WEBVTT

00:00:00.960 --> 00:00:03.936
This story begins in 1985,

00:00:03.960 --> 00:00:05.936
when at age 22,

00:00:05.960 --> 00:00:08.336
I became the world chess champion

00:00:08.360 --> 00:00:11.560
after beating Anatoly Karpov.

00:00:12.480 --> 00:00:13.736
Earlier that year,

00:00:13.760 --> 00:00:17.376
I played what is called
simultaneous exhibition

00:00:17.400 --> 00:00:21.696
against 32 of the world's
best chess-playing machines

00:00:21.720 --> 00:00:23.080
in Hamburg, Germany.

00:00:24.160 --> 00:00:25.360
I won all the games,

00:00:26.560 --> 00:00:29.736
and then it was not considered
much of a surprise

00:00:29.760 --> 00:00:33.800
that I could beat 32 computers
at the same time.

00:00:34.480 --> 00:00:37.056
To me, that was the golden age.

00:00:37.080 --> 00:00:39.176
(Laughter)

00:00:39.200 --> 00:00:40.720
Machines were weak

00:00:41.680 --> 00:00:43.016
and my hair was strong.

00:00:43.040 --> 00:00:45.240
(Laughter)

00:00:46.720 --> 00:00:48.776
Just 12 years later,

00:00:48.800 --> 00:00:53.416
I was fighting for my life
against just one computer

00:00:53.440 --> 00:00:54.640
in a match

00:00:55.360 --> 00:00:57.416
called by the cover of "Newsweek"

00:00:57.440 --> 00:00:59.216
"The Brain's Last Stand."

00:00:59.240 --> 00:01:00.456
No pressure.

00:01:00.480 --> 00:01:02.000
(Laughter)

00:01:03.040 --> 00:01:05.616
From mythology to science fiction,

00:01:05.640 --> 00:01:08.376
human versus machine

00:01:08.400 --> 00:01:11.160
has been often portrayed
as a matter of life and death.

00:01:11.960 --> 00:01:13.536
John Henry,

00:01:13.560 --> 00:01:15.256
called the steel-driving man

00:01:15.280 --> 00:01:19.056
in the 19th century
African American folk legend

00:01:19.080 --> 00:01:20.456
was pitted in a race

00:01:20.480 --> 00:01:23.216
against a steam-powered hammer

00:01:23.240 --> 00:01:25.600
bashing a tunnel through mountain rock.

00:01:26.800 --> 00:01:31.000
John Henry's legend
is a part of a long historical narrative

00:01:31.680 --> 00:01:34.760
pitting humanity versus technology.

00:01:36.200 --> 00:01:39.080
And this competitive rhetoric
is standard now.

00:01:39.560 --> 00:01:41.520
We are in a race against the machines,

00:01:42.360 --> 00:01:44.440
in a fight or even in a war.

00:01:45.880 --> 00:01:47.496
Jobs are being killed off.

00:01:47.520 --> 00:01:51.080
People are being replaced
as if they had vanished from the Earth.

00:01:52.240 --> 00:01:55.736
It's enough to think that the movies
like "The Terminator" or "The Matrix"

00:01:55.760 --> 00:01:56.960
are nonfiction.

00:01:59.640 --> 00:02:03.960
There are very few instances
of an arena where --

00:02:05.360 --> 00:02:09.616
the human body and mind
can compete on equal terms

00:02:09.639 --> 00:02:11.480
with a computer or a robot.

00:02:12.280 --> 00:02:14.138
Actually, I wish there were a few more.

00:02:15.760 --> 00:02:16.960
Instead,

00:02:17.840 --> 00:02:22.496
it was my blessing and my curse

00:02:22.520 --> 00:02:25.216
to literally become the proverbial man

00:02:25.240 --> 00:02:28.336
in the man versus machine competition

00:02:28.360 --> 00:02:30.240
that everybody is still talking about.

00:02:33.120 --> 00:02:38.136
In a most famous human-machine
competition since John Henry,

00:02:38.160 --> 00:02:40.736
I played two matches

00:02:40.760 --> 00:02:44.200
against the IBM supercomputer Deep Blue.

