0:00

The. Year was nineteen fifty six

0:02

and the place was Dartmouth College.

0:04

In. A research proposal: A math professor used

0:06

a term that was then entirely new.

0:09

And. Entirely fanciful. Artificial.

0:13

Intelligence. There's nothing fanciful

0:15

about A anymore. The

0:17

directors of the Stanford

0:19

Institute for Human Centered

0:21

Artificial Intelligence janitor Mandy

0:23

and fatally on uncommon

0:25

Knowledge Now. Welcome

0:37

to uncommon knowledge. I'm Peter Robinson.

0:39

Philosopher John A Tremendous serve from

0:41

two thousand to Twenty Seventeen. As

0:43

Provost here at Stanford University Doctor

0:46

Regiment, he received his undergraduate degree

0:48

from the University of Nevada before

0:50

earning his doctorate in Philosophy at

0:52

Stanford. He earned that doctorate in

0:55

Nineteen Eighty Three. And. Became

0:57

a member of the Stanford Philosophy Department

0:59

the very next year. He's the author

1:01

of a number of books, including the

1:04

Nineteen Ninety Volume The Concept of Logical

1:06

Consequence Since stepping down as Provost Doctor,

1:08

a tremendous Zelda number of positioned at

1:11

Stanford including and for our purposes today

1:13

this is the relevant position. Code.

1:15

Director. Of. The Stanford Institute

1:17

for Human Centered Artificial Intelligence. Born

1:20

in Beijing, doctor Faye Faye Li

1:22

moved to this country at the

1:24

age of fifteen. She received her

1:27

undergraduate degree from Princeton and a

1:29

doctorate in Electrical Engineering from the

1:31

California Institute of Technology. Now.

1:34

A professor of computer science here

1:36

at Stanford Doctor leaves the founder

1:39

once again the Stanford Institute for

1:41

Human Centered Artificial Intelligence. Doctor Lose

1:43

memoir published just last year. The

1:46

world's I see curiosity, exploration and

1:48

discovery at the dawn of a

1:51

I. John. That you Monday

1:53

and fifty li. Thank you for making the time to

1:55

join me. That measure of Rainer

1:57

L. adding S. I. would

1:59

say that i'm great to ask a dumb question, but

2:02

I'm actually going to ask a question that is right at

2:05

the top of my form. What

2:08

is artificial intelligence? I

2:10

have seen the term 100 times

2:12

a day for what, several years now.

2:15

I have yet to find a succinct

2:19

and satisfying explanation. Let's

2:22

see. Let's go to the philosophy. Here's a man who's

2:24

professionally rigorous. But here's a woman who actually

2:26

knows you. Yeah, I've seen those answers. Let's

2:29

take the answer and then I will give you a

2:31

different answer. Oh, really? All right. Okay.

2:34

Peter used the word succinct and I'm sweating here. Because

2:37

artificial intelligence by today

2:39

is already a collection

2:41

of methods

2:45

and tools that summarizes

2:47

the overall area

2:50

of computer science that has to

2:52

do with data, pattern

2:56

recognition, decision making in

2:59

natural language, in images,

3:02

in videos, in robotics,

3:04

in speech. So

3:06

it's really a collection. At the

3:08

heart of artificial intelligence is

3:11

statistical modeling such as machine

3:13

learning using computer programs. But

3:17

today, artificial intelligence truly is

3:19

an umbrella term

3:21

that covers many things that

3:24

we're starting to feel familiar

3:26

about. For example, language intelligence,

3:28

language modeling, or speech,

3:31

or vision. John, you

3:34

and I both knew John McCarthy who

3:36

came to Stanford after he wrote

3:38

that, used the term, coined the term artificial

3:41

intelligence, now the late John McCarthy. And

3:43

I confess to you who knew him as I

3:46

did that I'm a little suspicious

3:48

of the term because I knew John. And

3:50

John liked to be provocative. And

3:52

I am thinking to myself, wait a moment, we're

3:55

still dealing with ones and zeros.

3:58

Computers are calculating machines. Artificial.

4:01

Intelligence is a. Is

4:03

a marketing term so know.

4:07

It's It's not really a marketing term

4:09

so I will give it. give you

4:11

an answer is more like what John

4:13

would have given and forth and that

4:16

is. It's is the field. the sub

4:18

field of computer science that attempts to

4:20

create machines. That. A

4:22

can accomplish tasks that.

4:25

Seem. To require intelligence.

4:28

From. The. Early beginner

4:31

early artificial intelligence for system

4:33

since that play chess or

4:35

checkers even in a very

4:37

very simple things and John.

4:40

Who as you know who

4:42

knew him? Ah, was. Ah,

4:46

ambitious, And he thought that

4:48

in a summer conference at

4:51

Dartmouth. They. Could solve.

4:53

Most. Of the problems. For

4:56

it can. I'm going to come up

4:58

what nickname a couple of very same as events

5:00

what I'm looking for here. I'll

5:02

namely events we have in Nineteen Ninety Seven.

5:05

The computer to seats Garry Kasparov at chess

5:07

Big Moment for the first time. Big Blue

5:09

and I B M project to seats a

5:11

human being, a chest and not just a

5:14

human being, but Garry Kasparov who by some

5:16

measures is one of the half dozen greatest

5:18

chess players who ever met. And

5:22

as best I can tell computer science has

5:24

said he on. Things

5:26

are getting faster, but still

5:28

and then we have. In.

5:30

Twenty sistine. A computer

5:33

defeat go expert harm's way.

5:36

And the following year it defeats

5:38

go Grandmaster lead seat all I'm

5:40

not at all sure and pronouncing

5:43

that correctly, settle in a five

5:45

game match and people say whoa,

5:47

something just happened this time. So

5:49

what I'm looking for here is

5:52

something. Something. that

5:54

a layman like me can latch onto

5:56

the here's the discontinuity here's where we

5:58

entered a new moment here's artificial

6:00

intelligence. Am I looking for something that

6:02

doesn't exist? No,

6:05

no, I think you're not. So

6:09

the difference between Deep

6:11

Blue and which played

6:13

chess, which played chess, Deep Blue

6:15

was written using traditional

6:17

programming techniques, and

6:19

what Deep Blue did is it

6:22

would for each move, for each

6:24

position on the board, it would

6:26

look down to all the possible...

