CS 101: An Introduction to Computer

Science for Everyone

Computers and Computation
Fall 2015

Contents
1 The Antikythera mechanism: The clockwork computer 2

2 What is Computation? 4

3 Computational Thinking 10

4 The Many Facets of Natural Computing 13

5 Why Johnny Cant Steam: How video copyright went insane 25

6 Could your iPod be holding the greatest mystery in modern science 32

7 Scooping the Loop Snooper: a proof that the halting problem is undecid-
able 36

1
 11 THE ANTIKYTHERA MECHANISM: THE CLOCKWORK COMPUTER 2

The Antikythera mechanism

The clockwork computer
An ancient piece of clockwork shows the deep roots of modern technology
Sep 19th 2002 | from the print edition
WHEN a Greek sponge diver called Elias Stadiatos discovered the
wreck of a cargo ship off the tiny island of Antikythera in 1900, it
was the statues lying on the seabed that made the greatest
impression on him. He returned to the surface, removed his helmet,
and gabbled that he had found a heap of dead, naked women. The
ship's cargo of luxury goods also included jewellery, pottery, fine
furniture, wine and bronzes dating back to the first century BC. But
the most important finds proved to be a few green, corroded
lumps—the last remnants of an elaborate mechanical device.

The Antikythera mechanism, as it is now known, was originally housed in a wooden box about the
size of a shoebox, with dials on the outside and a complex assembly of bronze gear wheels within.
X-ray photographs of the fragments, in which around 30 separate gears can be distinguished, led
the late Derek Price, a science historian at Yale University, to conclude that the device was an
astronomical computer capable of predicting the positions of the sun and moon in the zodiac on
any given date. A new analysis, though, suggests that the device was cleverer than Price thought,
and reinforces the evidence for his theory of an ancient Greek tradition of complex mechanical
technology.

Michael Wright, the curator of mechanical engineering at the Science Museum in London, has
based his new analysis on detailed X-rays of the mechanism using a technique called linear
tomography. This involves moving an X-ray source, the film and the object being investigated
relative to one another, so that only features in a particular plane come into focus. Analysis of the
resulting images, carried out in conjunction with Allan Bromley, a computer scientist at Sydney
University, found the exact position of each gear, and suggested that Price was wrong in several
respects.

In some cases, says Mr Wright, Price seems to have “massaged” the number of teeth on particular
gears (most of which are, admittedly, incomplete) in order to arrive at significant astronomical
ratios. Price's account also, he says, displays internal contradictions, selective use of evidence and
unwarranted speculation. In particular, it postulates an elaborate reversal mechanism to get some
gears to turn in the right direction.

Since so little of the mechanism survives, some guesswork is unavoidable. But Mr Wright noticed a
fixed boss at the centre of the mechanism's main wheel. To his instrument-maker's eye, this was
suggestive of a fixed central gear around which other moving gears could rotate. This does away
with the need for Price's reversal mechanism and leads to the idea that the device was specifically
designed to model a particular form of “epicyclic” motion.

The Greeks believed in an earth-centric universe and accounted for celestial bodies' motions using
elaborate models based on epicycles, in which each body describes a circle (the epicycle) around a
point that itself moves in a circle around the earth. Mr Wright found evidence that the Antikythera
mechanism would have been able to reproduce the motions of the sun and moon accurately, using

“The Antikythera mechanism: The clockwork computer.” Economist, Sep 19th 2002. Copyright The Economist. Article provided by Library Reserves for private
study, scholarship, or research. Do not redistribute. This article can also be found at http://www.economist.com/node/1337165.
 1 THE ANTIKYTHERA MECHANISM: THE CLOCKWORK COMPUTER 3

an epicyclic model devised by Hipparchus, and of the planets Mercury and Venus, using an
epicyclic model derived by Apollonius of Perga. (These models, which predate the mechanism,
were subsequently incorporated into the work of Claudius Ptolemy in the second century AD.)

A device that just modelled the motions of the sun, moon, Mercury and Venus does not make
much sense. But if an upper layer of mechanism had been built, and lost, these extra gears could
have modelled the motions of the three other planets known at the time—Mars, Jupiter and
Saturn. In other words, the device may have been able to predict the positions of the known
celestial bodies for any given date with a respectable degree of accuracy, using bronze pointers on
a circular dial with the constellations of the zodiac running round its edge.

Mr Wright devised a putative model in which the mechanisms for each celestial body stack up like
layers in a sandwich, and started building it in his workshop. The completed reconstruction, details
of which appeared in an article in the Horological Journal in May, went on display this week at
Technopolis, a museum in Athens. By winding a knob on the side, celestial bodies can be made to
advance and retreat so that their positions on any chosen date can be determined. Mr Wright says
his device could have been built using ancient tools because the ancient Greeks had saws whose
teeth were cut using v-shaped files—a task that is similar to the cutting of teeth on a gear wheel.
He has even made several examples by hand.

How closely this reconstruction matches up to the original will never be known. The purpose of
two dials on the back of the device is still unclear, although one may indicate the year. Nor is the
device's purpose obvious: it may have been an astrological computer, used to speed up the
casting of horoscopes, though it might just as easily have been a luxury plaything. But Mr Wright
is convinced that his epicyclic interpretation is correct, and that the original device modelled the
entire known solar system.

The Greeks had a word for it

That tallies with ancient sources that refer to such devices. Cicero, writing in the first century BC,
mentions an instrument “recently constructed by our friend Poseidonius, which at each revolution
reproduces the same motions of the sun, the moon and the five planets.” Archimedes is also said
to have made a small planetarium, and two such devices were said to have been rescued from
Syracuse when it fell in 212BC. This reconstruction suggests such references can now be taken
literally.

It also provides strong support for Price's theory. He believed that the mechanism was strongly
suggestive of an ancient Greek tradition of complex mechanical technology which, transmitted via
the Arab world, formed the basis of European clockmaking techniques. This fits with another,
smaller device that was acquired in 1983 by the Science Museum, which models the motions of
the sun and moon. Dating from the sixth century AD, it provides a previously missing link between
the Antikythera mechanism and later Islamic calendar computers, such as the 13th century
example at the Museum of the History of Science in Oxford. That device, in turn, uses techniques
described in a manuscript written by al-Biruni, an Arab astronomer, around 1000AD.

The origins of much modern technology, from railway engines to robots, can be traced back to the
elaborate mechanical toys, or automata, that flourished in the 18th century. Those toys, in turn,
grew out of the craft of clockmaking. And that craft, like so many other aspects of the modern
world, seems to have roots that can be traced right back to ancient Greece.

Copyright © The Economist Newspaper Limited 2012. All rights reserved.

“The Antikythera mechanism: The clockwork computer.” Economist, Sep 19th 2002. Copyright The Economist. Article provided by Library Reserves for private
study, scholarship, or research. Do not redistribute. This article can also be found at http://www.economist.com/node/1337165.
 22 WHAT IS COMPUTATION? 4

E D C Z Y X
B A W

What is
B
A

V
C
Z

U
Computation? D
Y

T
E
X

S
F

R
W

G
Our culture is in the process of renegotiating what it thinks computation and
computer really mean.

Q
V

By Ian Horswill
H

DOI: 10.1145/2090276.2090283

I
P
U

n the 19th century, our society fundamentally changed the way it produced material
I

O
goods. Whereas production had previously depended on human and animal labor, the
T J
S steam engine allowed much of that labor to be transferred to machines. This made K N
the manufacture of many goods and services more efficient, and therefore cheaper. It
Q R L M
also made other kinds of products available that would otherwise have been impossible to
manufacture. Resulting in widespread economic dislocation, skilled tradesmen were no
longer needed. Much of the population were forced to leave their rural communities for
cities, radically changing the very environment they lived in. Consumer culture, television,
corporations, and the automobile would not have been possible—or at least as influential—
P O N without this automation of production, better knownZ as the Y
Industrial
X Revolution.
M
We now hear we’re in the midst of A
about the esoteric worship of acro- W
is “just” computation, then what isn’t
an “information evolution.” Like the nyms and punctuation. computation? And what does it even
Industrial Revolution, it involves auto- B mean to say that something is or isn’t
WHAT ISN’T COMPUTATION
V
mation—the transfer of human labor computation?
L

to machines. However physical labor Computation isn’t tied to numbers, Computation is an idea in flux. One
C

has been replaced by intellectual la- acronyms, punctuation, or syntax. But hundred years ago, nearly everyone
U
K

bor. While we cannot know the ulti- one of the things that makes computa- thought of computation as being a
mate consequences of these changes, tion so interesting is that, in all hones- mental operation involving numbers.
D

information technology has played ty, it is not entirely clear what compu- And being a mental operation, it could
T
J

an increasingly central role in our so- tation really is. We all generally agree only be done by people. Today, the
ciety over the past two decades. In the that when someone balances their overwhelming majority of what we con-
E

1950s, if you and your friends wanted checkbook, they’re doing computa- sider to be computation is done by ma-
I

to go to a movie, you wouldn’t have tion. But many people argue the brain chines. But at the same time, we still
started by walking to the library. But is fundamentally a computer. If that’s somehow associate it with thought. In
F

today you wouldn’t think twice about true, does it mean that (all) thought is Western culture, we tend to take our
H

checking the Internet for reviews and computation? Or that all computation capacity for thought as the central dis-
R

show times. Similarly, average people is thought? We know that the “virtual” tinction between ourselves and other
G

wouldn’t have published their diaries environments in computer games are animals. Moreover, we view our spe-
G

in the 1950s, but today people blog really just computational simulations. cific thoughts and feelings as being
Q

about everything from international But it has also been argued that the one of the major constituents of our
Illustration by James Joyce

politics to their pets’ favorite toys. real universe is effectively a computer, personal identity. So thought is consti-
H

Yet for all the importance of these that quantum physics is really about tutive both of our collective humanity
F

technologies, people still tend to think information, and that matter and en- and of our individual identities.
I

of computation as being about num-

E ergy are really just abstractions built Changing our ideas about computa-
O
bers, and computer science as being from information. But if the universe
J tion changes our ideas about thought
D N
B C X R D S Ł      
Ť Ţ ţ Ť Ł    ĩ ţ Ū  Ł   ĩ ť
K L M 9

Ian Horswill, “What is Computation?.” XRDS, March 2012. Copyright Association for Computing Machinery. Article provided by Library Reserves for private
study, scholarship, or research. Do not redistribute. This article can also be found at http://xrds.acm.org/article.cfm?aid=2090283.
 2 WHAT IS COMPUTATION?
and therefore our ideas about our-
selves. It reverberates through our cul-
ture, producing excitement, anxiety,
and countless B-movies about virtual
reality and cyber-thingies.
And yet for all our collective fascina-
tion with computation, what we mean by
the term is still somewhat mysterious.
QUESTIONS AND ANSWERS
Suppose you ask someone: “What’s
seven plus three?”If they reply
“10,” then they just performed a
computation. So we can provisionally
think of computation as a kind of
question-answering. In the case of
addition, the question always involves
a pair of numbers and the answer
is always another number. For any
particular addition question, there’s a
corresponding number, determined
by the pair of numbers, which is the
desired answer.
Now suppose you ask them “What’s
one million, seven hundred eighty-two
thousand, six hundred, and seventy-
eight plus three million, two hundred
ninety-two thousand, seven hundred
and four?” Unless they’re especially
good at mental arithmetic, they won’t
be able to keep all the numbers in their
head and so they’ll have to take out
pencil and paper, ask you to repeat the
question, write it down, perform the ad-
dition on paper, and read off the answer.
But now arithmetic is no longer just
a mental operation; it’s also a physical
operation. It involves manipulation
and change of physical objects (the
pencil and paper). We’re used to think-
ing of the arithmetic as a “mental”
thing that takes place “in” our heads.
But in this case, it’s spread out between
a person’s head, hands, pencil, and pa-
per. Another change from the seven-
plus-three example is that while we
presented the numbers to the person
as a set of sounds in spoken English,
they then represented them as symbols
on paper, presumably Arabic numer-
als. (In fact, “the numbers themselves”
never even made an appearance, since
they’re completely intangible; we only
have access to them through represen-
tations.)
What’s interesting here is that none
of these differences matter. So long
as the person comes up with the right
answer in whatever representational
system or systems they happened to be
using, we consider them to have suc-
cessfully done the arithmetic. Put an-
other way, it doesn’t matter which path
we take through the diagram below:
Writing Writing
add2
Write Read
Speech Speech
add1
We’ll call this the principle of be-
havioral equivalence: If a person or
system reliably produces the right an-
swer, they can be considered to have
solved the problem regardless of what
procedure or representation(s) they
used. Behavioral equivalence is abso-
lutely central to the modern notion of
computation. If we replace one compu-
tational system with another that has
the “same” behavior, the computation
will still “work.”
THE FUNCTIONAL MODEL
The question is what it means to have
the “same” behavior. For question-
answering, it means giving the same
answers to the same questions; that is,
generating the same output. We can
provisionally think of a computation-
al problem as a function from a set of
possible inputs (the questions) to their
desired outputs (the answers). If we as-
sume that the only important part of a
computation’s behavior is its output,
then we have a functional model of
computation. Procedures are just ways
of computing functions and so two
procedures are behaviorally equivalent
if and only if they compute the same
function, although they may differ in
efficiency.
Procedures differ from functions
in that functions only specify what
their outputs should be, whereas pro-
cedures specify how to compute them.
Computer science is in many ways the
study of procedural knowledge: How to
5
describe procedures, construct them,
and compare competing procedures
for solving the same computational
problem.
THE IMPERATIVE MODEL
The functional model is a pretty good
model of what we mean by computa-
tion; it covers all of math and a lot of
programming. But it assumes com-
putations (1) read an input, think for
a while, write an output, and halt;
and (2) the only important aspect of a
program’s behavior is the relation be-
tween its input and output.
The problem is our arithmetic ex-
ample, like modern von Neumann
computers, worked by gradually modi-
fying scratch paper—writing carries
down, partial sums, etc. And the act of
writing a carry mark on scratch paper,
like the act of storing a register to main
memory, isn’t best thought of as com-
puting a function, per se; it is an im-
perative. It’s an operation that works
by modifying the internal state of the
machine. So we have the embarrassing
situation that our model doesn’t con-
sider the steps of our computations to
be computations.
Moreover, many computations just
don’t fit into the functional model.
Does the delete file command on your
laptop produce an output? Not really.
Whereas the modification of memory
may be a side effect of an arithmetic
program, it’s the very purpose of de-
letion; and an animation program’s
purpose is to draw successive frames
to video memory. Nor do all programs
terminate; you actually want your op-
erating system or anti-lock brake con-
troller to be infinite loops! These are
all computations that:
š7h[f[h\ehc[Zc[Y^Wd_YWbbo$
š:eiec[j^_d]ki[\kb$
š:e_jki_d][nWYjboj^[iWc[fh_c_-
tive instructions that “normal”
procedures use.
But these aren’t considered com-
putations under the functional model
because they don’t compute functions.
There is another model of computa-
tion, the imperative model: Computa-
tions are sequences of imperatives that
manipulate representations. The ad-
vantage of this model is that it’s more
inclusive than the functional model.

10 X R D S Ł      
 Ť Ţ ţ Ť  Ł     ĩ ţ Ū  Ł   ĩ ť

Ian Horswill, “What is Computation?.” XRDS, March 2012. Copyright Association for Computing Machinery. Article provided by Library Reserves for private
study, scholarship, or research. Do not redistribute. This article can also be found at http://xrds.acm.org/article.cfm?aid=2090283.
 2 WHAT IS COMPUTATION?
The disadvantage is that it’s hard to rea-
son about programs at this level. Com-
putations in the functional model have
simple specifications—the function to
be computed—and while proving that a
program satisfies the specification may
be hard, it’s at least a clearly defined no-
tion. Under the imperative model, it can
be difficult even to write a specification,
much less determine whether an imple-
mentation satisfies it.
Now that we have a couple of differ-
ent ways to think about computation,
let’s think about what we mean by
computers and what makes them un-
usual as a kind of mechanism.
COMPUTERS
Although computation always involves
the manipulation of representations,
the modern digital computer has the
additional property that those proce-
dures are themselves representations.
They can be manipulated like any oth-
er data: copied, modified, erased, etc.
Procedures differ from other represen-
tations only in the machine is able to
interpret (execute) them. This provides
us with one of the two defining charac-
teristics of modern digital computers:
programmability.
Programmability allows a single
device to serve many different func-
tions, simply by changing a stored
representation inside the device. Al-
though easy to take for granted, it
is this programmability that makes
computers the social and economic
force that they are today. Computers
could never have become economi-
cally viable if consumers had to buy
separate machines for email, text pro-
cessing, and Web browsing.
The second defining property of
modern digital computers is their use
of a single uniform meta-representa-
tion for all types of data, namely text.
Computer text is like any other text—
strings of symbols drawn from a fixed
alphabet. The only interesting differ-
ence from any other text is that it uses
an alphabet with only two symbols,
conventionally called 0 and 1. The
choice of a binary alphabet is ultimate-
ly a matter of engineering convenience:
It’s easier to make a reliable electronic
memory circuit if it only has to reli-
ably distinguish between two sym-
bols. And over the last 50 years, we’ve
devised clever ways of encoding other
representations as binary text strings,
including human text, tree and graph
structures, pictures, and sound.
META-PROGRAMMING
Modern digital computers are charac-
terized by:
š7kd_Ò[Zc[jW#h[fh[i[djWj_edX_-
nary text strings) in which all other
representations can be encoded.
š7i[je\Xk_bj#_dYeccWdZi\ehcW-
nipulating binary text.
šJ^[ WX_b_jo je ijeh[ i[gk[dY[i e\
commands (i.e. procedures) in
memory as binary text.
šJ^[ WX_b_jo je _dj[hfh[j [n[Ykj[
previously stored procedures.
The fact that procedures are stored
in the same bit-string format as other
data has profound consequences; it al-
lows us to write meta-programs that
examine, manipulate, and even create
other procedures.
While programmability makes
computer hardware economically vi-
able, meta-programmability makes
software economically viable. Humans
are the scarcest and most expensive
resource in the computer industry. So
just as the Industrial Revolution re-
quired the use of machines to make
machines, software engineering re-
quires the use of software to make soft-
ware. Consequently, the overwhelming
historical trend in software develop-
ment has been to automate more and
more of the development task, even
at the expense of writing less efficient
programs that require more powerful
computers.
All the tools we take for granted as
programmers—compilers, debuggers,
profilers, static checkers, testing rigs,
and even source-code management
systems—are meta-programs. They au-
tomate mundane aspects of program-
ming, freeing you to focus on the tasks
that require human creativity.
SIMULATION
Although all modern computers share
the same broad traits, they can’t nec-
essarily run one another’s programs.
Yet any computer can at least store any
other computer’s programs, because
all modern computers store data as
6
bit strings. Given that, we can write a
meta-program for one computer to in-
terpret or emulate the instruction set
of another. The emulator makes the
machines behaviorally equivalent.
There are a few obvious issues with
simulation. The simulated machine
will be slower. And a computer with
64K of memory can’t simulate one with
100GB. But can one computer always
simulate another if it has enough mem-
ory? The question matters because if A
can simulate B but B can’t simulate A,
then we know A can perform at least
one computational task that B can’t—A
is in some sense more powerful than B.
So the question is: Is there an instruc-
tion set that’s sufficient for simulating
any kind of computer with any kind of
instruction set?
We have the answer, and that an-
swer is “probably.”
UNIVERSALITY
The first machines that could simu-
late other computers were designed by
British mathematician Alan Turing,
and so are now referred to as Turing
machines [1]. Most Turing machines
weren’t even programmable, but Tur-
ing showed there was a particular kind
of Turing machine, which he called the
Universal Machine, which could take
representations of other Turing ma-
chines and simulate them. That meant
the Universal Machine was not only
programmable, it could compute any-
thing that any Turing machine could
compute.
The details of Turing machines
aren’t important to this discussion.
What matters is that they can simulate
all the computers we’ve managed to
think up, and consequently can com-
pute anything those computers can
compute. This means any computer
that’s able to simulate Turing’s Uni-
versal Machine is also able to simu-
late any computer, and therefore able
to compute anything we know how to
compute. This property of being able
to simulate Turing machines, and
therefore being able to simulate any
known computer, is called “Turing
completeness,” “Turing equivalence,”
or just “universality.”
The amazing thing is that nearly
all digital computers are Turing com-
plete. In fact, some stunningly simple

