The Fractalist

by Benoit Mandelbrot

  On July 20, 1958, unaware of what I was doing, I cast a die never to be retrieved. A summer job at IBM in New York ended my apprenticeship.

  21

  At IBM Research Through Its Golden Age in the Sciences, 1958–93

  JUNE 20, 1958, was originally meant to begin a summer visit to IBM, located in Yorktown Heights, about an hour north of New York City. Aliette, baby Laurent, and I flew from Paris to New York. The three of us and our few suitcases easily fit into a taxicab. The difference in time zones made that day very long and tiring—yet almost routine and humdrum. But appearances were deceiving.

  A Summer Job Becomes a Life’s Work

  Chance took me to IBM for that summer job in 1958. But only weeks after my arrival, I changed my mind and decided to stay. When I retired in 1993, my formal link to IBM had lasted for close to the thirty-five-year stretch of that company’s concern with science. I stayed neither for the income nor for any subtle reason. I stayed by simple necessity—as I interpreted it after a staff meeting called by IBM’s first director of research, Emanuel Piore (1908–2000).

  Manny started the meeting with an observation: “I hear rumors of great unease among the troops. Many of you seem to wonder why we hired you, and you worry about the constant churning. When will the management stabilize, and when will you be told what we really want you to work on? In fact, there is no secret whatsoever.”

  As he continued, he confirmed my own interpretation of the situation.

  “Most of you are fresh Ph.D.s and believe that the highest calling is to compete with your former adviser in adding footnotes to your thesis. But you’ll soon find that on a daily basis, pure scientific research is a very difficult and in most cases unrewarding profession. At work you never have enough time to do what you want, and your wives complain that on Saturday mornings you go to the lab instead of taking the kids to the ball game. If you want to please your wife by being better paid and coming home in the evening without a briefcase, just tell us. The IBM development laboratories in upstate New York must grow, and their staff must be almost completely upgraded. Many prime jobs beg to be filled.

  “But you may be hooked on pure scientific research. Fine. Wonderful. Research can offer you the choice between any number of exciting and well-rewarded tasks. Some of you may even dream of becoming great scientists. Marvelous! We can easily afford a few great scientists doing their own thing.”

  I still remember an extraordinary thrill and feeling of relief upon hearing those words. Given my gambling mood at that point in life, IBM’s constant churning was a major attraction for me. The last thing I wanted was the order I knew in France. Right or wrong, I felt sufficiently stimulated by personal dreams. So I gambled on IBM and IBM gambled on me. We both succeeded and those successes were not unrelated.

  Manny Piore’s words convinced me—and, in turn, I convinced Aliette—that it would be best to linger at IBM for perhaps a year or so. Soon it became clear that I had set myself to remain there for an open-ended period.

  I think back to Manny’s words at that staff meeting: “Pure scientific research is a very difficult and in most cases unrewarding profession.… You never have enough time to do what you want … on Saturday mornings you go to the lab instead of taking the kids to the ball game.” These words would come true in my case. The Hippocratic oath, taken by physicians, says, “First, do no harm.” I strongly feel that it ought to apply to scientists as well. A father with a self-assigned and never-fulfilled mission is not a full-time father and can play havoc with his family. With Aliette in charge, I think I was able to abide by the oath. Let’s leave it at that.

  Settling in the United States

  Now that we had decided to stay, it was time to find a place to live. Our two-year honeymoon near Geneva in the peerless La Boverie was in our minds when Aliette and I went house hunting. Lady Luck helped, and for four years we lived in an uncannily similar place—above the garage on an estate belonging to David Swope, where our second son, Didier, was born.

  (Illustration Credit 21.4)

  When we moved in, we missed the open view, south over the Rhône River, that graced our first home. But—once again—fate was on our side. The night before Thanksgiving brought a dreadful storm, and the next morning—lo and behold!—enormous trees had been uprooted and an admirable view had opened down the Hudson River all the way south to the Tappan Zee Bridge and, on clear days, even to Manhattan!