6:28

Every conceivable decision tree. Every decision

6:30

tree to a certain

6:32

depth. I mean, obviously you can't

6:34

go all the way. And

6:36

it would have ways of

6:38

weighing which ones are best. And so

6:40

then it would say, this is the

6:42

best move for me at this time.

6:45

That's why in some sense it

6:48

was not theoretically very interesting. The

6:52

Go AlphaGo... AlphaGo

6:56

which was a Google project. It was a Google

6:58

project. This uses deep

7:03

learning. It's a neural net. It's

7:06

not explicit programming. We

7:09

don't know, we don't go into

7:11

it with an idea

7:13

of, here's the algorithm

7:15

we're going to use. Do this and then

7:17

do this and do this. So

7:20

it was actually quite a surprise, particularly

7:23

AlphaGo. Not

7:25

to me, but sure. To

7:28

the public, yes. To the public. But

7:31

our colleague, I'm going at this one more

7:33

time because I really want to understand this.

7:35

I really do. Our

7:37

colleague here at Stanford, Zhiyuan, who must be known

7:39

to both of you, a physicist here at Stanford,

7:41

and he said to me, Peter, what you need

7:43

to understand about the

7:45

moment when a computer defeated Go, which

7:48

is much

7:50

more complicated, at least in the decision space,

7:52

is much, much bigger, so to speak, than

7:55

chess. There are more pieces, more squares. And

7:58

Zhiyuan said to me... that

8:01

whereas chess just did more quickly what a

8:03

committee of grandmasters would have decided on, the

8:06

computer in Go was

8:09

creative. It was pursuing strategies

8:11

that human beings had never pursued before.

8:13

Is there something to that? Yeah,

8:15

so there's a famous... They think getting impatient

8:17

with me. I'm asking such a good question. No, no,

8:20

you're asking such a good question. So in the third

8:22

game of the... I think it was the third game

8:24

of the five games, there was a move, I think

8:26

it was move 32. 32 or 35. It's that the

8:28

computer program made a move

8:35

that really surprised every single

8:37

Go masters. Not only Lisa

8:39

Dole himself, but everybody who's

8:42

watching. That's a very surprising

8:46

move.

8:48

In fact, even post-anonymizing how

8:51

that move came about, the

8:54

human masters would say this

8:57

is completely unexpected. What

8:59

happens is that the computers

9:02

like John says,

9:04

right, has the

9:07

learning ability and has the inference

9:09

ability to think about patterns or

9:12

to decide on certain

9:14

movements even outside of

9:19

the trained, familiar

9:21

human masters domain

9:24

of knowledge. May I expand on that? The

9:29

thing is these deep neural nets

9:32

are supremely good

9:35

pattern recognition systems,

9:39

but the patterns they recognize, patterns

9:42

they learn to recognize

9:44

are not necessarily exactly the

9:46

patterns that humans recognize.

9:49

So it was seeing something

9:51

about that position and it

9:53

made a move that because

9:55

of the patterns that

9:57

it recognized in the book, in the board

10:00

that made

10:02

no sense from a human standpoint. In

10:06

fact, all of

10:08

the lessons in how to play Go tell

10:11

you never make a move that close to

10:13

the edge that quickly. And

10:17

so everybody thought it made a mistake, and

10:19

then it proceeded to win. And

10:22

I think the way to understand that is

10:24

it's just seeing patterns that

10:26

we don't see. It's

10:28

computing patterns

10:31

that is not traditionally human, and

10:34

it has the capacity to compute.

10:37

OK. I'm trying to... We're

10:39

already entering this territory, but

10:42

I am trying really hard to tease out

10:44

the, wait a

10:46

moment, these are still just machines

10:48

running zeros and ones, bigger

10:51

and bigger memory, faster and faster ability to

10:53

calculate, but we're still dealing with machines that

10:55

run zeros and ones. It's one strand. And

10:58

the other strand is, as you well know,

11:01

2001's Space Odyssey, where the computer takes over

11:03

the ship. Open the pod bay

11:05

doors, Hal. I'm sorry,

11:08

Dave. I'm afraid I

11:10

can't do that. OK. We'll

11:13

come to this soon enough. Fei-Fei

11:16

Li, in your memoir, The World I

11:18

See, quote, I believe our civilization stands

11:20

on the cusp of

11:22

a technological revolution with

11:25

the power to reshape life as we

11:27

know it. Revolution,

11:32

reshape life as we know it. Now

11:34

you're a man whose whole academic training

11:36

is in rigor. Are you going to

11:38

let her get away with this kind

11:40

of wild overstatement? No,

11:42

I don't think it's an overstatement. I

11:46

think she's right. He told me to write a book.

11:50

And you, Peter, it's a technology

11:53

that is extremely powerful,

11:55

that will allow us and is allowing us

11:58

to get to the point.

12:00

computers to do things we never

12:03

could have programmed them to do. And

12:06

it will change everything, but

12:08

it's like, what a lot of

12:10

people have said, it's like electricity, it's

12:12

like the steam revolution. It's

12:16

not something necessarily to be afraid of. It's

12:19

not that it's going to suddenly take over the

12:21

world. That's not what Fei-Fei was saying. Right.

12:25

It's a powerful tool

12:27

that will revolutionize industries

12:29

and humans the way

12:31

we live. But the word revolution

12:33

is not that it's a conscious

12:35

being. It's just a powerful tool

12:37

that changes things. I would find

12:39

that reassuring if a few pages later Fei-Fei

12:42

had not gone on to write. Oh no.