X R D S Ł      
Ť Ţ ţ Ť Ł    ĩ ţ Ū  Ł   ĩ ť 11

Ian Horswill, “What is Computation?.” XRDS, March 2012. Copyright Association for Computing Machinery. Article provided by Library Reserves for private
study, scholarship, or research. Do not redistribute. This article can also be found at http://xrds.acm.org/article.cfm?aid=2090283.
 2 WHAT IS COMPUTATION?
mechanisms are Turing complete [2].
One of the practical consequences of
this is that hardware designers can
largely ignore the issue of whether
their instruction sets are universal
(they are) and focus on finding instruc-
tion sets that can be made to run very
quickly in hardware.
Of course this doesn’t preclude the
possibility that someone might some-
day come up with a radically different
design for a computer that couldn’t be
simulated by a Turing machine. But
thus far, every computational system
that’s been devised has either been
Turing complete (can simulate and be
simulated by any other Turing com-
7
other machines could compute. It
turns out that not only are some prob-
lems provably uncomputable, but in a
sense it’s the very power of universal
machines that makes those problems
uncomputable.
If we return to the idea of offload-
ing parts of the software development
task to the computer, we can see that
people have developed meta-pro-
grams for detecting many different
kinds of common bugs: type check-
ers for finding type errors, memory
debuggers for finding leaks, deadlock
detectors, and so on. And yet, every
computer user has the experience
of having their machines hang on
a far too regular basis. So if we can
perform automated detection of type
errors, why can’t we perform automat-
ed detection of infinite loops?
The automated detection of infinite
loops is known as the halting problem:
Given a program and a potential input
for it, determine whether the program
would ever halt if it were run on that
input. It’s hard because running for a
long time doesn’t always mean that a
program is hung; it’s always possible
Image courtesy of Flickr user Dino Gravato
that it’s about to finish and give you an
answer.
One might think we’d be able to use a
simulator to solve the halting problem.
After all, a simulator tells us the exact
output a program generates. That’s ac-
tually more information than we need,
since we just care whether there will be
an output. Unfortunately, this doesn’t
work. The problem is that since simu-
lators run the programs they simulate,
if the simulated program runs forever,
so does the simulator.
plete system) or is weaker than a Tur- Let’s think for a moment about what
ing machine (i.e. can’t simulate a Tur- it would mean for the halting problem
ing machine but can still be simulated to be computable. If the halting prob-
by a Turing machine). So it is generally lem were computable, we could write
assumed that any function that can be a meta-procedure called halts() that
computed can be computed by a Tur- took two inputs, a procedure and an
ing machine. This is known as Church’s input to run it on, which would infalli-
Thesis, or sometimes as the Church- bly output “yes” or “no” depending on
Turing Hypothesis, after 20th century whether the specified procedure would
meta-mathematician Alonzo Church. halt if run on the specified input. That
turns out to be the difficulty with the
THE LIMITS OF COMPUTATION halting problem, because it leads to a
You may have noticed a certain care- paradox. For if we could solve the halt-
fulness of wording in our definition of ing problem—if we could write the
universality. We didn’t say that univer- halts procedure—we could write an-
sal machines could compute anything, other procedure, call it contrarian(),
only that they could compute anything that calls it as a subroutine:

12 X R D S Ł      
 Ť Ţ ţ Ť  Ł     ĩ ţ Ū  Ł   ĩ ť

Ian Horswill, “What is Computation?.” XRDS, March 2012. Copyright Association for Computing Machinery. Article provided by Library Reserves for private
study, scholarship, or research. Do not redistribute. This article can also be found at http://xrds.acm.org/article.cfm?aid=2090283.
 2 WHAT IS COMPUTATION?
void contrarian(int input) {
if (halts(contrarian, input))
while (TRUE) { /* loop infinitely */ }
} IMITATION, EQUIVALENCE, AND
But notice that contrarian is asking
halts whether it (contrarian) will halt it-
self. And being contrarian, it then does
exactly the opposite. So no matter what
halts returns, it will be wrong by defini-
tion; if halts() says it will halt, contrar-
ian runs the infinite while loop. And if
it says it will hang, contrarian returns
immediately.
The fact we can write a four line C
program to break any proposed solu-
tion to the halting problem means
that the halting problem must be un-
computable. This proof, or rather, a
much more rigorous version of it, was
due to Turing. It’s important because
it shows that:
šThere are problems that are un-
computable by Turing machines,
and so by any mechanical system
we’ve ever been able to devise.
šMehi[ o[j" j^[o _dYbkZ[ fheXb[ci
we actually care about.
šKbj_cWj[bo"j^[[n_ij[dY[e\kdYec-
putable problems is due to the very
power of general-purpose compu-
tation and, specifically, to the ca-
pacity for meta-computation.
However, the uncomputability of
the halting problem is often misun-
derstood, so let’s be very clear on what
it does and doesn’t entail. It doesn’t
mean we can never determine if a pro-
gram halts; in fact, we can solve the
halting problem for many of the pro-
grams that we care about in real life.
What we can’t do is write an infallible
program for determining whether an
arbitrary piece of program text will
halt for an arbitrary input text. Any
program we write for the halting prob-
lem must necessarily get it wrong some
of the time, even if it generally gets it
right in the cases we really care about.
The halting problem has also been
used to argue that the brain must not
be a computer because people can tell
whether programs will halt and com-
puters cannot. Regardless of whether
or not the brain is best thought of as a
computer, people are actually just like
computers on the halting problem: We
can solve it a lot of the time, but we also
get it wrong sometimes.
INTELLIGENCE
There’s an old party game, called “the
imitation game,” in which two people—
usually a man and a woman—hide in
different rooms while the other guests
try to determine who is whom by sub-
mitting written questions and receiv-
ing written answers. The idea is to see
whether the hiders can fool the rest of
the guests. In the terminology we’ve
used above, the people in the rooms
are trying to act behaviorally equiva-
lent to one another. Obviously, this
doesn’t mean that if the impersonators
win they’ve actually changed identities
or sexes. (That is, sex in the biological
sense. One might argue that they’ve
temporarily changed their socially de-
fined gender identities.)
But imagine we play the game not
with two people, but with one person
and one computer. What if the com-
puter could fool people into thinking it
was human? In an important 1950 pa-
per, “Computing Machinery and Intel-
ligence,” Turing argued if a computer
could fool humans into thinking it was
human, then it would have to be con-
sidered to be intelligent even though
it wasn’t actually human [3]. In other
words, intelligence is ultimately a be-
havioral phenomenon [4]. That is, an
entity is intelligent because it behaves
intelligently, not because it’s com-
posed of some special substance such
as living tissue or a soul. (Note that this
isn’t an argument as to whether people
have souls or not, simply that it’s pos-
sible to be intelligent without one. It’s
not a theological claim.) Moreover,
intelligence is a computational phe-
nomenon, amenable to computational
analysis.
Together with cybernetics, and ear-
lier work on the theory of computation,
Turing’s paper set the stage for the de-
velopment of artificial intelligence in
the 1950s and ‘60s. These provided a
scientific model for the study of mental
representations and the processes that
operate on them, which were then im-
ported into fields such as psychology
and linguistics, leading to the develop-
ment of the field of cognitive science in
the 1970s.
8
COMPUTATIONAL NEUROSCIENCE
We started with the idea of computation
being a process that humans perform
in their brains, and then looked at how
specific instances of that process could
be modeled at a coarse level by mechani-
cal processes. That led us eventually to
both the limits of mechanical proce-
dures, and the claim that it was none-
theless theoretically possible for a ma-
chine to be intelligent. Conversely, it’s
become common in our culture to talk
about the brain as being a kind of com-
puter. At this point, you may object, “I
don’t feel like a computer.” So let’s talk
about what it might mean to say that the
brain is a kind of computer. Certainly,
the brain can do computation. But that
doesn’t mean if you cut open someone’s
skull you’re going to see something
that’s structurally similar to a lap-
top or a Turing machine. But as we’ve
seen, computation is all about behav-
ioral equivalence. So the real question
is could a brain and a PC, in principle,
simulate one another? One side is obvi-
ous. A person can simulate a PC, by writ-
ing down the contents of the computer’s
memory on (a whole lot of) scratch pa-
per and then manually executing the
instructions in its program, one at a
time. It would be incredibly boring, but
it would be conceptually possible. So the
real question is whether a sufficiently
big PC could simulate the human brain.
The brain is a network of roughly
100 billion cells called neurons, to-
gether with other tissues that help the
neurons function. To a first approxima-
tion, each neuron has a set of inputs,
called dendrites, which receive signals
from neighboring neurons, and a sin-
gle output, the axon, which produces
a signal based on the cell’s inputs and
internal state. Each neuron’s axon is
connected to dendrites of many other
neurons. When one neuron stimulates
another neuron, it predictably increas-
es or decreases the rate or likelihood of
the second neuron stimulating other
neurons. If it’s predictable, then it
should be possible to write a computer
program that simulates it, and indeed
there are many computational models
of different kinds of neurons. But if
each individual neuron can be simu-
lated computationally, then it should
be possible in principle to simulate the
whole brain by simulating the individ-

X R D S Ł      
Ť Ţ ţ Ť Ł    ĩ ţ Ū  Ł   ĩ ť 13

Ian Horswill, “What is Computation?.” XRDS, March 2012. Copyright Association for Computing Machinery. Article provided by Library Reserves for private
study, scholarship, or research. Do not redistribute. This article can also be found at http://xrds.acm.org/article.cfm?aid=2090283.
 2 WHAT IS COMPUTATION?
ual neurons and connecting the simu-
lations together.
While there’s a difference between
“should be possible” and “is actually
practical,” this is the general idea be-
hind the claim that we can understand
the brain computationally. Assuming
it’s right, it has some important con-
sequences. If computers can simulate
brains (and by similar reasoning, hu-
mans in general), then they can solve
any problem humans can solve, if only
by simulating a human. But since we
know that computers can’t solve cer-
tain problems (the uncomputable prob-
lems), that would mean that humans
can’t solve uncomputable problems ei-
ther. In other words, it imposes a funda-
mental limit on human knowledge.
Although we’re a long way from being
able to fully simulate a brain, computa-
tional neuroscience—in which scien-
tists try to understand neural systems
as computational processes—is an im-
portant and growing area of biological
research. By modeling neural systems
as computational systems, we can bet-
ter understand their function. And in
some experimental treatments, such as
cochlear implants, we can actually re-
place damaged components with com-
putational systems that are behaviorally
equivalent to the biological tissue.
SO WHAT IS COMPUTATION?
Computation is an idea in flux. Our
culture is in the process of renegotiat-
ing what it thinks the concepts com-
putation and computer really mean.
Computation lies on a conceptual fault
line where small changes can have ma-
jor consequences to how we view the
world. That makes it a very interesting
field to follow right now.
It’s an interesting exercise to type
“define computation” into Google and
look at the definitions you get. Search
results include “determining some-
thing by mathematical or logical meth-
ods,” which doesn’t apply well to the
idea of DNA being a computer program,
or “finding a solution to a problem from
given inputs by means of an algorithm,”
which is essentially the functional mod-
el that, as we saw, doesn’t apply well to
things like computer games or anti-lock
brake systems. Another possible an-
swer is that computation is “informa-
tion processing,” although that simply
Turing argued if
a computer could
fool humans into
thinking it was
human, then it
would have to be
considered to be
intelligent even
though it wasn’t
actually human.
begs the question of what we mean by
information.
A pragmatic definition would be
that computation is what (modern,
digital) computers do, so a system is
“computational” if ideas from com-
puters are useful for understanding
it. This is probably the closest defini-
tion to how real people use the term in
practice. As we’ve seen, there are a lot
of different kinds of systems that are
computational under this view. Com-
putation is then a set of metaphors, a
lens through which we can view and
describe other phenomena.
But that isn’t a very satisfying defi-
nition. Another view would be that
computation is wrapped up with the
study of behavioral equivalence. Under
this view, computation is the process
of producing some desired behavior
without prejudice as to whether it is
implemented through silicon, neu-
rons, or clockwork.
Whatever definition we adopt, it’s
clear that our society is undergoing
profound changes as a result of both
computational technology and com-
putational metaphors. These raise a
number of psychological and social is-
sues we need to think about:
šWhat kinds of computational sys-
tems are easy for people to use?
š>em YWd m[ Z[i_]d ioij[ci j^Wj
people find more rewarding and
engaging?
š>em YWd m[ Z[i_]d bWh][ ioij[ci
that are easy for humans to under-
stand?
šM^WjWh[j^[h_iaie\_d\ehcWj_ed
9
technology? Do we really want to
make airplanes that are controlled
by computers, so that if program
crashes, the plane does too? Do
databases and surveillance tech-
nology give governments and cor-
porations too much power?
š>emYWdm[ki[_d\ehcWj_edj[Y^-
nology to improve people’s lives?
Computation is a broad, rich field.
It has had deep influences on our lives
and culture. If your PC, cell phone, and
Web access suddenly disappeared, you
would probably have to radically recon-
figure your life, even though these have
only been widely available in the last 15
years. It’s unlikely that the changes to
society are going to stop any time soon.
By better understanding computation,
we can help make sure those changes
are for the better.
Biography
Ian Horswill is an associate professor of computer
science at Northwestern University. He is a member
of the Department of Electrical Engineering and
Computer Science, where he is Director of the Division
of Graphics and Interactive Media, and was cofounder
of Northwestern’s Animate Arts Program. His research
interests include interactive entertainment technologies,
and cognitive modeling for virtual characters, particularly
modeling of emotion and personality. He received his Ph.D.
in computer science from the Massachusetts Institute of
Technology in 1993. He has been the chair of the standing
committee of the Association for the Advancement of
Artificial Intelligence’s Fall and Spring Symposium Series,
as well as the 2009 International Conference on the
Foundations of Digital Games.
References
[1] Turing, A. On computable numbers, with an application
to the Entscheidungsproblem. In P roceedings of the
London Mathematical Society, Series 2, 42 (1936),
230-265. Note that Turing machines were more
a conceptual design than a practical engineering
design. Although Turing did work on the design of
actual working hardware, his work on Turing machines
came before the construction of real computers was
practical.
[2] Minsky and Blum proved in the 1960s that a “two-
counter machine” was Turing complete. A two-counter
machine is a machine whose only data representation
is a pair of numbers (non-negative integers, in fact),
and whose only commands are to add or subtract 1
from one of the numbers and to check whether one of
them was zero.
[3] Turing, A. Computing machinery and intelligence.
Originally published by Oxford University Press on
behalf of MIND (the Journal of the Mind Association)
59, 236 (1959), 433-60; http://www.abelard.org/
turpap/turpap.htm
[4] Although widely accepted, the computational view is
not universally accepted. See for example, Searle, J.
R. Minds, brains, and programs. Behavioral and Brain
Sciences 3, 3 (1980), 417-457; http://www.bbsonline.
org/Preprints/OldArchive/bbs.searle2.html
© 2012 ACM 1528-4972/12/03 $10.00

14 X R D S Ł      
 Ť Ţ ţ Ť  Ł     ĩ ţ Ū  Ł   ĩ ť

Ian Horswill, “What is Computation?.” XRDS, March 2012. Copyright Association for Computing Machinery. Article provided by Library Reserves for private
study, scholarship, or research. Do not redistribute. This article can also be found at http://xrds.acm.org/article.cfm?aid=2090283.
 33 COMPUTATIONAL THINKING
Viewpoint Jeannette M. Wing
Computational Thinking
It represents a universally applicable attitude and skill set everyone, not just
computer scientists, would be eager to learn and use.
omputational thinking
C builds on the power and
limits of computing
processes, whether they are exe-
cuted by a human or by a
machine. Computational
methods and models give us
the courage to solve prob-
lems and design systems that no one of us would
be capable of tackling alone. Computational think-
ing confronts the riddle of machine intelligence:
What can humans do better than computers? and
What can computers do better than humans? Most
fundamentally it addresses the question: What is
computable? Today, we know only parts of the
answers to such questions.
Computational thinking is a fundamental skill for
everyone, not just for computer scientists. To read-
ing, writing, and arithmetic, we should add compu-
tational thinking to every child’s analytical ability.
Just as the printing press facilitated the spread of the
three Rs, what is appropriately incestuous about this
vision is that computing and computers facilitate the
spread of computational thinking.
Computational thinking involves solving prob-
lems, designing systems, and understanding human
behavior, by drawing on the concepts fundamental
to computer science. Computational thinking
includes a range of mental tools that reflect the
breadth of the field of computer science.
Having to solve a particular problem, we might
ask: How difficult is it to solve? and What’s the best
LISA HANEY
way to solve it? Computer science rests on solid the-
oretical underpinnings to answer such questions pre-
Jeannette Wing, “Computational Thinking.” CACM, March 2006. Copyright Association for Computing Machinery. Article provided by Library Reserves for
private study, scholarship, or research. Do not redistribute. This article can also be found at http://www.cs.cmu.edu/~wing/publications/Wing06.pdf.
10
cisely. Stating the difficulty of a problem accounts
for the underlying power of the machine—the com-
puting device that will run the solution. We must
consider the machine’s instruction set, its resource
constraints, and its operating environment.
In solving a problem efficiently,, we might further
ask whether an approximate solution is good
enough, whether we can use randomization to our
advantage, and whether false positives or false nega-
tives are allowed. Computational thinking is refor-
mulating a seemingly difficult problem into one we
know how to solve, perhaps by reduction, embed-
ding, transformation, or simulation.
Computational thinking is thinking recursively. It
is parallel processing. It is interpreting code as data
and data as code. It is type checking as the general-
ization of dimensional analysis. It is recognizing
both the virtues and the dangers of aliasing, or giv-
ing someone or something more than one name. It
is recognizing both the cost and power of indirect
addressing and procedure call. It is judging a pro-
gram not just for correctness and efficiency but for
aesthetics, and a system’s design for simplicity and
elegance.
Computational thinking is using abstraction and
decomposition when attacking a large complex task
or designing a large complex system. It is separation
of concerns. It is choosing an appropriate representa-
tion for a problem or modeling the relevant aspects
of a problem to make it tractable. It is using invari-
ants to describe a system’s behavior succinctly and
declaratively. It is having the confidence we can
safely use, modify, and influence a large complex
system without understanding its every detail. It is
COMMUNICATIONS OF THE ACM March 2006/Vol. 49, No. 3 33
 3 COMPUTATIONAL THINKING
Viewpoint
modularizing something in anticipation of multiple
users or prefetching and caching in anticipation of
future use.
Computational thinking is thinking in terms of
prevention, protection, and recovery from worst-case
scenarios through redundancy, damage containment,
and error correction. It is calling gridlock deadlock
and contracts interfaces. It is learning to avoid race
conditions when synchronizing meetings with one
another.
Computational thinking is using heuristic reason-
ing to discover a solution. It is planning, learning,
and scheduling in the presence of uncertainty. It is
search, search, and more search, resulting in a list of
Thinking like a computer scientist means more than being able to
program a computer. It requires thinking at multiple levels of abstraction.
Web pages, a strategy for winning a game, or a coun-
terexample. Computational thinking is using massive
amounts of data to speed up computation. It is mak-
ing trade-offs between time and space and between
processing power and storage capacity.
Consider these everyday examples: When your
daughter goes to school in the morning, she puts in
her backpack the things she needs for the day; that’s
prefetching and caching. When your son loses his
mittens, you suggest he retrace his steps; that’s back-
tracking. At what point do you stop renting skis and
buy yourself a pair?; that’s online algorithms. Which
line do you stand in at the supermarket?; that’s per-
formance modeling for multi-server systems. Why
does your telephone still work during a power out-
age?; that’s independence of failure and redundancy
in design. How do Completely Automated Public
Turing Test(s) to Tell Computers and Humans
Apart, or CAPTCHAs, authenticate humans?; that’s
exploiting the difficulty of solving hard AI problems
to foil computing agents.
Computational thinking will have become
ingrained in everyone’s lives when words like algo-
rithm and precondition are part of everyone’s vocab-
34 March 2006/Vol. 49, No. 3 COMMUNICATIONS OF THE ACM
Jeannette Wing, “Computational Thinking.” CACM, March 2006. Copyright Association for Computing Machinery. Article provided by Library Reserves for
private study, scholarship, or research. Do not redistribute. This article can also be found at http://www.cs.cmu.edu/~wing/publications/Wing06.pdf.
11
ulary; when nondeterminism and garbage collection
take on the meanings used by computer scientists;
and when trees are drawn upside down.
We have witnessed the influence of computa-
tional thinking on other disciplines. For example,
machine learning has transformed statistics. Statisti-
cal learning is being used for problems on a scale, in
terms of both data size and dimension, unimagin-
able only a few years ago. Statistics departments in
all kinds of organizations are hiring computer scien-
tists. Schools of computer science are embracing
existing or starting up new statistics departments.
Computer scientists’ recent interest in biology is
driven by their belief that biologists can benefit
from computational thinking. Computer science’s
contribution to biology goes beyond the ability to
search through vast amounts of sequence data look-
ing for patterns. The hope is that data structures
and algorithms—our computational abstractions
and methods—can represent the structure of pro-
teins in ways that elucidate their function. Compu-
tational biology is changing the way biologists
think. Similarly, computational game theory is
changing the way economists think; nanocomput-
ing, the way chemists think; and quantum comput-
ing, the way physicists think.
This kind of thinking will be part of the skill set
of not only other scientists but of everyone else.
Ubiquitous computing is to today as computational
thinking is to tomorrow. Ubiquitous computing was
yesterday’s dream that became today’s reality; com-
putational thinking is tomorrow’s reality.
WHAT IT IS, AND ISN’T
Computer science is the study of computation—
what can be computed and how to compute it.
Computational thinking thus has the following
characteristics:
 3 COMPUTATIONAL THINKING
Conceptualizing, not programming. Computer sci-
ence is not computer programming. Thinking
like a computer scientist means more than being
able to program a computer. It requires thinking
at multiple levels of abstraction;
Fundamental, not rote skill. A fundamental skill is
something every human being must know to
function in modern society. Rote means a
mechanical routine. Ironically, not until computer
science solves the AI Grand Challenge of making
computers think like humans will thinking be
rote;
A way that humans, not computers, think. Computa-
tional thinking is a way humans solve problems;
it is not trying to get humans to think like com-
puters. Computers are dull and boring; humans
are clever and imaginative. We humans make
computers exciting. Equipped with computing
devices, we use our cleverness to tackle problems
we would not dare take on before the age of com-
puting and build systems with functionality lim-
ited only by our imaginations;
Complements and combines mathematical and engi-
neering thinking. Computer science inherently
draws on mathematical thinking, given that, like
all sciences, its formal foundations rest on mathe-
matics. Computer science inherently draws on
engineering thinking, given that we build systems
that interact with the real world. The constraints
of the underlying computing device force com-
puter scientists to think computationally, not just
mathematically. Being free to build virtual worlds
enables us to engineer systems beyond the physi-
cal world;
Ideas, not artifacts. It’s not just the software and
hardware artifacts we produce that will be physi-
cally present everywhere and touch our lives all
the time, it will be the computational concepts
we use to approach and solve problems, manage
our daily lives, and communicate and interact
with other people; and
For everyone, everywhere. Computational thinking
will be a reality when it is so integral to human
endeavors it disappears as an explicit philosophy.
Jeannette Wing, “Computational Thinking.” CACM, March 2006. Copyright Association for Computing Machinery. Article provided by Library Reserves for
private study, scholarship, or research. Do not redistribute. This article can also be found at http://www.cs.cmu.edu/~wing/publications/Wing06.pdf.
12
Many people equate computer science with com-
puter programming. Some parents see only a narrow
range of job opportunities for their children who
major in computer science. Many people think the
fundamental research in computer science is done
and that only the engineering remains. Computa-
tional thinking is a grand vision to guide computer
science educators, researchers, and practitioners as we
act to change society’s image of the field. We espe-
cially need to reach the pre-college audience, includ-
ing teachers, parents, and students, sending them
two main messages:
Intellectually challenging and engaging scientific prob-
lems remain to be understood and solved. The prob-
lem domain and solution domain are limited only
by our own curiosity and creativity; and
One can major in computer science and do anything.
One can major in English or mathematics and go
on to a multitude of different careers. Ditto com-
puter science. One can major in computer science
and go on to a career in medicine, law, business,
politics, any type of science or engineering, and
even the arts.
Professors of computer science should teach a
course called “Ways to Think Like a Computer Sci-
entist” to college freshmen, making it available to
non-majors, not just to computer science majors. We
should expose pre-college students to computational
methods and models. Rather than bemoan the
decline of interest in computer science or the decline
in funding for research in computer science, we
should look to inspire the public’s interest in the
intellectual adventure of the field. We’ll thus spread
the joy, awe, and power of computer science, aiming
to make computational thinking commonplace. c
Jeannette M. Wing (wing@cs.cmu.edu) is the President’s
Professor of Computer Science in and head of the Computer Science
Department at Carnegie Mellon University, Pittsburgh, PA.
© 2006 ACM 0001-0782/06/0300 $5.00
COMMUNICATIONS OF THE ACM March 2006/Vol. 49, No. 3 35
 44 THE MANY FACETS OF NATURAL COMPUTING 13