  After the Swope estate, we bought a house in Chappaqua, chosen for its blandness—which delivered on its promise—and lived there for five years. It was a perfect setting for our young boys. In the photograph on the following page, I am flanked by Didier and Laurent in our Chappaqua living room.

  For the next thirty-five years, our home was a place matching La Boverie and the Swope estate: a wonderful old pile in Scarsdale, five minutes from shopping yet so completely isolated from the neighborhood as to remind one of an old-fashioned Japanese house—on a far bigger scale. After an anonymous entry, the road immediately took a sharp turn into a lot so private it felt like a reassuring womb. Near the top of a rocky outcrop that probably never seemed worth farming, it became lost in a grove of very old oaks. The oak nearest the house may well have dated to the arrival of the white man to those shores.

  (Illustration Credit 21.5)

  Because the thorough complication of this house—which enchanted me—scared off all “normal” bidders, we could afford it. It was built in stages between 1840 and 1940, so the height of the rooms on the first floor varied, and the second floor was filled with threatening steps. Not one window was a rectangle when we moved in, as we found out whenever work had to be done. No contractor dared estimate the cost of any change. But chance led me to a soul mate, a man named Robert Robillard, a schoolteacher who knew how to use every tool—the kind of man whom I visualize in a horse-driven wagon on the way to conquer America. He needed additional income and was looking for intellectual challenge, and so for many years, the two of us fixed basic wear and tear in that old pile.

  So the date June 20, 1958, has come to mark in my mind the midpoint of my life. That date has figured on thousands of forms and will not be forgotten. Next to the 1936 move from Warsaw to Paris, it witnessed the second major break in my life.

  Unlike the first, this break never became complete, insofar as we still speak French at home. Besides, how could I possibly forget the country that helped me survive the war, accepted me, offered me its culture, and made me a free man? I never regretted the move to the United States. However, increasingly—and unavoidably so, as time goes on and friends vanish—I feel in France like a visitor from far away and far back in time.

  Emanuel Piore

  Born in Lithuania, Manny Piore had received his Ph.D. when the Depression hit the United States. He survived the bad days, and when war broke out and scientists were suddenly needed, he went to Washington, D.C. He became a key man in the creation of a three-legged institutional system linking the National Science Foundation (NSF), the National Institutes of Health (NIH), and the Office of Naval Research (ONR) in a policy to support science well beyond topics of direct interest to the navy. Shortly afterward, he became director of research at IBM and built up the Thomas J. Watson Research Center.

  In other words, this one remarkable man was responsible for several major science-funding institutions in the United States. Neither a great scientist nor an engineering innovator, he was a shrewd operator with a sense of noblesse oblige. The system he built was far from perfect but had an extraordinary asset: each leg used different criteria. And IBM Research chose its own way.

  Since then, IBM and the ONR have abandoned pure research. The bulk of academic support comes from the NSF and the NIH. The process they use to select their beneficiaries applies the same criteria to everyone. This may explain why the NSF has always treated me abominably—almost invariably, the topic I proposed to investigate was perceived as incomprehensible, annoying, or worse.

  Ralph E. Gomory


  IBM’s research staff was mostly very young in 1958. Shortly after my discreet arrival, a new hire thundered in—Ralph E. Gomory, five years my junior. For me, he was the most important IBMer in every way.

  His Ph.D. from Princeton was also in a classic subfield of pure mathematics that he immediately abandoned as being too mature. He then solved a famous problem of applied mathematics by finding an algorithm that gives integer, not fractional, solutions to linear programming problems. Integer solutions are required in many actual applications, and solving this problem made Ralph extremely well known. IBM hired him and he achieved several other breakthroughs.

  We met not long after his arrival, when my elder son joined a play group run by his wife. We became good friends, and I moved into his department. He was named director of mathematical sciences, then director of research, before he left for IBM corporate headquarters. After retiring from IBM, he was the president of the Sloan Foundation for many years.

