You are on page 1of 2

STAN HINDA/GETTY

Grandmaster Garry Kasparov during the last of six games against Deep Blue in 1997; the computer won the match by 3.5 games to 2.5.

ART IFICIAL INTELLIGENCE

Chess match of the century


Demis Hassabis lauds Garry Kasparovs account of battling supercomputer Deep Blue.

N
early 20 years ago, I was fortunate crucial traits such as tasks that people consider incredibly difficult,
enough to play friendly blitz chess generality, adaptability but totally fail at commonsense tasks we find
against former world champion and learning. intuitive (a phenomenon called Moravecs
Garry Kasparov. It was quite an experience; As he details in paradox). It was also due to industry and
his competitive spirit and creative genius Deep Thinking, Kasp- research dynamics in the 1980s and 1990s:
were palpable. I had recently founded Elixir arov reached a simi- in pursuit of quick results, labs ditched gen-
Studios, which specialized in artificial lar conclusion. The eralizable, learning-based approaches in
intelligence (AI) games, and my ambition book is his first thor- favour of narrow, hand-coded solutions that
was to conduct cutting-edge research in the ough account of the exploited machines computational speed.
field. AI was on my mind that day: Kasparov match, and it offers Deep Thinking: The focus on brute-force approaches had
had played chess against IBMs supercom- thoughtful medita- Where Machine upsides, Kasparov explains. It may not have
puter Deep Blue just a few years before. Now, tions on technology. Intelligence delivered on the promise of general-purpose
he sets out the details of that titanic event in The title references Ends and Human AI, but it did result in very powerful chess
Creativity Begins
his memoir Deep Thinking. what he believes chess GARRY KASPAROV engines that soon became popularly avail-
The 1997 match was a watershed for AI engines cannot do: PublicAffairs: 2017. able. Today, anyone can practise for free
and an extraordinary technical feat. Strangely, they can calculate, but against software stronger than the greatest
although Kasparov lost, it left me more in awe not innovate or create. They cannot think human chess masters, enabling enthusiasts
of the incredible capabilities of the human in the deepest sense. In drawing out these worldwide to train at top levels. Before Deep
brain than of the machine. Kasparov was able distinctions, Kasparov provides an impres- Blue, pessimists predicted that the defeat of a
to compete against a computational leviathan sively researched history of AI and the fields world chess champion by a machine would
and to complete myriad other tasks that make ongoing obsession with chess. lead to the games death. In fact, more people
us all distinctly human. By contrast, Deep For decades, leading computer scientists play now than ever before, according to
Blue was hard-coded with a set of specialized believed that, given the traditional status World Chess Federation figures.
rules distilled from chess grandmasters, and of chess as an exemplary demonstration Chess engines have also given rise to
empowered with a brute-force search algo- of human intellect, a competent computer exciting variants of play. In 1998, Kasparov
rithm. It was programmed to do one thing chess player would soon also surpass all other introduced Advanced Chess, in which
only; it could not have played even a simpler human abilities. That proved not to be the humancomputer teams merge the calcu-
game such as noughts and crosses with- case. This has to do partly with differences lation abilities of machines with a persons
out being completely reprogrammed. I felt between human and machine cognition: pattern-matching insights. Kasparovs
that this brand of intelligence was missing computers can easily perform calculation embrace of the technology that defeated

2 7 A P R I L 2 0 1 7 | VO L 5 4 4 | NAT U R E | 4 1 3

2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
COMMENT BOOKS & ARTS

him shows how computers can inspire,

ZEPHYR/SPL
rather than obviate, human creativity.
In Deep Thinking, Kasparov also delves
into the renaissance of machine learning,
an AI subdomain focusing on general-
purpose algorithms that learn from data.
He highlights the radical differences
between Deep Blue and AlphaGo, a learn-
ing algorithm created by my company
DeepMind to play the massively complex
game of Go. Last year, AlphaGo defeated
Lee Sedol, widely hailed as the great-
est player of the past decade. Whereas
Deep Blue followed instructions care-
fully honed by a crack team of engineers
and chess professionals, AlphaGo played
against itself repeatedly, learning from its
mistakes and developing novel strategies.
Several of its moves against Lee had never
been seen in human games most nota-
bly move 37 in game 2, which upended
centuries of traditional Go wisdom by
playing on the fifth line early in the game. An angiogram and computed-tomography scan of a mans brain, used to locate his language centre.
Most excitingly, because its learning
algorithms can be generalized, AlphaGo PSYC H O LO GY