12:45

There's no separating the beauty of science

12:47

from something like, say, a

12:49

Manhattan project, close quote. Nuclear

12:52

science. We can produce

12:54

abundant energy, but it can also produce weapons

12:57

of indescribable horror. AI

13:01

has boogeymen of its own, whether

13:03

it's killer robots, widespread surveillance, or

13:05

even just automating all

13:07

eight billion of us out of our jobs. Now,

13:10

we could devote an entire program to each of those boogeymen,

13:13

and maybe at some point we should. But

13:17

now that you have scared me, even in

13:19

the act of reassuring me, and

13:21

in fact it throws me that you're so eager to reassure

13:23

me that I think maybe I really should be even more

13:26

scared than I am. Let me just go

13:28

right down. Here's the killer robots. Let me quote the

13:30

late Henry Kissinger. I'm just going to put these up

13:32

and let you... You

13:34

may calm me down if you can. Henry

13:37

Kissinger. If you imagine a war

13:39

between China and the United States, you have artificial

13:42

intelligence weapons. Nobody has

13:45

tested these things on a broad scale, and

13:48

nobody can tell exactly what will

13:50

happen when AI fighter planes

13:52

on both sides interact. So

13:55

you are then... I'm quoting Henry Kissinger, who is not

13:57

a fool, after all. So you are then

13:59

in a world of... potentially total

14:01

destructiveness." Fei-Fei?

14:05

So, like I said, I'm now denying

14:07

how powerful these tools are.

14:10

I mean, humanity, before AI,

14:12

has already created tools and

14:14

technology that are very destructive,

14:16

could be very destructive. We

14:18

talk about Mahatan Project, right?

14:21

But that doesn't mean that

14:24

we should collectively decide to use

14:26

this tool in this destructive way.

14:29

Okay, Peter, you know, think

14:31

back before you even

14:33

had heard about artificial intelligence. Which actually

14:35

was five years ago, maybe. No, I know.

14:38

This is all happening so fast. Just five

14:40

years ago. Or ten years ago. Remember

14:44

the tragic incident

14:46

where an Iranian

14:50

passenger plane was shot down flying

14:52

over the Persian Gulf by

14:55

an Aegis system? Yes, yes.

14:57

And one of our ships. One of

14:59

our ships, an automated system,

15:01

because it had to be automated

15:03

in order to be... Humans can't

15:05

react to that. Exactly.

15:09

And in this case, for reasons

15:11

that I think are quite understandable

15:13

now that you understand the incident, but it

15:17

did something that was horrible. That's

15:20

not different in kind from what you can do with AI,

15:22

right? So

15:24

we as creators

15:29

of these devices or as users

15:31

of AI have to be

15:34

vigilant about what

15:36

kind of use we put them to. And

15:39

when we decide to put them to one

15:41

particular use, and there may be uses,

15:45

the military has many good uses for them, we

15:47

have to be vigilant about

15:50

their doing what we intend

15:52

them to do rather than doing things

15:54

that we don't intend to do. So

15:56

you're announcing a great theme. And

15:59

that theme is... that what Dr.

16:01

Fei-Fei Li has invented makes

16:05

the discipline to which you have

16:07

dedicated your life, philosophy, even

16:10

more important, not less so. Yeah, that's why we're

16:13

co-directors. The power of the future makes the

16:15

human being more important, not less so. Am I

16:17

being glib? Or is that on to something

16:19

else? So let me tell you a

16:21

story about... So

16:24

Fei-Fei used to live next door to me, or close

16:26

to next door to me. And

16:29

I was talking to her... I'm not sure whether that would

16:31

make me feel more safe or more... And

16:34

I was talking to her... I was still

16:36

privileged. And she

16:38

said to me, you and

16:40

John Hennessy start a lot of institutes that

16:44

brought technology into other

16:46

parts of the university. We

16:49

need to start an institute

16:51

that brings philosophy and ethics

16:54

and the social sciences into

16:56

AI. Because

16:58

AI is too dangerous to

17:01

leave it to the computer scientists alone.

17:06

Nothing wrong with it. There are many stories about how

17:08

hard it was to persuade him when he was provost,

17:10

and you succeeded. Can I... just

17:12

one more boogeyman briefly? Yeah.

17:14

And we'll return to that theme that you just gave

17:17

us there, and then we'll get back to the Stanford Institute.

17:21

I'm quoting you again. This is from your memoir. The

17:24

prospect of just automating all billion of us

17:26

out of our jobs. That's the phrase you

17:28

used. Well, it turns out that

17:31

it took me mere seconds using

17:33

my AI-enabled search algorithm,

17:36

search device, to find

17:38

a Goldman Sachs study from last year, predicting

17:41

that in the United States and Europe,

17:43

some two-thirds of all jobs could

17:46

be automated, at least to some degree.

17:49

So why shouldn't

17:51

we all be terrified? Henry

17:53

Kissinger, world apocal... All right, maybe that's a

17:55

bit too much. But my job!

18:00

I think job change is

18:02

real. Job change is real

18:04

with every single technological advances

18:07

that humanity, human civilization has

18:09

faced. That

18:11

is real and that's not to be taken

18:13

lightly. We also have to

18:16

be careful with the word job. Job

18:18

tends to describe a holistic profession

18:21

or that a person attaches

18:25

his or her income as well. It's

18:27

not an identity rule. But there

18:29

is also within every job, pretty much

18:32

within every job, there are so many

18:34

tasks. It's hard to imagine

18:36

there's one job that has only

18:38

one singular task. Being

18:42

a professor, being a scholar, being a doctor,

18:44

being a cook, all

18:46

of this job has multiple tasks.

