by Elwyn Berlekamp
Early chronology
Earned PhD and became assistant professor six years after graduating high school
I was born on September 6, 1940. I graduated high school in June 1958, and enrolled as a freshman at MIT that fall. I had a National Merit Scholarship, but no advanced standing. At MIT I majored in electrical engineering, in a cooperative program which included work in industry, usually leading to both BS and MS at the end of five years. My cooperative industry was Bell Telephone Labs (BTL). At MIT, I took extra courses, more advanced courses, and several advanced standing exams. In December 1961 I was among five winners of the national William Lowell Putnam Mathematics Competition. I received both my BS and MS degrees in EE in September 1962. My master’s thesis, in artificial intelligence, was supervised at MIT by Prof. John McCarthy (1927–2011) (who soon thereafter left MIT for Stanford) and at Bell Labs by John L. Kelly, Jr., author of a seminal 1955 paper1 on asset allocation in portfolio theory.
I spent the summer of 1963 working at Bell Labs under David Slepian a renowned information and coding theorist. In the spring of 1964 Lotfi Zadeh, then chairman of Electrical Engineering at UC Berkeley (who later became very famous as the founder of fuzzy set theory) persuaded me to come to California during spring vacation to interview for an assistant professorship. They offered me the job as soon as I finished my PhD. I returned to MIT to work on my thesis in earnest. My committee was Bob Gallager, Claude Shannon, Peter Elias and Jack Wozencraft. They all paid attention, but they pulled in somewhat different directions. My PhD thesis, “Block Coding with Noiseless Feedback”, spawned several publications. The key to one of them was to recast the problem as an asymmetric combinatorial game between the Coder and the Noisemaker, and then to find asymptotically optimum strategies for playing that game.
As a student at MIT, I had also been very interested in courses in the design of digital circuits. Upon arriving at UCB in the fall of 1964, I agreed to teach a related undergraduate course on “Operating Systems”, a subject which was then considered avant garde. The class included a precocious student named Ken Thompson. He was then a senior looking for a permanent job. He asked me for a job reference, but I persuaded him to stay on another year at Berkeley. He received an MS under my supervision, after which I got him a job in the computer science research department at Bell Labs, where he had enormous success. He coinvented the Unix Operating System with Dennis Ritchie for which they received the 1983 Turing Award.
In fall 1964, I joined the Bay Area’s chapter of the Institute of Electrical and Electonic Engineer’s (IEEE) Information Theory Society. Other active members included Tom Kailath, Tom Cover, Norm Abramson, and Jim Omura from Stanford; Nelson Blachman and Phil Fire from GTE Sylvania; and Jim Spilker, an entrepreneur who started a company called Stanford Telecommunications.
In January 1965 Sol Golomb recruited me to become a consultant to the Space Communications program at the Jet Propulsion Lab (JPL) in Pasadena, CA. For the next year and a half, I flew down there one day every week, and became part of a very stimulating group including Andy Viterbi and Len Kleinrock (then at UCLA), Irwin Jacobs (then at UCSD), Lloyd Welch (at USC), Gus Solomon (at JPL), Marshall Hall, Richard Stanley, and Bob McEliece (at Caltech), and others.
I left Berkeley in the summer of 1966 and returned to Bell Labs in New Jersey. There I joined a group including Henry Pollak, Ron Graham, and Neil Sloane. In 1967–1968 I was recruited to also become a consultant to the Institute for Defense Analysis’ (IDA) cryptography research center in Princeton. I returned to UC Berkeley in 1971, while retaining many of my informal connections with people at BTL, JPL, IDA, and MIT. By 2010, I had publications with over 40 coauthors. Nearly all of them were from those institutions.
One of my publications2 in a math journal attracted the interest of Paul Erdős, the legendary itinerant Hungarian mathematician. It constructed partitions which addressed the finite Van der Waerden’s problem. Although my paper gives the best construction known for some appropriate values of the parameters, it is far from the best bound on the other side. I encountered Erdos once or twice every year from 1968 to 1995, and on every occasion he expressed disappointment that neither I nor anyone else had significantly narrowed this gap in Van der Waerden’s problem.
Coding (and synchronization)
Invented a series of algorithms and implementations which made powerful error-correcting codes useful in many applications
Prof. Avi Widgerson, head of computer science at the Institute for Advanced Study in Princeton, has declared these among the four most important algorithms of the 20th century in all of computer science.
Claude Shannon’s classic 1948 paper3 created the subject of Information and Coding Theory. Shannon’s approach was statistical: he proved by his now classical random coding arguments that very good error-correcting codes exist, but, except under very restrictive circumstances such as those addressed by Hamming, there were no constructions or practical decoding algorithms. Bounds on the asymptotic performance of the best error-correcting codes were then quite weak. Shannon himself led studies to improve them.