  Ralph’s being my manager may well have been the longest-lasting direct or near-direct reporting relationship at IBM. Of course, as his schedule grew ever more demanding, our actual meetings became increasingly rare. But there is no question that, after Lady Luck, it is to Ralph that I owe the most: being made an IBM Fellow and, more important, being allowed the freedom to either wander off or dig in, without which the gamble I took would not have lasted nor borne the fruit it did.

  Golden Ages—Mythical and Real

  The golden age of IBM in the sciences coincided with the thirty-five years of my own belated golden period. It began roughly when I came to IBM and ended on a day when half of the staff was asked to leave and the other half was asked to get practical. Having come by chance in 1958, I stayed because nobody offered a better fit, and quickly thrived. The gamble for both parties was very successful. Within three years I had made two major discoveries—like hitting two jackpots in a row! Each one brought a visiting professorship in a field I knew almost nothing about: first in economics (later called finance) and next in the field of exotic noises.

  My consulting role for “down-to-earth” colleagues was varied and mostly enjoyable. In many cases, I was solely a passing helper. But on one unforgettable occasion, a colleague and friend, Jay M. Berger, asked for help with some unglamorous but pesky noises on telephone lines. Our solution saved IBM from investing heavily in a development that was bound to fail. This episode also steered me to a scientific topic that I would have certainly missed.

  Otherwise, consulting hardly interfered with my contributions to diverse existing sciences—or perhaps to a new science of roughness. IBM gave me the base that made everything else possible. It allowed essential visits to Harvard and Yale and other arrangements of brick-and-mortar. It also allowed me to move effortlessly between nonoverlapping fields of knowledge and organized science.

  Does it follow that those years were a paradise on earth? Of course not! During IBM’s years of glory as a scientific powerhouse, daily life saw many avoidable negatives. Of course! But remarkably, the overall balance was very positive. It may be that—to last—pure gold must be alloyed, and useful alloys produced in great haste may contain noxious elements. There was a constant element of pointless silliness, bad apples at different levels of the hierarchy, and the like. A paradise IBM’s golden age never was. But I always felt that personal freedom, if not priceless, was very expensive. I paid the price and got something in return. Fair bargain.

  How IBM Came to Invite Me

  By my life’s peculiar standards, the causal chain that led to IBM was short, with no unusual wrinkles. When I was at Princeton as a postdoc, I met Manfred Kochen (1928–84) at the Institute’s cafeteria. He was a younger man with a background and ambitions similar to mine. He went on to join IBM when the brand-new research division was recruiting relentlessly and indulging in active public relations. Hearing I was going to spend the summer of 1957 at Cornell University in upstate New York, “Fred” brought me to a temporary site—ostensibly for a lecture but in reality for a combined interview and sales pitch.

  His manager was the physicist Michael Watanabe (1910–93), whose Ph.D. chairman in Paris had been Louis de Broglie. The manager above that was Nathaniel Rochester (1919–2001), a career IBM engineer credited with a near replica of von Neumann’s pioneering Princeton computer. Staff was needed for a machine translation project—very premature but well supported—and I was a rare warm body with good name recognition for my work in linguistics. I told them that a very nice job was waiting for me in Lille, and that my interests had shifted. Their answer was that they needed good people in every area, and they reluctantly changed their offer to regular summer visits, beginning in 1958.

  Coming to Terms with Being an IBMer

  In 1958, IBM was weighed down by an old and once carefully groomed reputation for extremely provincial and paternalistic human relations: company songs, compulsory white shirt and proper tie, and the like. Out of the blue, it set out to hire an entirely different technical workforce. As indicated by Piore’s words early in this chapter, the research division often felt like a hiring office for the development sites in upstate New York. In earlier years, most IBM hires had come from small colleges or trade schools. A new flood from competitive schools of engineering created extraordinary change on a daily basis.

  While IBM’s permanent home was being constructed, the staff was moved to several temporary locations. The largest was in the village of Yorktown Heights. I was assigned to the much smaller Lamb Estate, Tudor-style buildings scattered around an incredibly beautiful site overlooking the Hudson River.