review articles
DOI :1 0 .1 1 45/ 140 0 1 81 .140 0 20 0
applications, as well as biology, chem-
Natural computing builds a bridge between istry, and physics experimental labora-
tory research.
computer science and natural sciences. In this review we describe com-
puting paradigms abstracted from
BY LILA KARI AND GRZEGORZ ROZENBERG natural phenomena as diverse as
self-reproduction, the functioning of

The Many
the brain, Darwinian evolution, group
behavior, the immune system, the char-
acteristics of life, cell membranes, and
morphogenesis. These paradigms can

Facets of
be implemented either on traditional
electronic hardware or on alternative
physical media such as biomolecular
(DNA, RNA) computing, or trapped-ion

Natural
quantum computing devices. Dually,
we describe several natural processes
that can be viewed as information pro-
cessing, such as gene regulatory net-

Computing
works, protein-protein interaction net-
works, biological transport networks,
and gene assembly in unicellular or-
ganisms. In the same vein, we list ef-
forts to understand biological systems
by engineering semi-synthetic organ-
isms, and to understand the universe
from the point of view of information
processing.
This review was written with the ex-
“Biology and computer science—life and pectation that the reader is a computer
scientist with limited knowledge of
computation—are related. I am confident that natural sciences, and it avoids dwell-
at their interface great discoveries await those ing on the minute details of various
natural phenomena. Thus, rather than
who seek them.” being overwhelmed by particulars, it is
— Leonard Adleman, our hope that readers see this article
Scientific American, Aug. 1998 as simply a window into the profound
relationship that exists between nature
and computation.
Natural computing is the field of research that There is information processing in
nature, and the natural sciences are al-
investigates models and computational techniques ready adapting by incorporating tools
inspired by nature and, dually, attempts to under- and concepts from computer science
stand the world around us in terms of information at a rapid pace. Conversely, a closer
look at nature from the point of view
processing. It is a highly interdisciplinary field that of information processing can and will
connects the natural sciences with computing
science, both at the level of information technology The vivid images peppered throughout this
story offer glimpses of what can happen when
and at the level of fundamental research.33 nature, art, and computer science join forces.

As a matter of fact, natural computing areas and While not directly referenced in this article,
these images serve to offer readers some
topics come in many flavors, including pure startling perspectives of nature up close as
only technology can provide.
theoretical research, algorithms and software
72 C O M M UN I C AT I ON S O F T H E AC M | O C TO B E R 2 00 8 | VO L . 51 | NO. 1 0