My contributions to information theory began in 1963. I found constructions for asymptotically optimum codes for three special cases:
- the binary symmetric channel with noiseless feedback;
- binary erasure-burst channels, and
- arbitrary discrete memoryless channels in the limiting case of zero rate.
My proof of the optimality of the zero-rate case resolved one of Shannon’s favorite open problems, and led to my inclusion as a coauthor, with Shannon and Gallager, of parts one and two of our classic paper on performance bounds of long block codes. An extension of some of that work, which I coauthored with Irwin Jacobs proved the limitations of a then popular coding technique using convolutional codes with sequential decoding.
In 1960, Reed and Solomon had introduced a class of codes, which became known as RS codes. These codes are optimum when the alphabet size matches the block size. As this implies large alphabets for long block sizes, these codes were initially less popular than BCH codes, a mathematically similar class of codes introduced at about the same time by Bose, Chaudhuri and Hocquenghem. BCH codes are typically binary. RS codes can also be effective on binary channels when concatenated with other codes as proposed by Forney in 1965, and/or when bursts are common. Both BCH codes and RS codes are called “Algebraic codes” because the locations of errors can be expressed as roots of a polynomial over a large finite field.
I became interested in Algebraic Coding in the mid-1960s. My work in that field provided a long series of major improvements to that subject. Best known among these numerous results are my algorithm to factor polynomials over finite fields,4 and the Berlekamp–Massey algorithm to find linear recursions in long data streams. I published both of those algorithms in my 1968 book, Algebraic Coding Theory (New York: McGraw-Hill, 1968, MR0238597). Another major new result in that book was the enumeration of the number of information bits in long binary BCH codes. This book won the IEEE Information Theory’s annual best research paper award. The applicability of some of these algorithms to problems in cryptography attracted the attention of the National Security Agency (NSA), who in 1967–1968 recruited me to become a consultant to their research group at the Institute for Defense Analysis in Princeton, while I retained my primary job at Bell Labs. My work there led to my recognition by (now IEEE-) Eta Kappa Nu as their “Outstanding Young Electrical Engineer” in 1971. I was nominated for that award by John R. Pierce (1910–2002), then a Bell Labs executive best known for his earlier work inventing microwave communications.
My algorithm to factor polynomials over finite fields was soon extended to become the first step of several algorithms to factor polynomials over the rational integers. Joel Moses at MIT included these algorithms in his early symbolic manipulation system called MACSYMA, a predecessor of today’s Mathematica and Maple.
After I left Bell Labs and returned to the Berkeley faculty, I expanded my consulting activities, providing more improvements in algorithms for error-correcting codes. Some of them were then poorly implemented, leading to rumors that my algorithms were impractical. To counter this rumor, I founded a small company called Cyclotomics, hired three people part-time, and built a small special purpose computer christened the “GF1”, (the world’s foremost Galois Field computer). It met the specifications of a frequency hopping antijam military communications system called the Joint Tactical Information Distribution System (JTIDS). An RS code over the alphabet of 32 symbols lay at the heart of this system. The Defense Department’s JTIDS program funded four big prime contractors. Cyclotomics became a subcontractor to one of them, ITT. Cyclotomics’ RS decoders beat their competitors by wide margins in all dimensions: Less than half the number of components, rack space, weight, and power consumption, and several times their speed. All this even though some of the competitors used the same “Berlekamp decoding algorithms” published in my 1968 book, but lacking the further improvements and embellishments. Cyclotomics’ GF1 computer shattered the then prevalent myth that powerful highspeed algebraic error-correction was not feasible.
The GF1 architecture featured a special Galois Field arithmetic unit, as well as a different Galois Field based indexing system and more traditional units managing input/output and data’s random access memory (RAM), and the Program’s Read-Only memory (ROM). All of these units operated concurrently. GF1 programming was done in a novel language which simplified both simulation and testing, as well as direct compilation for the ROM. It also facilitated hardware debugging and maintenance via a then state-of-the-art methodology called signature analysis. The GF1 was documented in Cyclotomics’ first patent, (now summarized on my website via a link through the line of microcode at the bottom of my home page). It also became the basis of my election to the National Academy of Engineering (NAE) in 1977. I was then their youngest member.
The many embellishments and improvements introduced by Cyclotomics included the concepts of readable erasures and helical interleaving with burst forecasting. We also developed other models of GF1 computers for other applications. We designed and built the error-correction decoders for the downlink of the Hubble Space Telescope. By the 1980s, NASA Goddard and JPL had each developed its own competing “NASA standard” for space communications. I worked with both of them, and tried to bring them together. But my efforts failed, and NASA adopted two competing standards, both written by me. Another important innovation that became a NASA standard was my “bit-serial Reed–Solomon encoder”.5 To the great surprise of nearly everyone, it showed that Reed–Solomon codes, based on large finite fields, can be encoded with hardware even simpler than binary codes of the same redundancy.
