in a religious store in the Bronx, the kind of store that sold statues of angels in the front and potions in the back. He’d mix the potions himself and thought it was just ridiculous that people would come in and buy them as if a jar of water and food coloring could really work magic. Bell Labs arranged a bus that would take us from the dorms to the work in the morning and return us at 5 in the afternoon. So we all worked long hours. Sometimes I’d hang out in the little Bell Labs library. I think that’s where I read Exploratory Data Analysis and the Federalist Papers books.

One weekend night a bunch of us drove to some faraway place to go to a club that wouldn’t let us all in—they checked id’s and I was only 20. So we ended up driving back—the guy driving me had had a few drunks and was going way too fast on the freeway ramp. After that I’ve been pretty careful to avoid getting in the car with a drunk. (A couple years later a different friend got mad at me because I insisted on taking the subway rather than get in a car he was driving from Manhattan to Brooklyn. This was a dude who’d once fallen asleep for a few seconds while driving on the New Jersey Turnpike—I’d been in the car at the time!—so I think I can be excused for my caution.)

I had some roommates in 1986 as well, but it wasn’t quite so much fun, maybe it was the different group or maybe just that the second year is never quite the same. All four of us were MIT students. One of my three roommates was Rick, a very nice guy from southern Ohio, where his family ran a car-repair place. One of the others was a flamboyant Latino gay guy named Art who I would sometimes give a lift to a club in New Brunswick. Art had a gay guidebook to the U.S., the bandanna code, the whole deal. One weekend Rick had a friend visiting, his roommate from MIT, a buttoned-down white guy who was a real homophobe. Not in an obnoxious way but you could tell that the whole gay thing made him uncomfortable. The funny thing was that Art was out of town so the ‘phobe had to sleep in his bed! He didn’t want to use Art’s pillows, though.

I didn’t really fit in so well with some of these people. I liked them but didn’t have much in common with them, but I was too shy to mix and find the other summer students who had interests closer to mine. (Here I’m thinking of literature etc., not statistical graphics and MCMC convergence.) In junior high and high school I ended up having a great group of friends without ever having to put in effort to make the connections, and it took me a long time to figure out how to make these sorts of connections on my own.

I returned to Bell Labs to work in the statistics group for six weeks in 1990, but this time was different. I’d completed my Ph.D. and was set to go to Berkeley in the fall. During the previous winter, I’d applied for several jobs, and my first interview was at Bell Labs. I loved the place and really wanted to work there, but I have to admit I was worried that I wouldn’t fit in. Everybody had a picture on their desk of wife and 3 kids, and I didn’t feel ready for that lifestyle. (But, let me assure you, I have nothing at all against people who have pictures of their wife and 3 kids in their office!). I think I would’ve taken the Bell Labs job, except that (a) it was located in the New Jersey suburbs, and (b) they didn’t actually offer me the job until two days before I had to decide between my other job offers. Bell Labs was my first interview but the last place to offer me a job. They actually called me at 9am on a Wednesday with the job offer. I sleepily turned them down, saying that I was deciding between Chicago and Berkeley. (I chose the place where not all the people were nice but the bread was excellent. And who among you would say I made the wrong decision?)

After that, I resolved to never, never turn down a job offer in my sleep.

Anyway, I called up the Bell Labs people around May and told them I was finishing early, could I work there for a few weeks during the summer? They said sure. But this time I worked on my own. I came in every day—I stayed with a couple other stat/CS guys in Trevor Hastie’s house—and worked, then hung out in NY on the weekends, staying with my sister in Brooklyn. At this point my car wasn’t working so well. It had some problem with the starter motor. Sometimes when I was parked or stopped at a light the car wouldn’t start and I’d open the door, start pushing, then pop it into second gear to get it going. I got in the habit of never parking at the bottom of a hill. Not long after, I got it fixed, but at that point I’d already bought a one-way flight to California so I sold it, for about the cost of the repair bill I’d just paid. I sometimes have dreams that take place in the present time but in which I still have my car. But I’m almost certain it’s been junked—how many 35-year-old Corollas are still running out there?

Back at the Labs, I had various ideas of what to work on but got almost nothing done. But then, somehow—I’m not sure how it happened—I came up with a useful way of monitoring the mixing of iterative simulations. I wrote up my idea—it was about 9 pages single-spaced—and I knew it was the best idea I would ever have. It took about 2 years for the paper to actually get published, but I knew it was dynamite right away. I showed it to Don Rubin, he had some suggestions, and we wrote it up more formally. In the end, then, I had very little interaction, intellectually or socially, with the others in the statistics group that summer, but something useful came out of it.