18:49

What we're seeing as technology is

18:52

changing how some of these tasks can be

18:55

done. And it's true as

18:57

it changes these tasks, some of them,

18:59

some part of them could

19:01

be automated. It's starting

19:03

to change how the jobs are and eventually

19:05

it's going to impact jobs. So this is

19:08

going to be a gradual process and it's

19:10

very important we stay on top

19:12

of this. This is why Human

19:15

Center AI Institute was founded is

19:17

these questions are profound. They're

19:19

by definition multidisciplinary. Computer

19:23

scientists alone cannot do all

19:25

the economic analysis, but economists

19:27

now understanding what these

19:30

computer science programs

19:33

do will not by themselves understand

19:35

the shift of the jobs. Okay,

19:38

John, may I tell you? Go ahead. But

19:40

let me just point something out. The

19:43

Goldman Sachs study said

19:45

that such and

19:47

such percentage of jobs will be

19:49

automated or can be automated at

19:51

least in part. Yes. Now

19:54

what they're saying is that a certain number

19:56

of the tasks that go into a particular

19:58

job. So

20:00

Peter, you said it only

20:03

took me a few seconds to

20:06

go to the computer and find

20:08

that article. Guess

20:10

what? That's one

20:13

of the tasks that would have taken you

20:15

a lot of time. So

20:17

part of your job has

20:20

been automated. Okay,

20:22

now let me tell you a story. But

20:24

also empowered. Empowered, okay fine, thank

20:27

you, thank you, thank you, you're making me feel good. Now

20:29

let me tell you a story. All

20:31

three of us live in California, which means all three

20:34

of us probably have some friends down in Hollywood. And

20:37

I have a friend who was involved in the writers

20:39

strike. Yeah. Okay,

20:41

and here's the problem. To

20:44

run a sitcom, you

20:46

used to run a writers room. And

20:49

the writers room would employ seven, a dozen,

20:52

on the Simpsons show, the cartoon show. They'd keep,

20:54

they'd had two, a couple of writers rooms running.

20:56

They were employing 20. And these

20:59

were the last kind of person you'd

21:01

imagine a computer could replace

21:03

because they were well educated and

21:05

witty and quick with words. And

21:09

you think of computers as just

21:11

running calculations. Maybe spreadsheets, maybe someday

21:13

they can eliminate accountants, but writers,

21:15

Hollywood writers. And

21:18

it turns out, and my friend illustrated

21:20

this for me by saying, doing

21:24

the artificial intelligence thing where it had

21:26

a prompt, draft a

21:29

skit for Saturday

21:31

Night Live in which

21:33

Joe Biden and Donald Trump are

21:35

playing beer pong. 15

21:39

seconds. Now professionals

21:41

could have tightened it up but

21:43

it was pretty funny and it was instantaneous.

21:45

And you know what that means? That

21:48

means you don't need four

21:50

or five of the seven writers. You need a senior

21:52

writer to assign intelligence

21:55

the artificial, and you need maybe one other writer

21:57

or two other writers to tighten it up. redraft

22:00

it, it is

22:02

upon us. And your artificial

22:04

intelligence is going to get bad press when

22:07

it starts eliminating the jobs of

22:09

the chattering classes, and that has

22:11

already begun. Tell me I'm wrong.

22:13

Do you know, before the

22:16

agricultural revolution, something

22:18

like 80, 90 percent of all the people

22:22

in the United States were

22:25

employed on farms. Now

22:29

it's down to 2 percent or 3 percent,

22:33

and those same farms, that same

22:35

land, is far, far more productive.

22:38

Now, would you say that your

22:41

life or anybody's life now

22:43

was worse off than it

22:45

was in the 1890s when

22:47

everybody was

22:51

working on the farm? No. So

22:53

yes, you're right. It

22:56

will change jobs, it will

22:58

make some jobs easier, it

23:00

will allow us to do things

23:02

that we could not do before, and yes,

23:05

it will allow fewer people to

23:07

do more of what they were doing before, and

23:09

consequently there will

23:17

be fewer people in that line of work.

23:20

That's true. I also want

23:22

to just point out two things. One is

23:24

that jobs is always changing,

23:26

and that change is always painful. And

23:28

as computer

23:30

scientists, as philosophers, also as citizens of

23:33

the world, we should be empathetic of

23:35

that, and nobody is saying we should

23:37

just ignore that change in

23:40

pain. So this is why we're studying

23:42

this, we're trying to talk to policymakers,

23:45

we're educating the population. In the

23:47

meantime, I think we should give

23:49

more credit to human creativity in

23:52

the face of AI. I

23:54

start to use this example

23:57

that's not even AI. Think

23:59

about the advanced speaking of

24:02

Hollywood graphics technology

24:05

CGI and all that right

24:08

the video gaming industry or animation and all that

24:10

right one of many

24:13

of our including our children's

24:16

favorite animation series is by

24:18

Ghibli studio you know

24:21

princess no Mononaki my neighbor

24:24

Totoro spirited

24:26

a wall all of

24:28

these were made during

24:30

a period where computer graphics

24:32

technology is far more advanced

24:34

than these hand-drawn animations

24:38

yet they're the beauty

24:40

the creativity the emotion the

24:42

uniqueness in this film continue

24:44

to inspire and just entertain

24:47

humanity so I think we

24:49

need to still

24:51

have that pride and also give

24:54

the credit to humans let's

24:56

not forget our creativity

24:58

and the emotion and intelligence is unique

25:01

it's not going to be taken away

25:03

by technology thank you I feel

25:05

slightly reassured I'm still

25:07

nervous about my job but I feel slightly reassured but

25:10

you mentioned government a moment ago which

25:12

leads us to how we should regulate

25:15

AI let me

25:17

give you two quotations I'll begin

25:19

I'm coming to the quotation from the two of you

25:21

but I'm going to start with

25:23

a recent article in the Wall Street Journal

25:25

by Senator Ted Cruz of Texas and former

25:28

senator Phil Graham also of Texas quote the

25:31

Clinton administration took a hands-off approach

25:34

to regulating the early in the

25:36

internet in so doing it unleashed

25:38

extraordinary economic growth and prosperity the

25:42

Biden administration by contrast is

25:44

impeding innovation in

25:46

artificial intelligence with aggressive

25:48

regulation close quote that's

25:50

them this is you also

25:53

a recent article in the Wall Street

25:56

Journal John Echamendi and Fei-Fei Li quote

25:59

President Biden signed an executive

26:01

order on artificial intelligence that

26:03

demonstrates his administration's commitment to harness

26:06

and govern the technology. President Biden

26:08

has set the stage and now

26:10

it is time for Congress to

26:12

act. Cruz and Graham,

26:15

less regulation. Echamendi and

26:17

Lee, Biden administration has done well. Now

26:19

Congress needs to give us even more.