A more refined decoding algorithm which I coinvented with Lloyd Welch provided several potential advantages over Berlekamp–Massey. Under some circumstances, the benefits of these new algorithms exceeded another factor of 10. A Berkeley graduate student named Madhu Sudan invented further extensions of the Berlekamp–Welch algorithm. Sudan’s decoding provided more powerful error-correction for codes of very low rates. He applied this to probabilistic proofs, an important topic in theoretical computer science. Sudan became a professor of computer science at MIT and the winner of the International Math Society’s 2002 Nevanlinna Prize for outstanding contributions in Mathematical Aspects of Information Sciences. It is considered the Applied Mathematics Equivalent of a Fields Medal in Pure Mathematics.
Cyclotomics never had any angel investors or venture capital. It bootstrapped its growth by profitable sales, and eventually grew to a peak of 40 people. We eventually received 12 patents of which I was inventor or coinventor. Some of these addressed issues of synchronization and soft decisions as well as traditional detection and correction. Beyond communications, we also pioneered applications of algebraic error-correction technology to several data storage technologies, including magnetic tapes, magnetic discs, and optical discs. Relevant types of magnetic tapes ranged from a few early import/export records made by the U.S. Census Bureau in the early 1950s to digitized video recorded and played at 680 Megabits per second in the mid-1980s. Attempts to license the most advanced coding techniques to manufacturers of read-only compact disks were unsuccessful; they chose to rely instead on less powerful but royalty free algorithms in my 1968 book. By the mid-1980s there were over 40 US companies trying to develop read-write optical memories. Cyclotomics became involved in developing controllers for several of them. The biggest of these was Eastman Kodak. Kodak acquired Cyclotomics in 1985.
Soon after the Kodak acquisition Cyclotomics launched the first project to record enhanced quality digitized sound on movie film rather than on a separate magnetic strip, as was then the current standard. This project, which became known as Cinema Digital Sound (CDS), (a.k.a. “Digital Optical Sound System”) entailed challenging synchronization issues due to mechanical jitter and weave. It also provided a very challenging noise environment. Lloyd Welch and I invented and patented a solution to these problems. At Cyclotomics we designed and built a few prototypes. Our business partners In the movie-projector division at Optical Radiation Corp (ORC), led by engineers Howard Flemming and Ron Uhlig, combined them with new speakers, replicated and deployed them to 60 movie theaters. CDS received a 1995 Scientific and Engineering Academy Award (an Oscar) from the Academy of Motion Picture Arts and Sciences. But by then I had left Kodak/ Cyclotomics; ORC had suffered a big scandal in another division; and the business partnership between Kodak and ORC had fallen apart.
Jack McCauley’s first job was a Cyclotomics engineer working on CDS. He continued his career at Guitar Hero, and later cofounded Oculus Rift. By 2015, he was considered by many to be the world’s foremost engineering expert on virtual reality.
In the late 1980s my colleagues at Cyclotomics and I designed and built a prototype “hypersystolic” decoder, implementing a radically new architecture which avoided problems of clock skew at very high speeds then attainable with Gallium Arsenide components.
In 2017, Kevin Neilson at Xelic, is now implementing some of my algebraic decoding algorithms.6 Using a fully pipelined architecture, he is implementing the Berlekamp–Rumsey–Solomon algorithm in an iterative decoder using 6-error-correcting BCH component codes. This is used on fiber optic channels with rates exceeding 100 Gbps.
In the early 1980s, Cyclotomics had a few commercial cryptography contracts, including one from General Motors seeking to improve their electronic car keys. In 1984 Cyclotomics spun off its consulting contracts in cryptography to a new startup which became known as Cylink. Its cofounders were Jim Omura (90%) and I (10%). Cylink obtained venture funding from a group formed by Jim Simons, whom I had met at an IDA interview in 1967. In 1996 Cylink went public on NASDAQ. Adjusted for inflation, it became a unicorn with over 400 employees. Its products encrypted all of Swift’s foreign exchange transactions, totaling several trillion dollars per day. Cylink endured a boom, a scandal, a bust, and a partial recovery, all a couple years ahead of many other Silicon Valley companies of that “Dotcom” era. Cylink was acquired by Safenet in 2002.
Cyclotomics had a small division led by Steve Pope which became an early designer of custom ICs, fabricated by outside foundries. We received technical and business advice from Cyclotomics’ stockholders Dave Hodges and John Torode. In the late 1980s John Torode and his wife formed a Seattle-based company named IC Designs, in which I became an active investor and the third board member. We developed some novel clock-generating circuits which included both analog and digital on the same chip, and the business boomed. In 1993 we were acquired by Cypress Semiconductors and became “Cypress Timing Technology.”
