0:00:00.767,0:00:04.585
When I was a kid, I was[br]the quintessential nerd.

0:00:05.773,0:00:07.633
I think some of you were too.

0:00:07.633,0:00:09.160
(Laughter)

0:00:09.160,0:00:12.275
And you sir, who laughed the loudest,[br]you probably still are.

0:00:12.275,0:00:14.455
(Laughter)

0:00:14.455,0:00:16.056
I grew up in a small town

0:00:16.056,0:00:17.902
in the dusty plains of north Texas,

0:00:17.902,0:00:21.078
the son of a sheriff who was[br]the son of a pastor.

0:00:21.078,0:00:23.647
Getting into trouble was not an option.

0:00:23.647,0:00:28.048
And so I started reading[br]calculus books for fun.

0:00:28.048,0:00:29.202
(Laughter)

0:00:29.202,0:00:30.872
You did too.

0:00:30.872,0:00:34.560
That led me to building a laser[br]and a computer and model rockets,

0:00:34.560,0:00:36.998
and that led me to making rocket fuel

0:00:36.998,0:00:38.344
in my bedroom.

0:00:38.344,0:00:41.683
Now, in scientific terms,

0:00:41.683,0:00:45.014
we call this a very bad idea.

0:00:45.014,0:00:46.375
(Laughter)

0:00:46.375,0:00:48.063
Around that same time,

0:00:48.063,0:00:51.078
Stanley Kubrick's "2001: A Space Odyssey"[br]came to the theaters,

0:00:51.078,0:00:53.850
and my life was forever changed.

0:00:53.850,0:00:56.368
I loved everything about that movie,

0:00:56.368,0:00:58.972
especially the HAL 9000.

0:00:58.972,0:01:01.244
Now, HAL was a sentient computer

0:01:01.244,0:01:04.009
designed to guide the Discovery spacecraft

0:01:04.009,0:01:06.314
from the Earth to Jupiter.

0:01:06.314,0:01:08.314
HAL was also a flawed character,

0:01:08.314,0:01:13.195
for in the end he chose[br]to value the mission over human life.

0:01:13.195,0:01:14.947
Now, HAL was a fictional character,

0:01:14.947,0:01:17.827
but nonetheless he speaks to our fears,

0:01:17.827,0:01:20.014
our fears of being subjugated

0:01:20.014,0:01:22.795
by some unfeeling, artificial intelligence

0:01:22.795,0:01:25.869
who is indifferent to our humanity.

0:01:25.869,0:01:28.657
I believe that such fears are unfounded.

0:01:28.657,0:01:31.806
Indeed, we stand at a remarkable time

0:01:31.806,0:01:32.895
in human history,

0:01:32.895,0:01:38.051
where, driven by refusal to accept[br]the limits of our bodies and our minds,

0:01:38.051,0:01:40.081
we are building machines

0:01:40.081,0:01:43.300
of exquisite, beautiful[br]complexity and grace

0:01:43.300,0:01:45.483
that will extend the human experience

0:01:45.483,0:01:48.005
in ways beyond our imagining.

0:01:48.005,0:01:50.615
After a career that led me[br]from the Air Force Academy

0:01:50.615,0:01:52.455
to Space Command to now,

0:01:52.455,0:01:54.339
I became a systems engineer,

0:01:54.339,0:01:56.878
and recently I was drawn[br]into an engineering problem

0:01:56.878,0:01:59.437
associated with NASA's mission to Mars.

0:01:59.437,0:02:02.121
Now, in space flights to the Moon,

0:02:02.121,0:02:04.286
we can rely upon mission control

0:02:04.286,0:02:07.263
in Houston to watch over[br]all aspects of the flight.

0:02:07.263,0:02:10.648
However, Mars is 200 times further away,

0:02:10.648,0:02:13.830
and as a result it takes on average

0:02:13.830,0:02:17.295
13 minutes for a signal to travel[br]from the Earth to Mars.

0:02:17.295,0:02:20.884
If there's trouble,[br]there's not enough time.

0:02:20.884,0:02:23.531
And so a reasonable engineering solution

0:02:23.531,0:02:26.164
calls for us to put mission control

0:02:26.164,0:02:29.135
inside the walls of the Orion Spacecraft.