26:22

No. All right, John. No,

26:24

I don't agree with that. So I

26:27

believe regulating any kind of technology

26:29

is very difficult and

26:32

you have to be careful not

26:34

to regulate too soon or

26:38

not to regulate too late. Let

26:41

me give you another example. You talked

26:43

about the Internet and it's true. The

26:45

government really was quite hands off and

26:47

that's good. That's good. It worked out.

26:50

It worked out. But now

26:52

let's also think about social media. Social

26:55

media has not worked

26:58

exactly, worked out exactly the

27:00

way we want

27:03

it. We originally believed that we

27:05

were going to enter

27:08

a golden age in which

27:10

friendship, comity, well, and everybody

27:12

would have a voice and

27:15

we could all live

27:18

together, kumbaya and so forth. That's not

27:20

what happened. Jonathan

27:22

Haight has a new book out on

27:24

the particular pathologies among young people from

27:26

all of these social media. Not an

27:28

argument, it's an argument, but it's based

27:30

on lots of data. So

27:34

it seems to me that I'm

27:37

in favor of very light

27:40

handed and

27:42

informed regulation

27:45

to try to put up sort of

27:47

bumpers, I don't know what

27:49

the analogy is, for the technology. I

27:53

am not for heavy

27:56

handed top down regulation

27:58

that stifles innovation. Okay,

28:00

here's another, let me get on

28:02

to this, I'm sure

28:04

you'll be able to adapt your answers to this question. Okay.

28:07

I'm continuing your Wall Street Journal piece. Big

28:10

tech companies can't be left to govern

28:12

themselves. Around here, Silicon

28:14

Valley, those are fighting words. Academic

28:17

institutions should play a leading role in

28:20

providing trustworthy assessments and benchmarking of

28:22

these advanced technologies. We

28:24

encourage an investment in

28:26

human capital to bring more talent to the field of

28:28

AI with academia and the government,

28:30

close quote. Okay, now, it

28:33

is mandatory for me to say this,

28:35

so please forgive me my fellow Stanford

28:39

employees, apart from anything else. Why

28:42

should academic institutions be trusted? Half

28:44

the country has lost faith in

28:46

academic institutions. DEI, the

28:49

whole woke agenda, anti-Semitism

28:52

on campus. We've got a Gallup, recent

28:54

Gallup poll showing the proportion of Americans who

28:56

expressed a great deal or quite a lot

28:58

of confidence in higher education. This

29:00

year came in at just 36%, and

29:04

that is down in the last eight years from 57%.

29:08

You are asking us to trust you

29:10

at the very moment when we believe we have good

29:12

reasons and knock it off. Trust

29:14

you? Okay, Faith. So,

29:16

I'll start with this first half

29:18

of the answer. I'm sure John has a

29:21

lot to say. I do want to make

29:23

sure, especially we're in the heads of co-directors

29:25

of HAI, when we

29:28

talk about the relationship between government

29:30

and technology, we tend to use

29:32

the word regulation. I really, really

29:34

want to double-click. I

29:36

want to use the word policy. And

29:39

policy and regulations are

29:41

related but not the same. When

29:44

John and I wrote that Wall

29:46

Street Journal opinion piece, we really

29:48

are focusing on a piece of

29:50

policy that is to resource

29:53

public sector AI, to resource academia,

29:55

because we believe that AI is

29:57

the only way to do it.

30:00

such a powerful technology and

30:02

science, and academia and public

30:04

sector still has a role

30:06

to play to create public

30:09

good. And public

30:11

goods are a curiosity-driven

30:13

knowledge exploration, are cures

30:16

for cancers, are

30:18

the maps of biodiversity

30:20

of our globe, are

30:22

discovery of nanomaterials that

30:25

we haven't seen before,

30:27

are different ways of

30:29

expressing in theater,

30:31

in writing, in music. These

30:33

are public goods. And when

30:35

we are looking, when we are

30:37

collaborating with the government on policy,

30:40

we're focusing on that. So

30:42

I really want to make sure. Regulation

30:44

we all have personal opinion, but

30:46

there's more than regulation in policy.

30:48

Yeah. So, yeah. We,

30:51

yeah, I, let me make one last run

30:53

at you. In my theory

30:55

here, although I'm asking questions that

30:57

you'd, I'm quite sure you'd like

31:00

to take me out and swap me around at this point,

31:02

John. But this is

31:04

serious. You've got the Stanford Institute for

31:06

Human Centered Artificial Intelligence, and that's because

31:09

you really think this is important. But

31:12

we live in a democracy, and you're going

31:14

to have to convince a whole lot of people. So let me

31:16

take one more run at you and then hand it back to

31:18

you, John. Your article in the

31:21

Wall Street Journal, again, let me repeat this. We

31:23

encourage an investment in human capital to

31:25

bring more talent to the field of AI with academia

31:27

and the government. That means money.

31:29

An investment means money, and it means

31:32

taxpayers' money. Here's what Cruz

31:34

and Graham say in the Wall Street Journal. The

31:36

Biden regulatory policy on AI has everything to do

31:38

with special interest rent seeking. Stanford

31:42

faculty make well above the national

31:44

average income. We are sitting at

31:46

a university with an endowment of

31:48

tens of billions of dollars. John,

31:52

why is not your article in the Wall

31:54

Street Journal the very

31:56

kind of rent seeking that

31:58

Senator Cruz and Graham have? and Senator

32:00

Graham are saying, are you kidding? Peter,

32:04

let's take another example.

