FOCUS May/June 2007

An Interview with Doron Zeilberger

By Joe Gallian and Michael Pearson

T o celebrate the opening of the Car-

riage House Conference Center at MAA
headquarters, made possible by a gift
from Paul and Virgina Halmos, the MAA
received a grant from the National Secu-
rity Agency to support a Distinguished
Lecture series intended to appeal to a
general audience.

The second lecture in the series was given

by Doron Zeilberger on February 20,
2007. Zeilberger is the Board of Gover-
nors Professor of Mathematics at Rutgers
University. He is widely known for the
development of “WZ” (Wilf-Zeilberger)
Theory and Zeilberger’s algorithm which
are used extensively in modern computer
algebra software. Zeilberger was the first Joe Gallian and Doron Zeilberger
to prove the elusive result in combinato-
rial theory known as the alternating sign
special functions are just sources for ex- and less-great, mathematicians through
matrix conjecture. Among his honors
amples and case studies of a methodology the centuries, using pencil-and-paper.
are: the American Mathematical Society
with the aim of training the computer to Of course, with computers you can do
Steele Prize for seminal contributions to
discover conjectures and then try to prove so much more, and you can be very sys-
research (co-recipient with Herb Wilf);
them all by itself, without any human tematic, and the great power of today’s
the Institute of Combinatorics and Its
intervention. computers, guided wisely, can take you a
Applications Euler Medal for “Outstand-
very long way. However, their emphasis
ing Contributions to Combinatorics”; the
JG: Are there now a number of people is still on using computers to find interest-
Laura H. Carnell Professorship at Temple
who are doing experimental mathemat- ing conjectures and phenomena, but not
University; and the MAA Lester R. Ford
ics? to prove them. The proof itself (when
award for a paper in The American Math-
feasible) is still done largely by human
ematical Monthly.
DZ: Experimental mathematics is a beings, although sometimes with the as-
rapidly growing field, both explicitly sistance of computers.
The citation for the Euler Medal de-
and implicitly. Explicitly, there is a very
scribes Zeilberger as “a champion of
good journal by that name, an Institute As for computer-generated proofs, there
using computers and algorithms to do
in Simon Fraser University, and at the is a whole community of Automatic
mathematics quickly and efficiently.” In
recent annual meeting at New Orleans Theorem Proving, which is very success-
his opinion “programming is even more
there was a special session dedicated ful. Their methodology is logic-based,
fun than proving, and, more importantly
to it. This is a good start, but still at its and they try to teach the computer formal
it gives as much, if not more, insight and
infancy. However, implicitly, more and reasoning, using axioms and “laws of
understanding.”
more mathematicians, even pure ones, deduction.” My approach is more akin
use the computer daily to formulate and to what the great algebraic geometer
While at the MAA for his lecture, Profes-
test conjectures, in a mode that George Shreeram Abhyankar calls “high-school
sor Zeilberger talked with Joe Gallian
Andrews calls “pencil with power-steer- algebra” and physicist Richard Feyn-
about his research and teaching. Follow-
ing.” However, more often than not, the man calls “Babylonian mathematics”:
ing are excerpts from the interview.
computer’s often crucial contribution is using algorithmic frameworks, that I call
not mentioned, or grossly understated. “ansatzes.” But since the algorithms are
JG: How do you describe your research
Human beings are such ingrates. symbolic rather than numeric, one can
area?
hope to prove genuinely new general
Also, Jon Borwein and David Bailey theorems, valid for infinitely many cases.
DZ: I work mainly in combinatorics and
wrote a very nice monograph on Experi- In this kind of research, we try to teach
the theory of special functions, but in the
mental Mathematics that I highly recom- the computer to first make a conjecture,
last 10–15 years I have considered myself
mend. But their emphasis is more on the all by itself, and then automatically prove
mainly an experimental mathematician,
traditional experimental mathematics its own conjectures, also all by itself!
and combinatorics and the theory of
that has been pursued by all the great, Of course, the only way we can do it,