I received the IEEE’s Koji Kobayashi Award in 1990 and the IEEE Hamming Medal in 1991. In 1993 I received the IEEE Information Theory Society’s Shannon Award. What turned out to be the last of my major publications on new algebraic decoding algorithms was published in the IEEE-IT Transactions in 1996. (In some special circumstances, it achieved yet another order of magnitude performance improvement.) I received an IEEE Golden Jubilee Award in 1998. I was elected to the National Academy of Sciences in 1999. The third edition of my book, Algebraic Coding Theory was published by World Scientific Publishing in 2014.
Quantitative finance
Pioneered algorithmic trading
In 1986 I became a consultant to Axcom Trading Advisors, a small company located in Newport Beach, CA. Its founder and CEO was Jim Ax. They relied on computer algorithms to manage a pool of investments in futures contracts called the “Medallion Fund”. Typical holding periods for their assets were one or two weeks. The pool manager was another then small company called Renaissance Technologies, owned and managed by James H. Simons. Although I visited Axcom only one or two days per month, I became interested in Axcom’s algorithms for predicting short-term price movements based entirely on analysis of their recent price movements. Ax formulated several related mathematical problems, some of which I solved. Following a big gain in the spring of 1988, Ax bought a big house in Malibu, more than an hour’s drive away from Newport Beach. Ax tried to manage the company by video conferencing. This proved difficult, even though the company always had less than a dozen people. Markets turned unfriendly and the Medallion Fund began to drift downwards. In fall 1988 Ax resisted my efforts to implement modest improvements in the trading algorithms, favoring instead more work on grander schemes in hopes of getting even bigger improvements. After being out of touch for several months in the spring of 1989 during my friendly exit from Kodak/Cyclotomics, I showed up at Axcom in early summer 1989. I learned that the fund had continued its downward drift and that Ax, the manager of the trading, was in a big dispute with Simons, the manager of all of their investors. I resolved this dispute by buying Ax out, restructuring the contracts with Simons to give Simons a 25% interest in Axcom in return for his not taking 25% of the fees off the top. With the approval of the other Axcom partner, Sandor Straus, the company and its remaining three employees moved to Berkeley, implemented the proposed embellishments in trading algorithms, and resumed trading. The markets were friendly. In the fall of 1989, Axcom recovered all that it had lost during the first half of that year. More improvements and a few more employees were added. In calendar 1990, Medallion investors enjoyed a gain of over 55%, net after big fees which then consisted of 5% of assets plus 20% of profits.
In January 1991 I sold all of my ownership interests in Axcom to Jim Simons and returned to academic pursuits.
Details of algorithms in finance, like those in government cryptography, are best kept secret, although some vague high-level guidelines have emerged in such articles as the cover story of Bloomberg Magazine’s December 2007 issue. I believe the first key to success is to hire people trained as engineers, mathematicians, statisticians, or physical scientists, NOT traditional money managers who were trained as MBAs.
Although 1990’s net gain of 55% was then Medallion’s best yet, they continued to do extraordinarily well. They continued hiring mathematicians, engineers, and scientists. They expanded into trading stocks as well as futures. Every calendar year has been profitable (although from mid-1988 to mid-1989 was down over 20%). In some years net gains were over 100%. Incentive fees were increased to over 40% of gross profits. Despite this, over several decades, this fund’s annualized net returns to investors have exceeded 30%. It has made Jim Simons the most successful fund manager on Wall Street, by far.
Under Ax the holding time of a position was typically a week or two. Under me, it became shorter. Because of this rapid turnover, Medallion’s trading volume was much higher than the size of the fund. So as the fund grew larger, it became so big that its own trading moved the markets. To mitigate this problem, Simons began redeeming all investors other than Renaissance employees. The rest of us were all squeezed out in the early 00s.
In the late 00s, four new partners and I launched another quantitatively managed fund, called “Berkeley Quantitative” (BQ). After a couple years of development and studies, we accepted money from outside investors and began trading. As more investors entered, the fund size grew. Net performance reached 17%, and the fund size reached \$250 million about 1.5 years after trading had begun. But the next couple months saw unfriendly market conditions aggravated by BQ’s internal organizational problems. Investors suffered a 15% drawdown. Some big investors withdrew their money. Berkeley Quantitative redeemed its remaining investors, paid all of its debts, and closed down. The initial investors realized a net return of about 2% after about three years. Definitely not a success, but much better than the wipeouts experienced by most startups.