Science in spite of itself


holds promise far beyond the game for
which it was created. Kasparov relishes
this potential, discussing applications from
machine translation to automated medical
diagnoses. AI will not replace humans, he
argues, but will enlighten and enrich us, Barbara A. Spellman hails an analysis of reproducibility
much as chess engines did 20 years ago. in psychology by a champion for change.
His position is especially notable coming
from someone who would have every

C
reason to be bitter about AIs advances. hris Chamberss portrait should (D.J.Bem J.Pers. Soc.
His account of the Deep Blue match sit high on the wall of heroes in the Psychol. 100, 407425;
itself is fascinating. Famously, Kasparov movement to reform science. A 2011). Published in
stormed out of one game and gave antag- cognitive neuroscientist and psychologist, the American Psycho-
onistic press conferences in which he Chambers has had an important role as an logical Associations
protested against IBMs secrecy around editor and advocate in identifying, challeng- prestigious Journal
the Deep Blue team and its methods, and ing and changing practices responsible for of Personality and
insinuated that the company might have the reproducibility crisis. Social Psychology, it
cheated. In Deep Thinking, Kasparov offers Many fields of science social, life, physi- left many psycholo-
an engaging insight into his psychological cal and medical have had to acknowledge The Seven gists outraged. The
state during the match. To a degree, he in recent years that much of their published Deadly Sins of article had followed
walks back on his earlier claims, conclud- research is not replicable (see M.Munaf Psychology: the rules of scientific
ing that although IBM probably did not Nature 543, 619620; 2017). Psychological A Manifesto practice. Its stud-
for Reforming
cheat, it violated the spirit of fair competi- science was hit hard by that problem early this the Culture of
ies all supported the
tion by obscuring useful information. He decade. But it quickly joined the vanguard of Scientific Practice same hypothesis; its
also provides a detailed commentary on reform. In The Seven Deadly Sins of Psychol- CHRIS CHAMBERS methods included
several crucial moments; for instance, he ogy part history, part autobiography, largely Princeton University randomization and
dispels the myth that Deep Blues bizarre manifesto Chambers identifies some sins, Press: 2017. standard-looking data
move 44 in the first game of the match left from biased reasoning to outright fraud, that analyses. But close
him unrecoverably flummoxed. led us to this point. And he describes specific scrutiny of the paper and subsequent failures
Kasparov includes enough detail to reforms, some already well under way, that to replicate the studies (plus the unwilling-
satisfy chess enthusiasts, while providing will make science more transparent, acces- ness of journals to publish those failures)
a thrilling narrative for the casual reader. sible and reproducible. As he shows, the sins revealed many of the sins.
Deep Thinking delivers a rare balance of are (mostly) not those of individual scientists, The sin that makes the biggest news splash
analysis and narrative, weaving commen- but of the processes and incentive structures is outright fraud changing or fabricating
tary about technological progress with an under which scientists work. data, or making up an entire study. It may
inside look at one of the most important Chambers ably illustrates these failings be the least interesting (and, we hope, the
chess matches ever played. with tales from psychological science. The least prevalent) sin, but is illustrated by
first chapter describes a 2011 paper by social Chamberss tale of a psychologist who did
Demis Hassabis is the founder and chief psychologist Daryl Bem that reported nine just that at least 58 times. If you stop to
executive of DeepMind, a neuroscience- experiments demonstrating evidence of pre- ask why a scientist would commit fraud,
inspired AI company based in London. cognition the ability to predict the future the perverse nature of scientific processes

4 1 4 | NAT U R E | VO L 5 4 4 | 2 7 A P R I L 2 0 1 7

2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.

You might also like