1
00:00:01,175 --> 00:00:04,215
This story begins in 1985,

2
00:00:04,215 --> 00:00:06,157
when at age 22,

3
00:00:06,157 --> 00:00:08,708
I became the world chess champion

4
00:00:08,708 --> 00:00:12,636
after beating Anatoly Karpov.

5
00:00:12,636 --> 00:00:13,994
Earlier that year,

6
00:00:13,994 --> 00:00:17,586
I played what is called
simultaneous exhibition

7
00:00:17,586 --> 00:00:22,001
against 32 of the world's
best chess playing machines

8
00:00:22,001 --> 00:00:24,179
in Hamburg, Germany.

9
00:00:24,179 --> 00:00:26,386
I won all the games,

10
00:00:26,386 --> 00:00:29,686
and then it was not considered
much of a surprise

11
00:00:29,686 --> 00:00:34,752
that I could beat 32 computers
at the same time.

12
00:00:34,752 --> 00:00:37,291
To me, that was the golden age.

13
00:00:37,291 --> 00:00:39,847
(Laughter)

14
00:00:39,847 --> 00:00:41,671
Machines were weak,

15
00:00:41,671 --> 00:00:43,684
and my hair was strong.

16
00:00:43,684 --> 00:00:46,030
(Laughter)

17
00:00:46,030 --> 00:00:49,076
Just 12 years later,

18
00:00:49,076 --> 00:00:53,706
I was fighting for my life
against just one computer

19
00:00:53,706 --> 00:00:55,041
in a match

20
00:00:55,041 --> 00:00:57,676
called by the cover of Newsweek

21
00:00:57,676 --> 00:00:59,573
"The Brain's Last Stand."

22
00:00:59,573 --> 00:01:00,971
No pressure.

23
00:01:00,971 --> 00:01:03,140
(Laughter)

24
00:01:03,140 --> 00:01:06,018
From mythology to science fiction,

25
00:01:06,018 --> 00:01:08,436
human versus machine

26
00:01:08,436 --> 00:01:12,274
has been often portrayed
as a matter of life and death.

27
00:01:12,274 --> 00:01:15,611
John Henry called the steel-driving man

28
00:01:15,611 --> 00:01:19,394
in the 19th century
African American folk legend

29
00:01:19,394 --> 00:01:20,889
was pitted in a race

30
00:01:20,889 --> 00:01:23,748
against steam-powered hammer

31
00:01:23,748 --> 00:01:27,247
bashing a tunnel through mountain rock.

32
00:01:27,247 --> 00:01:31,883
John Henry's legend is a part
of a long historical narrative

33
00:01:31,883 --> 00:01:36,026
pitting humanity versus technology.

34
00:01:36,026 --> 00:01:39,840
And this competitive rhetoric
is standard now.

35
00:01:39,840 --> 00:01:42,452
We are in a race against the machines,

36
00:01:42,452 --> 00:01:46,146
in a fight or even in a war.

37
00:01:46,146 --> 00:01:47,779
Jobs are being killed off.

38
00:01:47,779 --> 00:01:51,902
People are being replaced
as if they had vanished from the Earth.

39
00:01:51,902 --> 00:01:55,992
It's enough to think that the movies
like "The Terminator" or "The Matrix"

40
00:01:55,992 --> 00:01:58,180
are nonfiction.

41
00:01:58,180 --> 00:02:05,577
There are very few instances in an arena

42
00:02:05,577 --> 00:02:09,715
where the human body and mind
can compete on equal terms

43
00:02:09,715 --> 00:02:12,454
with a computer or a robot.

44
00:02:12,454 --> 00:02:14,841
Actually, I wish there were a few more.

45
00:02:14,841 --> 00:02:22,971
Instead, it was my blessing and my curse

46
00:02:22,971 --> 00:02:25,500
to literally become the proverbial man

47
00:02:25,500 --> 00:02:28,582
in the man versus machine competition

48
00:02:28,582 --> 00:02:33,362
that everybody is still talking about.

49
00:02:33,362 --> 00:02:38,433
In a most famous human-machine
competition since John Henry,

50
00:02:38,433 --> 00:02:41,003
I played two matches

51
00:02:41,003 --> 00:02:44,616
against the IBM supercomputer Deep Blue.