00:02:47.040 --> 00:02:49.216
Nobody remembers
that I won the first match --

00:02:49.240 --> 00:02:51.576
(Laughter)

00:02:51.600 --> 00:02:55.000
(Applause)

00:02:55.920 --> 00:03:00.896
In Philadelphia, before losing the rematch
the following year in New York.

00:03:00.920 --> 00:03:02.680
But I guess that's fair.

00:03:04.320 --> 00:03:09.416
There is no day in history,
special calendar entry

00:03:09.440 --> 00:03:12.936
for all the people
who failed to climb Mt. Everest

00:03:12.960 --> 00:03:15.696
before Sir Edmund Hillary
and Tenzing Norgay

00:03:15.720 --> 00:03:16.920
made it to the top.

00:03:17.960 --> 00:03:21.720
And in 1997, I was still
the world champion

00:03:24.520 --> 00:03:28.720
when chess computers finally came of age.

00:03:29.520 --> 00:03:31.496
I was Mt. Everest,

00:03:31.520 --> 00:03:33.120
and Deep Blue reached the summit.

00:03:34.600 --> 00:03:38.656
I should stay of course,
not that Deep Blue did it,

00:03:38.680 --> 00:03:40.816
but its human creators --

00:03:40.840 --> 00:03:44.176
Anantharaman, Campbell, Hoane, Hsu.

00:03:44.200 --> 00:03:45.400
Hats off to them.

00:03:46.840 --> 00:03:51.256
As always, machine's triumph
was a human triumph,

00:03:51.280 --> 00:03:56.040
something we tend to forget when humans
are surpassed by our own creations.

00:03:58.360 --> 00:03:59.800
Deep Blue was victorious,

00:04:01.400 --> 00:04:02.600
but was it intelligent?

00:04:03.360 --> 00:04:05.120
No, no it wasn't,

00:04:06.200 --> 00:04:11.256
at least not in the way Alan Turing
and other founders of computer science

00:04:11.280 --> 00:04:12.480
had hoped.

00:04:13.240 --> 00:04:18.016
It turned out that chess
could be crunched by brute force,

00:04:18.040 --> 00:04:22.296
once hardware got fast enough

00:04:22.320 --> 00:04:25.280
and algorithms got smart enough.

00:04:26.760 --> 00:04:30.456
Although by the definition of the output,

00:04:30.480 --> 00:04:33.696
grandmaster-level chess,

00:04:33.720 --> 00:04:35.000
Deep Blue was intelligent.

00:04:37.320 --> 00:04:39.720
But even at the incredible speed,

00:04:40.560 --> 00:04:43.760
200 million positions per second,

00:04:45.360 --> 00:04:46.560
Deep Blue's method

00:04:47.360 --> 00:04:53.960
provided little of the dreamed-of insight
into the mysteries of human intelligence.

00:04:56.960 --> 00:04:58.776
Soon,

00:04:58.800 --> 00:05:01.376
machines will be taxi drivers

00:05:01.400 --> 00:05:03.816
and doctors and professors,

00:05:03.840 --> 00:05:06.440
but will they be "intelligent"?

00:05:07.840 --> 00:05:10.336
I would rather leave these definitions

00:05:10.360 --> 00:05:13.920
to the philosophers and to the dictionary.

00:05:15.440 --> 00:05:19.320
What really matters is how we humans

00:05:20.320 --> 00:05:23.920
feel about living and working
with these machines.

00:05:26.160 --> 00:05:31.416
When I first met Deep Blue
in 1996 in February,

00:05:31.440 --> 00:05:34.040
I had been the world champion
for more than 10 years,

00:05:36.080 --> 00:05:40.096
and I had played 182
world championship games

00:05:40.120 --> 00:05:45.216
and hundreds of games against
other top players in other competitions.

00:05:45.240 --> 00:05:50.296
I knew what to expect from my opponents

00:05:50.320 --> 00:05:52.000
and what to expect from myself.