Kari and Rozenberg, “The Many Facets of Natural Computing.” CACM, October 2008. Copyright Association for Computing Machinery. Article provided by
Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://www.csd.uwo.ca/~lila/p72.pdf.
 4 THE MANY FACETS OF NATURAL COMPUTING
change what we mean by computation.
Our invitation to you, fellow computer
scientists, is to take part in the uncov-
ering of this wondrous connection.a
Nature as Inspiration
Among the oldest examples of nature-
inspired models of computation are
the cellular automata conceived by
Ulam and von Neumann in the 1940s.
a A few words are in order about the organization
of this article. The classifications and labels
we use for various fields of research are purely
for the purpose of organizing the discourse. In
reality, far from being clear-cut, many of the
fields of research mentioned here overlap, or
fit under more than one category. The general
audience for whom this article is intended, our
respective fields of expertise, and especially
the limited space available for this review af-
fected both the depth and breadth of our expo-
sition. In particular, we did not discuss some
fields of research that have large overlaps with
natural computing, such as bioinformatics,
computational molecular biology, and their
roles in, for example, genomics and proteom-
ics. In addition, our explanations of various
aspects, themes, and paradigms had to be
necessarily oversimplified. As well, the space
we devoted to various fields and topics was
influenced by several factors and, as such, has
no relation to the respective importance of the
field or the relative size of the body of research
in that field.
Kari and Rozenberg, “The Many Facets of Natural Computing.” CACM, October 2008. Copyright Association for Computing Machinery. Article provided by
Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://www.csd.uwo.ca/~lila/p72.pdf.
John von Neumann, who was trained
in both mathematics and chemistry,
investigated cellular automata as a
framework for the understanding of
the behavior of complex systems. In
particular, he believed that self-repro-
duction was a feature essential to both
biological organisms and computers.40
A cellular automaton is a dynami-
cal system consisting of a regular grid
of cells, in which space and time are
discrete. Each of the cells can be in one
of a finite number of states. Each cell
changes its state according to a list of
given transition rules that determine
its future state, based on its current
state and the current states of some of
its neighbors. The entire grid of cells
updates its configuration synchro-
nously according to the a priori given
transition rules.
Cellular automata have been ap-
plied to the study of phenomena as
diverse as communication, computa-
tion, construction, growth, reproduc-
tion, competition, and evolution. One
of the best known examples of cellular
automata—the “game of life” invented
by Conway—was shown to be compu-
tationally universal. Cellular automata
have been extensively studied as an al-
O C TOBE R 20 0 8 | VO L. 51 | NO. 10 | COM M UN I CATI ONS OF THE ACM
14
Neri Oxman, an architect
and researcher currently
working for her Ph.D. in
design and computation
at MIT, formed an
interdisciplinary
research initiative
called Materialecology
that undertakes
design research in the
intersection between
architecture, engineering,
computation, biology
and ecology. Here, she
illustrates how plants
often grow in fashion to
maximize the surface
area of their branching
geometries while
maintaining structural
support. This work was
done in collaboration
with W. Craig Carter,
a professor in MIT’s
Department of
Material Science and
Engineering. For more
images, see http://www.
materialecology.com/.
ternative explanation to the phenome-
non of emergence of complexity in the
natural world, and used, among others,
for modeling in physics and biology.
In parallel to early comparisons39
between computing machines and the
human nervous system, McCulloch and
Pitts proposed the first model of artifi-
cial neurons. This research eventually
gave rise to the field of neural computa-
tion, and it also had a profound influ-
ence on the foundations of automata
theory. The goal of neural computa-
tion was twofold. On one hand, it was
hoped that it would help unravel the
structure of computation in nervous
systems of living organisms (How does
the brain work?). On the other hand, it
was predicted that, by using the princi-
ples of how the human brain process-
es information, neural computation
would yield significant computational
advances (How can we build an intel-
ligent computer?). The first goal has
been pursued mainly within the neu-
rosciences under the name of brain
theory or computational neuroscience,
while the quest for the second goal has
become mainly a computer science
discipline known as artificial neural
networks or simply neural networks.5
73
 4 THE MANY FACETS OF NATURAL COMPUTING
review articles
An artificial neural network consists
of interconnected artificial neurons.31
Modeled after the natural neurons,
each artificial neuron A has n real-val-
ued inputs, x1, x2, …, xn, and it computes
its own primitive function fA as follows.
Usually, the inputs have associated
weights, w1, w2, …, wn. Upon receiving
the n inputs, the artificial neuron A
produces the output fA(w1x1 + w2x2 + …
+ wnxn). An artificial neural network is
a network of such neurons, and thus
a network of their respective primitive
functions. Some neurons are selected to
be the output neurons, and the network
function is a vectorial function that, for
n input values, associates the outputs of
the m output neurons. Note that differ-
ent selections of the weights produce
different network functions for the
same inputs. Based on given input-out-
put pairs, the network can “learn” the
weights w1, …, wn. Thus, there are three
important features of any artificial neu-
ral network: the primitive function of
each neuron, the topology of the net-
work, and the learning algorithm used
to find the weights of the network. One
of the many examples of such learning
algorithms is the “backwards propaga-
tion of errors.” Back-propagation is a
supervised learning method by which
the weights of the connections in the
network are repeatedly adjusted so as
to minimize the difference between the
actual output vector of the net and the
desired output vector. Artificial neural
networks have proved to be a fruitful
paradigm, leading to successful novel
applications in both new and estab-
lished application areas.
74 C O M M UN I C AT I O N S O F T H E AC M
Kari and Rozenberg, “The Many Facets of Natural Computing.” CACM, October 2008. Copyright Association for Computing Machinery. Article provided by
Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://www.csd.uwo.ca/~lila/p72.pdf.
While Turing and von Neumann
dreamed of understanding the brain,
and possibly designing an intelligent
computer that works like the brain, evo-
lutionary computation6 emerged as an-
other computation paradigm that drew
its inspiration from a completely dif-
ferent part of biology: Darwinian evolu-
tion. Rather than emulating features of
a single biological organism, evolution-
ary computation draws its inspiration
from the dynamics of an entire species
of organisms. An artificial evolution-
ary system is a computational system
based on the notion of simulated evo-
lution. It features a constant- or vari-
able-size population of individuals, a
fitness criterion according to which the
individuals of the population are being
From Archimorph,
where work is
continuing on their
L-System and
Evolutionary
Algorithm, including
new images of
L-Systems growths
as well as diagrams
explaining the process
of the overall design.
For more images,
see archimorph.
wordpress.com/.
evaluated, and genetically inspired op-
erators that produce the next genera-
tion from the current one. In an evolu-
tionary system, the initial population of
individuals is generated at random or
heuristically. At each evolutionary step,
the individuals are evaluated according
to a given fitness function. To form the
next generation, offspring are first gen-
erated from selected individuals by us-
ing operators such as mutation of a par-
ent, or recombination of pairs or larger
subsets of parents. The choice of par-
ents for recombination can be guided
by a fitness-based selection operator,
thus reflecting the biological principle
of mate selection. Secondly, individu-
als of the next generation are selected
from the set of newly created offspring,
sometimes also including the old par-
ents, according to their fitness—a pro-
cess reflecting the biological concept of
| O C TO B E R 2 00 8 | VO L . 51 | NO. 10
15
environmental selection.
Evolutionary systems have first been
viewed as optimization processes in the
1930s. The basic idea of viewing evolu-
tion as a computational process gained
momentum in the 1960s, and evolved
along three main branches.13 Evolution
strategies use evolutionary processes
to solve parameter optimization prob-
lems, and are today used for real-val-
ued as well as discrete and mixed types
of parameters. Evolutionary program-
ming originally aimed at achieving the
goals of artificial intelligence via evo-
lutionary techniques, namely by evolv-
ing populations of intelligent agents
modeled, for example, as finite-state
machines. Today, these algorithms
are also often used for real-valued pa-
rameter optimization problems. Ge-
netic algorithms originally featured a
population of individuals encoded as
fixed-length bit strings, wherein muta-
tions consisted of bit-flips according
to a typically small, uniform mutation
rate, the recombination of two parents
consisted of a cut-and-paste of a prefix
of one parent with a suffix of the other,
and the fitness function was problem-
dependent. If the initial individuals
were to encode possible solutions to
a given problem, and the fitness func-
tion were designed to measure the op-
timality of a candidate solution, then
such a system would, in time, evolve
to produce a near-optimal solution to
the initial problem. Today, genetic al-
gorithms are also modified heavily for
applications to real-valued parameter
optimization problems as well as many
types of combinatorial tasks such as,
for example, permutation-based prob-
lems. As another application, if the
individuals were computer programs,
then the genetic algorithm technique
would result in “the fittest” computer
programs, as is the goal of genetic pro-
gramming.22
Cellular automata, neural compu-
tation, and evolutionary computation
are the most established “classical”
areas of natural computing. Several
other bio-inspired paradigms emerged
more recently, among them swarm in-
telligence, artificial immune systems,
artificial life, membrane computing,
and amorphous computing.
A computational paradigm strad-
dling at times evolutionary computa-
tion and neural computation is swarm
 4 THE MANY FACETS OF NATURAL COMPUTING
intelligence.16 A swarm is a group of mo-
bile biological organisms (such as bac-
teria, ants, termites, bees, spiders, fish,
birds) wherein each individual com-
municates with others either directly or
indirectly by acting on its local environ-
ment. These interactions contribute to
distributive collective problem solving.
Swarm intelligence, sometimes re-
ferred to as collective intelligence, is de-
fined as the problem-solving behavior
that emerges from the interaction of
such a collection of individual agents.
For example, in research simulating
flocking behavior, each individual was
endowed with three simple possible
behaviors: to act as to avoid collision,
to match velocity with neighbors, and
to stay close to nearby flock mates. The
simulations showed that flocking was
an emergent behavior that arose from
the interaction of these simple rules.
Particle swarm optimization was in-
troduced as a new approach to optimi-
zation that had developed from simple
models of social interactions, as well as
of flocking behavior in birds and other
organisms. A particle swarm optimiza-
tion algorithm starts with a swarm of
“particles,” each representing a poten-
tial solution to a problem, similar to
the population of individuals in evolu-
tionary computation.
Particles move through a multidi-
mensional search space and their po-
sitions are updated according to their
own experience and that of their neigh-
bors, by adding “velocity” to their cur-
rent positions. The velocity of a particle
depends on its previous velocity (the
“inertia” component), the tendency
towards the past personal best posi-
tion (the cognitive, “nostalgia” compo-
nent), and the move toward a global or
local neighborhood best (the “social”
component). The cumulative effect is
that each particle converges towards a
point between the global best and its
personal best. Particle Swarm Optimi-
zation algorithms have been used to
solve various optimization problems,
and have been applied to unsupervised
learning, game learning, scheduling
and planning applications, and design
applications.
Ant algorithms were introduced to
model the foraging behavior of ant
colonies. In finding the best path be-
tween their nest and a source of food,
ants rely on indirect communication
Kari and Rozenberg, “The Many Facets of Natural Computing.” CACM, October 2008. Copyright Association for Computing Machinery. Article provided by
Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://www.csd.uwo.ca/~lila/p72.pdf.
16
review articles
by laying a pheromone trail on the way
back to the nest if they found food, and
following the concentration of phero-
mones in the environment if they are
A closer look looking for food. This foraging behav-
ior has inspired a large number of ant
at nature from algorithms used to solve mainly com-
the point of view
binatorial optimization problems de-
fined over discrete search spaces.
of information Artificial immune systems are compu-
tational systems devised starting in the
processing can late 1980s and early 1990s as computa-
and will change tionally interesting abstractions of the
natural immune system of biological
what we mean by organisms. Viewed as an information
computation. Our processing system, the immune sys-
tem performs many complex computa-
invitation to you, tions in a highly parallel and distribut-
fellow computer ed fashion.11 It uses learning, memory,
associative retrieval, and other mecha-
scientists, is to nisms to solve recognition and classi-
fication problems such as distinction
take part in the between self and nonself cells, and
uncovering of neutralization of nonself pathogenic
agents. Indeed, the natural immune
this wondrous system has sometimes been called the
connection. “second brain” because of its powerful
information processing capabilities.
The natural immune system’s main
function is to protect our bodies against
the constant attack of external patho-
gens (viruses, bacteria, fungi, and para-
sites). The main role of the immune
system is to recognize cells in the body
and categorize them as self or nonself.12
There are two parts of the immune sys-
tem: innate (non-specific) and adaptive
(acquired). The cells of the innate im-
mune system are immediately avail-
able to combat against a wide variety
of antigens, without requiring previous
exposure to them. These cells possess
the ability of ingesting and digesting
several “known” pathogens. In con-
trast, the adaptive immune response
is the antibody production in response
to a specific new infectious agent. Our
body maintains a large “combinatorial
database” of immune cells that circu-
late throughout the body. When a for-
eign antigen invades the body, only a
few of these immune cells can detect
the invaders and physically bind to
them. This detection triggers the pri-
mary immune response: the genera-
tion of a large population of cells that
produce matching antibodies that aid
in the destruction or neutralization of
the antigen. The immune system also
retains some of these specific-anti-
O C TO BE R 2 0 0 8 | VOL . 51 | N O. 10 | COM M UN I CATI ON S OF TH E ACM 75
 4 THE MANY FACETS OF NATURAL COMPUTING
review articles
body-producing cells in immunologi-
cal memory, so that any subsequent
exposure to a similar antigen can lead
to a rapid, and thus more effective, im-
mune response (secondary response).
The computational aspects of the
immune system, such as distinguish-
ing of self from nonself, feature extrac-
tion, learning, memory, self-regulation,
and fault tolerance, have been exploit-
ed in the design of artificial immune
systems that have been successfully
used in applications. The applications
are varied and include computer virus
detection, anomaly detection in a time
series of data, fault diagnosis, pattern
recognition, machine learning, bioin-
formatics, optimization, robotics, and
control. Recent research in immunol-
ogy departs from the self-nonself dis-
crimination model to develop what is
known as the “danger theory,” wherein
it is believed that the immune system
differentiates between dangerous and
non-dangerous entities, regardless of
whether they belong to self or to non-
self. These ideas have started to be ex-
ploited in artificial immune systems in
the context of computer security.
While artificial immune systems
(a.k.a. immunological computation,
immunocomputing) constitute an ex-
ample of a computational paradigm
inspired by a very specific subsystem
of a biological organism, artificial life
takes the opposite approach. Artificial
life (ALife) attempts to understand the
very essence of what it means to be
alive by building ab initio, within in
silico computers and other “artificial”
media, artificial systems that exhibit
properties normally associated only
with living organisms.24 Lindenmayer
systems (L-systems), introduced in 1968,
can be considered as an early example
of artificial life.
L-systems are parallel rewriting sys-
tems that, starting with an initial word,
proceed by applying rewriting rules in
parallel to all the letters of the word,
and thus generate new words.34 They
have been most famously used to mod-
el plant growth and development,29 but
also for modeling the morphology of
other organisms.
Building on the ideas of evolution-
ary computation, other pioneers of ar-
tificial life experimented with evolving
populations of “artificial creatures”
in simulated environments.9 One ex-
76 CO MM U N I CATI O NS O F T H E AC M
Kari and Rozenberg, “The Many Facets of Natural Computing.” CACM, October 2008. Copyright Association for Computing Machinery. Article provided by
Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://www.csd.uwo.ca/~lila/p72.pdf.
17
ample was the design36 of evolving vir-
tual block creatures that were selected
for their ability to swim (or walk, or
jump), and that competed for a com-
While artificial mon resource (controlling a cube) in
a physically simulated world endowed
immune systems with realistic features such as kinemat-
ics, dynamics, gravity, collisions, and
constitute an friction. The result was that creatures
example of a evolved which would extend arms to-
wards the cube, while others would
computational crawl or roll to reach it, and some even
paradigm inspired developed legs that they used to walk
towards the cube. These ideas were
by a very specific taken one step further25 by combining
subsystem of the computational and experimental
approaches, and using rapid manufac-
a biological turing technology to fabricate physical
organism, artificial robots that were materializations of
their virtually evolved computational
life attempts to counterparts. In spite of the simplic-
ity of the task at hand (horizontal lo-
understand the very comotion), surprisingly different and
essence of what it complex robots evolved: many of them
exhibited symmetry, some moved side-
means to be alive. ways in a crab-like fashion, and others
crawled on two evolved limbs. This
marked the emergence of mechanical
artificial life, while the nascent field
of synthetic biology, discussed later,
explores a biological implementation
of similar ideas. At the same time,
the field of Artificial Life continues to
explore directions such as artificial
chemistry (abstractions of natural mo-
lecular processes), as well as tradition-
ally biological phenomena in artificial
systems, ranging from computational
processes such as co-evolutionary ad-
aptation and development, to physical
processes such as growth, self-replica-
tion, and self-repair.
Membrane computing investigates
computing models abstracted from
the structure and the functioning of
living cells, as well as from the way the
cells are organized in tissues or higher
order structures.26 More specifically,
the feature of the living cells that is
abstracted by membrane computing
is their compartmentalized internal
structure effected by membranes. A
generic membrane system is essen-
tially a nested hierarchical structure
of cell-like compartments or regions,
delimited by “membranes.” The entire
system is enclosed in an external mem-
brane, called the skin membrane, and
everything outside the skin membrane
is considered to be the environment.
| O C TO B E R 2 00 8 | VO L . 5 1 | N O. 1 0
 4 THE MANY FACETS OF NATURAL COMPUTING
Each membrane-enveloped region con-
tains objects and transformation rules
which modify these objects, as well as
specify whether they will be transferred
outside or stay inside the region. The
transfer thus provides for communica-
tion between regions. Various formal
mechanisms were developed that re-
flect the selective manner in which bio-
logical membranes allow molecules to
pass through them.
Another biologically inspired fea-
ture of membrane systems as math-
ematical constructs is the fact that,
instead of dealing with sets of objects,
one uses multisets wherein one keeps
track of the multiplicity of each ob-
ject. The computational behavior of a
membrane system starts with an initial
input configuration and proceeds in a
maximally parallel manner by the non-
deterministic choice of application
of the transformation rules, as well as
of the objects to which they are to be
applied. The output of the computa-
tion is then collected from an a priori
determined output membrane. Next
to the basic features indicated previ-
ously, many alternatives of membrane
systems have been considered, among
them ones that allow for membranes to
be dissolved and created. Typical appli-
cations of membrane systems include
biology (modeling photosynthesis and
certain signaling pathways, quorum
sensing in bacteria, modeling cell-me-
diated immunity), computer science
(computer graphics, public-key cryp-
tography, approximation and sorting
algorithms, and solving computation-
ally hard problems), and linguistics.
Amorphous computing is a paradigm
that draws inspiration from the de-
velopment of form (morphogenesis)
in biological organisms, wherein in-
teractions of cells guided by a genet-
ic program give rise to well-defined
shapes and functional structures.
Analogously, an amorphous comput-
ing medium comprises a multitude of
irregularly placed, asynchronous, lo-
cally interacting computing elements.1
These identically programmed “com-
putational particles” communicate
only with particles situated within a
small given radius, and may give rise
to certain shapes and patterns such as,
for example, any pre-specified planar
graph. The goal of amorphous com-
puting is to engineer specified coher-
Kari and Rozenberg, “The Many Facets of Natural Computing.” CACM, October 2008. Copyright Association for Computing Machinery. Article provided by
Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://www.csd.uwo.ca/~lila/p72.pdf.
ent computational behaviors from the
interaction of large quantities of such
unreliable computational particles in-
terconnected in unknown, irregular,
and time-varying ways. At the same
time, the emphasis is on devising new
programming abstractions that would
work well for amorphous computing
environments. Amorphous computing
has been used both as a programming
paradigm using traditional hardware,
and as the basis for “cellular comput-
ing,” discussed later, under the topics
synthetic biology, and computation in
living cells.
Nature as Implementation
Substrate
In the preceding section we saw cel-
lular automata inspired by self-repro-
duction, neural computation by the
functioning of the brain, evolutionary
computation by the Darwinian evolu-
tion of species, swarm intelligence by
the behavior of groups of organisms,
artificial immune systems by the natu-
ral immune system, artificial life by
properties of life in general, membrane
computing by the compartmentalized
organization of the cells, and amor-
phous computing by morphogenesis.
All these are computational techniques
that, while inspired by nature, have
been implemented until now mostly
on traditional electronic hardware.
An entirely distinct category is that of
computing paradigms that use a radi-
OC TO BER 20 0 8 | VOL . 51 | N O. 10 | COM M U NI CATI ON S OF TH E AC M
18
review articles
cally different type of “hardware.” This
category includes molecular comput-
ing and quantum computing.b
Molecular computing (known also
as biomolecular computing, biocom-
puting, biochemical computing, DNA
computing), is based on the idea that
data can be encoded as biomolecules —
such as DNA strands — and molecular
biology tools can be used to transform
this data to perform, for example, arith-
metic or logic operations. The birth of
this field was the 1994 breakthrough
experiment by Leonard Adleman who
solved a small instance of the Hamil-
tonian Path Problem solely by manipu-
lating DNA strands in test tubes.2
DNA (deoxyribonucleic acid) is a
linear chain made up of four different
types of nucleotides, each consisting
of a base (Adenine, Cytosine, Guanine,
or Thymine) and a sugar-phosphate
unit. The sugar-phosphate units are
linked together by covalent bonds to
b There are several research areas that, because
of the limited space, we could not discuss
here. Thus, for example, non-classical, uncon-
ventional computation38 focuses on carefully
examining and possibly breaking the classi-
cal (Turing, von Neumann) computation as-
sumptions, and developing a more general
science of computation. A substantial part of
this research is concerned with implementing
computation on new physical substrates, ex-
ploiting in this way computational properties
of various physical, chemical, and biological
media. A majority of this research is entwined
with, and motivated by, natural computing.
McGill University’s
Laboratory for Natural
and Simulated Cognition
(LNSC) investigates
human cognition
through a combination
of psychological
and computational
approaches. Using the
Cascade-correlation
algorithm, LNSC
researchers created a
program that outputs
a 2D display of random
output values of neural
networks. The results
are sometimes quite
phenomenal and
artistic. For more, see
www.psych.mcgill.ca/
labs/lnsc/.
77
 4 THE MANY FACETS OF NATURAL COMPUTING
review articles
form the backbone of the DNA single
strand. Since nucleotides may differ
only by their bases, a DNA strand can
be viewed as simply a word over the
four-letter alphabet {A,C,G,T}. A DNA
single strand has an orientation, with
one end known as the 5a end, and the
other as the 3a end, based on their
chemical properties. By convention,
a word over the DNA alphabet repre-
sents the corresponding DNA single
strand in the 5a to 3a orientation, that
is, the word GGTTTTT stands for the
DNA single strand 5a– GGTTTTT –3a. A
crucial feature of DNA single strands is
their Watson-Crick complementarity:
A is complementary to T, G is comple-
mentary to C, and two complementary
DNA single strands with opposite ori-
entation bind to each other by hydro-
gen bonds between their individual
bases. In so doing, they form a stable
DNA double strand resembling a heli-
cal ladder, with the backbones at the
outside and the bound pairs of bases
lying inside. For example, the DNA sin-
gle strand 5a– AAAAACC – 3a will bind
to the DNA single strand 5a– GGTTTTT
– 3a to form the 7 base-pair-long (7bp)
double strand
5a − AAAAACC − 3a
3a − TTTTTGG − 5a
Another molecule that can be used
for computation is ribonucleic acid,
78 CO M M U N I C AT I ON S OF T H E AC M
Kari and Rozenberg, “The Many Facets of Natural Computing.” CACM, October 2008. Copyright Association for Computing Machinery. Article provided by
Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://www.csd.uwo.ca/~lila/p72.pdf.
RNA. While similar to DNA, RNA dif-
fers in three main aspects: RNA is usu-
ally single-stranded while DNA is usu-
ally double-stranded, RNA nucleotides
contain the sugar ribose, while DNA
nucleotides contain the sugar deoxyri-
bose, and in RNA the nucleotide Ura-
cil, U, substitutes for Thymine, which
is present in DNA.
There are many possible DNA bio-
operations that one can use for com-
putations,21 such as: cut-and-paste
operations achievable by enzymes, syn-
thesizing desired DNA strands up to a
certain length, making exponentially
many copies of a DNA strand, and read-
ing out the sequence of a DNA strand.
These bio-operations and the Watson-
Crick complementary binding have
all been used to control DNA compu-
tations and DNA robotic operations.
While initial experiments solved simple
instances of computational problems,
more recent experiments tackled suc-
cessfully sophisticated computational
problems, such as a 20-variable in-
stance of the 3-Satisfiability-Problem.
The efforts toward building an auton-
omous molecular computer include
implementations of computational
state transitions with biomolecules,
and a DNA implementation of a finite
automaton with potential applications
to the design of smart drugs.
More importantly, since 1994, re-
search in molecular computing has
| O C TO B E R 2 00 8 | VO L. 5 1 | NO. 1 0
19
gained several new dimensions. One
of the most significant achievements
of molecular computing has been its
contribution to the massive stream of
research in nanosciences, by providing
computational insights into a number
of fundamental issues. Perhaps the
most notable is its contribution to the
understanding of self-assembly, which
is among the key concepts in nanosci-
ences.30 Recent experimental research
into programmable molecular-scale
devices has produced impressive self-
assembled DNA nanostructures35 such
as cubes, octahedra, Sierpinski trian-
gles,32 DNA origami, or intricate nano-
structures that achieve computation
Paul W.K. Rothemund,
a senior research
associate at California
Institute of Technology,
has developed a method
of creating nanoscale
shapes and patterns
using DNA. The smiley
faces are actually
giant DNA complexes
called “scaffolded DNA
origami.” Rothemund
notes that while the
smiley face shape may
appear silly, there is
serious science behind
it. He hopes to use this
DNA origami (and other
DNA nanotechnologies)
to build smaller, faster
computers and devices.
For more on his work,
visit http://www.dna.
caltech.edu/~pwkr/.
such as binary counting, or bit-wise
cumulative XOR. Other experiments
include the construction of DNA-based
logic circuits, and ribozymes that can
be used to perform logical operations
and simple computations. In addition,
an array of ingenious DNA nanoma-
chines8 were built with potential uses
to nanofabrication, engineering, and
computation: molecular switches that
can be driven between two conforma-
tions, DNA “tweezers,” DNA “walkers”
that can be moved along a track, and
autonomous molecular motors.
A significant amount of research in
molecular computing has been dedi-
cated to the study of theoretical models
of DNA computation and their proper-
ties. The model of DNA computing in-
 4 THE MANY FACETS OF NATURAL COMPUTING
troduced by Head, based on splicing (a
combination of cut-and-paste opera-
tions achievable by enzymes), predates
the experimental proof-of-principle of
DNA computing by almost 10 years.
Subsequently, studies on the compu-
tational power of such models proved
that various subsets of bio-operations
can achieve the computational power
of a Turing machine, showing thus that
molecular computers are in principle
possible.27 Overall, molecular comput-
ing has created many novel theoretical
questions, and has considerably en-
riched the theory of computation.
Quantum Computing is another par-
adigm that uses an alternative “hard-
ware” for performing computations.19
Already in 1980 Benioff introduced
simulations of classical Turing Ma-
chines on quantum mechanical sys-
tems. However the idea of a quantum
computer that would run according
to the laws of quantum physics and
operate exponentially faster than a
deterministic electronic computer to
simulate physics, was first suggested
by Feynman in 1982. Subsequently,
Deutsch introduced a formal model
of quantum computing using a Turing
machine formalism, and described a
universal quantum computer.
A quantum computer uses distinc-
tively quantum mechanical phenom-
ena, such as superposition and en-
tanglement, to perform operations on
data stored as quantum bits (qubits).
A qubit can hold a 1, a 0, or a quantum
superposition of these. A quantum
computer operates by manipulating
those qubits with quantum logic gates.
The notion of information is different
when studied at the quantum level. For
instance, quantum information cannot
be measured reliably, and any attempt
at measuring it entails an unavoidable
and irreversible disturbance.
The 1980s saw an abundance of
research in quantum information
processing, such as applications to
quantum cryptography which, unlike
its classical counterpart, is not usu-
ally based on the complexity of com-
putation but on the special properties
of quantum information. Recently an
open air experiment was reported in
quantum cryptography (not involv-
ing optical cable) over a distance of
144km, conducted between two Ca-
nary islands.
Kari and Rozenberg, “The Many Facets of Natural Computing.” CACM, October 2008. Copyright Association for Computing Machinery. Article provided by
Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://www.csd.uwo.ca/~lila/p72.pdf.
20
review articles
The theoretical results that cata-
pulted quantum computing to the
forefront of computing research were
Shor’s quantum algorithms for factor-
It is indeed ing integers and extracting discrete log-
arithms in polynomial time, obtained
believed that one in 1994 — the same year that saw the
first DNA computing experiment by
of the possible Adleman. A problem where quantum
contributions of computers were shown to have a qua-
dratic time advantage when compared
computer science to classical computers is quantum da-
to biology could be tabase search that can be solved by Gro-
ver’s algorithm. Possible applications
the development of of Shor’s algorithm include breaking
a suitable language RSA exponentially faster than an elec-
tronic computer. This joined other ex-
to accurately and citing applications, such as quantum
succinctly describe, teleportation (a technique that trans-
fers a quantum state, but not matter
and reason about, or energy, to an arbitrarily distant loca-
tion), in sustaining the general interest
biological concepts in quantum information processing.
and phenomena. So far, the theory of quantum com-
puting has been far more developed
than the practice. Practical quantum
computations use a variety of imple-
mentation methods such as ion-traps,
superconductors, nuclear magnetic
resonance techniques, to name just a
few. To date, the largest quantum com-
puting experiment uses liquid state
nuclear magnetic resonance quantum
information processors that can oper-
ate on up to 12 qubits.
Nature as Computation
The preceding sections describe re-
search on the theory, applications and
experimental implementations of na-
ture-inspired computational models
and techniques. A dual direction of re-
search in natural computing is one in
which the main goal becomes under-
standing nature by viewing processes
that take place in nature as informa-
tion processing.
This dual aspect can be seen in sys-
tems biology, and especially in compu-
tational systems biology, wherein the
adjective “computational” has two
meanings. On one hand it means the
use of quantitative algorithms for com-
putations, or simulations that comple-
ment experiments in hypothesis gen-
eration and validation. On the other
hand, it means a qualitative approach
that investigates processes taking place
in cells through the prism of commu-
nications and interactions, and thus of
O C TO BER 20 0 8 | VO L. 51 | N O. 1 0 | COMM U NI CAT ION S OF TH E ACM 79
 4 THE MANY FACETS OF NATURAL COMPUTING
review articles
computations. We shall herein address
mostly the second aspect, whereby sys-
tems biology aims to understand the
complex interactions in biological sys-
tems by using an integrative as opposed
to a reductionist approach. The re-
ductionist approach to biology tries to
identify all the individual components
of functional processes that take place
in an organism, in such a way that the
processes and the interactions between
the components can be understood. In
contrast, systems biology takes a sys-
temic approach in focusing instead on
the interaction networks themselves,
and on the properties of the biological
systems that arise because of these in-
teraction networks. Hence, for exam-
ple, at the cell level, scientific research
on organic components has focused
strongly on four different interdepen-
dent interaction networks, based on
four different “biochemical toolkits:”
nucleic acids (DNA and RNA), proteins,
lipids, carbohydrates, and their build-
ing blocks (see Cardelli,10 whose cat-
egorization we follow here).
The genome consists of DNA se-
quences, some of which are genes that
can be transcribed into messenger
RNA (mRNA), and then translated into
proteins according to the genetic code
that maps 3-letter DNA segments into
amino acids. A protein is a sequence
over the 20-letter alphabet of amino ac-
ids. Each gene is associated with other
DNA segments (promoters, enhancers,
or silencers) that act as binding sites
for proteins that activate or repress
the gene’s transcription. Genes inter-
act with each other indirectly, either
through their gene products (mRNA,
proteins), which can act as transcrip-
tion factors to regulate gene transcrip-
tion — either as activators or repres-
sors — or through small RNA species
that directly regulate genes.
These gene-gene interactions, to-
gether with the genes’ interactions with
other substances in the cell, form the
most basic interaction network of an
organism, the gene regulatory network.
Gene regulatory networks perform
information processing tasks within
the cell, including the assembly and
maintenance of the other networks.
Research into modeling gene regu-
latory networks includes qualitative
models such as random and probabi-
listic Boolean networks, asynchronous
80 CO M M U N I CAT I O N S O F T H E AC M
Kari and Rozenberg, “The Many Facets of Natural Computing.” CACM, October 2008. Copyright Association for Computing Machinery. Article provided by
Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://www.csd.uwo.ca/~lila/p72.pdf.
21
automata, and network motifs.
Another point of view,20 is that the
entire genomic regulatory system can
be thought of as a computational sys-
As the natural tem, the “genomic computer.” Such a
perspective has the potential to yield
sciences are rapidly insights into both computation as hu-
mans historically designed it, and com-
absorbing ideas putation as it occurs in nature. There
of information are both similarities and significant
differences between the genomic com-
processing, and puter and an electronic computer. Both
the meaning of perform computations, the genomic
computer on a much larger scale. How-
computation is ever, in a genomic computer, molecular
changing as it transport and movement of ions through
electrochemical gradients replace wires,
embraces concepts causal coordination replaces imposed
from the natural temporal synchrony, changeable ar-
chitecture replaces rigid structure, and
sciences, we have communication channels are formed
on an as-needed basis. Both comput-
the rare privilege ers have a passive memory, but the ge-
to take part in nomic computer does not place it in an
a priori dedicated and rigidly defined
several such place; in addition, the genomic com-
metamorphoses. puter has a dynamic memory in which,
for example, trancriptional subcircuits
maintain given regulatory states. In a ge-
nomic computer robustness is achieved
by different means, such as by rigorous
selection: non (or poorly)-functional
processes are rapidly degraded by vari-
ous feedback mechanisms or, at the cell
level, non (or poorly)-functional cells are
rapidly killed by apoptosis, and, at the or-
ganism level, non (or poorly)-functional
organisms are rapidly out-competed
by more fit species. Finally, in the case
of a genomic computer, the distinction
between hardware and software breaks
down: the genomic DNA provides both
the hardware and the digital regulatory
code (software).
Proteins and their interactions form
another interaction network in a cell,
that of biochemical networks, which
perform all mechanical and metabolic
tasks inside a cell. Proteins are folded-
up strings of amino acids that take
three-dimensional shapes, with pos-
sible characteristic interaction sites ac-
cessible to other molecules. If the bind-
ing of interaction sites is energetically
favorable, two or more proteins may spe-
cifically bind to each other to form a
dynamic protein complex by a process
called complexation. A protein complex
may act as a catalyst by bringing togeth-
er other compounds and facilitating
| O C TO B E R 20 08 | VO L . 51 | N O. 1 0
 4 THE MANY FACETS OF NATURAL COMPUTING
chemical reactions between them. Pro-
teins may also chemically modify each
other by attaching or removing modify-
ing groups, such as phosphate groups,
at specific sites. Each such modification
may reveal new interaction surfaces.
There are tens of thousands of proteins
in a cell. At any given moment, each of
them has certain available binding sites
(which means that they can bind to oth-
er proteins, DNA, or membranes), and
each of them has modifying groups at
specific sites either present or absent.
Protein-protein interaction networks
are large and complex, and finding a
language to describe them is a difficult
task. Significant progress in this direc-
tion was made by the introduction of
Kohn-maps, a graphical notation that
resulted in succinct pictures depict-
ing molecular interactions. Other ap-
proaches include the textual bio-calcu-
lus, or the recent use of existing process
calculi (π-calculus), enriched with sto-
chastic features, as the language to de-
scribe chemical interactions.
Yet another biological interaction
network, and the last that we discuss
here, is that of transport networks medi-
ated by lipid membranes. Some lipids
can self-assemble into membranes and
contribute to the separation and trans-
Kari and Rozenberg, “The Many Facets of Natural Computing.” CACM, October 2008. Copyright Association for Computing Machinery. Article provided by
Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://www.csd.uwo.ca/~lila/p72.pdf.
port of substances, forming transport
networks. A biological membrane is
more than a container: it consists of a
lipid bilayer in which proteins and oth-
er molecules, such as glycolipids, are
embedded. The membrane structural
components, as well as the embedded
proteins or glycolipids, can travel along
this lipid bilayer. Proteins can inter-
act with free-floating molecules, and
some of these interactions trigger sig-
nal transduction pathways, leading to
gene transcription. Basic operations
of membranes include fusion of two
membranes into one, and fission of a
membrane into two. Other operations
involve transport, for example trans-
porting an object to an interior compart-
ment where it can be degraded. Formal-
isms that depict the transport networks
are few, and include membrane systems
described earlier, and brane calculi.
The gene regulatory networks, the
protein-protein interaction networks,
and the transport networks are all in-
terlinked and interdependent. Genes
code for proteins which, in turn, can
regulate the transcription of other
genes, membranes are separators but
also embed active proteins in their sur-
faces. Currently there is no single for-
mal general framework and notation
O CTO BER 2 0 08 | VO L. 51 | N O. 1 0 | COM M UN ICATION S OF THE ACM
22
review articles
Artist Jonathan
McCabe’s interests
include theories of
biological pattern
formation and evolution
and their application to
computer art. He writes
computer programs
that measure statistical
properties of images
for use in artificial
evolution of computer
art. For more, see www.
jonathanmccabe.com/.
able to describe all these networks and
their interactions. Process calculus has
been proposed for this purpose, but a
generally accepted common language
to describe these biological phenom-
ena is still to be developed and uni-
versally accepted. It is indeed believed
that one of the possible contributions
of computer science to biology could
be the development of a suitable lan-
guage to accurately and succinctly de-
scribe, and reason about, biological
concepts and phenomena.18
While systems biology studies
complex biological organisms as inte-
grated wholes, synthetic biology is an
effort to engineer artificial biological
systems from their constituent parts.
The mantra of synthetic biology is that
one can understand only what one can
construct. Thus, the main focus of syn-
thetic biology is to take parts of natu-
ral biological systems and use them to
build an artificial biological system for
the purpose of understanding natural
phenomena, or for a variety of possible
applications. In this sense, one can
make an analogy between synthetic
biology and computer engineering.3
The history of synthetic biology can
be arguably traced back to the discov-
ery in the 1960s, by Jacob and Monod,
81
 4 THE MANY FACETS OF NATURAL COMPUTING
review articles
of mathematical logic in gene regula-
tion. Early achievements in genetic
engineering using recombinant DNA
technology (the insertion, deletion,
or combination of different segments
of DNA strands) can be viewed as the
experimental precursors of today’s
synthetic biology, which now extends
these techniques to entire systems of
genes and gene products. One goal can
be constructing specific synthetic bio-
logical modules such as, for example,
pulse generator circuits that display a
transient response to variations in in-
put stimulus.
Advances in DNA synthesis of lon-
ger and longer strands of DNA are pav-
ing the way for the construction of
synthetic genomes with the purpose of
building an entirely artificial organism.
Progress includes the generation of a
5,386bp synthetic genome of a virus,
by rapid (14-day) assembly of chemi-
cally synthesized short DNA strands.37
Recently an announcement was made
of the near completion of the assem-
bly of an entire “minimal genome” of
a bacterium, Mycoplasma Genitalium.7
Smith and others indeed found about
100 dispensable genes that can be re-
moved individually from the original
genome. They hope to assemble a mini-
mal genome consisting of essential
genes only, that would be still viable but
shorter than the 528-gene, 580,000bp
genome of M.Genitalium. This human-
made genome could then be inserted
82 C OM M U N I C AT I O NS O F T H E AC M
Kari and Rozenberg, “The Many Facets of Natural Computing.” CACM, October 2008. Copyright Association for Computing Machinery. Article provided by
Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://www.csd.uwo.ca/~lila/p72.pdf.
into a Mycoplasma bacterium using a
technique wherein a whole genome can
be transplanted from one species into
another, such that the resulting prog-
eny is the same species as the donor ge-
nome. Counterbalancing objections to
assembling a semi-synthetic cell with-
out fully understanding its functioning,
the creation of a functionally and struc-
turally understood synthetic genome
was proposed,17 containing 151 genes
(113,000bp) that would produce all the
basic molecular machinery for protein
synthesis and DNA replication. A third
approach to create a human-made cell
is the one pursued by Szostak and oth-
ers, who would construct a single type of
RNA-like molecule capable of self-repli-
cating, possibly housed in a single lipid
membrane. Such molecules can be ob-
tained by guiding the rapid evolution of
an initial population of RNA-like mol-
ecules, by selecting for desired traits.
Lastly, another effort in synthetic
biology is toward engineering multi-
cellular systems by designing, for ex-
ample, cell-to-cell communication
modules that could be used to coordi-
nate living bacterial cell populations.
Research in synthetic biology faces
many challenges, some of them of an
information processing nature. There
arguably is a pressing need for stan-
dardization, modularization, and ab-
straction, to allow focusing on design
principles without reference to lower-
level details.15
European artist Leonel
Moura works with
AI and robotics. The
Swarm Paintings,
produced in 2001, were
the result of several
experiments with
an “Ant Algorithm”
where he tried to
apply virtual emergent
pheromone trails to
a real space pictorial
expression. In this case,
a computer running
an ant algorithm was
connected to a robotic
arm that “translated” in
pencil or brush strokes
the trails generated by
the artificial swarm of
ants. For more images,
see www.leonelmoura.
com/.
| O C TO B E R 2 00 8 | VO L . 5 1 | NO. 1 0
23
Besides systems biology that tries
to understand biological organisms as
networks of interactions, and synthet-
ic biology that seeks to engineer and
build artificial biological systems, an-
other approach to understanding na-
ture as computation is the research on
computation in living cells. This is also
sometimes called cellular computing,
or in vivo computing, and one particular
study in this area is that of the computa-
tional capabilities of gene assembly in
unicellular organisms called ciliates.
Ciliates possess two copies of their
DNA: one copy encoding functional
genes, in the macronucleus, and an-
other “encrypted” copy in the micro-
nucleus. In the process of conjugation,
after two ciliates exchange genetic in-
formation and form new micronuclei,
they use the new micronuclei to as-
semble in real-time new macronuclei
necessary for their survival. This is ac-
complished by a process that involves
re-ordering some fragments of DNA
(permutations and possibly inversions),
and deleting other fragments from the
micronuclear copy. The process of gene
assembly is fascinating from both the
biological and the computational point
of view. From the computational point
of view, this study led to many novel and
challenging research themes.14 Among
others, it was proved that various mod-
els of gene assembly have full Turing
machine capabilities.23 From the bio-
logical point of view, the joint effort of
computer scientists and biologists led
to a plausible hypothesis (supported
already by some experimental data)
about the “bioware” that implements
the process of gene assembly, which is
based on the new concept of template-
guided recombination.4, 28
Other approaches to cellular com-
puting include developing an in vivo
programmable and autonomous finite-
state automaton within E.Coli, and de-
signing and constructing in vivo cellu-
lar logic gates and genetic circuits that
harness the cell’s existing biochemical
processes.
At the end of this spectrum of views
of nature as computation, the idea was
even advanced by Zuse and Fredkin
in the 1960s that information is more
fundamental than matter or energy.
The Zuse-Fredkin thesis stated that the
entire universe is some kind of compu-
tational device, namely a huge cellular
 4 THE MANY FACETS OF NATURAL COMPUTING
automaton continuously updating its
rules. Along the same lines, it has been
recently suggested that the universe is
a quantum computer that computes it-
self and its own behavior.
Natural Sciences: Ours to Discover
Science advances in ever-widening cir-
cles of knowledge. Sometimes it metic-
ulously crawls. Other times it leaps to a
new dimension of understanding and,
in the process, it reinvents itself. As the
natural sciences are rapidly absorbing
ideas of information processing, and
the meaning of computation is chang-
ing as it embraces concepts from the
natural sciences, we have the rare privi-
lege to take part in several such meta-
morphoses.
At this moment we and our natural
scientist fellows are awash in wave after
gigantic wave of experimental, especial-
ly biological, data. Just underneath this
tumultuous surface lie ingenious algo-
rithms waiting to be designed, elegant
theorems waiting to be proven, natural
laws waiting to be discovered that will
put order into chaos. For, as Spinoza
wrote, “nothing happens in nature that
does not follow from her laws.”
Conversely, as this review shows,
there is an abundance of natural phe-
nomena that can inspire computing
paradigms, alternative physical sub-
strates on which to implement compu-
tations, while viewing various natural
processes as computations has become
more and more essential, desirable,
and inevitable. All these developments
are challenging our assumptions about
computation, and indeed, our very def-
inition of it.
In these times brimming with ex-
citement, our task is nothing less than
to discover a new, broader, notion of
computation, and to understand the
world around us in terms of informa-
tion processing.
Let us step up to this challenge. Let
us befriend our fellow the biologist, our
fellow the chemist, our fellow the phys-
icist, and let us together explore this
new world. Let us, as computers in the
future will, embrace uncertainty. Let us
dare to ask afresh: “What is computa-
tion?”, “What is complexity?”, “What
are the axioms that define life?”
Let us relax our hardened ways of
thinking and, with deference to our sci-
entific forebears, let us begin anew.
Kari and Rozenberg, “The Many Facets of Natural Computing.” CACM, October 2008. Copyright Association for Computing Machinery. Article provided by
Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://www.csd.uwo.ca/~lila/p72.pdf.
Literature
The upper-bound placed on the num-
ber of references was a real limitation
for this review, since the literature on
natural computing is vast. For a more
complete list of references the reader
is referred to the full version of this ar-
ticle at www.csd.uwo.ca/˜lila/Natural-
Computing-Review.pdf.
Almost each of the areas we men-
tioned here has an extensive scien-
tific literature as well as a number of
specialized journals and book series.
There are also journals and book se-
ries aimed at the general natural com-
puting community, among them the
journals Natural Computing, Springer,
Theoretical Computer Science, Series C:
Theory of Natural Computing, Elsevier,
the Natural Computing book series,
Springer, and the upcoming Handbook
of Natural Computing (G. Rozenberg, T.
Bäck, J. Kok, editors, Springer).
Acknowledgments
We gratefully acknowledge comments
on early drafts of this paper by T. Bäck,
D. Bentley, G. Brassard, D. Corne, M.
Hirvensalo, J. Kari, P. Krishna, H. Lip-
son, R. Mercer, A. Salomaa, K. Sims, H.
Spaink, J. Timmis, C. Torras, S. Watt,
R. Weiss.
This work was supported by NSERC
Discovery Grant and Canada Research
Chair Award to L.K., and NSF grant
0622112 to G.R.
References
1. Abelson, H., Allen, D., Coore, D., Hanson, C., Homsy, G.,
Knight Jr., T., Nagpal, R., Rauch, E., Sussman, G., and
Weiss, R. Amorphous computing. Commun. ACM 43, 5
(May 2000), 74–82.
2. Adleman, L. Molecular computation of solutions
to combinatorial problems. Science 266 (1994),
1021–1024.
3. Andrianantoandro, E., Basu, S., Karig, D., and Weiss,
R. Synthetic biology: new engineering rules for an
emerging discipline. Molecular Systems Biology 2
(2006), 1–14.
4. Angeleska, A., Jonoska, N., Saito, M., and Landweber,
L. RNA-guided DNA assembly. J. Theoretical Biology
248 (2007), 706–720.
5. Arbib, M., editor. The Handbook of Brain Theory and
Neural Networks. MIT Press, 2003.
6. Bäck, T., Fogel, D., and Michalewicz, Z., editors.
Handbook of Evolutionary Computation. IOP
Publishing, U.K., 1997.
7. Barry, P. Life from scratch: learning to make synthetic
cells. Science News, 173, 2 (2008), 27.
8. Bath, J. and Turberfield, A. DNA nanomachines.
Nature Nanotechnology 2 (May 2007), 275–284
9. Brooks, R. Artificial life: From robot dreams to reality.
Nature 406 (2000), 945–947.
10. Cardelli, L. Machines of systems biology. Bulletin of
the EATCS 93 (2007), 176–204.
11. Dasgupta, D. editor. Artificial Immune Systems and
Their Applications. Springer, 1998.
12. de Castro, L. and Timmis, J. Artificial Immune
Systems: A New Computational Intelligence
Approach. Springer, 2002.
13. De Jong, K. Evolutionary Computation: A Unified
O C TO BER 20 0 8 | VO L. 51 | NO. 10 | COM M UN I CATI ON S OF THE ACM
24
review articles
Approach. MIT Press, 2006.
14. Ehrenfeucht, A., Harju, T., Petre, I., Prescott, D., and
Rozenberg, G. Computation in Living Cells: Gene
Assembly in Ciliates. Springer, 2004.
15. Endy, D. Foundations for engineering biology. Nature
438 (2005), 449–453.
16. Engelbrecht, A. Fundamentals of Computational
Swarm Intelligence. Wiley and Sons, 2005.
17. Forster, A. and Church, G. Towards synthesis of a
minimal cell. Molecular Systems Biology 2, 45 (Aug.
2006).
18. Fox Keller, E. and Harel, D. Beyond the gene. PLoS
ONE 2, 11 (2007), e1231.
19. Hirvensalo, M. Quantum Computing, 2nd Ed. Springer,
2004.
20. Istrail, S., De-Leon, B-T., and Davidson, E. The
regulatory genome and the computer. Developmental
Biology 310 (2007), 187–195.
21. Kari, L. DNA computing—the arrival of biological
mathematics. The Math. Intelligencer 19, 2 (1997),
9–22.
22. Koza, J. Genetic Programming: On the Programming of
Computers by Means of Natural Selection. MIT Press,
1992.
23. Landweber, L. and Kari, L. The evolution of cellular
computing: Nature’s solution to a computational
problem. Biosystems 52, 1/3 (1999), 3–13.
24. Langton, C., editor. Artificial Life. Addison-Wesley
Longman, 1990.
25. Lipson, H. and Pollack, J. Automatic design and
manufacture of robotic lifeforms. Nature 406, (2000),
974–978.
26. Paun, G. Membrane Computing: An Introduction.
Springer, 2002.
27. Paun, G., Rozenberg, G., and Salomaa, A. DNA
Computing: New Computing Paradigms. Springer,
1998.
28. Prescott, D., Ehrenfeucht, A., and Rozenberg, G.
Template guided recombination for IES elimination
and unscrambling of genes in stichotrichous ciliates. J.
Theoretical Biology 222, 3 (2003), 323–330.
29. Prusinkiewicz, P. and Lindenmayer, A. The Algorithmic
Beauty of Plants. Springer, 1990.
30. Reif, J. and LaBean, T. Autonomous programmable
biomolecular devices using self-assembled DNA
nanostructures. Commun. ACM 50, 9 (Sept. 2007),
46–53.
31. Rojas, R. Neural Networks: A Systematic Introduction.
Springer, 1996.
32. Rothemund, P., Papadakis, N., and Winfree, E.
Algorithmic self-assembly of DNA Sierpinski triangles.
PLoS Biology 2, 12 (Dec. 2004).
33. Rozenberg, G. Computer science, informatics and
natural computing—personal reflections. In New
Computational Paradigms: Changing Conceptions of
What Is Computable. Springer, 2008, 373–379.
34. Rozenberg, G. and Salomaa, A. The Mathematical
Theory of L Systems. Academic Press, 1980.
35. Seeman, N. Nanotechnology and the double helix.
Scientific American Reports, 17, 3 (2007), 30–39.
36. Sims, K. Evolving 3D morphology and behavior by
competition. In Proceedings of Artificial Life IV. MIT
Press, 1994, 28–39.
37. Smith, H., Hutchison III, C., Pfannkoch, C., and Venter,
C. Generating a synthetic genome by whole genome
assembly: IX174 bacteriophage from synthetic
oligonucleotides. PNAS 100, 26 (2003), 15440–15445.
38. Stepney, S. et al. Journeys in non-classical
computation I: A grand challenge for computing
research. Int. J. Parallel, Emergent and Distributed
Systems 20, 1 (2005), 5–19.
39. von Neumann, J. The Computer and the Brain. Yale
University Press, 1958.
40. von Neumann, J. Theory of Self-Reproducing
Automata. U. Illinois Press, 1966. Edited and
completed by A.W.Burks.
Lila Kari (lila@csd.uwo.ca) is Professor and Canada
Research Chair in Biocomputing in the Department of
Computer Science at the University of Western Ontario,
London, Canada.
Grzegorz Rozenberg (rozenber@liacs.nl) is Professor
at the Leiden Institute of Advanced Computer Science,
Leiden University, Leiden, The Netherlands, and Adjunct
Professor in the Department of Computer Science at the
University of Colorado at Boulder, USA.
© 2008 ACM 0001-0782/08/1000 $5.00
83
 55 WHY JOHNNY CANT STEAM: HOW VIDEO COPYRIGHT WENT INSANE 25