32:07

So one of the greatest policy

32:10

decisions that this country has ever

32:12

made was when Vannevar

32:15

Bush, advisor to

32:18

at the time President Truman, convinced,

32:21

he stayed on through Eisenhower as I recall,

32:24

is by partisan. Exactly. No, no,

32:26

it was not a partisan issue

32:29

at all, but convinced Truman

32:33

to set up the NSF for

32:37

funding, Curiosity-based

32:40

research, advanced research

32:43

at the universities, and

32:46

then not to say

32:49

that companies don't have any role, not to

32:52

say that government has no role, they both

32:54

have roles, but they're different

32:56

roles. And companies

33:00

are, tend to be better

33:02

at development, better at producing

33:04

products, and tapping into

33:06

things that can, within a

33:08

year or two or three, can

33:10

be a product that will be useful. Scientists

33:15

at universities don't

33:17

have that constraint. They don't have to worry

33:19

about when is this going to be commercial.

33:21

And that has, I

33:25

think, had such

33:28

an incalculable effect

33:31

on the prosperity of

33:33

this country, on the fact that

33:36

we are the leader in every

33:38

technology field. It's

33:40

not an accident that we're the leader in

33:42

every technology field. We weren't, we didn't use,

33:45

and, and does it affect your argument if

33:47

I add, it also enabled us,

33:49

or contributed to a victory in

33:52

the Cold War, the weapons systems

33:55

that came out of universities? All

33:57

right. Well, no, absolutely. And, you

33:59

know, In

34:01

other words, it ended up being a defensive demand.

34:03

You could argue from all kinds of points of

34:06

view that it was a good ROI

34:08

for taxpayers' money. So

34:10

we're not arguing for higher

34:13

salaries for faculty or anything of that

34:15

sort. But we

34:17

think, particularly in AI,

34:20

it's gotten to the

34:22

point where scientists at

34:24

universities no

34:26

longer play in the game

34:29

because of the cost of the

34:31

computing, the cost, the inaccessibility of

34:33

the data. That's why

34:35

you see all of these developments coming out of companies. That's

34:38

great. Those are great developments. But

34:42

we need to have also

34:44

people who are exploring

34:46

these technologies without looking

34:49

at the product, without being driven

34:52

by the profit motive. And

34:54

then eventually, hopefully, they will develop

34:57

discoveries, they will make discoveries, will

35:00

then be commercializable. Okay. I

35:02

noticed in your book, Feifei, I was very struck

35:04

that you said, I think it was about a

35:06

decade ago, 2015, I

35:08

think was the, that you noticed

35:10

that you were beginning to lose colleagues to

35:12

the private sector. Yeah. Presumably,

35:15

because they just pay so phenomenally well around here

35:18

in Silicon Valley. But then there's also the point

35:20

that to get to make progress in AI,

35:23

you need an enormous amount of

35:25

computational power. And

35:27

assembling all those ones and

35:30

zeros is extremely expensive. So

35:33

chat GPT, what is the parent company? OpenAI.

35:36

OpenAI got started with an

35:38

initial investment of a billion dollars. An

35:42

initial, friends and family capital of a billion

35:44

dollars is a lot of money even around

35:46

here. Okay. Yes.

35:50

All right. It

35:52

feels to me as though every one of these topics is worth

35:54

a day long. Actually,

35:57

I think they are. And by the way.

36:00

This has happened before, where the

36:03

science has become so expensive

36:06

that it could no longer ... that

36:09

university-level research and researchers could no longer

36:11

afford to do the science.

36:14

It happened in high-energy physics.

36:17

High-energy physics used to mean you had

36:19

a Van de Graaff generator in your

36:22

office, and that was your accelerator. You

36:24

could get it. You could do what

36:26

you needed to do. And

36:30

then it no longer was ... the

36:33

energy levels were higher and

36:35

higher. And what happened? Well,

36:38

the federal government stepped in and said,

36:40

we're going to help. We're going to

36:42

build an accelerator. Stanford

36:45

linear accelerator. Exactly. Sandia Labs, Lawrence

36:47

Livermore, all these are at least

36:49

in part federal established. CERN. CERN,

36:53

which is European. Well, Fermilab.

36:55

So the first accelerator was

36:58

Slack, Stanford linear accelerator center,

37:00

then Fermilab, and

37:03

so on and so forth. Now, right. CERN

37:05

is late ... actually late in the

37:07

game, and it's European

37:09

consortium. But the thing

37:12

is, we

37:14

could not continue the science

37:18

without the help of

37:20

the government, in the government. Well,

37:22

there is another ... and then in addition to

37:24

high energy physics, and then

37:27

bio, right? Especially

37:29

with genetic sequencing and high

37:31

throughput genomics, and biotech

37:34

is also changing. And now you

37:36

see a new wave of

37:40

biology labs that are actually heavily

37:42

funded by the combination of government

37:44

and philanthropy and all that, and

37:47

that stepped in to supplement

37:49

what the traditional

37:53

university model is. And so we're

37:55

now here with AI and computer

37:57

science. We

38:01

have to do another show on that one alone, I think. The

38:05

Singularity. Oh, good! This

38:07

is good. Reassuring. You're both

38:10

rolling your eyes. Wonderful. I

38:12

feel better about this already. Good. Ray

38:15

Kurzweil, you know exactly where this is going. Ray Kurzweil

38:17

writes a book in 2005. This

38:19

gets everybody's attention and still scares lots of

38:21

people to death, including me.

38:24

The book is called The Singularity is Near.

38:27

And Kurzweil predicts a singularity that will

38:30

involve, and I'm quoting him, the merger

38:33

of human technology with human

38:35

intelligence. He's not saying

38:38

the tech will mimic more and more closely

38:40

human intelligence. He is saying they will merge.

38:43

I set the date for the singularity

38:45

representing a profound and disruptive transformation in

38:47

human capability as 2045. Okay.

38:52

That's the first quotation. Here's the

38:54

second. It comes from the Stanford

38:56

course catalog's description of the philosophy

38:58

of artificial intelligence. A

39:01

freshman seminar that was taught

39:03

last quarter, as I recall,

39:05

by one John Echimendi. Here's

39:09

from the description. Is it really

39:11

possible for an artificial

39:14

system to achieve genuine intelligence,

39:16

thoughts, consciousness, emotions? What

39:19

would that mean? John,

39:21

is it possible? What would it mean? I

39:27

think the answer is actually no. And

39:30

thank goodness. You kept me

39:33

waiting for a moment.