Here’s the job announcement:

Research Scientist in Statistics and Learning at Bell Labs

We invite applications for a research position in statistics, or Computer Science with a concentration in Machine Learning, Data Mining, or Pattern Recognition, and be knowledgeable in fundamental theory, classical and modern methods, and standard analysis tools in statistics and learning.

Candidates must demonstrate a potential for performing original, pioneering research, with promises for strong, practical impact. Also expected are a collaborative spirit, broad technical and application interests, and strong communication skills.

We invite applications from both new and experienced graduates. Women and underrepresented minorities are especially encouraged to apply. Alcatel-Lucent is an Equal Opportunity Employer.

See http://ect.bell-labs.com/sl for more details about the department.

Interested applicants please forward a CV, a Statement of Interests, samples of significant publications, and the names and contact information of 3 referees to Dr. Tin Kam Ho at tin.ho@alcatel-lucent.com.

Evaluation of candidates begins immediately and will continue until the position is filled.

As you might know, statistics at Bell Labs has an impressive history, with highlights including the fast Fourier transform, lowess, S (which has become R), generalized additive models, and . . . R-hat! Bell Labs isn’t what it once was—which just makes me want to cry—but maybe I’m just sad about the passage of time. The last time I was there was a bit over five years ago, when Diane Lambert was nice enough to invite me to a mini-conference they’d arranged on the occasion of John Chambers’s retirement. I was surprised and pleased to have been invited, as I did not know Chambers well—he has a quiet demeanor and when I was working at Bell I had not yet started thinking seriously about statistical graphics. The conference was interesting and I followed it up by writing a several-paged single-spaced letter in my notebook (my paper notebook—this was before I had the portable laptop computer) to John, conveying my thoughts on the connections between statistical graphics and complex modeling. Nowadays I’d put this sort of thing on the blog, but in the event I put the letter in an envelope addressed to John Chambers, Member of Technical Staff (retired), Bell Labs, Murray Hill, N.J. I was told that John went on a long vacation upon his retirement. I hope the letter eventually reached him. Writing it occupied my entire train ride back to New York.

P.S. I looked up Gowton Achaibar on the web and he seems to be doing pretty well.

P.P.S. I looked up Loren to see what he was doing now and I found this (see also here). When I was applying to grad schools in physics, Loren was one of my letter writers, but I was a little afraid of what he might write—he was a very nice guy but a bit personally distant—so as a safeguard I had one of his letters sent to me. (I can’t remember how I arranged this.) I read the letter and it was really positive! So then I felt like bad about doing that.

P.P.P.S. Look carefully at the list of speakers at the program for the Chambers conference!

11 Comments

  1. John Mashey says:

    Yes, and here I link to some of the people who were around. I left in 1983, just before the Bell System split up, when Bell Labs was ~25,000 people, after having a decade or two of steady growth (especially with big expansion of software efforts.)

    Sadly, when organizations go through long, slow downsizing, the demographics do change.

    See also: R2-D2 on nature of serious R&D management.

    • Andrew says:

      This reminds me of the mystery of how I, as one person, can make any contributions at all, given the existence of various big-budget organizations (e.g., Microsoft Research, the University of X, where X is any large state) that throw enough money at statistics to hire lots of me’s.

      Now that I am going through millions of dollars of government grants (and, as a result, producing Stan, mi, blme, etc.), this is less of a question, but I could certainly have raised it 10 or 20 years ago.

      In statistics (although maybe not anymore in physics) there are so many different possible ideas and so many different interactions between basic research and applications, that a single researcher or small group can make real progress by being in the right place at the right time. Hence the benefit of diversity.

      For example, consider some famous Bell Labs research: S. S (which became R) is great, but it’s traditionally been focused on graphics and exploratory data analysis, since that’s what Tukey etc. liked to do. Much of the statistical modeling in S and R is an afterthought, as can be seen for example in the Venables and Ripley book, which devotes many loving chapters to nonparametric methods but then throws off linear models, Anova, etc., as a sort of obligation. In my letter to John Chambers, I discussed the potential benefits of applying exploratory-data-analysis thinking to complex modeling.

      It’s the nature of successful projects to get stuck, I think, hence helpful for outsiders to be able to make connections.

      (With my Harvard degree etc., I’m certainly no “outsider” in statistics, but I do think that in many of my research projects, I’m coming from the outside, whether from my non-subjectivist Bayesianism or my model-based graphics.)

      All the above is not really about me. I think the same general idea would apply to any individual researcher who is working near, but not part of, a larger group.