52
00:02:47,539 --> 00:02:49,758
Nobody remembers
that I won the first match --

53
00:02:49,758 --> 00:02:51,596
(Laughter)

54
00:02:51,596 --> 00:02:57,750
(Applause) --

55
00:02:57,750 --> 00:03:00,849
in Philadelphia, before losing the rematch
the following year in New York.

56
00:03:00,849 --> 00:03:04,286
But I guess that's fair.

57
00:03:04,286 --> 00:03:09,841
There is no day in history
special calendar entry

58
00:03:09,841 --> 00:03:13,348
for all the people who failed
to climb Mt. Everest

59
00:03:13,348 --> 00:03:16,033
before Sir Edmund Hillary
and Tenzing Norgay

60
00:03:16,033 --> 00:03:18,152
made it to the top.

61
00:03:18,470 --> 00:03:22,683
And in 1997, I was still
the world champion

62
00:03:22,683 --> 00:03:30,012
when chess computers finally came of age.

63
00:03:30,012 --> 00:03:32,020
I was Mt. Everest,

64
00:03:32,020 --> 00:03:34,264
and Deep Blue reached the summit.

65
00:03:34,264 --> 00:03:38,946
I should stay of course,
not that Deep Blue did it,

66
00:03:38,946 --> 00:03:41,391
but its human creators --

67
00:03:41,391 --> 00:03:44,748
Anantharaman, Campbell, Hoane, Hsu.

68
00:03:44,748 --> 00:03:46,374
Hats off to them.

69
00:03:46,374 --> 00:03:51,626
As always, machine's triumph
was a human triumph,

70
00:03:51,626 --> 00:03:58,585
something we tend to forget when humans
are surpassed by our own creations.

71
00:03:58,585 --> 00:04:00,780
Deep Blue was victorious,

72
00:04:00,780 --> 00:04:03,687
but was it intelligent?

73
00:04:03,687 --> 00:04:06,600
No, no it wasn't,

74
00:04:06,600 --> 00:04:11,539
at least not in the way Alan Turing
and other founders of computer science

75
00:04:11,539 --> 00:04:13,486
had hoped.

76
00:04:13,486 --> 00:04:18,048
It turned out that chess could be
crunched by brute force,

77
00:04:18,048 --> 00:04:22,625
once hardware got fast enough

78
00:04:22,625 --> 00:04:26,637
and algorithms got smart enough.

79
00:04:26,637 --> 00:04:30,453
Although by the definition of the output,

80
00:04:30,453 --> 00:04:36,772
grandmaster-level chess,
Deep Blue was intelligent,

81
00:04:36,772 --> 00:04:40,525
but even at the incredible speed,

82
00:04:40,525 --> 00:04:45,549
200 million positions per second,

83
00:04:45,549 --> 00:04:47,833
Deep Blue's method

84
00:04:47,833 --> 00:04:54,986
provided little of the dreamed-of insight
into the mysteries of human intelligence.

85
00:04:57,425 --> 00:05:01,266
Soon, machines will be taxi drivers

86
00:05:01,266 --> 00:05:03,860
and doctors and professors,

87
00:05:03,860 --> 00:05:06,145
but will they be "intelligent?"

88
00:05:06,145 --> 00:05:10,115
I would rather leave these definitions

89
00:05:10,115 --> 00:05:15,472
to the philosophers and to the dictionary.

90
00:05:15,472 --> 00:05:19,809
What really matters is how we humans

91
00:05:19,809 --> 00:05:25,703
feel about living and working
with these machines.

92
00:05:25,703 --> 00:05:31,929
When I first met Deep Blue
in 1996 in February,

93
00:05:31,929 --> 00:05:35,340
I had been the world champion
for more than 10 years,

94
00:05:35,340 --> 00:05:40,634
and I had played 182
world championship games

95
00:05:40,634 --> 00:05:45,672
and hundreds of games against
other top players in other competitions.

96
00:05:45,672 --> 00:05:50,640
I knew what to expect
from my opponents

97
00:05:50,640 --> 00:05:53,246
and what to expect from myself.