39:35

I think the fantasies that

39:38

Ray Kurzweil and others have

39:43

been spinning up, I guess

39:45

that's the way to put it, stem

39:48

from a lack of understanding of

39:52

how the human being really

39:54

works and don't

39:56

understand how crucial biology

39:59

is. is to the

40:01

way we work, the way

40:03

we are motivated, how we

40:05

get desires, how we get

40:07

goals, how we get how

40:09

we become

40:12

humans, become people. And

40:14

what AI has done so

40:16

far, AI is capturing what

40:19

you might think of as the

40:22

information processing piece

40:26

of what we do. So part of

40:28

what we do is information processing. So

40:31

it's got the right frontal cortex but hasn't got

40:33

the left frontal cortex yet. Yeah,

40:35

that's an oversimplification. But yeah, imagine that

40:37

on television. So

40:41

I actually think it is, first

40:43

of all, the date. 2045

40:46

is insane. That

40:52

will not happen. And secondly, it's not even clear

40:54

to me that we will ever go back.

40:57

I can't believe I'm saying this.

40:59

In his defense, I don't think

41:01

he's saying that 2045 is the

41:03

day that the machines become conscious

41:06

beings like humans. It's

41:10

more an inflection point of the

41:12

power of the technology that is

41:14

disrupting the society. He's

41:18

late. We're already there. Exactly.

41:21

That's what I'm saying. I think you're

41:23

being overly generous.

41:29

I think that what he means by the singularity

41:31

is the date at which we

41:33

create an artificial intelligence system

41:36

that can improve itself

41:39

and then get into a cycle,

41:41

a recursive cycle, where it

41:43

becomes a super intelligence.

41:46

And I deny that. He's

41:48

playing the 2001 Space Odyssey game here. Different

41:53

question but related question. In some ways, this

41:55

is a more serious question, I think. Although

41:58

that's series two. Here's the

42:01

late Henry Kissinger again. Quote, we

42:03

live in a world which

42:05

has no philosophy.

42:08

There is no dominant philosophical

42:11

view. So

42:13

the technologists can run wild. They

42:16

can develop world-changing things and there's

42:18

nobody to say, we've got to integrate

42:20

this into something. All

42:23

right, I'm going to put it crudely again. But

42:26

in China a century

42:28

ago, we still had Confucian thought,

42:31

dominant among, at least among the educated

42:33

classes on my very thin understanding of

42:35

Chinese history. In

42:37

this country until the day before

42:40

yesterday, we still spoke without irony

42:42

of the Judeo-Christian tradition,

42:44

which involved certain concepts

42:46

about morality, what it meant

42:49

to be human. It

42:52

assumed a belief in God, but it turned out you

42:54

could actually get pretty far along, even

42:56

if you didn't believe in it. And

42:59

Kissinger is now saying it's all fallen

43:01

apart. There is no dominant

43:03

philosophy. This

43:05

is a serious problem. Is it not? There's

43:08

nothing to integrate AI into.

43:11

You take his point. You're the philosopher. You're

43:13

the philosopher. I think this is a great,

43:16

first of all, thank you for that quote. I

43:26

didn't read that quote from Henry

43:29

Kissinger. This is

43:31

why we founded the Schumann Center

43:33

AI Institute. These are the fundamental

43:35

questions that our

43:37

generation needs to figure out. That's not

43:40

just a question, that's the question. It

43:42

was one of the fundamental questions. That's

43:44

also one of the fundamental questions that

43:46

illustrates why universities are

43:48

still relevant today. One

43:53

of the things that Henry Kissinger said

43:55

in that quote is that there is

43:57

no dominant philosophy.

44:00

no one dominant philosophy

44:02

like the Judeo-Christian tradition, which

44:04

used to be the dominant

44:06

tradition in the... This was

44:08

a different conversation in Paris in the 12th century, for

44:10

example. The university in Paris... In order

44:12

to have... In order

44:14

to take values into account

44:16

when you're creating an AI

44:18

system, you don't need a

44:20

dominant tradition.

44:22

I mean, there's... What

44:25

you need, for example, for

44:27

most ethical traditions, is the

44:29

golden rule. Okay,

44:32

so we can still get along with each

44:34

other, even when it comes

44:36

to deep, deep questions of values such as

44:38

this. We still have enough common ground. I

44:42

believe so. I have

44:45

yet another sigh of relief. Okay,

44:47

let's talk a little bit. We're talking a little

44:49

bit about a lot of things here, but so

44:52

it is. Let us speak of many things as

44:55

it is written in Alice in Wonderland, the Stanford

44:57

Institute. The Stanford

45:00

Institute for Human Centered Artificial

45:02

Intelligence, of which you are co-directors,

45:04

and I just have two questions and

45:07

respond as you'd like. Can you give

45:09

me some taste, some feel for what you're

45:11

doing now, and

45:15

in some ways more important, but more elusive, where

45:17

you'd like to be in just five years, say.

45:19

Everything in this field is moving. So if I

45:21

would... My impulse is to say 10 years because

45:23

it's a rounder number. It's too far

45:26

off in this field. Fei-Fei? I

45:29

think what really has happened in the

45:31

past five years by Stanford High, among

45:33

many things... I just want to make

45:35

sure everybody is following you. HAI, Stanford High,

45:37

is the way it's known on this campus.

45:39

Yes. All right, go ahead. Is

45:41

that we have put

45:43

a stick on the ground for

45:46

Stanford as well as for everybody

45:48

that this is an interdisciplinary study.