Combinatorial game theory
Cofounded combinatorial game theory, an important new branch of mathematics
My fascination with mathematical games began when I learned to play Dots-and-Boxes in the first grade. Since then, I have discovered several mathematical theorems which underlie this game and others. In 1967 I met Richard K. Guy who shared these interests. I proposed jointly writing a book on that subject. Guy recruited John H. Conway to join the effort, and our collaboration began when we convened at a conference at Oxford in 1969. Conway then lived in England; Guy in Canada, and I in California. Conway spent most of 1971–72 visiting Caltech, and I met him there one evening every week after spending the day at JPL. New ideas and new results appeared at a rapid pace which continued throughout the 1970s. During this decade all three of us converged every December and June at Richard Guy’s home in Calgary. Our title was chosen to be Winning Ways for your Mathematical Plays. Among many new discoveries, Conway invented surreal numbers, a classification of infinities more relevant to ivory-tower mathematical logic than to any playable board game. I viewed it as a digression from the purpose of the book, so Conway spun off and published it as a separate book called On Numbers and Games. Joint work on Winning Ways resumed not long after that disruption, and the scope of the book expanded again and again. Its first edition, published in 1982, comprised two volumes with two parts each. Its second edition expanded each of these parts into its own separate volume, published separately in the easily recalled years of 2001, 2002, 2003, and 2004.
According to Martin Gardner, author of the extremely popular “Mathematical Games” column in the Scientific American magazine from 1957 to 1982, Winning Ways is the “greatest contribution of the 20th century to the burgeoning field of recreational mathematics. No other work has been so packed with completely new and significant material, or presented with so much wit, depth, and clarity.” Winning Ways also became the foundation of a less recreational subject called “Combinatorial Game Theory”, CGT. It was officially accepted by Math Reviews (now MathSciNet) as a new branch of mathematics. It has attracted the interest and contributions of many mathematicians and computer scientists.
In 1987 the American Mathematical Society sponsored a symposium commemorating the multifaceted works of John von Neumann the legendary mathematician, economic game theorist, computer architect, quantum mechanics theorist, etc. I gave the invited lecture on game theory. Someone noted that most of the over 200 games in Winning Ways were invented by the authors, and asked me whether this theory applied to any “real” games. My answer was that it certainly applied to the popular kids’ game of “Dots-and-Boxes”, and possibly also to the classic East Asian board game known as “Go”. But the question and my answer left me uncomfortable, and provoked my renewed study of Go. Over the next year, I discovered that late-stage Go endgames could be simplified by “chilling” (a special case of the “cooling” operator in Winning Ways), and that although cooling was generally not invertible, in several special cases, including Go, chilling could be inverted by a new operator I called “warming”. David Wolfe (1965(?)-), then my graduate student, joined this research effort. We used these techniques to find provable strategies for playing many late-stage Go endgame problems. We published the book, Mathematical Go. It received rave comments from a wide spectrum of reviewers. It included several problems we composed, some of whose solutions entailed sequences lasting over 100 moves. The problems were difficult enough to stump virtually all professional Go players, among whom Wolfe and I acquired a reputation as composers of very hard endgame problems. But, as it turns out, these problems all quickly yield to the application of the relevant CGT.
In efforts to quantize the importance of various choices of moves in earlier-stage Go endgames, I devised a variation of the game, which I called “Coupon Go”. Multiday Coupon Go tournaments among some of the world’s best professional Go players have been held in Seoul and Beijing. Some have adapted this as an important teaching technique. These include Naiwei Rui, many time world’s women’s champion, and her husband Jujo Jiang, by far the best and most famous Go player in North America in the 1990s.
By the turn of the century, CGT had been productively applied to an increasingly wide collection of popular board games. These include the ancient game of Hawaiian checkers, known as Konane. They also include a modern game called Amazons, which was invented by Walter Zamkauskas in Argentina in 1988. It has become popular on the web.
In 2014, the American Mathematical Society published a brilliant treatise entitled Combinatorial Game Theory. Its author is Aaron N. Siegel, a former PhD student of mine. It contains nearly all that was known, and much that wasn’t known before then.
STEM education (and its popularization)
Contributed to the popularization of recreational mathematics and its use as a vehicle to increase interest in STEM education
I have also been active in the popularization of Science, Technology, Engineering and Mathematics, directed both at K–12 education and at adults.
Since 1999, my efforts to popularize these subjects include my coauthorship with Joe Buhler of the “Puzzles” column” in the semiannual newsletter, Emissary, published by the Mathematical Sciences Research Institute (MSRI) . One problem I published in 2012 sheds light on the causes of political gridlock.