14
May/June 2007 FOCUS
at present, is by focusing on narrowly- JG: You seem to have an evangelical-like gorithm for deciding whether a sequence
defined areas. faith that computers will make obsolete is closed-form.
the way mathematics has been done for
JG: Are you getting some resistance from over 2000 years. Is that a fair assessment JG: It was quite a coup to get Donald
the traditional math community with your of your view? Knuth to write the preface. How did that
approach? come about?
DZ: Yes, but people who know me well
DZ: Definitely. Many people are not very know that they should not take me too DZ: First, and foremost, Knuth is a
comfortable with this approach. First, seriously (laughing). In the Talmud it good friend of Herb Wilf. Also, Knuth
they claim that you can’t trust comput- says that if you have a talit (garment) loves binomial coefficient identities. He
ers (as if humans were so trustworthy!), and one guy says that it’s all his and the dedicated quite a few pages to them in
and they also feel that it’s not fun to other guy says that it’s all his, then they his classic Art of Computer Program-
have a computer do your work, or that each get half. But if somebody only ming, which is the “bible” of computer
it is “cheating.” For me, teaching the claims a half and the other guy claims science, and later, in much more detail, in
computer how to discover and prove new a whole, then the guy who claims a half his beautiful book with Ron Graham and
theorems is even more fun than discover- will only get a quarter and the other guy Oren Patashnik, Concrete Mathematics:
ing and proving them myself. As for the gets three quarters. So if you want a A Foundation for Computer Science. He
cheating part, this is science not sport, half, you have to claim a whole. So you still loves them. In the latest issue of The
and besides we can always change the have to overstate your case. Then again, American Mathematical Monthly (Febru-
rules of the game. sometimes overstating can backfire and ary 2007) there is a problem proposed by
turn people off. him that could be easily done using the
JG: The Wilf-Zeilberger algorithmic WZ method. I was a little taken aback. If
proof theory is contained in all major JG: Tell me about your frequent co-au- somebody submitted a problem to prove
computer algebra systems. What does thor Shalosh B. Ekhad. that 11x13 = 122–1, they wouldn’t accept
it do? it, since there are nowadays (and have
DZ: It is a charming individual. Of been for the last 5000 years) algorithms
DZ: It’s a collection of algorithms that course it is made of silicon, and it is not that can routinely prove this. Similarly,
can discover, and then prove, binomial really one body, but it is definitely one for Knuth’s Monthly problem there are
coefficient identities, and identities in- soul (software). The body has just been (and have been for the last 15 years) al-
volving sums or integrals of special reincarnated many times. As we know, gorithms that can routinely find a proof.
functions. As you know, the binomial co- computers are very powerful, but their But then again, Knuth is Knuth, and if he
efficient “n choose k” counts the number life expectancy is much shorter than that proposed as a problem to prove that 1 + 1
of k-element subsets of an n-element set. of humans, since computers get better = 3 – 1, it might be accepted. Of course,
Since many objects in combinatorics boil and faster so quickly, you have to get a Knuth is very much aware of WZ, and
down to unions and cartesian products new one every three years, but you can he probably had some hidden agenda in
of such choosings, it turns out that many always upload all the software from one proposing this problem. Maybe he meant
enumeration problems can be expressed Shalosh to the next, thereby guaranteeing some cute combinatorial argument.
as such binomial coefficient sums, and the immortality of its soul.
often we get surprising identities. Before JG: I notice that the preface of your book
W-Z theory, every such identity required JG: Where did you get that particular A = B begins with Knuth’s famous quote
its own ad hoc proof, and any such new name? “Science is what we understand well
identity, or even a new proof of an old enough to explain to a computer. Art is
identity, was publishable. But thanks to DZ: The original Shalosh B. Ekhad was everything else we do.” Was A = B the
Wilf-Zeilberger theory, any such identity actually a Hebrew translation of the first first place that quote appeared in print?
can now be proved automatically. No one PC that I owned, called AT&T 3B1. At
today would submit a paper stating and the time it was a very innovative machine, DZ: Yes, I think so.
proving the “theorem” that “ten plus five the first UNIX PC, that was manufactured
equals three times five.” Analogously, an by AT&T in the 80s. The Hebrew transla- JG: I was very surprised that a com-
identity like tion of 3B1 is Shalosh B. Ekhad. mercial publisher would agree to permit
the book to be downloaded free over the
n k
 2n 
3
( )
3n ! JG: Tell me about your book A = B. internet.
∑( ) −1  
 n+ k
= 3
n!
k =0 DZ: It was written by Marko Petkovsek, DZ: Herb Wilf is a great advocate of free
Herb Wilf, and myself, and is an el- publishing and AK Peters is not a typical
first proved by Dixon in 1904, is now ementary introduction to the so-called publisher. Klaus Peters, and his wife Al-
completely routinely and automatically Wilf-Zeilberger algorithmic proof theory ice, are very good people, who care more
provable thanks to the Wilf-Zeilberger mentioned above . It also has a long chap- about mathematics than making a quick
algorithmic proof theory. ter on the very important Petkovsek al- buck. Commercially it seems to have

15
FOCUS May/June 2007

been a wash, some people who would

have bought it, didn’t, and vice versa.