      • John mashey says:

        Well, Tukey did have some influence :-)

        But beyond that, certainly at BTL (but also elsewhere), small numbers of people have managed to create huge effects.
        Read that R2-D2 article: in a big R&D organization, one needs to run a research funnel with progressive commitment, but the R1 projects generally start very small. Some big organizations have spent much money on hiring many people and not necessarily gotten great payback, i.e., throwing money at this is often the wrong thing if done too early.

        Recall that UNIX was originally written by 2 guys (one of whom, Dennis Ritchie died recently.) When I started in 1973, there were ~20 UNIX systems in existence.

        As for possible limitations of S/R, recall that the first UNIX version of S was on the VAX 11/780, a 5MHz (not GHz) machine with typical early memory sizes of 1-2MB (not GB). An iPhone has vastly more … everything, but at least a VAX was a step up from a PDP-11. Early origins of languages matter a great deal, but of course S=>R is one of a relatively successful survivors and generally, at any given language level, there are only so many niches. C wiped out assembler language and other languages around its level. You may want something above R.

        See Languages, Levels, Libraries, and Longevity,
        or what happens if a UNIX C timer routine is running your starship 5,000 years from now.

  2. ymb says:

    Just be aware that there is no library anymore to hang out.

  3. Cathy O'Neil says:

    Wow, you have a lot of energy. Just reading that essay makes me exhausted.

    I visited Bell Labs once for a week, I was invited by Joel Spencer but he wasn’t there, and instead Ron Graham hosted me and my husband. It was the only week I ever did math with my husband. It was really great and productive, but I can’t think about it without also remembering that I was pregnant with my first child, and going through really bad morning sickness, and it seemed like every food option in suburban New Jersey was overly salted and greasy. I ended up finding one, exactly one, place I could stand eating, I think it was a T.G.I.Friday’s or something. It was like an alien planet. But Bell Labs was nice. I also remember meeting Jeff Lagarias there, who told me I could use a Bell Labs notebook but then they would own any ideas I wrote in it, even if I didn’t write anything in it for fifteen years. I still have that notebook somewhere, and I have nearly filled it up with shopping lists.

    Cathy

    • John Mashey says:

      “Ron Graham hosted me and my husband”
      Did either of you juggle?

      “food option” yes, that was an issue, 5 of my years were spent a mile from BTL-Murray Hill and food might have been lighter.

  4. Brent Buckner says:

    You wrote:
    “Bell Labs arranged a bus that would take us from the dorms to the work in the morning and return us at 5 in the afternoon. So we all worked long hours.”

    So the bus ran 7 days a week? Or the morning trip from the dorms to the work was at 4?
    ;-)

    • Andrew says:

      Just 5 days a week. But the bus arrived at the Labs before 9 and left around 5. That was a pretty long day, considering that there was nothing to do inside but work.

      Don’t forget, back then there was no email, web, etc. Work was a place to work, not to goof off.

      • Brent Buckner says:

        I didn’t think that a 40-hour (full) work week would have met your comparables for “long hours”; I stand corrected.

      • John Mashey says:

        No web and email? there was no web, but there certainly was email inside BTL and elsewhere via UUCP, there were various other networks, Of course, unlike today, where people should be assigned a permanent email address at birth :-), only a relatively small fraction of people had access to convenient email that could easily reach enough people.

        This is obvious example fo a strong network effect, akin to earlier telephones, especially ion the early transition from unconnected local nets to being able to reach a huge network.

        Search Google Images of “600 Mountain Ave, Murray Hill, New Jersey”.
        That’s a ~5,000-person complex in the woods. The nearest builtup are is New Providence and as noted by Cathy, its restaurants were not exciting.
        People went to work, had lunch and stayed there.

  5. Bell Labs sure was different when I got there after Lucent took over in 1996. By 2000, it was dying, by 2002, completely dead (my whole group in speech and image processing was disbanded, which really bummed me out because we had just hired one of my grad students from CMU and he was just starting as it died).

    But we did have busloads of interns over the summer. I’d hitch a ride back with them to Newark, where I could catch the PATH back to NYC, where there’s lots of good food.

    Of course there was e-mail in the 1980s at places like Bell Labs. There were also newsgroups. I used to tell everyone that e-mail would be the only way they’d communicate once everyone got networked computers. Everyone just stared blankly and said things like “Why would I ever want to do that”. I was hardly visionary enough to imagine mobile texting phones from my vantage of the mid-1980s when Sun Workstations had just replaced those VAX-11-750s and PDP-11s.

Leave a Reply