I have served on the governing boards of each of the two foremost private schools in the East Bay: College Prep and Head-Royce. But my bigger efforts have been directed towards extracurricular education. In the late 1970s, I recruited a team of graduate students who beefed up the intellectual content of the very popular computer games at the Lawrence Hall of Science. As chairman of the governing board of MSRI in the late 1990s, I helped finance Zvezdelina Stankova’s initiative to start the Berkeley Math Circle, where I have also been an annual guest teacher. The circle has a few dozen students, now typically junior high school age, who meet in a classroom one evening per week to share and enjoy solving math or logic problems not covered in their school curricula. Sam Vandervelde WROTE a book, Circle in a Box, explaining how to organize and run such a circle. There are now over a hundred of these circles active in the United States.
One key to the success of such efforts is to eradicate (or at least blur) the line many people imagine between “real” mathematics/engineering and “recreational mathematics”. There are many games which straddle this imagined divide.
Mathematics lies all around us. It lies unnoticed just below the surface of many things, such as the game of Dots-and-Boxes.
Since 2015, I have devoted much of my time to the preparation of short introductory videos aimed at getting more junior high school students interested in combinatorial games and some of the mathematics behind them. Links to these videos can be found on my website and/or my channel on YouTube.
In 2017, I became the initial donor to fund “America Asks, Science Answers”, a new public information campaign by the National Academies.
Leadership
In the early 1970s, I was chairman of a group called FOCUS, which was the science and academic advisory panel for the Institute of Defense Analysis’ cryptography research group in Princeton. Prof. Creighton Buck from the University of Wisconsin was another active member of that group. Following this, in 1974, J. C. R. Licklider appointed me to chair a high-level year-long study of future mass memory technologies, funded by the Defense Advanced Research Projects Agency (DARPA).
In 1973 I was President of the IEEE Information Theory Society. We became active participants in the Nixon–Kissinger era of detente. We invited R. L. Dobrushin, a well known Russian information theorist, to write a survey of information theory in the Soviet Union. When his lengthy paper arrived, our attempts to find an experienced translator failed, so I translated it myself. It was published in IEEE-IT.7 We gave our annual best paper award to a young Russian named Valery Goppa AND I presented it to him in Moscow. His later work founded the specialty of algebraic geometry codes. We IDA also inaugurated our now annual Shannon Award, and I presented the first such award to Shannon himself at our International Symposium in Ashkelon, Israel.
In 1974, the IEEE Press published my book, Key Papers in the Development of Coding Theory (New York: IEEE Press, 1974, MR0384299).
In 1974, a search committee led by Creighton Buck offered me the directorship of the University of Wisconsin’s Army Mathematics Research Center, a strong research group that had become famous for being bombed by four antiwar activists in 1970. It was a tempting offer, but I decided to remain at Berkeley.
In 1973, I became one of a dozen faculty cofounders of the Computer Science Division within UC Berkeley’s Department of Electrical Engineering and Computer Science. I was then also a member of the MIT Corporation’s Visiting Committee for their Electrical Engineering Department, where I played a minor role in getting MIT’s EE Department to follow Berkeley by changing their name to EECS in 1974.
In 1975–1977, I was UCB’s Associate Chairman of EECS for their CS Division. It was a stellar group, whose faculty and alumni have subsequently received nine Turing Awards. During this period we recruited some stellar visiting professors, including Ken Thompson of Bell Labs (1983 Turing awardee), Jan Rajchman (a prolific inventor of over 100 patents from RCA, who had served on Berlekamp’s DARPA panel), Alan Kotok (My MIT classmate who had become the lead designer of DEC’s PDP8, PDP10, and KL10 computers), and Gus Solomon (of JPL, the coinventor of Reed–Solomon codes). I personally recruited several new faculty, including David Patterson, THEN fresh out of graduate school at UCLA. During this period, the CS Division also promoted to tenure several notables including Mike Stonebraker (2014 Turing awardee) and Susan Graham (who CS became chairman of Harvard’s Overseers shortly before the end of Larry Summers’ presidency there). Among many CS graduate students at UCB in the mid-1970s who went on to notable accomplishment and fame were Mike Sipser (now Dean of Science at MIT), Bill Joy (a software superguru and cofounder of Sun Microsystems), and Eric Schmidt (later CEO of Google).
In 1979, I joined MSRI founders Chern, Singer and Moore as well as UC math department chairman Alberto Grunbaum, in a meeting which persuaded UC Chancellor Bowker to “support” (verbally but not financially) MSRI as an allied but independent off-campus entity, with its own board of governors, neither controlled nor overseen by UCB.
Like others with Top Secret clearances in cryptography, I have refrained from publishing any papers directly relating to cryptography, whether they are classified or not.