JG: I notice that you prefer Maple over

other software. Why is that?

DZ: Originally, Maple was much cheaper

and more accessible than its competitors.
Mathematica was much more commer-
cial. Unfortunately, now Maple is very
commercial too, and Mathematica’s
prices have gone down, and many people
prefer the latter for its elegant syntax and
beautiful graphics. So I use Maple for
“old times sake” and mainly because I
am used to it.

JG: Describe your Experimental Math-

Doron Zeilberger and Ivars Peterson.
ematics course.
90s. He did good research, but not as should be open refereeing. Also people
DZ: I really enjoy teaching it. It is a much as he could have, because he was should post their PhD theses, especially
graduate class (with one or two advanced so busy developing and testing his card the introduction, that often gives a very
undergraduate students) that is held in game. After his PhD, he got a one-year good overview of a field.
a “smart” classroom where everyone is job in some small college in Washington
connected to a terminal. It is a hands-on State, with a salary of about 22K. When JG: Another thing that surprises me is
course where the main point is to teach the year was almost up, and it was not that you slip a bit of humor in your grant
students how to program in Maple, in clear what his prospect for the future in proposals. One is “There is a delicate
order to explore new mathematics. The academia would be, his card game caught balance and trade-off between the general
actual topic changes every year, so that on, and the rest is history. and the specific, the abstract and the con-
I won’t get bored, and sometimes I even crete, the strategical and the tactical, the
learn a new subject myself, since the JG: There must be a story behind the sacred and the profane.” Another is the
best way of learning a new subject is by “Who you gonna call” T-shirt. line “My particular shtick is experimental
teaching it. Of course, in this class we mathematics.” A third is “The computer
also learn how to program everything, DZ: This is a T-shirt that I am wearing in is a powerful tool, go forth and use it!”
and an even better way to learn a new my picture on my website. It features a Your proposals are less formal than I
subject is by teaching it to a computer, i.e. certain binomial-coefficient identity, with would expect.
programming. So both my students and the caption “Who you gonna call.” The
I learn the substance of the course very back of that T-shirt has the few lines of DZ: That’s my style. I don’t like to be
well, and at the same time my students Macsyma code needed to prove it, and the too serious. But I think that it has cost me
learn how to program in Maple. caption “the binomial-coefficient-iden- some funding. In my last grant proposal,
tity-busters.” My kids told me that it is an I was hoping to get co-funded by the
JG: You have Erdös number 2, Einstein allusion to the film Ghost Busters. This computer science division of the NSF.
number 3, Wiles number 3, and Knuth cute design was made by Herb Wilf’s son, I had no problem getting funding from
number 2 but you say on your website David, who is a lawyer by profession, and the mathematics division, but of course
that you are most proud of your Garfield Herb gave me one of them. mathematics doesn’t have much funds,
number 2. Why is that? so I was hoping to get some additional
JG: Something that I found surpris- funding from computer science. So I also
DZ: Richard Garfield is one of the most ing is that the narratives for your grant sent my grant proposal to the computer
talented and creative combinatorialists proposals are posted at your website. science division. That was a disaster.
alive, but he used his talents in a non- That certainly helps people learn about One of the panelists gave me a lecture,
standard way. He is also a very nice your work. saying this might make a good essay,
guy, and I am fortunate to have known but it’s not what one would expect in a
him, although only briefly. During the DZ: I think that it’s a waste to write grant proposal, and they refused to give
mid-90s, he was a teenage idol thanks something just for five or six people and me funding.
to his innovative and lucrative Magic: for no one else to be able to see it. I think
The Gathering card game. He designed that everything should be public; I don’t JG: You are known for your opinions on
the game while he was Herb Wilf’s PhD like secrecy. I also hate the tradition of your web page, some of which are a bit
student at Penn in the late 80s and early anonymous refereeing. I think that there over-the-top, especially those on April 1.