00:05:52.680 --> 00:05:57.856
I was used to measure their moves

00:05:57.880 --> 00:06:01.496
and to gauge their emotional state

00:06:01.520 --> 00:06:05.360
by watching their body language
and looking into their eyes.

00:06:05.880 --> 00:06:09.880
And then I sat across
the chessboard from Deep Blue.

00:06:12.960 --> 00:06:15.816
I immediately sensed something new,

00:06:15.840 --> 00:06:17.160
something unsettling.

00:06:19.440 --> 00:06:22.240
You might experience a similar feeling

00:06:23.320 --> 00:06:25.856
the first time you ride
in a driverless car

00:06:25.880 --> 00:06:30.720
or the first time your new computer
manager issues an order at work.

00:06:33.800 --> 00:06:36.920
But when I sat at that first game,

00:06:38.080 --> 00:06:40.216
I couldn't be sure

00:06:40.240 --> 00:06:43.920
what is this thing capable of.

00:06:44.920 --> 00:06:48.080
Technology can advance in leaps,
and IBM had invested heavily.

00:06:48.680 --> 00:06:49.880
I lost that game

00:06:52.320 --> 00:06:54.096
and I couldn't help wondering,

00:06:54.120 --> 00:06:55.680
might it be invincible?

00:06:56.600 --> 00:06:58.960
Was my beloved game of chess over?

00:07:00.800 --> 00:07:04.936
These were human doubts, human fears,

00:07:04.960 --> 00:07:06.640
and the only thing I knew for sure

00:07:07.400 --> 00:07:10.296
was that my opponent Deep Blue
had no such worries at all.

00:07:10.320 --> 00:07:12.080
(Laughter)

00:07:13.920 --> 00:07:15.320
I fought back

00:07:16.400 --> 00:07:18.080
after this devastating blow

00:07:19.000 --> 00:07:20.200
to win the first match,

00:07:20.960 --> 00:07:22.600
but the writing was on the wall.

00:07:24.400 --> 00:07:26.536
I eventually lost to the machine

00:07:26.560 --> 00:07:29.616
but I didn't suffer the fate of John Henry

00:07:29.640 --> 00:07:32.680
who won but died
with his hammer in his hand.

00:07:37.720 --> 00:07:40.256
It turned out that the world of chess

00:07:40.280 --> 00:07:43.520
still wanted to have
a human chess champion.

00:07:44.920 --> 00:07:46.600
And even today,

00:07:48.080 --> 00:07:51.536
when a free chess app
on the latest mobile phone

00:07:51.560 --> 00:07:53.576
is stronger than Deep Blue,

00:07:53.600 --> 00:07:55.080
people are still playing chess,

00:07:56.680 --> 00:07:58.920
even more than ever before.

00:07:59.800 --> 00:08:03.016
Doomsayers predicted
that nobody would touch the game

00:08:03.040 --> 00:08:05.296
that could be conquered by the machine,

00:08:05.320 --> 00:08:07.536
and they were wrong, proven wrong,

00:08:07.560 --> 00:08:11.016
but doomsaying has always been
a popular pastime

00:08:11.040 --> 00:08:12.400
when it comes to technology.

00:08:14.360 --> 00:08:17.096
What I learned from my own experience

00:08:17.120 --> 00:08:21.776
is that we must face our fears

00:08:21.800 --> 00:08:25.520
if we want to get the most
out of our technology,

00:08:26.360 --> 00:08:28.736
and we must conquer those fears

00:08:28.760 --> 00:08:34.000
if we want to get the best
out of our humanity.

00:08:36.120 --> 00:08:37.895
While licking my wounds,

00:08:37.919 --> 00:08:39.880
I got a lot of inspiration

00:08:41.080 --> 00:08:43.775
from my battles against Deep Blue.

00:08:43.799 --> 00:08:46.920
As the old Russian saying goes,
if you can't beat them, join them.