0:02:29.135,0:02:31.386
Another fascinating idea

0:02:31.386,0:02:32.138
in the mission profile

0:02:32.138,0:02:33.827
places humanoid robots[br]on the surface of Mars

0:02:33.827,0:02:36.849
before the humans themselves arrive,

0:02:36.849,0:02:38.886
first to build facilities

0:02:38.886,0:02:42.786
and later to serve as collaborative[br]members of the science team.

0:02:42.786,0:02:46.444
Now, as I looked at this[br]from an engineering perspective,

0:02:46.444,0:02:49.346
it became very clear to me[br]that what I needed to architect

0:02:49.346,0:02:51.726
was a smart, collaborative,

0:02:51.726,0:02:54.366
socially intelligent[br]artificial intelligence.

0:02:54.366,0:02:57.904
In other words, I needed to build[br]something very much like a HAL

0:02:57.904,0:03:01.144
but without the homicidal tendencies.

0:03:01.144,0:03:03.117
(Laughter)

0:03:03.117,0:03:04.807
Let's pause for a moment.

0:03:04.807,0:03:08.761
Is it really possible to build[br]an artificial intelligence like that?

0:03:08.761,0:03:10.231
Actually, it is.

0:03:10.231,0:03:11.448
In many ways,

0:03:11.448,0:03:13.522
this is a hard engineering problem

0:03:13.522,0:03:15.357
with elements of AI,

0:03:15.357,0:03:18.593
not some wet hairball of an AI problem[br]that needs to be engineered.

0:03:18.593,0:03:22.362
To paraphrase Alan Turing,

0:03:22.362,0:03:25.109
I'm not interested in a sentient machine.

0:03:25.109,0:03:26.488
I'm not building a HAL.

0:03:26.488,0:03:29.110
All I'm after is a simple brain,

0:03:29.110,0:03:32.871
something that offers[br]the illusion of intelligence.

0:03:32.871,0:03:36.031
The art and the science of computing

0:03:36.031,0:03:36.795
have come a long way

0:03:36.795,0:03:37.853
since HAL was onscreen,

0:03:37.853,0:03:41.024
and I'd imagine if his inventor[br]Dr. Chandra were here today,

0:03:41.024,0:03:43.380
he'd have a whole lot of questions for us.

0:03:43.380,0:03:45.310
Is it really possible for us

0:03:45.310,0:03:49.382
to take a system of millions[br]upon millions of devices

0:03:49.382,0:03:51.331
to read in their data streams,

0:03:51.331,0:03:53.359
to predict their failures[br]and act in advance?

0:03:53.359,0:03:54.644
Yes.

0:03:54.644,0:03:57.816
Can we build systems that converse[br]with humans in natural language?

0:03:57.816,0:03:58.603
Yes.

0:03:58.603,0:04:00.898
Can we build systems that recognize[br]objects, identify emotions,

0:04:00.898,0:04:05.325
emote themselves, play games,[br]and even read lips?

0:04:05.325,0:04:06.377
Yes.

0:04:06.377,0:04:08.787
Can we build a system that sets goals,

0:04:08.787,0:04:12.331
that carries out plans against those goals[br]and learns along the way?

0:04:12.331,0:04:13.609
Yes.

0:04:13.609,0:04:15.469
Can we build systems

0:04:15.469,0:04:17.110
that have a theory of mind?

0:04:17.110,0:04:18.581
This we are learning to do.

0:04:18.581,0:04:22.566
Can we build systems that have[br]an ethical and moral foundation?

0:04:22.566,0:04:25.360
This we must learn how to do.

0:04:25.360,0:04:28.238
So let's accept for a moment that it's[br]possible to build such

0:04:28.238,0:04:29.768
an artificial intelligence

0:04:29.768,0:04:31.968
for this kind of mission and others.

0:04:31.968,0:04:34.613
The next question you[br]must ask yourself is,

0:04:34.613,0:04:36.390
should we fear it?

0:04:36.390,0:04:38.138
Now, every new technology

0:04:38.138,0:04:40.406
brings with it some[br]measure of trepidation.

0:04:40.406,0:04:42.952
When we first saw cars,

0:04:42.952,0:04:46.900
people lamented that we would see[br]the destruction of the family.

0:04:46.900,0:04:49.566
When we first saw telephones come in,

0:04:49.566,0:04:52.363
people would worry it would destroy[br]all civil conversation.

0:04:52.363,0:04:56.328
At the point in time we saw[br]the written word become pervasive,

0:04:56.328,0:04:58.812
people thought we would lose[br]our ability to memorize.