Why Johnny can't stream: How video copyright

went insane
Deploying 10,000 tiny antennas makes no technical sense—but the law demands it.

by James Grimmelmann - Aug 30 2012, 8:00am CDT

Aurich Lawson / Thinkstock

Suppose I could offer you a choice of two technologies for watching TV online. Behind Door Number
One sits a free-to-watch service that uses off-the-shelf technology and that buffers just enough of each
show to put the live stream on the Internet. Behind Door Number Two lies a subscription service that
requires custom-designed hardware and makes dozens of copies of each show. Which sounds easier
to build—and to use? More importantly, which is more likely to be legal?

If you went with Door Number One, then you are a sane person, untainted by the depravity of modern
copyright law. But you are also wrong. The company behind Door Number One, iCraveTV, was
enjoined out of existence a decade ago. The company behind Door Number Two, Aereo, just survived
its first round in court and is still going strong.

The difference between them—and the reason for Aereo's willfully perverse design—originated in a
critical 2008 DVR decision by the federal Second Circuit Court of Appeals in Cartoon Network v. CSC
Holdings (which everyone just calls "Cablevision"). The tech at issue in Cablevision was a "DVR in the
cloud," and because of the way the Second Circuit answered the question of whether a DVR
"performs" a copyrighted TV show when the user hits "play," the decision opened a whole range of
possibilities for entrepreneurs willing to mash up technologies in ways God never intended.

This is the story of Cablevision, the companies that followed in its wake, and how we got to the strange
place where wasting resources on thousands of tiny antennas made you legal—but where using one
antenna broke the law.

Backdrop
To understand what Cablevision decided and why it matters, we need to understand a bit about how
copyright law treats broadcasting and streaming. For a very long time, copyright has covered more
than just making copies. In 1856, for example, Congress gave playwrights an exclusive right to stage
their plays in public. That "public performance right" has gradually expanded to cover almost everything
that can be copyrighted and performed, from movies to musicals.

The meaning of a "public performance," however, has been surprisingly hard to pin down. For more
than a century, technologists have been coming up with unexpected ways of bringing media to people.
Some cases were easy: the courts quickly decided that showing movies in theaters and broadcasting

James Grimmelmann, “Why Johnny Cant Steam: How video copyright went insane.” Ars Technica, October 2012. Copyright Ars Technica. Article provided
by Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://arstechnica.com/tech-policy/2012/08/
why-johnny-cant-stream-how-video-copyright-went-insane/.
 5 WHY JOHNNY CANT STEAM: HOW VIDEO COPYRIGHT WENT INSANE 26

songs on the radio were public performances requiring permission of the rightsholder, while mere
audition (in English, that's "listening to the radio") was not.

Cable TV, though, broke the mold. Back when the typical cable TV
operator was an edgy upstart, the cable business model was
retransmission without the express written consent of anyone. A If the Internet is a
cable network would put up an antenna somewhere with good TV copyright minefield,
reception, lay wires over the hills and far away to communities with Cablevision had just
terrible TV reception, and relay the signals to paying subscribers
right-clicked a new
there.
safe square for
In a pair of decisions, 1968's Fortnightly v. United Artists and 1974's startups to stand on.
Teleprompter v. CBS, the Supreme Court held that cable
retransmission was not a "performance" for copyright purposes and
so didn't need copyright owners' permission.

"Broadcasters perform. Viewers do not perform," the court explained in Fortnightly. A cable network
was just a way to help viewers receive distant TV broadcasts, like a gigantic pair of rabbit ears.

But the Supreme Court had meddled with the primal forces of nature, and Congress promptly swung
into action, revising the law to override Fortnightly and Teleprompter. The 1976 Copyright Act added a
"transmit clause" to its definitions to make clear that whether a work was performed "by means of any
device or process" and whether the public received it "in the same place or in separate places and at
the same time or at different times," it would still infringe if transmitted without permission.