98
00:05:53,246 --> 00:05:55,600
I was used to measure

99
00:05:55,600 --> 00:06:01,889
their moves and to gauge
their emotional state

100
00:06:01,889 --> 00:06:04,233
by watching their body language
and looking into their eyes.

101
00:06:04,233 --> 00:06:12,192
And then I sat across
the chess board from Deep Blue.

102
00:06:12,192 --> 00:06:16,375
I immediately sensed something new,

103
00:06:16,375 --> 00:06:18,631
something unsettling.

104
00:06:18,631 --> 00:06:26,390
You might experience a similar feeling

105
00:06:26,390 --> 00:06:28,339
the first time you ride
in a driverless car

106
00:06:28,339 --> 00:06:32,499
or the first time your new computer
manager issues an order at work.

107
00:06:32,499 --> 00:06:38,745
But when I sat at that first game,

108
00:06:38,745 --> 00:06:42,804
I couldn't be sure

109
00:06:42,804 --> 00:06:45,485
what is the thing capable of.

110
00:06:45,485 --> 00:06:49,102
Technology can advance in leaps,
and IBM had invested heavily.

111
00:06:49,102 --> 00:06:52,816
I lost that game,

112
00:06:52,816 --> 00:06:54,821
and I couldn't help wondering,

113
00:06:54,821 --> 00:06:57,427
might it be invincible?

114
00:06:57,427 --> 00:07:02,027
Was my beloved game of chess over?

115
00:07:02,027 --> 00:07:05,904
These were human doubts, human fears,

116
00:07:05,904 --> 00:07:07,743
and the only thing I knew for sure

117
00:07:07,743 --> 00:07:13,133
was that my opponent Deep Blue
had no such worries at all.

118
00:07:13,133 --> 00:07:14,189
(Laughter)

119
00:07:14,189 --> 00:07:16,768
I fought back

120
00:07:16,768 --> 00:07:19,451
after this devastating blow
to win the first match,

121
00:07:19,451 --> 00:07:24,877
but the writing was on the wall.

122
00:07:24,877 --> 00:07:28,113
I eventually lost to the machine

123
00:07:28,113 --> 00:07:30,217
but I didn't suffer the fate of John Henry

124
00:07:30,217 --> 00:07:37,927
who won but died
with his hammer in his hand.

125
00:07:37,927 --> 00:07:40,717
It turned out that the world of chess

126
00:07:40,717 --> 00:07:45,293
still wanted to have
a human chess champion.

127
00:07:45,293 --> 00:07:50,069
And even today, when a free chess app

128
00:07:50,069 --> 00:07:54,360
on the latest mobile phone
is stronger than Deep Blue,

129
00:07:54,360 --> 00:07:57,090
people are still playing chess,

130
00:07:57,090 --> 00:08:01,741
even more than ever before.

131
00:08:01,741 --> 00:08:04,013
Doomsayers predicted that nobody
would touch the game

132
00:08:04,013 --> 00:08:05,742
that could be conquered by the machine,

133
00:08:05,742 --> 00:08:08,099
and they were wrong, proving wrong,

134
00:08:08,099 --> 00:08:11,695
but doomsaying has always been
a popular pasttime

135
00:08:11,695 --> 00:08:14,840
when it comes to technology.

136
00:08:14,840 --> 00:08:17,660
What I learned from my own experience

137
00:08:17,660 --> 00:08:22,380
is that we must face our fears

138
00:08:22,380 --> 00:08:25,343
if we want to get the most
out of our technology,

139
00:08:25,343 --> 00:08:29,102
and we must conquer those fears

140
00:08:29,102 --> 00:08:35,698
if we want to get the best
out of our humanity.

141
00:08:35,698 --> 00:08:38,242
While licking my wounds,

142
00:08:38,242 --> 00:08:41,726
I got a lot of inspiration

143
00:08:41,726 --> 00:08:44,556
from my battles against Deep Blue.

144
00:08:44,556 --> 00:08:49,367
As the old Russian saying goes,
if you can't beat them, join them.