16
May/June 2007 FOCUS

What changes — if any — are you hoping my students love me. In my computer they are wrong. First, teaching is great
to encourage in the mathematics commu- algebra class, I gave my students some fun, and secondly, it is very important,
nity by posting your opinions online? extremely complicated differential equa- since this is the future!
tions that they had to use Maple to solve.
DZ: I’m not trying to change people’s The solution happened to be a cardioid. So already today teaching is at least as
views, at least not consciously. I just Then they had to draw it (using the Maple important as research. But in years to
like to express my opinions. It’s only plot program), and cut it out. come, when more and more original
the Internet, which is a free-for-all, and mathematical research will be conducted
one doesn’t have to be too uptight. So I JG: You have been known to sponta- by computers, the importance of the hu-
don’t really worry about whether or not neously hand out money to students man research mathematician will dimin-
my pieces always make sense. Hopefully for solving problems or pointing out a ish, and the importance of teaching, at all
they do most, or least some, of the time. correction in class. What is the largest levels, will increase tremendously. Also
prize you have ever awarded, and what for a long time to come, we still need
JG: You have been known to celebrate was it for? humans to program the computer, but
both Valentine’s Day and April Fools what is programming? It is teaching com-
Day in the classes you teach. What is the DZ: For calculation errors the most I’ve puters, and I am sure that being a good
most memorable thing you have done to given out is $1. But for conceptual errors programmer and being a good teacher
mathematically celebrate a holiday? in graduate classes, I think I once gave (for humans) are strongly correlated. One
$10. I also offer prizes for really chal- of the reasons that I believe that I am a
DZ: In my calculus class, I assign a lenging problems. good teacher is that I do so much pro-
homework “project” due February 14, gramming, and am used to spelling out
to graph the parametric equation for a JG: Do you have any concluding each and every step. In short, the future
cardioid. Some people came with very words? of mathematics is in good teaching, both
strange shapes, but some people realized to machines and to humans.
what was going on and just drew a regu- DZ: Yes, I believe that teaching is at
lar heart shape. But they made a pointy least as important as research, and I put Joe Gallian is President of the MAA.
one, which is wrong, because a cardioid lots of effort into my teaching and take Michael Pearson is the MAA Director of
is more rounded. It was nice to get 150 great pride when I do a good job. Many Programs and Services.
valentines (some rounded, some not), research mathematicians dislike teaching
and I could brag to my wife how much and view it as an unavoidable chore, but

In Memoriam W e are deeply grateful for the

generosity of the following individual,
Paul Joseph Cohen died in Stanford side, in 1969. He taught at California
on March 23. Born in 1934, Cohen got State University, Chico, from 1968 who has made a bequest to the
his graduate training at the University to 2004. Burk was the author of two Mathematical Association of America.
of Chicago and eventually became books, both on the theory of integra-
a professor at Stanford University. tion. The second, entitled A Garden
Though he made contributions to anal- of Integrals, will be published by the Every bequest is a powerful
ysis and the theory of partial differen- MAA in May.
tial equations, he was best known for expression of loyalty,
his development of “forcing,” now a James Eells died on February 14 at their lifetime involvement, and
fundamental tool of set theory, and us- the age of 80. Eels graduated from
ing the new technique to prove that the Bowdoin College and then from Har- their faith in the future of the MAA.

b
Continuum Hypothesis and the Axiom vard University. His doctoral work We remember each of them fondly
of Choice are independent from the was completed in 1954, under Hassler
axioms of set theory. This ground- Whitney. After holding several presti- and with deep personal and
breaking work won him the Fields gious positions in the U.S., he moved professional respect.
Medal in 1966, and he received many to the United Kingdom and became
other honors and prizes. FOCUS will Professor of Analysis at the University
have a longer article about Cohen in of Warwick. He was later named the
its August/September issue. first head of the mathematics group at
the International Center of Theoretical
Francis D. Parker
Frank Burk died on March 17. Born Physics in Trieste, Italy. His research Member for 55 years
in 1942, Burk received a Ph.D. from dealt with harmonic maps, geometric
the University of California, River- evolutions, and stochastic analysis.

17