0:04:58.812,0:05:00.916
These things are all true to a degree,

0:05:00.916,0:05:03.005
but it's also true that these technologies

0:05:03.005,0:05:06.408
brought to us things that extended[br]the human experience

0:05:06.408,0:05:09.928
in some profound ways.

0:05:09.928,0:05:12.591
So let's take this a little further.

0:05:12.591,0:05:17.722
I do not fear the creation[br]of an AI like this,

0:05:17.722,0:05:21.864
because it will eventually embody[br]some of our values.

0:05:21.864,0:05:24.233
Consider this: building a cognitive system

0:05:24.233,0:05:27.911
is fundamentally different[br]than building a traditional

0:05:27.911,0:05:29.217
software-intensive system of the past.

0:05:29.217,0:05:31.418
We don't program them. We teach them.

0:05:31.418,0:05:33.875
In order to teach a system[br]how to recognize flowers,

0:05:33.875,0:05:36.934
I show it thousands of flowers[br]of the kinds I like.

0:05:36.934,0:05:39.308
In order to teach a system[br]how to play a game --

0:05:39.308,0:05:42.114
Well, I would. You would too.

0:05:42.114,0:05:45.611
I like flowers. Come on.

0:05:45.611,0:05:48.543
To teach a system how[br]to play a game like go,

0:05:48.543,0:05:50.392
I'd have it play thousands of games of go,

0:05:50.392,0:05:52.563
but in the process I also teach it[br]how to discern

0:05:52.563,0:05:54.715
a good game from a bad game.

0:05:54.715,0:05:58.318
If I want to create an artificially[br]intelligent legal assistant,

0:05:58.318,0:06:01.019
I will teach it some corpus of law[br]but at the same time

0:06:01.019,0:06:02.998
I am fusing with it

0:06:02.998,0:06:06.315
the sense of mercy and justice[br]that is part of that law.

0:06:06.315,0:06:09.824
In scientific terms, this is what[br]we call ground truth,

0:06:09.824,0:06:11.900
and here's the important point:

0:06:11.900,0:06:16.378
in producing these machines,[br]we are therefore teaching them

0:06:16.378,0:06:16.761
a sense of our values.

0:06:16.761,0:06:19.638
To that end, I trust[br]in artificial intelligence

0:06:19.638,0:06:24.291
the same, if not more, as a human[br]who is well-trained.

0:06:24.291,0:06:25.685
But, you may ask,

0:06:25.685,0:06:28.551
what about rogue agents,

0:06:28.551,0:06:31.230
some well-funded[br]non-government organization?

0:06:31.230,0:06:35.224
I do not fear an artificial intelligence[br]in the hand of a lone wolf.

0:06:35.224,0:06:37.392
Clearly, we cannot protect ourselves

0:06:37.392,0:06:39.958
against all random acts of violence,

0:06:39.958,0:06:41.841
but the reality is such a system

0:06:41.841,0:06:45.591
requires substantial training[br]and subtle training

0:06:45.591,0:06:47.606
far beyond the resources of an individual,

0:06:47.606,0:06:51.738
and furthermore, it's far more[br]than just injecting an Internet virus

0:06:51.738,0:06:53.410
to the world where you push a button,

0:06:53.410,0:06:55.068
all of a sudden it's in a million places

0:06:55.068,0:06:57.474
and laptops start blowing up[br]all over the place.

0:06:57.474,0:07:00.452
Now, these kinds of substances[br]are much larger

0:07:00.452,0:07:02.596
and we'll certainly see them coming.

0:07:02.596,0:07:05.289
Do I fear that such[br]an artificial intelligence

0:07:05.289,0:07:08.263
might threaten all of humanity?

0:07:08.263,0:07:09.801
If you look at movies

0:07:09.801,0:07:12.731
such as "The Matrix," "Metropolis,"

0:07:12.731,0:07:15.777
"The Terminator," shows[br]such as "Westworld,"

0:07:15.777,0:07:17.969
they all speak of this kind of fear.

0:07:17.969,0:07:22.965
Indeed, in the book "Superintelligence"[br]by the philosopher Nick Bostrom,

0:07:22.965,0:07:24.392
he picks up on this theme

0:07:24.392,0:07:28.164
and observes that a superintelligence[br]might not only be dangerous,

0:07:28.164,0:07:32.000
it could represent an existential threat[br]to all of humanity.