  To someone who had lived through major events and read many history books, the atmosphere at IBM then recalled an aspect of France during the Revolution and empire. Since the ancien régime upper crust had mostly emigrated or holed up in provincial estates, very few promising individuals were available for promotion. Therefore, the selection rules had to change, and various old restrictions on inclusion were loosened.

  For example, I think of Lazare Carnot, a French leader whose importance is often underestimated. An engineering officer trained before the Revolution, he was underused and feared early retirement. When the Revolution came, he was put in charge of building up the army and had to choose between timeservers who had not emigrated and men who in peacetime would have been passed over for their nonconformism, nonaristocratic or ethnic background, or other conspicuous “flaws.” The Revolution succeeded because Carnot hired men such as the Corsican Napoléon Bonaparte.

  For different reasons, IBM found itself in a similar situation. Competitors like MIT, Bell Labs, and General Electric, flush with money triggered by the Soviet Sputnik, were free to hire or import anybody with impeccable credentials. What made IBM Research a unique experiment—historically very significant if not always flawlessly planned? For one thing, relaxed hiring rules brought in many individuals for whom other institutions did not compete: “oddballs,” “wild geese,” scientists whose high-class record was marred by some fault or another or by disputes with faculty advisers.

  I think of John Backus (1924–2007), who probably never had an adviser because he attended many schools—and none for very long. He contributed mightily to IBM’s hegemony at one time. Using a computer early on was extraordinarily difficult and time-consuming. Every problem had to be broken down by hand into a multitude of very precise instructions that had been wired into the dumb machine. With a small group, no fuss, and ahead of schedule, John developed a “high-level” programming language dubbed FORTRAN (from “formula translator”), which was never a work of art or admired, but had an unarguable virtue: it existed. Compared to the earlier assembly languages, it was nirvana. IBM was lucky that he did not work for a competitor.

  I think of John Cocke (1925–2002), who sounded and looked like a bad film’s take on a rich senator from North Carolina. He arguably stood next to Seymour Cray in the tiny cohort of people who understood everything about computers; in particular, he originated something quite important
called RISC computer architecture.

  And of Gerd Binnig (very much alive), whose school record had been spotty but who impressed Alex Müller of the IBM branch in Zurich as having a mind capable of “lifting the heavy Swiss dough.” He went on to invent the microscopes that can see atoms, to bring to IBM a flow of licensing fees, and to create nanoscience. He received for his efforts a Nobel in physics.

  He was followed the very next year by Müller, who was awarded the Nobel for discovering high-temperature superconductivity—triggering a “physics Woodstock.”

  Many of these oddballs eventually settled down or left, but a remarkable several dozen remained. Their flawed, inadequate, or unconventional early résumés were forgotten, and for their contributions they harvested academy memberships, five Nobels, and other honors beyond counting.

  Did this unplanned experiment prove anything? I can’t be accused of envying those who do well at exams (nor of biting the hand that fed me) by noting that the IBM experiment confirmed my longstanding lack of respect for exam rankings.

  Programming Before a Pervasive Concern with Security

  For years, the research division owned no computer. A few hours a day, it could borrow one in Poughkeepsie, New York. The programs were punched on computer cards and transported—a several-hour trip—in a station wagon shuttle. A painfully awkward process actually worked. Your deck went in the morning to Poughkeepsie, and in the evening it returned—mostly with the message that some dreadful programming error had to be fixed. You sent it off again the next day, and so on.

  One colleague spent an incredible amount of time on this process. To compute astronomical tables according to the Babylonians’ model of the heavens, Bryant Tuckerman was sweating blood. I dared to wonder what the rush was. This calculation had waited for several millennia, and waiting for the faster computer soon to be installed locally would barely make a difference. Bryant stuck to his guns and produced an enormous document. Finally, a number of secretaries retyped the computer output in a form suitable for printing, and a huge book was published by the American Philosophical Society. I fear that very few copies were either sold or used. But these comments are anything but critical. The tenacity of my colleagues who first tamed the beast was an object of pure wonder.