In 1976 the IEEE-IT Transactions published a paper by Diffie and Hellman which introduced a system for public exchange of cryptography keys.8 Other academic papers on cryptography soon appeared in other journals. This aroused concerns at NSA that this information might be useful to U.S. adversaries. To address these concerns, Admiral Bobby Inman, then Director of NSA, and his chief lawyer, Dan Silver, invited me to convene a meeting of cognizant academics. I convened that meeting at the UC Berkeley Faculty Club in November 1979. Key attendees included Bobby Inman himself and Berkeley law professor Michael Heyman. The meeting led to a commission appointed by the American Council on Education, whose members were representatives of the relevant scholarly societies. Mike Heyman was chairman. I represented the IEEE. During the course of that commission’s work, Heyman was promoted to become Chancellor of UCB.
In 1978, McEliece, van Tilborg and I coauthored a paper on the inherent intractability of certain coding problems.9 The main result was that the problem of decoding certain variants of Goppa codes is provably hard, in a technical sense known in computer science as “NP Complete”. Although apparently unrelated to cryptography, a subsequent paper by McEliece used this result to prove that the cracking of a new cryptographic system he designed is also NP-hard. Theoretically, that’s a stronger standard of security than any commercial system in wide usage, either then or now.
In the mid-1980s, I served on the UC President’s “Science and Academic Advisory Group” which oversaw the Department of Energy’s classified nuclear energy Labs at Livermore and Los Alamos.
I have been an active member of the Finance Committee of the National Academy of Sciences (NAS) since 2000. From 2007–2013, I also served on the separate Finance Committee of the National Academy of Engineering (NAE) with then NAE treasurer Dan Mote. I was the only person to have served on both.
Throughout the 1980s, 97% of MSRI’s support came from NSF and other federal agencies. Near the end of that decade, MSRI Director Irving Kaplansky appointed a three-person group to raise additional funding from private sources. Its members were William Randolph Hearst, III, Jim Simons, and I. After several unsuccessful forays over about a year, Hearst and Simons quit, but I persisted. I expanded the group to become MSRI’s “Friends”. There were many more meetings and forays, but negligible success. This situation continued through Kaplansky’s retirement and replacement by Bill Thurston, a legendary topologist and Fields Medalist. He shared my belief that MSRI needed attention-getting physical objects rather than only pictures of mathematicians writing formulas on blackboards. A sculpture called the “7-fold way” was financed by a \$25,000 gift from Mitsubishi Electric Research Laboratories, for whom I had consulted the prior year. It was MSRI’s largest donation up until then.
MSRI’s board of governors initially consisted of the chairmen of the math departments of MSRI’s sponsoring institutions. These were about a dozen west coast research universities including U. Washington, U. Oregon, UC Berkeley, Stanford, UCSC, UCLA, Caltech, USC, UC Irvine, UCSD. Thurston nominated me to become chairman of this board in 1994, in hopes of broadening it to include some prospective donors. But donations remained near zero. When Thurston’s term as director expired in 1995, I chaired the search committee for his successor. I kept some key prospective donors aware of its progress. David Eisenbud was hired as a new director in 1997. Simons and Hearst joined the board soon thereafter, and so did Sandor Straus, Roger Strauch, Ed Baker and others. My term as chairman ended in 1998, but Eisenbud then raised over \$10 million in less than a decade. This mostly financed a new auditorium. Several named postdocs were established, and a growing endowment campaign began. I remain an active MSRI trustee, involved in several of MSRI’s outreach programs, including National Math Circles, the Bay Area Mathematics Olympiad for high school students, and the National Math Festivals in Washington, D.C. The first one, in spring 2015, included a program which attracted 10 sitting US Senators and a few staffers and members of the House of Representatives. The second festival in spring 2017, like the first one in 2015, attracted over 20,000 people to public events in Washington, D.C.
In 2000, Dave Hodges persuaded me to join the board of the International Computer Science Institute (ICSI), another off-campus institute in Berkeley. Its budget was about double MSRI’s, but mathematics research is much cheaper than CS, and ICSI is not as well known. In most of the 1990s, ICSI’s major funding came from a consortium of European governments led by the Germans, who used it as a postdoctoral training institute. But a change of German policy led to financial problems at ICSI, which were mitigated primarily by ICSI Director Nelson Morgan’s successful applications for increased contract research support from American companies such as at AT&T Labs. Hodges nominated me to be his successor as chairman of ICSI’s board. As chairman, I traveled to Helsinki and Madrid in support of efforts to persuade other European countries to take advantage of the deficit left by the German cutbacks. I also recruited a new board member, Cliff Higgerson. A Berkeley alum living in Palo Alto, he was a very successful venture capitalist whose seed-stage “Vanguard IV” fund included Cienna, Advanced Fiber Optics, and Network Appliance. These three big hits of the dot.com era made this fund among the most successful venture funds ever. (In a surprisingly short time frame, it returned to investors over 15 times their paid-in capital). Higgerson became my successor as chairman of ICSI.