Meanwhile, Congress bought off the cable companies—who didn't like this at all—with a complicated
licensing scheme. In essence, copyright law was sucked into telecommunications law's gravity well. In
time, satellite TV broadcasters got the same deal: they were subject to copyright, but with their own
crazy-intricate licensing system spelling out exactly what they could do and how much it would cost.

Harmony and order returned to the universe—until the Internet came along.

Son of cable
The Internet has never played nice with carefully crafted regulatory schemes. Since streaming became
practical in the 1990s, a series of adventuresome dot-com entrepreneurs have been searching for a
way to repeat the cable systems' original legal coup, bringing live TV to Joe User—preferably without
paying to do so. (It's hard to make a living by streaming video when copyright owners can always turn
around and grab back your profits by demanding higher licensing fees. Exhibit 1: Netflix. Exhibit B:
Hulu.)

The first high-profile try was the aforementioned service behind Door Number One, the Canadian
website iCraveTV. It hoisted an antenna in Toronto, picked up TV signals from Buffalo, New York then
turned around and streamed those signals on the Internet (surrounded, of course, with ads). It tried to
lock out Americans, but it didn't try very hard (it asked users to enter a Canadian area code) and nearly
half of its viewership came from the United States. When the movie studios sued in the United States,
the court had little difficulty claiming jurisdiction over iCraveTV and its officers. An injunction was
entered; goodbye, iCraveTV.

More recently, websites like FilmOn and ivi have tried to solve the copyright problem by fleeing back
into telecommunications law. Both of them restream over-the-air TV on the Internet. ivi argued that it
was a "cable system" and was therefore entitled to the Copyright Act's special license for cable
companies; FilmOn called itself a "carrier" with a similar goal.

Neither ended well. A federal court shut down ivi, giving a long history of the cable license and pointing
out, rather inconveniently for ivi, that the company hadn't even attempted to comply with any of the
FCC's extensive rules for cable licensing. The appeals court agreed—an Internet streaming server is
not a "cable system." (While the FCC is slowly considering whether to start treating pure TV-over-IP
services like cable systems, it's not there yet.)

FilmOn agreed to an injunction and to pay $1.6 million.

Hope for the rebroadcasters, however, came from an unlikely place: the cable industry.

James Grimmelmann, “Why Johnny Cant Steam: How video copyright went insane.” Ars Technica, October 2012. Copyright Ars Technica. Article provided
by Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://arstechnica.com/tech-policy/2012/08/
why-johnny-cant-stream-how-video-copyright-went-insane/.
 5 WHY JOHNNY CANT STEAM: HOW VIDEO COPYRIGHT WENT INSANE 27

Like a DVR... in the cloud!

In the half-century since Fortnightly, the cable companies have grown from scrappy underdogs into
dominant media empires. Uneasy lies the head that wears the crown, however, and one of the
technologies that makes cable executives sweat is the DVR. Recording TV and watching it at one's
leisure has the potential to wrest control of the viewing experience away from the wire and put it in the
viewer's hands. The cable response—straight out of a business-school textbook—has been to embrace
the threat and get into the DVR business. Cable companies soon integrated DVRs into set-top boxes,
offering subscribers one-stop recording (for a fee, of course).

The boffins at Cablevision, a New York-area cable company, had an even better idea. They proposed
sucking the DVR out of the set-top box and moving it up into the cloud, placing the hard drive in a
Cablevision head-end facility where the TV signals are modulated for cable transmission.

As you might expect, Cablevision's announcement of its planned RS-DVR (short for "remote storage
DVR") drew the wrath of cable's traditional frenemies: the TV networks who supply most cable content.
The networks were largely powerless against traditional DVRs, thanks to the Supreme Court's decision
in the landmark Sony case. VCRs, the Court held, are legal to sell because customers could put them
to "substantial noninfringing uses," such as "time-shifting" live TV for later viewing. Nor is home
playback covered by copyright: that's a completely legal private performance unless you invite your
neighbors over and charge admission.

The networks liked their odds against the RS-DVR. The essence of the Sony defense had been that
the VCR left Sony's control when it was sold. The RS-DVR, however, was sitting right there in a
Cablevision head-end, on Cablevision property, where Cablevision techs could switch it off or
reprogram it at any moment. And unlike a VCR, where the signal starts in the home and stays there,
the RS-DVR transmits signals from the Cablevision head-end to the user's home. To the networks, this
seemed just like the question Congress settled when it overruled Fortnightly and Teleprompter:
transmissions by cable companies to viewers are public performances, and permission is required.

The networks sued to prevent Cablevision from rolling out the RS-DVR service and won at the trial
level.

But on appeal, Cablevision and the RS-DVR emerged triumphant. The Second Circuit seized on two
facts to justify its holding that watching a show on the RS-DVR wasn't a public performance. First, each
time a user recorded a program, the RS-DVR made a separate copy of it for her, storing it on her own
dedicated hard drive space. Second, each time she played back a program, it came from her own
stored copy. This, the court concluded, meant that while Cablevision might "perform" the TV shows by
streaming them, it didn't perform them for "the public." One person does not the public make, not if she
has her own copy.

The TV networks argued in vain that all of Cablevision's customers should be aggregated together as
"the public," since they were all receiving the broadcast the networks sent to Cablevision. But the court
concluded that the individual copies in the RS-DVR broke the "chain of transmissions" that took a TV
show from broadcaster to viewers. The transmission from NBC to Cablevision's RS-DVR was a (public)
performance; each transmission from an RS-DVR to its user was a separate (private) performance to
an audience of one. The RS-DVR was legal.

For Cablevision's customers, the consequences have been underwhelming. The RS-DVR rolled out
last year in a trial in the Bronx with the oh-so-catchy name of "DVR Plus" and a $10.95/month price tag.
(<crickets>) But for companies interested in building Internet businesses with a copyright angle,
Cablevision was a godsend. If the Internet is a copyright minefield, Cablevision had just right-clicked a
new safe square for startups to stand on.

James Grimmelmann, “Why Johnny Cant Steam: How video copyright went insane.” Ars Technica, October 2012. Copyright Ars Technica. Article provided
by Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://arstechnica.com/tech-policy/2012/08/
why-johnny-cant-stream-how-video-copyright-went-insane/.
 5 WHY JOHNNY CANT STEAM: HOW VIDEO COPYRIGHT WENT INSANE 28

Aereo—what an odd way to do things.

Thousands of tiny antennas

Now we're ready to look behind Door Number Two and see what a difference Cablevision makes.
Aereo is a New York startup backed by media baron Barry Diller, and its play is the same one cable
companies made in their Wild West days: rebroadcasting live over-the-air TV without anyone's
permission. Quietly stepping over the arrow-strewn bodies of its predecessors, Aereo filled its Brooklyn
data center with dime-sized antennas—80 on each circuit board, with 16 boards to a rack. When a user
is logged in, Aereo designates one of the antennas as "hers" and starts recording the chosen channel
to a unique copy on a hard drive, Cablevision-style. Then, just like with Cablevision's RS-DVR, she can
stream the stored video over the Internet.

Two aspects of this business model stand out. The first is how
precisely it hews to Cablevision. Each TV stream on Aereo, in
Cablevision's words, "is made to a single subscriber using a single Thousands of tiny
unique copy produced by that subscriber." The only significant antennas are a
difference is that Aereo starts with TV signals that come in over the ridiculous way of
air rather than on a wire. (As a sign of how dependent Aereo is on
capturing over-the-air
Cablevision, the service is only offered in New York, where the
Second Circuit's decisions apply. Cross the river to New Jersey,
TV. Storing a
which is part of the Third Circuit, and Aereo is unavailable in your permanent copy
area.) rather than a buffer
just large enough for
Second, the striking thing about how Aereo works is just how
gratuitously profligate it is with technology. Thousands of tiny streaming is a
antennas are a ridiculous way of capturing over-the-air TV. Storing pessimization, not an
a permanent copy rather than a buffer just large enough for optimization.
streaming is a pessimization, not an optimization. And of course
Aereo keeps as many separate copies of each program as there
are viewers who want it. If copyright law made sense, copyright
owners themselves would offer TV streaming on the Internet. But copyright law hasn't made sense for
years, and Aereo embraced the madness.

It worked. In July, a federal judge denied broadcasters' request for an injunction to shut down Aereo.
The broadcasters tried to get around Cablevision by arguing that the individual copies didn't really
count because—I am not making this up—the viewer could start streaming a show before it ended. It's
bizarre that buffering a TV show through a copy being recorded to a hard drive should let Aereo escape
liability, but the argument that viewers should have to wait until the end of the show is just as strange.
As the judge explained, the broadcasters' view would mean that a user "who begins watching a
recording of the Academy Awards, initially broadcast at 6:00 pm, one minute before the program ends
at 11:00 pm" would be watching an infringing public performance, while a viewer "who begins watching
a standard half-hour sit-com just a minute after its initial broadcast ends" would be watching a legal
private performance from her own copy.

In a bizarre postscript to the Aereo lawsuit, the entrepreneur behind FilmOn, Alki David, announced a
new service using the Aereo model, called BarryDriller.com. The subject of this off-color pun, Aereo's
backer Barry Diller, was unamused and promptly sued. The BarryDriller.com domain now redirects to
David's CBSYouSuck.com, a site that promotes David's ongoing suit against CBS and CNET for
distributing LimeWire.

James Grimmelmann, “Why Johnny Cant Steam: How video copyright went insane.” Ars Technica, October 2012. Copyright Ars Technica. Article provided
by Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://arstechnica.com/tech-policy/2012/08/
why-johnny-cant-stream-how-video-copyright-went-insane/.
 5 WHY JOHNNY CANT STEAM: HOW VIDEO COPYRIGHT WENT INSANE 29

The world's longest video cable

It's not just TV-on-the-Internet that got a jump-start from Cablevision. All kinds of other streaming-based
business models suddenly seemed possible, too. The crucial question was where to get the copy you
were streaming to a customer. Cablevision and Aereo got the copies when customers hit "record."
Cablevision had successfully argued that "the person who actually presses the button to make the
recording supplies the necessary element of volition [to infringe], not the person who manufactures,
maintains, or, if distinct from the operator, owns the machine." (Similarly, in home taping with VCRs and
DVRs, viewers make their own copies, which are often legal under fair use as time-shifting.)

A California startup called Zediva had another idea. It bought DVDs, taking advantage of the first sale
doctrine to become the undisputed owner of copies of movies. It filled a Santa Clara data center with
hundreds of DVD players. When a customer hit play on, say, Little Fockers, Zediva's servers reserved
a DVD player with the DeNiro/Stiller turkey in it, started the DVD playing, and streamed the video to the
customer over the Internet. Zediva called it "a DVD with a very long cable attached."

Once again, the business model looked engineered to take

advantage of Cablevision, and Zediva argued that its streams
weren't public performances because they went to individual Having an individual
viewers and no one else. And once again, it's a business model file for each user is
that would not exist in a world with copyright policy that was not crucial; using
demonstrably insane. Rapidly spinning optical discs make sense as
deduplication would
a distribution technology because they're compact and durable. But
they're a hassle and a half for playback, because they scratch, skip,
mean crossing out of
and make random access a pain. If you're going to use the Internet Cablevision country.
for distribution, better to take the DVDs out of the picture and use
them as coasters. But since Cablevision had opened up what
seemed like a gap in copyright law, Zediva poured shiny lacquered
discs into the breach.

Unfortunately for Zediva, it ran headlong into one of Cablevision's major limitations. Each Zediva
stream went to a unique subscriber—but it didn't come from a completely unique copy. After one user
finished watching Little Fockers, the DVD and player would go back into the pool for others to use. This
meant that the same DVD would be shown repeatedly to different users over time.

This mattered because of a case law that Cablevision had delicately tiptoed around. Long before DVRs
and Internet streaming video, there were video stores. One, Maxwell's Video Showcase, had tried to
compete with movie theaters by also renting video booths for two to four people. You would pick your
movie, get some popcorn, then sit on a luxurious upholstered bench and watch Raiders of the Lost Ark
on a gigantic nineteen-inch screen. In a case called Columbia Pictures v. Redd Horne, the court held
that Maxwell's engaged in infringing public performances because "it shows each copy [of a film]
repeatedly to different members of the public," even if not simultaneously.

Other courts reached similar results in cases involving hotels and adult theaters, so Cablevision
adopted a distinction between one copy and many. A million viewers and a million copies—OK. A
million viewers but only one copy—not OK. This reasoning doomed Zediva, because of course the
whole point of its business model was that it would reuse DVDs and stream them to different users in
rapid succession.

Pause to note what a silly distinction this is. Making separate copies for each user is a massive waste
of storage. Systems engineers would say that Cablevision should make only as many copies as it
needs to meet demand. Making more does nothing to improve the experience for users; it does nothing
to change the impact on copyright owners. All it does is drive up costs. But courts have to play the

James Grimmelmann, “Why Johnny Cant Steam: How video copyright went insane.” Ars Technica, October 2012. Copyright Ars Technica. Article provided
by Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://arstechnica.com/tech-policy/2012/08/
why-johnny-cant-stream-how-video-copyright-went-insane/.
 5 WHY JOHNNY CANT STEAM: HOW VIDEO COPYRIGHT WENT INSANE 30

hands they're dealt, and the Cablevision court was working with precedents that made the use of
individual copies highly significant. If it is fair to say that Cablevision won on a technicality, then it is
also fair to say that Zediva lost on one—and vice versa. These are precisely the kinds of technicalities
that matter in modern copyright law.

Rock lockers
Cablevision is even spilling over into the simmering dispute over file lockers. It's not entirely a
coincidence that Amazon, Google, and Apple—big companies with a lot to lose if they get their
copyright compliance wrong—all launched cloud-based music players in 2011. Their first line of legal
defense is presumably the DMCA safe harbor for user-uploaded content, but there are traces of
Cablevision in their policies.

The free versions of these music lockers require users to upload each and every song they want, no
matter how many others have uploaded the same song. (Apple and Amazon's paid versions, which will
match files in the user's collection, are licensed.) This might not be necessary for DMCA compliance,
but from a Cablevision perspective, having an individual file for each user is crucial; using deduplication
would mean crossing out of Cablevision country.

Or perhaps not. MP3tunes is (or maybe was) a music locker that deduplicates at a file-block level.
Music companies sued, citing Cablevision. But a federal judge thought that this deduplication wasn't a
problem, explaining, "The record demonstrates that MP3tunes does not use a 'master copy' to store or
play back songs stored in its lockers. Instead, MP3tunes uses a standard data compression algorithm
that eliminates redundant digital data." So either the judge didn't think that individual copies were
necessary, or he misunderstood what a "master copy" was.

The most ambitious music locker service doesn't even describe itself as one. Instead, ReDigi calls itself
"The World's First Pre-Owned Digital Marketplace." (The claim may be a bit overstated; similar ideas
include the Digital Content Exchange and this patent application.) Why buy "Rumor Has It" from iTunes
for $1.29 when someone else who already did is willing to sell it to you for 79 cents? ReDigi's offline
model is the used CD store—which of course is completely legal because of first sale.

But wait, you may be saying, at a used CD store the previous owner gives up the CD when the new
owner gets one. Online, don't they now both have a copy? Great question. ReDigi's answer involves
some DRM gymnastics. First, it uploads the track to its servers, giving the seller streaming access.
When the owner "sells" the track, ReDigi changes the ownership bits on the file, locking the seller out
and giving the buyer exclusive streaming access. (To keep the seller from simply uploading a copy and
keeping the original, ReDigi deletes the track from the seller's computer and uses DRM to make sure
it's never copied back.) In other words, ReDigi is a streaming music locker with a key that can be
transferred from one user to another.

ReDigi is a complicated bank shot off of Cablevision. It strings together Cablevision's public
performance rule—to justify streaming music originally uploaded by user A to user B—with a series of
other arguments. Uploads are personal fair use by users, not infringing copies made by ReDigi. A
change in the ownership bit is not an infringing "distribution" of anything to anyone. Downloads by the
new owner are personal fair use. And so on.

There are serious questions about whether all of the pieces really hang together. (A skeptic would say
that when the file changes "owners," it fails the individual-copies test in the same way that Zediva did.)
Still, the general picture should be familiar by now: a business using legal stepping stones to create a
path from Step One (online streaming) to Step Three (profit).

James Grimmelmann, “Why Johnny Cant Steam: How video copyright went insane.” Ars Technica, October 2012. Copyright Ars Technica. Article provided
by Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://arstechnica.com/tech-policy/2012/08/
why-johnny-cant-stream-how-video-copyright-went-insane/.
 5 WHY JOHNNY CANT STEAM: HOW VIDEO COPYRIGHT WENT INSANE 31

A complex legacy
The post-Cablevision cases are almost comically formalistic about technical details. Instead of looking
at the front-end user experience, they focus on the back-end hardware and software. Sooner or later,
someone is going to argue to a court that it makes a difference for copyright purposes whether a video
stream is decoded by the main CPU or by a dedicated graphics card—or some other distinction equally
remote from anything the typical viewer thinks about when trying to catch up on last week's episode of
Breaking Bad.

This technological formalism has real costs and real benefits for all concerned. On the upside, Lawful
Good technologists and investors need bright-line guidance. Imagine being a cloud computing vendor,
watching the file locker litigation and worrying that one judge could scuttle your entire business model.
Or worse, imagine being a cloud computing customer facing the risk that one judge could consign your
files to Davy Jones's locker.

Actually, there's no need to imagine: this is precisely what happened to MegaUpload's customers. Their
data was eaten by the cloud of legal uncertainty that is criminal copyright infringement. Faced with the
huge overhang of liability created by copyright doctrines that treat even evanescent "copies" in RAM as
potential infringements, courts may be crafting exceptions, like Cablevision's public performance rule,
to preserve some breathing room for innovation. Having clear rules can be helpful.

On the downside, the contortions required to fit through Cablevision's hoops can be cringe-inducing,
and not just for the technology companies blowing through their angel money to buy extra storage and
bandwidth. One man's "hoop" is another man's "loophole," and to copyright owners, Cablevision looks
like a copyright-evasion cookbook: How To Succeed in Piracy Without Really Infringing. From their
perspective, the post-Cablevision burst of innovation consists entirely of Frankenstein-style attempts to
resurrect business models already condemned by the courts.

The road not taken

Perhaps we can think about the problem of copyright on the Internet another way. Instead of asking
which back-end technologies are legal, it might make more sense to ask what it is legal for users to do
with computers on the front end. This approach would let people spend less time worrying about the
exact definitions of "reproduction" and "performance" and more time thinking about users' rights,
especially under fair use.

Cablevision itself illustrates what might have been. The whole point of the RS-DVR was that it was a
perfect substitute for a home DVR. Reasoning by analogy, then, we might say that the two ought to
either both be legal or both be illegal. And since home DVRs seem here to stay, it ought to be
permissible for Cablevision to offer its customers exactly the same service they could have gotten by
buying a gizmo. Call it "noninfringing personal fair use" and we can all go home.

The strangest thing about Cablevision is that the court didn't even get a chance consider that
argument. The parties agreed not to litigate the fair use issue. Yes, you read that right. It was a quid
pro quo: Cablevision didn't invoke users' fair use rights and the cable networks didn't try to hold
Cablevision liable for users' infringements. That turned a case about users' uses into a case about
Cablevision's technologies, changing a common sense debate over how far viewers can go in storing
TV programs and watching them later into an abstruse legal disputation over the minutiae of primary
ingest buffers and chains of transmission.

Another way of thinking about the decision is that the court wanted to provide a fair use ruling for
viewers but wasn't given the option, so it settled for much narrower rulings on the technologies
involved.

Copyrighted content is the nuclear fuel of the Internet. It powers high-energy innovation, but can cause
catastrophic legal meltdown if mishandled. Prolonged exposure has been scientifically proven to cause
business-model mutations. Cablevision gave risk-tolerant entrepreneurs an inanimate carbon rod:
enough to save the day for some of them, but hardly a long-term solution.

James Grimmelmann is a Professor at New York Law School.