00:08:48.880 --> 00:08:50.256
Then I thought,

00:08:50.280 --> 00:08:52.616
what if I could play with a computer --

00:08:52.640 --> 00:08:55.800
together with a computer at my side,
combining our strengths,

00:08:57.160 --> 00:09:00.936
human intuition
plus machine's calculation,

00:09:00.960 --> 00:09:03.656
human strategy, machine tactics,

00:09:03.680 --> 00:09:06.096
human experience, machine's memory.

00:09:06.120 --> 00:09:08.320
Could it be the perfect game ever played?

00:09:10.000 --> 00:09:11.680
My idea came to life

00:09:12.920 --> 00:09:16.296
in 1998 under the name of Advanced Chess

00:09:16.320 --> 00:09:22.000
when I played this human-plus-machine
competition against another elite player.

00:09:23.280 --> 00:09:25.176
But in this first experiment,

00:09:25.200 --> 00:09:31.560
we both failed to combine
human and machine skills effectively.

00:09:34.920 --> 00:09:37.160
Advanced Chess found
its home on the Internet,

00:09:38.160 --> 00:09:43.016
and in 2005, a so-called
freestyle chess tournament

00:09:43.040 --> 00:09:44.400
produced a revelation.

00:09:47.240 --> 00:09:50.776
A team of grandmasters
and top machines participated,

00:09:50.800 --> 00:09:53.536
but the winners were not grandmasters,

00:09:53.560 --> 00:09:54.920
not a supercomputer.

00:09:55.680 --> 00:10:00.016
The winners were a pair
of amateur American chess players

00:10:00.040 --> 00:10:03.200
operating three ordinary PCs
at the same time.

00:10:05.560 --> 00:10:08.576
Their skill of coaching their machines

00:10:08.600 --> 00:10:14.376
effectively counteracted
the superior chess knowledge

00:10:14.400 --> 00:10:15.976
of their grandmaster opponents

00:10:16.000 --> 00:10:20.160
and much greater
computational power of others.

00:10:21.600 --> 00:10:23.560
And I reached this formulation.

00:10:24.560 --> 00:10:27.936
A weak human player plus a machine

00:10:27.960 --> 00:10:31.216
plus a better process is superior

00:10:31.240 --> 00:10:33.656
to a very powerful machine alone,

00:10:33.680 --> 00:10:37.576
but more remarkably,
is superior to a strong human player

00:10:37.600 --> 00:10:39.560
plus machine

00:10:41.120 --> 00:10:43.520
and an inferior process.

00:10:46.360 --> 00:10:48.480
This convinced me that we would need

00:10:50.000 --> 00:10:53.680
better interfaces
to help us coach our machines

00:10:54.520 --> 00:10:56.240
towards more useful intelligence.

00:10:58.320 --> 00:11:01.616
Human plus machine isn't the future,

00:11:01.640 --> 00:11:02.856
it's the present.

00:11:02.880 --> 00:11:07.016
Everybody that's used online translation

00:11:07.040 --> 00:11:11.336
to get the gist of a news article
from a foreign newspaper,

00:11:11.360 --> 00:11:13.000
knowing its far from perfect.

00:11:13.680 --> 00:11:15.776
Then we use our human experience

00:11:15.800 --> 00:11:17.896
to make sense out of that,

00:11:17.920 --> 00:11:20.696
and then machine
learns from our corrections.

00:11:20.720 --> 00:11:25.680
This model is spreading and investing
in medical diagnosis, security analysis.

00:11:26.440 --> 00:11:28.560
The machine crunches data,

00:11:29.320 --> 00:11:31.056
calculates probabilities,

00:11:31.080 --> 00:11:34.736
gets 80 percent of the way, 90 percent,

00:11:34.760 --> 00:11:39.136
making it easier for analysis

00:11:39.160 --> 00:11:41.760
and decision-making of the human party.

00:11:42.280 --> 00:11:47.120
But you are not going to send your kids

00:11:48.000 --> 00:11:51.560
to school in a self-driving car
with 90 percent accuracy,

00:11:52.600 --> 00:11:54.200
even with 99 percent.