0:07:32.000,0:07:33.789
Dr. Bostrom's basic argument

0:07:33.789,0:07:37.591
is that such systems will eventually

0:07:37.591,0:07:40.276
have such an insatiable[br]thirst for information

0:07:40.276,0:07:42.920
that they will perhaps learn how to learn

0:07:42.920,0:07:46.604
and eventually discover that they[br]may have goals

0:07:46.604,0:07:48.324
that are contrary to human needs.

0:07:48.324,0:07:50.206
Dr. Bostrom has a number of followers.

0:07:50.206,0:07:52.766
He is supported by people[br]such as Elon Musk

0:07:52.766,0:07:54.706
and Stephen Hawking.

0:07:54.706,0:07:57.846
With all due respect

0:07:57.846,0:08:00.251
to these brilliant minds,

0:08:00.251,0:08:02.600
I believe that they[br]are fundamentally wrong.

0:08:02.600,0:08:06.650
Now, there are a lot of pieces[br]of Dr. Bostrom's argument to unpack,

0:08:06.650,0:08:08.254
and I don't have time to unpack them all,

0:08:08.254,0:08:10.564
but very briefly, consider this:

0:08:10.564,0:08:14.228
super knowing is very different[br]than super doing.

0:08:14.228,0:08:17.394
HAL was a threat to the Discovery crew[br]only insofar as HAL commanded

0:08:17.394,0:08:20.651
all aspects of the Discovery.

0:08:20.651,0:08:23.166
So it would have to be[br]with a superintelligence.

0:08:23.166,0:08:25.697
It would have to have dominion[br]over all of our world.

0:08:25.697,0:08:28.529
This is the stuff of Skynet[br]from the movie "The Terminator"

0:08:28.529,0:08:30.467
in which we had a superintelligence

0:08:30.467,0:08:32.059
that commanded human will,[br]that directed every device

0:08:32.059,0:08:35.404
that was in every corner of the world.

0:08:35.404,0:08:37.057
Practically speaking,

0:08:37.057,0:08:39.345
it ain't gonna happen.

0:08:39.345,0:08:42.671
We are not building AIs[br]that control the weather,

0:08:42.671,0:08:44.021
that direct the tides,

0:08:44.021,0:08:47.466
that command us capricious,[br]chaotic humans.

0:08:47.466,0:08:50.813
And furthermore, if such[br]an artificial intelligence existed,

0:08:50.813,0:08:53.712
it would have to compete[br]with human economies,

0:08:53.712,0:08:57.357
and thereby compete for resources with us.

0:08:57.357,0:08:58.666
And in the end --

0:08:58.666,0:09:00.162
don't tell Siri this --

0:09:00.162,0:09:02.674
we can always unplug them.

0:09:02.674,0:09:04.943
(Laughter)

0:09:04.943,0:09:08.110
We are on an incredible journey

0:09:08.110,0:09:10.443
of coevolution with our machines.

0:09:10.443,0:09:12.999
The humans we are today

0:09:12.999,0:09:15.465
are not the humans we will be then.

0:09:15.465,0:09:18.904
To worry now about the rise[br]of a superintelligence

0:09:18.904,0:09:21.872
is in many ways a dangerous distraction

0:09:21.872,0:09:24.040
because the rise of computing itself

0:09:24.040,0:09:27.116
brings to us a number of human[br]and societal issues

0:09:27.116,0:09:29.456
to which we must now attend.

0:09:29.456,0:09:32.305
How shall I best organize society

0:09:32.305,0:09:34.731
when the need for human labor diminishes?

0:09:34.731,0:09:36.642
How can I bring understanding

0:09:36.642,0:09:40.370
and education throughout the globe[br]and still respect our differences?

0:09:40.370,0:09:44.575
How might I extend and enhance human life[br]through cognitive healthcare?

0:09:44.575,0:09:47.264
How might I use computing

0:09:47.264,0:09:50.049
to help take us to the stars?

0:09:50.049,0:09:52.507
And that's the exciting thing.

0:09:52.507,0:09:55.553
The opportunities to use computing

0:09:55.553,0:09:57.075
to advance the human experience

0:09:57.075,0:09:58.123
are within our reach,

0:09:58.123,0:09:59.960
here and now,

0:09:59.960,0:10:02.404
and we are just beginning.

0:10:02.404,0:10:04.150
Thank you very much.

0:10:04.150,0:10:07.806
(Applause)