James Grimmelmann, “Why Johnny Cant Steam: How video copyright went insane.” Ars Technica, October 2012. Copyright Ars Technica. Article provided
by Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://arstechnica.com/tech-policy/2012/08/
why-johnny-cant-stream-how-video-copyright-went-insane/.
 66 COULD YOUR IPOD BE HOLDING THE GREATEST MYSTERY IN MODERN SCIENCE32

$OJRULWKPLFWKLQNLQJLVOLNHO\WRFDXVH WKHPRVW GLVUXSWLYH

SDUDGLJPVKLIWLQWKHVFLHQFHVVLQFHTXDQWXPPHFKDQLFV

&RXOG <RXU L3RG %H WKH

+ROGLQJ

*UHDWHVW LQ0RGHUQ
0\VWHU\

6FLHQFH"
%HUQDUG &KD]HOOH
3ULQFHWRQ
8QLYHUVLW\

)XWXURORJLVWV DUHDQ DPLDEOH EXQFK VR LWLVDSX]]OH ZK\ JUHDW/RUG .HOYLQ EORYLDWHG WKLVJHP IRUWKHDJHV 7KHUH LV
WKHIXWXUHKDVEHHQ VRFUXHOWRWKHP)URP IO\LQJFDUVDQG QRWKLQJQHZ WREH GLVFRYHUHG LQSK\VLFV QRZ1RW KLV ORUG
VHOIFOHDQLQJKRXVHV WRWKDW
EXJDERR RIZRUNDKROLFV"WKH VKLS
VILQHVWKRXU
OHLVXUH VRFLHW\"WKH VRRWKVD\HU
V WUDLO LV OLWWHUHGZLWK WKH 0RRUH
V /DZ KDV IXHOHG FRPSXWHU VFLHQFH
V VL]]OH DQG
FDUFDVVHV RISHWSUHGLFWLRQV WXUQHGURDGNLOO VSDUNOHEXW LW
PD\ KDYH REVFXUHG LWVXQFDQQ\ UHVHPEODQFH WR
*RUGRQ 0RRUH QHHG QRWZRUU\7KH FRIRXQGHU RI ,QWHO SUH(LQVWHLQSK\VLFV KHDOWK\DQGSOXPS"DQG ULSHIRUDUHYR
WULHGKLVKDQGDWFU\VWDOJD]LQJ RQFH"DQG VWUXFNJROG+LV FHO OXWLRQ&RPSXWLQJ SURPLVHV WREH WKH
PRVW GLVUXSWLYH VFLHQWLI
HEUDWHGODZ PDNHV WKHRXWODQGLVK SUHGLFWLRQ WKDWHYHU\  LFSDUDGLJP VLQFHTXDQWXPPHFKDQLFV 8QIRUWXQDWHO\ LWLVWKH
PRQWKV OLNHFORFNZRUNZKLWHFODG WHFKQLFLDQVZLOO KXGGOH LQ SURYHUELDO ULGGOHZUDSSHG LQDP\VWHU\ LQVLGHDQHQLJPD7KH
D VLOLFRQZDIHU FOHDQ URRPDQGFUDPWZLFHDVPDQ\ WUDQVLVWRUV VWDNHVDUHKLJKIRURXU LQDELOLW\WRJHWZKDW FRPSXWLQJ LVDOO
RQWRDPLFURFKLS DERXWPD\ ZHOO SOD\ LFHEHUJ WRWKH7LWDQLF RIPRGHUQ VFLHQFH
0RRUH
V /DZ KDV UXOHGWKHURRVWIRUWKHODVWIRUW\ \HDUV$OO %ULOOLDQW IRUHVLJKWRU ODWHVWWULSHIURPWKH
.HOYLQ VFKRRO RI
WKHRRKV DQG DDKV \RX KHDU DERXW WKHGLJLWDO UHYROXWLRQ DUH SURSKHF\"&RPSXWLQJ LVWKH PHHWLQJ SRLQWRIWKUHH%LJ ,GHDV
QRWKLQJ EXW WKHVTXHDOV KXPDQV HPLWZKHQ WLFNOHGSLQNE\ XQLYHUVDOLW\GXDOLW\ VHOIUHIHUHQFH7R WKLVWULDGWKH
PRGHUQ
0RRUH
V /DZ )URP WKHQLFH PHGLFDO LPDJLQJHFRPPHUFH YLHZ DGGV WKHFRQFHSW RI WUDFWDELOLW\DQG WKHUHYROXWLRQDU\

ZKROHJHQRPH VHTXHQFLQJ WRWKHYLWDO ;ER[ ,0 L3RG LWV DOJRULWKPLFSDUDGLJP+HUH

V KRZ LW ZRUNV
UXOHKDVEHHQ DYHULWDEOH WLFNOHIHVW
0RRUH
V /DZ KDVEHHQ WKH
VL]]OLQJ FDXOGURQ LQZKLFK VDYY\ FRRNV KDYH ZKLSSHG XS D 8QLYHUVDOLW\
GD]]OLQJ YDULHW\RI WDVW\GLVKHV:LWKRXW LWWKH,QIRUPDWLRQ )HZ ZRXOG PLVWDNH \RXU L3RG IRUDQ ,%0 %OXH *HQH/"
6XSHUKLJKZD\ ZRXOG EH D EDFN DOOH\ WR6QRR]HYLOOH WKH WKHZRUOG
V IDVWHVWFRPSXWHU<HW IXQGDPHQWDOO\ WKHWZRDUH
FRROHVW WKLQJDERXW D FRPSXWHUZRXOG VWLOOEH WKHEOLQNLQJ WKHVDPH:K\ LVWKDW"$W WKHKHDUWRI\RXU L3RG LVDZULWWHQ
OLJKWV GRFXPHQWPDGH RIWZRSDUWVSURJUDPGDWD 7KH GDWD VHF
0RRUH
V /DZ KDVKDGDJRRG UXQEXWDODV LWVGD\V DUHQXP WLRQVWRUHVWKHVRQJV DV ORQJVHTXHQFHV RI2VDQG ,V7KH SUR
EHUHG %\PLGFHQWXU\ D UHSHDO LVDOO EXWFHUWDLQ:LWK WKH JUDP VHFWLRQH[SODLQV LQZRUGV DJDLQ2VDQG ,V KRZ WRUHDG
KHDG\ GD\V RI WKH,QFUHGLEOH 6KULQNLQJ&KLS UHFHGLQJ LQWKH PL[ D VPDWWHULQJRI
WKHGDWDDQG WXUQLWLQWRVRXQG$GG WRWKLV
SDVWH[SHFW WKHUHYROXWLRQWRJULQGWRDKDOWH[SHFWSLRQHHUV WR KDUGZDUH WKHFRQWURO WRUHDG WKHSURJUDP DQG IROORZ LWV
JLYHZD\ WRWLQNHUHUV %\HE\H WLFNOHIHVWKHOOR VOXPEHUSDUW\ LQVWUXFWLRQVDQGYRLOD \RX
YH JRW\RXUVHOIDQ L3RG7KH EHDX
1R WHDUVSOHDVH 3HUFKHG DWRS WKHLUWRZHULQJDFKLHYHPHQWV RIWKHVFKHPH LVWKDWWKHFRQWUROQHHGQRWNQRZ D WKLQJDERXW
W\
FRPSXWHU VFLHQWLVWV WKHFRRNV UHPHPEHU"ZLOO EDVN LQWKH PXVLF ,Q IDFWVLPSO\E\GRZQORDGLQJ WKHDSSURSULDWH SUR
VRRWKLQJFHUWDLQW\WKDWWKHLUJORULRXV VFLHQFH GLHG DW LWVSHDN JUDPGDWDGRFXPHQW \RX FDQ WXUQ\RXU L3RG LQWRDQ HDUWK
:LWK D WLQJHRIVDGQHVV EXWQRWD OLWWOHSULGH WKH\
OOFKLPH LQ TXDNH VLPXODWRUDZRUG SURFHVVRU DZHE EURZVHURUDSDSHU
XQLVRQ 7KHUH LVQRWKLQJ QHZ WREH GLVFRYHUHG LQFRPSXWHU ZHLJKW<RXU GDLQW\ OLWWOH 03 SOD\HU LVDXQLYHUVDO FRPSXWHU
VFLHQFHQRZ 6HSDUDWLQJ FRQWURO WKHKDUGZDUH IURPSURJUDP WKHVRIW
PRYLH EHIRUH\RXKDYH$ IHZ ZDUH ZDV WKH
,I\RX WKLQN\RX
YH VHHQWKLV PDMRU LQVLJKWRI$ODQ 7XULQJ"ZHOO EHVLGHV WKLV
VKRUW\HDUVEHIRUH(LQVWHLQ WXUQHGRXUZRUOG XSVLGH GRZQ WKH OLWWOHFRGHEUHDNLQJ WKLQJKHGLG LQ%OHWFKOH\ 3DUN WKDWKHOSHG

 $35,/

Bernard Chazelle, “Could your iPod be holding the greatest mystery in modern science.” Math Horizons, April 2006. Copyright Mathematical Association
of America. Article provided by Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http:
//www.cs.princeton.edu/~chazelle/pubs/ipod.pdf.
 6 COULD YOUR IPOD BE HOLDING THE GREATEST MYSTERY IN MODERN SCIENCE33

0$7+ +25,=216

ZLQ:RUOG:DU ,,7KH VHSDUDWLRQZDV WKHNH\ WRXQLYHUVDOLW\

1R RQHKDG VHHQDQ\WKLQJTXLWH OLNHLW
EHIRUH$W OHDVWQRWVLQFH
WKH&KLQHVH SKLORVRSKHU RSLQHG *LYH DPDQ D ILVKDQG \RX
IHHGKLPIRUDGD\7HDFK DPDQ WRILVKDQG\RX IHHGKLPIRUD
OLIHWLPH

,Q&RQIXFLXV
V KDQGV WKHVSHFLDOL]HG YLHZRI

ILVKLQJ ULYHUILVKHUPDQ
ILQGVLWVHOIUHSODFHGE\DXQLYHUVDORQH

ILVKLQJ ULYHU ILVKLQJPDQXDO  \RX
7KHUH \RXKDYH LW

FRPSXWLQJ GDWD SURJUDP FRQWURO
7KH FRQWUROSDUWRI\RXU L3RG LVDPDUYHO RIHOHFWURQLFV EXW
WKHVKRFNHULVWKDWLWQHHGQRWEH VRXQLYHUVDO FRPSXWHUVFDQ
EH EXLOWZLWK FRQWUROER[HV YDVWO\ VLPSOHU WKDQD FXFNRR
FORFN)RUDOOSXUSRVHV

FRPSXWLQJ GDWD SURJUDP
&DUWRRQ E\-RKQ-RKQVRQ
'XDOLW\
/HW
HPHDWFDNH
&RQVLGHU WKHL3RGGRFXPHQW 3ULQW WKLV
3XVK WKHVWDUWEXWWRQDQGZDWFK WKHZRUGV /HW
HPHDWFDNH
IODVKDFURVV WKHVFUHHQ1RWH KRZ WKHSURJUDPSDUWRIWKHGRF RQHDIWHUWKHRWKHUVR LWORRNV OLNHDUHJXODUSURJUDPGDWD L3RG
XPHQW3ULQW WKLVLV LQWHUSUHWHGDV D FRPPDQG"SULQWLQJ LV GRFXPHQW ELOOLRQV RIOHWWHUVORQJWKRXJK
ZDQWV DQGSULQWLQJLV
ZKDW LW ZLWK WKH
ZKDW LWJHWV&RQWUDVW WKLV
$&$$*$7 *&&$77* 7*77&7$ &**7$$&
GDWD SDUW/HW
HPHDW FDNHZKLFK LVWUHDWHGDV SODLQ WH[WQR
7KH EDVH SDLULQJV $7 DQG &* HQVXUH WKDWWKHWZR
RQH
V HDWLQJ DQ\WKLQJWR0DULH$QWRLQHWWH
V ODWHUFKDJULQ
VWUDQGVVSHOO WKHVDPHZRUGZLWKGLIIHUHQWOHWWHUV6RZH ORVH
6WULQJVRI2VDQG ,VDUH LQWHUSUHWHGLQRQHRIWZRZD\V DV IRUP QR JHQRPLF LQIRUPDWLRQLIZH WUDQVODWH WKHGDWD SDUW DQG
GDWDRUDVFRQWHQWSURJUDP7DSSLQJ LQWRWKHFRPLFDUWLVWLF ZKROH GRFXPHQW DV WKHGXSOLFDWHG WH[W
UHZULWHWKH
DQG DFDGHPLF SRWHQWLDO RI WKLVGXDOLW\JUHDWPLQGV ZHQW WR
$&$$*$7*&&$77* $&$$*$7 *&&$77*
ZRUN $EERWW DQG&RVWHOOR :KR
V RQ )LUVW" 0DJULWWH
7KLV LVWKHELRORJLFDO DQDORJ RI3ULQW WKLVWZLFH3ULQW WKLV
&HFL Q
HVWSDV XQHSLSH DQG6DXVVXUH VLJQLILHGYV VLJQL WZLFH/LIH
V EXWDZDONLQJ VKDGRZ0DFEHWK ZDUQHG XV1RW
ILHU 6WDULQJDWWKHVXEOLPHZLOO RIFRXUVH VHQG WKHGHHSHU
TXLWH/LIH
V EXWD VHOISULQWLQJL3RG2IIHQGHG VRXOVZLOO EDQJ
WKLQNHUVDPRQJ XV UXVKLQJIRUWKHFODVVLFV"VXFK DV+RPHU
RQ SUHDFKLO\ DERXW WKHUHEHLQJPRUH WRKXPDQ OLIHWKDQWKH
6LPSVRQ
V LPPRUWDOTXLS 2K0DUJH FDUWRRQVGRQ
WKDYH DQ\ EOLQGSXUVXLWRI VHOIUHSOLFDWLRQ"WKRXJK +ROO\ZRRG
V W\SLFDO
GHHSPHDQLQJ WKH\
UHMXVW VWXSLGGUDZLQJV WKDWJLYH \RX D IDUHZRXOG VHHP WRUHIXWHWKDW([LVWHQWLDO DQJVWDVLGHGXDOLW\
FKHDS ODXJK LV WKH RSWLRQ ZH KDYH WR LQWHUSUHW WKH ZRUG
$&$$*$7*&&$77* HLWKHUDV JHQHV WKHIRUPHQFRG
6HOI5HIHUHQFH
LQJRXU JHQRPH RUDV SURWHLQV WKHFRQWHQWPHGLDWLQJ WKH
:ULWH WKHL3RGGRFXPHQW 3ULQWWKLVWZLFH3ULQW WKLVWZLFH
'1$ UHSOLFDWLRQ 6HOIUHIHUHQFH LVWKHGXSOLFDWLRQ HPERGLHG
DQGSUHVV WKHVWDUWEXWWRQ7KH VFUHHQ OLJKWVXSZLWK WKH
ZRUGV
LQWKHEDVH SDLULQJV9LHZHG WKURXJKWKHFRPSXWLQJ OHQV
3ULQW WKLVWZLFH3ULQW WKLVWZLFH/R DQG EHKROG WKHWKLQJ
OLIH  GXDOLW\ VHOIUHIHUHQFH
SULQWV LWVHOI-XVWOLNHD FRPSXWHU YLUXV UHPHPEHU ,GLG QRW
WHDFK\RX WKLV7KHPDJLF ZRUG LVWZLFH )RU H[DPSOH WKH
L3RG GRFXPHQW 3ULQW WKLV3ULQW WKLV SULQWV WKLV 3ULQW
WKLV"PRUH 'U 6HXVV WKDQ VHOIUHSOLFDWLRQ
7KH %LJ ,GHDVZHUH WKHDLUWKDWWKH
*DQJ RI)RXU3ULQFH
WRQEUDQFKEUHDWKHGDOO GD\"WKDW ZRXOG EH$ORQ]R &KXUFK
$ODQ 7XULQJ.XUW *RGHO DQG -RKQYRQ1HXPDQQ 0RWKHU
1DWXUH RIFRXUVH ILJXUHGLWDOORXWDIHZELOOLRQ \HDUV HDUOLHU
V HDUV WKLV
5HIRUPDW \RXUJHQRPH E\ OLQLQJXS WKHWZRVWUDQGVRI'1$
,Q WKH 1DWXUH DUWLFOH WKDWXQYHLOHG WRWKHZRUOG WKH
VWUXFWXUHRI'1$ :DWVRQ DQG&ULFN VLJQHGRIIZLWK WKLVORYH
O\XQGHUVWDWHPHQW,WKDVQRWHVFDSHG RXUQRWLFH WKDWWKHVSH
FLILFSDLULQJ ZH KDYHSRVWXODWHG LPPHGLDWHO\VXJJHVWVDSRVVL
EOH FRS\LQJPHFKDQLVPIRU WKHJHQHWLFPDWHULDO 'XDOLW\ DQG
VHOIUHIHUHQFHHPEHGGHG LQPROHFXOHV ZKDW VZHHWPXVLF WR
PXVWKDYHEHHQ$ODV RXUZDU KHURZDV D OLW
7XULQJ
WOHGLVWUDFWHGDWWKHWLPHEXV\ DVKHZDV HQMR\LQJ WKHUHZDUGV

&RQWLQXHG RQSDJH 

:::0$$25*0$7++5,=16 

Bernard Chazelle, “Could your iPod be holding the greatest mystery in modern science.” Math Horizons, April 2006. Copyright Mathematical Association
of America. Article provided by Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http:
//www.cs.princeton.edu/~chazelle/pubs/ipod.pdf.
 6 COULD YOUR IPOD BE HOLDING THE GREATEST MYSTERY IN MODERN SCIENCE34

0$7+ +25,=216

&RQWLQXHGIURPSDJH  HFRPPHUFH 7KDW GD\ZLOO VHH ELRORJ\ FRQTXHU LWVKLJKHVW

WKDWWKH%ULWLVK DXWKRULWLHVKDG ODYLVKHG XSRQ KLP IRUVDYLQJ SHDN SURWHLQIROGLQJDQGPDWKHPDWLFLDQV FRQWHPSODWH HDUO\
:RUOG:DU ,,"JHQHURXV UHZDUGV OLNH
PLOOLRQV RIOLYHVGXULQJ UHWLUHPHQW,QGHHG WKHGD\ WKH-XUDVVLF,. DUH VKRZQ WREH
DFRXUWFRQYLFWLRQ IRUKRPRVH[XDOLW\ ZLWKDVHQWHQFHRIIRUFHG WUDFWDEOH 3 13 LQFRPSXWHUSDUODQFH SURRIILQGLQJZLOO EH
HVWURJHQ LQMHFWLRQV$OPRVW RQH \HDU WRWKHGD\ RI
:DWVRQ DQG UHYHDOHG WREHQRPRUH GLIILFXOWWKDQSURRIFKHFNLQJ$QGUHZ
&ULFN
V WULXPSK $ODQ 7XULQJZHQW KRPH LQMHFWHGF\DQLGH LQWR :LOHV WKHFRQTXHURU RI)HUPDW
V /DVW 7KHRUHP ZLOO EH IRXQG
ZDV
DQDSSOH DWH LWDQGGLHG+LVPRWKHUSUHIHUUHGWREHOLHYH LW WRGHVHUYH QRPRUH FUHGLWWKDQKLVUHIHUHHV1RWH WKDWWKLVVD\V
DQ DFFLGHQW
QRWKLQJDERXWXQGHUVWDQGLQJ WKHSURRI7R EH3RUQRWWREH3
WKDWLV13
V TXHVWLRQ ,WLVOLNHO\WKDW3 13 ZRXOG GR IRUVFL
7UDFWDELOLW\ HQFHZKDW WKHGLVFRYHU\ RI WKHZKHHO GLG IRUODQG WUDQVSRUWD

%LJ ,GHDZDV WKHKRKXPREVHU

7KH JHQHVLV RIWKLVIRXUWK WLRQ/LWWOHZRQGHU QRRQHEHOLHYHV LW
YDWLRQ WKDWFKHFNLQJ WKHYDOLGLW\ RIDPDWK SURRI WHQGVWREH ZKHHO LVDOZD\V QLFHEXWWRUROO ORJV LQWKH
7R GLVFRYHU WKH
PXFK HDVLHU WKDQILQGLQJWKHSURRI LQWKHILUVWSODFH %XW LVLW PXG KDV LWVFKDUPV WRR/LNHZLVH WKHLQWUDFWDELOLW\RISURRI

UHDOO\"$PD]LQJO\ QRRQHNQRZV ILQGLQJZRXOG KDYH LWVEHQHILWV:KHQ \RX SXUFKDVH DERRN

:HOFRPH WRWKH PRVW LPSRUWDQW RSHQTXHVWLRQ LQDOORIFRP IURP$PD]RQ WKHDVVXUDQFH WKDW\RXU WUDQVDFWLRQLVVHFXUH LV
SXWHUVFLHQFH(YHUZRQGHUHG LI\RXU L3RG
V WXQH OLEUDU\ SUHGLFDWHG RQPRUH WKDQ\RXUHQGHDULQJQDLYHWH)RURQH WKLQJ
LVULFKHQRXJK WROHW\RX FRPSLOH D SOD\OLVW RI D WKRXVDQG LWUHOLHV FULWLFDOO\RQ WKHLQWUDFWDELOLW\RI IDFWRULQJDQXPEHU
VRQJVQR WZRZKLFK KDYH HYHUEHHQ SOD\HG EDFNWREDFN RQ LQWRSULPHV
079" /HW
V KRSH QRWEHFDXVH QRWHYHQ DQ,%0 %OXH *HQH/ -XVWDVPRGHUQ SK\VLFV VKDWWHUVWKHSODWRQLF YLHZ RIDUHDO
FRXOG GR WKHMRE LQOHVV WLPHWKDQKDV SDVVHG VLQFHGLQRVDXUV LW\DPHQDEOH WRQRQLQYDVLYH REVHUYDWLRQ WUDFWDELOLW\FOREEHUV
ZHUH ODVWVHHQURDPLQJWKHHDUWK7R ILQGVXFKDSOD\OLVW SURRI FODVVLFDO QRWLRQVRINQRZOHGJH WUXVW DQGEHOLHI1R OHVV)RU