145
00:08:49,367 --> 00:08:50,676
Then I thought,

146
00:08:50,676 --> 00:08:54,119
what if I could play with a computer,

147
00:08:54,119 --> 00:08:57,545
together with a computer at my side,
combining our strengths,

148
00:08:57,545 --> 00:09:01,998
human intuition
plus machine's calculation,

149
00:09:01,998 --> 00:09:04,252
human strategy, machine tactics,

150
00:09:04,252 --> 00:09:06,513
human experience, machine's memory.

151
00:09:06,513 --> 00:09:10,556
Could it be the perfect game ever played?

152
00:09:10,556 --> 00:09:13,245
My idea came to life

153
00:09:13,245 --> 00:09:16,960
in 1998 on the name of advanced chess

154
00:09:16,960 --> 00:09:20,907
when I played this human-plus-machine
competition against another elite player,

155
00:09:20,907 --> 00:09:25,516
but in this first experiment,

156
00:09:25,516 --> 00:09:35,307
we both failed to combine human
and machine skills effectively.

157
00:09:35,307 --> 00:09:38,798
Advanced chess found
its home on the Internet,

158
00:09:38,798 --> 00:09:43,623
and in 2005, a so-called freestyle
chess tournament

159
00:09:43,623 --> 00:09:46,206
produced a revelation.

160
00:09:46,206 --> 00:09:51,055
A team of grandmasters
and top machines participated,

161
00:09:51,055 --> 00:09:54,632
but the winners were not grandmasters,

162
00:09:54,632 --> 00:09:55,986
not a supercomputer.

163
00:09:55,986 --> 00:10:01,138
The winners were a pair of amateur
American chess players

164
00:10:01,138 --> 00:10:05,944
operating three ordinary PCs
at the same time.

165
00:10:05,944 --> 00:10:09,101
Their skill of coaching their machines

166
00:10:09,101 --> 00:10:12,466
effectively counteracted
the superior chess knowledge

167
00:10:12,466 --> 00:10:16,593
of their grandmaster opponents

168
00:10:16,593 --> 00:10:21,964
and much greater computational
power of others,

169
00:10:21,964 --> 00:10:24,860
and I reached this formulation.

170
00:10:24,860 --> 00:10:28,349
A weak human player plus a machine

171
00:10:28,349 --> 00:10:31,550
plus a better process is superior

172
00:10:31,550 --> 00:10:34,007
to a very power machine alone,

173
00:10:34,007 --> 00:10:37,655
but more remarkably, is superior
to a strong human player

174
00:10:37,655 --> 00:10:46,803
plus machine and an inferior process.

175
00:10:46,803 --> 00:10:49,999
This convinced me that we would need

176
00:10:49,999 --> 00:10:54,716
better interfaces to help us
coach our machines

177
00:10:54,716 --> 00:10:57,672
towards more useful intelligence.

178
00:10:57,672 --> 00:11:01,935
Human plus machine isn't the future,

179
00:11:01,935 --> 00:11:03,108
it's present.

180
00:11:03,108 --> 00:11:09,620
Everybody that's used online translation
to get the gist of a news article

181
00:11:09,620 --> 00:11:11,755
from a foreign newspaper

182
00:11:11,755 --> 00:11:14,041
knowing its far from perfect.

183
00:11:14,041 --> 00:11:16,142
Then we use our human experience

184
00:11:16,142 --> 00:11:18,252
to make sense out of that,

185
00:11:18,252 --> 00:11:21,061
and then machine learns
from our corrections.

186
00:11:21,061 --> 00:11:27,045
This model is spreading and investing
in medical diagnosis, security analysis.

187
00:11:27,045 --> 00:11:30,527
The machine crunches data,

188
00:11:30,527 --> 00:11:31,576
calculates probabilities,

189
00:11:31,576 --> 00:11:35,047
gets 80 percent of the way, 90 percent,

190
00:11:35,047 --> 00:11:39,173
making it easier for analysis

191
00:11:39,173 --> 00:11:42,725
and decision-making of the human party.

192
00:11:42,725 --> 00:11:48,764
But you are not going to send your kids

193
00:11:48,764 --> 00:11:53,113
to school in self-driving car
with 90 percent accuracy,

194
00:11:53,113 --> 00:11:55,827
even with 99 percent.