00:11:55.560 --> 00:11:58.416
So we need a leap forward

00:11:58.440 --> 00:12:04.600
to add a few more crucial decimal places.

00:12:07.160 --> 00:12:11.200
Twenty years after
my match with Deep Blue,

00:12:12.200 --> 00:12:13.816
second match,

00:12:13.840 --> 00:12:20.136
this sensational
"The Brain's Last Stand" headline

00:12:20.160 --> 00:12:21.736
has become commonplace

00:12:21.760 --> 00:12:24.296
as intelligent machines

00:12:24.320 --> 00:12:25.520
move

00:12:26.560 --> 00:12:28.760
in every sector, seemingly every day.

00:12:30.160 --> 00:12:33.256
But unlike in the past,

00:12:33.280 --> 00:12:34.920
when machines replaced

00:12:36.480 --> 00:12:38.856
farm animals, manual labor,

00:12:38.880 --> 00:12:41.376
now they are coming
after people with college degrees

00:12:41.400 --> 00:12:42.680
and political influence,

00:12:44.120 --> 00:12:46.216
and as someone
who fought machines and lost,

00:12:46.240 --> 00:12:48.880
I am here to tell you
this is excellent, excellent news.

00:12:51.000 --> 00:12:53.216
Eventually, every profession

00:12:53.240 --> 00:12:55.336
will have to feel these pressures

00:12:55.360 --> 00:13:00.960
or else it will mean humanity
has ceased to make progress.

00:13:02.760 --> 00:13:03.960
We don't

00:13:05.440 --> 00:13:07.160
get to choose

00:13:08.480 --> 00:13:11.200
when and where
technological progress stops.

00:13:13.160 --> 00:13:14.520
We cannot

00:13:15.960 --> 00:13:17.456
slow down.

00:13:17.480 --> 00:13:19.296
In fact,

00:13:19.320 --> 00:13:21.240
we have to speed up.

00:13:24.600 --> 00:13:27.240
Our technology excels at removing

00:13:29.200 --> 00:13:32.560
difficulties and uncertainties
from our lives,

00:13:35.000 --> 00:13:37.816
and so we must seek out

00:13:37.840 --> 00:13:39.696
ever more difficult,

00:13:39.720 --> 00:13:43.800
ever more uncertain challenges.

00:13:48.200 --> 00:13:49.400
Machines have

00:13:51.880 --> 00:13:53.696
calculations.

00:13:53.720 --> 00:13:55.296
We have understanding.

00:13:55.320 --> 00:13:57.360
Machines have instructions.

00:13:58.840 --> 00:14:00.696
We have purpose.

00:14:00.720 --> 00:14:03.000
Machines have

00:14:05.080 --> 00:14:06.296
objectivity.

00:14:06.320 --> 00:14:07.520
We have passion.

00:14:08.600 --> 00:14:14.576
We should not worry
about what our machines can do today.

00:14:14.600 --> 00:14:19.176
Instead, we should worry
about what they still cannot do today,

00:14:19.200 --> 00:14:24.696
because we will need the help
of the new, intelligent machines

00:14:24.720 --> 00:14:28.800
to turn our grandest dreams into reality.

00:14:30.000 --> 00:14:31.320
And if we fail,

00:14:32.240 --> 00:14:36.896
if we fail, it's not because our machines
are too intelligent,

00:14:36.920 --> 00:14:38.320
or not intelligent enough.

00:14:39.200 --> 00:14:42.280
If we fail, it's because
we grew complacent

00:14:43.680 --> 00:14:45.240
and limited our ambitions.

00:14:46.520 --> 00:14:49.560
Our humanity is not defined by any skill,

00:14:51.280 --> 00:14:53.960
like swinging a hammer
or even playing chess.

00:14:54.560 --> 00:14:57.576
There's one thing only a human can do.

00:14:57.600 --> 00:14:58.800
That's dream.

00:15:00.120 --> 00:15:02.656
So let us dream big.

00:15:02.680 --> 00:15:03.896
Thank you.

00:15:03.920 --> 00:15:06.320
(Applause)