ILQGLQJ VHHPVKRSHOHVVO\ KDUGHYHQRQDFRPSXWHUEXWWRWHVW D WDVWHRI LWFRQVLGHU WKHJUHDW]HURNQRZOHGJH =. SDUDGR[

ZKHWKHU D WHQWDWLYHSOD\OLVWILWV WKHELOO SURRIFKHFNLQJ LVD WZRPXWXDOO\ GLVWUXVWLQJSDUWLHV FDQ FRQYLQFH HDFK RWKHUWKDW
PDWFK DOOSRVVLEOH SDLUVDJDLQVW079
V FRPSOHWH
FLQFK VLPSO\ HDFK RQH KROGV D SDUWLFXODU SLHFH RI LQIRUPDWLRQZLWKRXW

SOD\OLVWZKLFK LVUHDGLO\DFFHVVLEOH RQ WKHZHE UHYHDOLQJ D WKLQJDERXW LW3LFWXUH WZRILOWK\ULFK

EXVLQHVVPHQ
7KH WZLQ UHDOLW\ RIKDUG SURRIILQGLQJ DQG HDV\ SURRI VWXFNLQDQHOHYDWRU7KHLU LPPHGLDWHJRDO LVZKDWHOVH" ILQG

FKHFNLQJ LVKDUGO\DQ079LQGXFHG DEHUUDWLRQ&RPSXWHU VFL LQJRXWZKR

V WKHZHDOWKLHU =. GLDORJXHV SURYLGH WKHPZLWK
HQWLVWVKDYH FDWDORJXHG RYHU  SUREOHPV MXVW OLNH LW2I WKHPHDQV WRGR VRZKLOH UHYHDOLQJ ]HUR LQIRUPDWLRQDERXW
FRXUVH FRXUWHV\ RI0XUSK\
V /DZ WKHVH -XUDVVLF,. WKHLURZQZRUWK PDWHULDOZRUWK WKDWLV"WKH RWKHUNLQG LV
LQFOXGHDOORIWKHTXHVWLRQV KXPDQLW\ LVGHVSHUDWH WRDQVZHU"
LQDUWLILFLDO LQWHOOLJHQFHFRPSXWDWLRQDO ELRORJ\ UHVRXUFHDOOR
FDWLRQ UDWLRQDOGUXJGHVLJQ HWF
2. VR OLIHLV WRXJK%XW VLQFHZKHQ KDV WKDWREVHUYDWLRQ
TXDOLILHG DVD%LJ ,GHD"
6LQFH  URXJKO\ -XVWDV(LQVWHLQ UHEXLOW1HZWRQLDQ
W EH VXUSULVHG WRVHH=.
PHFKDQLFV DURXQG WKHFRQVWDQF\ RI WKHVSHHG RI OLJKW&RRN
(GPRQGV .DUS DQG /HYLQ VHWRXW WR UHEXLOG FRPSXWLQJ
DURXQG WKHQRWLRQRI WUDFWDELOLW\$ SUREOHP LVWUDFWDEOHLILW
FDQEH VROYHG LQWLPHJURZLQJSRO\QRUPDOO\ LQWKHLQSXWVL]H
ZKLFK LVD IDQF\ZD\ RI VD\LQJUHDVRQDEO\ IDVW1RQH RIWKH
-XUDVVLF,. DSSHDU WREH WUDFWDEOH$W OHDVWWKRVH LQWKHNQRZ
EHOLHYH WKH\DUHQRW"RI FRXUVH QRWVR ORQJDJR WKRVHLQWKH
NQRZ EHOLHYHG WKHHDUWKZDV IODW6DGO\ WKHJUHDWSURPLVH RI
FRPSXWLQJ VHHPV WROLHZLWKSUREOHPV DIIOLFWHGZLWK H[SRQHQ
WLDOLWLVWKHGUHDGHG DLOPHQW WKDWSODFHV HYHQ VPDOOVL]H SURE
 'RQ
W WU\WKLVDWKRPH LI\RX
UH
DOUHDG\ LQIXOOYLHZ
+HUH LVD=. TXHVWLRQ IRUWKH6WDWH'HSDUWPHQW FDQ D VLJ
QDWRU\ WRWKH1XFOHDU 1RQ3UROLIHUDWLRQ 7UHDW\ GHPRQVWUDWH
FRPSOLDQFH ZLWKRXW UHYHDOLQJ DQ\ LQIRUPDWLRQZKDWVRHYHU
DERXW LWVQXFOHDU IDFLOLWLHV"-XVWDVJDPH WKHRU\LQIOXHQFHGWKH
WKLQNLQJRIFROGZDU VWUDWHJLVWVGRQ
WKHRU\EHFRPH WKHUDJH LQLQWHUQDWLRQDOUHODWLRQVFLUFOHV
7UDFWDELOLW\ UHDFKHV IDUEH\RQG WKHUDFHWUDFNZKHUH FRP
SXWLQJFRPSHWHV IRUVSHHG ,WOLWHUDOO\IRUFHVXV WRWKLQNGLIIHU
HQWO\7KH DJHQWRIFKDQJH LVWKHXELTXLWRXV DOJRULWKP
7KH 5HYROXWLRQ
$OJRULWKPLF
$Q DOJRULWKP LVDQ L3RGSURJUDPZLWK DKXPDQ IDFH ,ID
FRPSXWHU FRXOGZDVK \RXUKDLU LWVSURJUDPZRXOG ORRN OLNH
 EXW WKHDOJRULWKP EHKLQG LWPLJKW
UHDG 5LQVH ODWKHUUHSHDW"
OHPVEH\RQG WKHUHDFKRIDQ\FRPSXWHU DFRPSXWHUVFLHQWLVW$Q DOJRULWKP LVD OLVWRILQVWUXFWLRQVWKDW
7KLV PXFK ZH NQRZ LW
VJHQHWLF ,ID VLQJOH RQH RI WKH WHOOVWKHFRPSXWHUZKDW WRGR ,W
PD\ ORRSDURXQGDQGHQWHUWDLQ
ZRQGHU RIZRQGHUV DOORIWKHP
-XUDVVLF,. LVWUDFWDEOHWKHQ DOWHUQDWLYHVDV LQ5LQVH ODWKHUUHSHDW LIXQKDSS\GU\JR WR
DUH7KHVH WRXJKSX]]OHV DUHQRWKLQJEXWGLIIHUHQWWUDQVODWLRQV RIILFHDQVZHU TXHVWLRQ ZK\GLGQ
W\RX ULQVH WKHVKDPSRR RII
RI WKH VDPH 6KDNHVSHDUH SOD\+HDG\ VWXII7KH GD\ \RXU \RXUKDLU9 $Q DOJRULWKP LVLQHVVHQFH DZRUN RIOLWHUDWXUH
SOD\OLVW TXHVWLRQ FDQEH DQVZHUHG LQD IHZKRXUVZLOO EH WKH 7KH OLEUDU\
V ERWWRP VKHOYHV PLJKW VWDFN WKHRQHOLQH

GD\SXEOLFNH\ FU\SWRJUDSK\GLHV EULQJLQJGRZQZLWK LWDOORI ]LQJHUV"DOJRULWKPLF PLQLDWXUHV WKDWORRS WKURXJKD WULYLDO

 $35,/

Bernard Chazelle, “Could your iPod be holding the greatest mystery in modern science.” Math Horizons, April 2006. Copyright Mathematical Association
of America. Article provided by Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http:
//www.cs.princeton.edu/~chazelle/pubs/ipod.pdf.
 6 COULD YOUR IPOD BE HOLDING THE GREATEST MYSTERY IN MODERN SCIENCE35

0$7+ +25,=216

DOJHEUDLF FDOFXODWLRQ WRSURGXFHIUDFWDOV SLFWXUHVRIGD]]OLQJ D YHULILHUSLFN WHQOLQHVDWUDQGRP2Q WKHEDVLV RIWKHVHWHQ

EHDXW\ DQG LQILQLWHLQWULFDF\ RUSULQW WKH WUDQVFHQGHQWDOO\ OLQHV DORQH WKHYHULILHUZLOO GHFLGH ZKHWKHU \RXUSURRI LV
P\VWHULRXV GLJLWVRI$OJRULWKPLF ]LQJHUVFDQGR HYHU\WKLQJ FRUUHFW7KH VKRFNHUEH\RQG DQ\UHDVRQDEOH GRXEWKHZLOO EH
ZH KDYH WKHVRQQHWVRQ WKH
)RU WKHUHVW PLGGOH VKHOYHV:LWK ULJKW
QDPHV OLNH))756$ /// $.6 WKH\DUHVKRUWDQGFULVSDQG 7KHPLQG UHHOV ,I\RXUSURRI LVILQHWKHQLW ZLOO SDVV DQ\
WHQGWRSDFNPRUH LQJHQXLW\SHUVTXDUH LQFKWKDQDQ\WKLQJHOVH WHVWWKHYHULILHUFDQ WKURZDW LW%XW EDVHG RQ RQO\ WHQOLQHV
LQWKHFRPSXWLQJZRUOG 7KH WRSVKHOYHVKROG WKHOXVKULFKO\ KRZ FDQKHNQRZ WKDW\RX
YH SURYHQ WKH5LHPDQQ K\SRWKHVLV
PXOWLOD\HUHG QRYHOV
WH[WXUHG DQGQRWDEDE\ FRXVLQ OLNH " ,I\RXUSURRI LVERJXV WKH
*LYH LWWRWKHPDOJRULWKPLF]LQJHUVNQRZ KRZ WRPDNH D LQWXLWLRQGRHV QRWKHOSPXFK HLWKHU3UHVXPDEO\ WKHDJUHHG
VFLHQWLVWVZRRQ1R RQHZKR
V HYHU WULHGWRFRPSXWH WKHGLJLWV XSRQ IRUPDWLVGHVLJQHG WRVPHDUDQ\EXJ DOODFURVV WKHSURRI
RI 7FE\KDQG FDQ UHPDLQXQPRYHG DW WKHVLJKWRI LWVGHFLPDO %XW KRZZLOO WKHYHULILHUEH VXUHWKDW\RXGLGQ
WSOD\ IDVWDQG
H[SDQVLRQ IORRGLQJDFRPSXWHU VFUHHQ OLNHODYD IORZLQJGRZQ ORRVHZLWK WKHIRUPDWWLQJUXOHV"7KH 3&3 DOJRULWKPXSHQGV
DYROFDQR$QG WKDW
VQRWHYHQ WKHDZHVRPH SDUW)RU WKDW
RQH EDVLF QRWLRQV RIHYLGHQFH DQG SHUVXDVLRQ DQG DFFRPSOLVKHV
PXVW WXUQWRWKHLQIDPRXV%UD]LOLDQ EXWWHUIO\
ZKRVH HYLOZLQJ ZKDW LVXVXDOO\ SKLORVRSK\
V SUHURJDWLYH WRWXUQWKHFRPSUH
IODSVFDXVH W\SKRRQVLQ&KLQD =LQJHUV HPERG\ WKHSRWHQWLDO KHQGHG LQWRWKHLQFRPSUHKHQVLEOH 6RPHZKHUH:LWWJHQVWHLQ
RID ORFDO DFWLRQ WRXQOHDVK FRORVVDO IRUFHVRQ DJOREDO VFDOH PXVW EH VPLOLQJ
FRPSOH[LW\HPHUJLQJRXWRIWULYLDOLW\
&HOOXODU DXWRPDWDFKDRV 0RRUH
V /DZ KDV SXWFRPSXWLQJ RQ WKHPDS DOJRULWKPV
G\QDPLFDO V\VWHPVDQGDOO WKDW
WKHRU\ ZLOO QRZ XQOHDVK LWVWUXHSRWHQWLDO 3K\VLFV DVWURQRP\DQG
)RU DJOLPSVH RIWKHILFWLRQJHQUHRQ WKHWRSVKHOYHV FKHFN FKHPLVWU\DUH DOO VFLHQFHV RIIRUPXODH &KDRV WKHRU\PRYHG
RXW3&3 6XSSRVH WKDWDIWHUSRSSLQJ WKHJHQLXVSLOO\RXZDNH WKHDOJRULWKPLF]LQJHU WRFHQWHUVWDJH7KH TXDQWLWDWLYH VFLHQFHV
XSRQH EULJKW
PRUQLQJZLWKD FRPSOHWHSURRIRIWKH5LHPDQQ RI WKHVW FHQWXU\HJJHQRPLFV QHXURELRORJ\ ZLOO FRP
K\SRWKHVLV \RXUKHDG WKDW
VWKH1RWRULRXV %,* RIPDWK
LQ SOHWH WKHGHWKURQHPHQWRIWKHIRUPXOD E\SODFLQJ WKHDOJRULWKP
UDS WKHELJJHVWRSHQSUREOHP LQWKHILHOG)HZ QXPEHU WKHR DW WKHFRUH RI WKHLU
PRGXV RSHUDQGL$OJRULWKPLF WKLQNLQJLV
ULVWVDUH OLNHO\WROLVWHQWR\RXUVWRU\
7KDW LVXQWLO\RX RIIHU OLNHO\WRFDXVH WKHPRVW GLVUXSWLYHSDUDGLJP VKLIWLQWKHVFL
WKHPWKH3&3 3UREDELOLVWLFDOO\&KHFNDEOH 3URRI GHDO<RX
OO HQFHV VLQFH TXDQWXPPHFKDQLFV $QG\HV \RXPD\ WUXVWWKH
ZULWHGRZQ \RXUSURRI LQDQDJUHHGXSRQ IRUPDWDQG WKHQOHW IXWXUHWREHNLQG WRWKLVSUHGLFWLRQ

L"7KH ,QGH[LQJ &RQWLQXHV "A"L"

WR
,QGH[  
3UREOHPV
0DWKHPDWLFDO
44+2S+
,QFOXGHVSUREOHPV IURP MRXUQDOVVXFKDV
A

?<A!A-
 
7ZR<HDU&ROOHJH 0DWKHPDWLFV-RXUQDO
";AA7A9A  M
DQG
0DUN
%RZURQ , 7KH $PHULFDQ 0DWKHPDWLFDO 0RQWKO\ ,,
A!AA
0DWKHPDWLFV 0DJD]LQH
3UREOHPV

RIRYWU 7KH 1HZ <RUN 6 DWH 0DWKHPDWLFV 7HDFKHUV -RXUQDO ,,
ZLWK $XWKRU
.QYZRUG
6XEMHFW DQG
&LWDWLRQ
,QGH[HV
P
 &UX[ 0DWKHPDWLFRUXP
A"D 6AN -RXUQDO RI5HFUHDWLRQDO 0DWKHPDWLFV HWF

9+++++O SS9[
+DUGERXQG 
,6%1 
,QGH[ WR0DWKHPDWLFDO 3UREOHPV FRQWDLQV WKHWH[WIURPWKRXVDQGV RISUREOHPV WKDW ZHUH SXEOLVKHG LQMRXUQDO L ,
SUREOHPFROXPQV GXULQJWKH\HDUV  7KH SUREOHPV DUH FODVVLILHG DQG WKHQVRUWHG E\WRSLF5HIHUHQFHV D `1A
ZKHUH WKHVROXWLRQWRWKHSUREOHPFDQ EH IRXQGLQWKHOLWHUDWXUH$OVR LQFOXGHGDUH FRPSUHKHQVLYH O0DWK3UR3UHO
DUH JLYHQ IRU
DXWKRU WLWOHFLWDWLRQDQG NH\ZRUG LQGH[HV DVZHOODV SUREOHPV IURPQDWLRQDODQG LQWHUQDWLRQDO PDWKHPDWLFDO RO\PSLDGV
%URZVH WKHSUREOHPV IRULQVSLUDWLRQRU ORFDWHRQH RQD VSHFLILF WRSLF ,QFOXGHVD OLVW
RIXQVROYHG SUREOHPV

7KHUH LVORWV
RIIUHHLQIRUPDWLRQ
DERXW SUREOHPVSUREOHPERRNV SUREOHPVROYLQJDQG SUREOHP MRXUQDOVRQRXUZHE VLWH

BZZZPDWKSURSUHVV
7R RUGHUFDOO WROOIUHH
FRPB
LQWKH86 RU&DQDGD RUHDV\ RUGHURQOLQHDWZZZ0DWK3UR3UHVVFRP

:::0$$25*0$7++25,=216

Bernard Chazelle, “Could your iPod be holding the greatest mystery in modern science.” Math Horizons, April 2006. Copyright Mathematical Association
of America. Article provided by Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http:
//www.cs.princeton.edu/~chazelle/pubs/ipod.pdf.
 77 SCOOPING THE LOOP SNOOPER: A PROOF THAT THE HALTING PROBLEM IS UNDECIDABL

SCOOPING THE LOOP SNOOPER

A proof that the Halting Problem is undecidable
Geoffrey K. Pullum
(School of Philosophy, Psychology and Language Sciences, University of Edinburgh)

No general procedure for bug checks will do.

Now, I won’t just assert that, I’ll prove it to you.
I will prove that although you might work till you drop,
you cannot tell if computation will stop.

For imagine we have a procedure called P

that for specified input permits you to see
whether specified source code, with all of its faults,
defines a routine that eventually halts.

You feed in your program, with suitable data,

and P gets to work, and a little while later
(in finite compute time) correctly infers
whether infinite looping behavior occurs.

If there will be no looping, then P prints out ‘Good.’

That means work on this input will halt, as it should.
But if it detects an unstoppable loop,
then P reports ‘Bad!’ — which means you’re in the soup.

Well, the truth is that P cannot possibly be,

because if you wrote it and gave it to me,
I could use it to set up a logical bind
that would shatter your reason and scramble your mind.

Here’s the trick that I’ll use — and it’s simple to do.
I’ll define a procedure, which I will call Q,
that will use P’s predictions of halting success
to stir up a terrible logical mess.

For a specified program, say A, one supplies,

the first step of this program called Q I devise
is to find out from P what’s the right thing to say
of the looping behavior of A run on A.

Geoffrey K. Pullum, “Scooping the Loop Snooper: a proof that the halting problem is undecidable.” Mathematics Magazine, October 2000. Copyright Mathematics
Magazine. Article provided by Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://www.lel.
ed.ac.uk/~gpullum/loopsnoop.html.
 7 SCOOPING THE LOOP SNOOPER: A PROOF THAT THE HALTING PROBLEM IS UNDECIDABL

If P’s answer is ‘Bad!’, Q will suddenly stop.

But otherwise, Q will go back to the top,
and start off again, looping endlessly back,
till the universe dies and turns frozen and black.

And this program called Q wouldn’t stay on the shelf;

I would ask it to forecast its run on itself.
When it reads its own source code, just what will it do?
What’s the looping behavior of Q run on Q?

If P warns of infinite loops, Q will quit;

yet P is supposed to speak truly of it!
And if Q’s going to quit, then P should say ‘Good.’
Which makes Q start to loop! (P denied that it would.)

No matter how P might perform, Q will scoop it:

Q uses P’s output to make P look stupid.
Whatever P says, it cannot predict Q:
P is right when it’s wrong, and is false when it’s true!

I’ve created a paradox, neat as can be —

and simply by using your putative P.
When you posited P you stepped into a snare;
Your assumption has led you right into my lair.

So where can this argument possibly go?

I don’t have to tell you; I’m sure you must know.
A reductio: There cannot possibly be
a procedure that acts like the mythical P.

You can never find general mechanical means

for predicting the acts of computing machines;
it’s something that cannot be done. So we users
must find our own bugs. Our computers are losers!

In October 2000, after a refereeing delay of nearly a year, an earlier and incorrect version of this poetic proof was published in Mathematics
Magazine (73, no. 4, 319–320). But it had an error. I am very grateful to Philip Wadler (Informatics, University of Edinburgh) and Larry
Moss (Mathematics, Indiana University) for helping with the development of this corrected version, which is now free of bugs (trust me;
you can check it). Thanks also to the late Dr. Seuss for the style, and of course to the pioneering work of Alan Turing (and Martin Davis’s
nice simplified presentation) for the content. Copyright © 2008 by Geoffrey K. Pullum. Permission is hereby granted to reproduce or
distribute this work for non-commercial, educational purposes relating to the teaching of computer science, mathematics, or logic, provided
this attribution is included.

Geoffrey K. Pullum, “Scooping the Loop Snooper: a proof that the halting problem is undecidable.” Mathematics Magazine, October 2000. Copyright Mathematics
Magazine. Article provided by Library Reserves for private study, scholarship, or research. Do not redistribute. This article can also be found at http://www.lel.
ed.ac.uk/~gpullum/loopsnoop.html.