195
00:11:55,827 --> 00:11:58,712
So we need a leap forward

196
00:11:58,712 --> 00:12:07,236
to add a few more crucial decimal places.

197
00:12:07,236 --> 00:12:12,450
Twenty years after
my match with Deep Blue,

198
00:12:12,450 --> 00:12:14,363
second match,

199
00:12:14,363 --> 00:12:19,365
this sensational "The Brain's
Last Stand" headline

200
00:12:19,365 --> 00:12:22,239
has become commonplace

201
00:12:22,239 --> 00:12:24,630
as intelligent machines

202
00:12:24,630 --> 00:12:30,721
move in every sector, seemingly every day.

203
00:12:30,721 --> 00:12:33,514
But unlike in the past,

204
00:12:33,514 --> 00:12:38,186
when machines replaced farm animals,

205
00:12:38,186 --> 00:12:42,637
manual labor, now they are coming after
people with college degrees

206
00:12:42,637 --> 00:12:44,637
and political influence,

207
00:12:44,637 --> 00:12:47,639
and as someone who fought machines
and lost, I am here to tell you

208
00:12:47,639 --> 00:12:51,323
this is excellent, excellent news.

209
00:12:51,323 --> 00:12:53,692
Eventually, every profession

210
00:12:53,692 --> 00:12:55,768
will have to feel these pressures

211
00:12:55,768 --> 00:13:02,583
or else it will mean humanity
has ceased to make progress.

212
00:13:02,583 --> 00:13:08,749
We don't get to choose

213
00:13:08,749 --> 00:13:13,484
when and where
technological progress stops.

214
00:13:13,484 --> 00:13:18,215
We cannot slow down.

215
00:13:18,215 --> 00:13:25,183
In fact, we have to speed up.

216
00:13:25,183 --> 00:13:29,764
Our technology excels at removing

217
00:13:29,764 --> 00:13:32,057
difficulties and uncertainties

218
00:13:32,057 --> 00:13:35,341
from our lives,

219
00:13:35,341 --> 00:13:38,361
and so we must seek out

220
00:13:38,361 --> 00:13:40,403
ever more difficult,

221
00:13:40,403 --> 00:13:48,766
ever more uncertain challenges.

222
00:13:48,766 --> 00:13:54,044
Machines have calculations.

223
00:13:54,044 --> 00:13:56,032
We have understanding.

224
00:13:56,032 --> 00:13:58,316
Machines have instructions.

225
00:13:58,316 --> 00:14:00,890
We have purpose.

226
00:14:00,890 --> 00:14:04,253
Machines have objectivity.

227
00:14:04,253 --> 00:14:08,781
We have passion.

228
00:14:08,781 --> 00:14:15,038
We should not worry about what
our machines can do today.

229
00:14:15,038 --> 00:14:19,457
Instead, we should worry about
what they still cannot do today,

230
00:14:19,457 --> 00:14:25,064
because we will need the help
of the new, intelligent machines

231
00:14:25,064 --> 00:14:30,410
to turn our grandest dreams into reality.

232
00:14:30,410 --> 00:14:32,597
And if we fail,

233
00:14:32,597 --> 00:14:36,299
if we fail, it's not because our machines

234
00:14:36,299 --> 00:14:39,631
are too intelligent,
or not intelligent enough.

235
00:14:39,631 --> 00:14:44,084
If we fail, it's because
we grew complacent

236
00:14:44,084 --> 00:14:47,110
and limited our ambitions.

237
00:14:47,110 --> 00:14:51,815
Our humanity is not defined by any skill,

238
00:14:51,815 --> 00:14:54,888
like swinging a hammer
or even playing chess.

239
00:14:54,888 --> 00:14:58,034
There's one thing only a human can do.

240
00:14:58,034 --> 00:15:00,049
That's dream.

241
00:15:00,049 --> 00:15:02,851
So let us dream big.

242
00:15:02,851 --> 00:15:04,869
Thank you.

243
00:15:04,869 --> 00:15:07,807
(Applause)