Back in 1993, Thomas M. Rodgers, an Atlanta-based businessman and mechanical puzzle enthusiast, recruited me and a famous magician named Mark Setteducati to help him organize a “Gathering” to honor Martin Gardner, the prolific author who had succeeded in popularizing recreational mathematics, tabletop physics, mechanical puzzles, magic tricks, and other such topics through his 299 monthly Scientific American columns and many books. Unlike a conventional conference or symposium, where most attendees come to hear lectures from a small set of speakers, Rodgers wanted much more audience involvement in this “Gathering for Gardner” (G4G). Most attendees were required to make a short presentation and bring a few hundred copies of an original puzzle or mathematical art object, to be distributed among all other attendees. The event was a big success. It was repeated as G4G2 in 1996, G4G3 in 1998, …, continuing in every other subsequent even-numbered year. Its attendance includes some of the world’s best mathematicians and magicians. It was originally funded only by registration fees plus Rodgers’ personal contributions. Eventually it grew into a nonprofit corporation, although still run largely out of Rodgers’ pocket. Following Martin Gardner’s death in 2010, G4G initiated a plethora of widely distributed local events known as “Celebration of Mind” (CoM). They were originally intended to be held simultaneously on Gardner’s birthday, October 21. But that requirement was later loosened to “late October”. When Rodgers died in 2012, I became chairman of the G4G board. I guided the organization through its transition to greater stability. In 2014, there were over 150 Celebration of Mind events, on all seven continents (including Antarctica!). Although I retired from this position in 2016, G4G continues to thrive. Since 2016–2017, its chairman has been Nancy Blachman and its presidents have been Mark Mitton (the best magician on the east coast of the US), and Erik Demaine (a superstar computer science professor at MIT, widely known for his solution of “Origami” and numerous clever algorithms). The most recent Gathering, G4G13 on April 11–15, 2018, continued this grand tradition. Much more information about G4G and CoM, including many pictures and videos, can be found on the web at celebrationofmind.org.
Service
Teaching
Supervised and mentored students who have had a significant impact on electrical engineering and computer science
A list of my thesis students appears on my website.10 Many have gone on to happy, productive, and distinguished careers. Here are a few which might be considered especially noteworthy:
Ken Thompson (MS 1966) coinvented the Unix operating system. He also wrote “Belle”, the world’s computer chess champion in the mid-1980s, and mentored Eric Schmidt.
Oscar Moreno (PhD 1973) spent most of his career as a professor in Puerto Rico. He also founded a company which managed the Puerto Rican internet domain.
Robert Li’s PhD thesis (1974) pioneered the study of loopy combinatorial games. He later wrote the book on Algebraic Switching Theory,11 and became professor and chairman of the Department of Information Engineering at the Chinese University of Hong Kong. Late in his career, he coauthored his first paper in information and coding theory, called “Network Coding”.12 It won the IEEE-IT 2005 annual “best research paper” award.
Larry Carter (PhD 1974) became professor and chairman of the Department of Computer Science and Engineering at UC San Diego.
Po Tong (PhD 1982) coinvented hypersystolic architecture at Cyclotomics, and later became lead designer of Reed–Solomon decoding chips at LSI logic and Texas Instruments.
Jorge-Nuno Silva (MS 1991) became professor at University of Lisbon, where with his colleague Carlos Santos (a grandstudent of Richard Guy), he showed that the Hawaiian game Konane includes all combinatorial games. They have also had an enormous impact on education: in all public schools in Portugal, introductory combinatorial games are now a standard part of the 5th grade curriculum.
David Wolfe (PhD 1992) copioneered the application of Combinatorial Game Theory to Go.
David Moews (PhD 1993) is a leading cryptanalyst at IDA.
Dan Garcia (MS 1995) has been teaching the largest and most popular computer science classes at UC Berkeley.
Martin Mueller (PhD 1995 and ICSI postdoc 1996–1997) is professor of computer science at U. Alberta. His students have included David Silver (PhD 2009) and Aja Huang (Postdoc 2011–2), leaders of the major Artificial Intelligence breakthrough, AlphaGo.
Amit Sahai (Honors BS 1996) is Director of the UCLA Center for Encrypted Functionalities.
Julia Kempe (PhD 2001) earned concurrent PhDs from Berkeley and U. Paris! She published over 100 papers on Quantum Computation and Information. She is now the Director of Center for Data Science at NYU.
David DesJardins (PhD 2001) became a leading cryptanalyst at IDA while on leave from graduate school, then later an early employee of Google. He now leads his large philanthropic family foundation.
Aaron Siegel (PhD 2005 and MSRI postdoc 2005–2006) wrote the prevailing treatise on Combinatorial Game Theory. He now leads a group at AirBnB.