45:53

AI, artificial intelligence, is

45:55

a science of its own. It's a

45:57

powerful tool. What

46:00

happens is that you can welcome

46:02

so many disciplines to

46:04

cross-pollinate around the topic of

46:07

AI, or use the

46:09

tools of AI to make

46:11

other sciences happen, or to

46:14

explore other new ideas. And

46:16

that concept of making this

46:18

an interdisciplinary and multidisciplinary field

46:21

is what I think Stanford High

46:24

brought to Stanford, and also hopefully

46:26

to the world. And just

46:28

like you said, computer science is kind of

46:30

a new field. Only the

46:33

late John McCarthy coined the term

46:35

in the late 50s. Now

46:39

it's moving so fast. Everybody feels it's

46:42

just a niche computer science field

46:45

that's just making its way into the

46:47

future. But we're saying,

46:49

no, look abroad. There's

46:52

so many disciplines that can be put

46:54

here. Who competes with the Stanford Institute

46:56

and Human-Centered Design? Is there such an institute

46:58

at Harvard or Oxford or Beijing? I

47:01

just don't know what this is. In

47:03

the five years since we launched, there have been

47:05

a number of similar institutes

47:07

that have been created

47:10

at other universities. We don't see that as competition

47:12

in any way. If these arguments you've been making

47:14

are valid, then we need them. We need them.

47:16

We should walk you back there. We need them

47:19

as a movement. We need them. And

47:21

part of what I think we've succeeded

47:23

to a certain extent doing is

47:26

communicating this vision of

47:29

the importance of keeping the

47:31

human and human

47:34

values at the center when

47:36

we are developing this

47:38

technology, when we are

47:41

applying this technology. And

47:44

we want to communicate that to the

47:46

world. We want other centers

47:48

that adopt a similar standpoint.

47:52

And importantly, one

47:55

of the things that they didn't mention is, one

47:58

of the things we try to do is edge the world. and

48:00

educate, for example, legislators

48:05

so that they understand

48:07

what this technology is, what it

48:09

can do, what it can't

48:11

do. So you're traveling to

48:13

Washington or the very generous

48:16

trustees of this institution are bringing

48:18

congressional staff and they're both? Both.

48:20

Both are happening. So,

48:22

Feifei, first of all, did you teach that

48:25

course in Stanford HAI or was

48:27

the course located in the philosophy department or cross-listed?

48:29

I'm just trying to get a feel for what's

48:31

actually taking place there now. Yeah,

48:33

I actually taught it in the

48:35

confines of the HAI building. Okay,

48:37

so it's an HAI? No,

48:40

it's a philosophy. It's listed as

48:42

a philosophy course but taught in the HAI.

48:44

He's the former provost. He's an

48:47

inter-flood disciplinary, walking wonder. And

48:50

your work in AI-assisted

48:52

healthcare, is that taking

48:54

place in HAI or is it at

48:57

the medical school? Well, that's the beauty.

48:59

It's taking place in HAI, computer

49:01

science department, the medical school, even

49:04

has collaborators from the law school,

49:06

from the political science

49:08

department. So that's the beauty.

49:10

It's deeply interdisciplinary. If

49:13

I were the provost, I'd say this is starting to

49:15

sound like something that's about to run amok. Doesn't

49:17

that sound a little too interdisciplinary, John? Don't

49:21

we need to define things a little bit here? Let

49:23

me tell you, let me say something. So,

49:26

Steve Denning, who was the

49:29

chair of our board of trustees for

49:31

many years, and has been a long,

49:33

long time supporter of the

49:36

university in many, many ways. In

49:39

fact, we are the Denning co-directors

49:41

of Stanford HAI. Steve

49:46

saw five, six years

49:48

ago, he said, you know, AI

49:51

is going to impact

49:53

in a free department at

49:55

this university. And

49:58

we need to have an institute

50:00

that makes sure that that

50:03

happens the right way. That

50:05

that impact does

50:08

not run amok. Where

50:12

would you like to be in five years? What's

50:14

a course you'd like to be teaching in five years? What's

50:17

a special project? I

50:19

would like to teach a freshman

50:21

seminar called The Greatest Discoveries by

50:23

AI. Oh, alright.

50:27

Okay. A

50:30

last question. I

50:33

have one last question, but that does not mean that

50:35

each of you has to hold yourself to one last

50:38

answer, because it's a kind of open-ended question. I

50:42

have a theory, but

50:44

all I do is wander around this campus. The

50:46

two of you are deeply impeded here, and you ran the place

50:48

for 17 years, so you'll know more than

50:51

I will, including you may know that

50:53

my theory is wrong, but I'm going to trot

50:55

it out, modest though it may be even so.

51:00

Milton Friedman, the late Milton Friedman, who when

51:02

I first arrived here, was a colleague at

51:04

the Hoover Institution. In fact, by some miracle,

51:06

his office was on the same hallway as

51:08

mine, and I used to stop in on

51:10

him from time to time. He

51:13

told me that he went into economics because

51:16

he grew up during the Depression, and

51:19

the overriding question in

51:22

the country at that time was, how

51:24

do we satisfy our material needs? There

51:27

were millions of people without jobs. There

51:30

really were people who had trouble feeding

51:32

their families. Alright. I

51:35

think of my own generation, which

51:38

is more or less John's generation. You'll come

51:40

much later, Fez-Fetti. Thank you. And

51:43

for us, I

51:45

don't know what kind of discussions you had in the dorm room, but

51:47

when I was in college, there were both sessions

51:49

about the Cold War. Were the Russians

51:51

going... The Cold War was real

51:53

to our generation. That

51:56

was the overriding question. How

52:00

can we defend our way of life? How can we defend our

52:02

fundamental principles? All right. Here's

52:05

my theory. For

52:08

current students, they've

52:11

grown up in a period

52:13

of unimaginable prosperity. Material

52:16

needs are just not the problem.

52:19

They have also grown up during

52:22

a period of relative peace. The

52:24

Cold War ended, you could put different – the

52:27

Soviet Union declared itself defunct in 1991. Cold

52:31

War is over at that moment at the latest. The

52:35

overriding question for these kids today

52:39

is meaning. What is

52:41

it all for? Why

52:44

are we here? What does it

52:46

mean to be human? What's the

52:48

difference between us and

52:51

the machines? And if my

52:53

little theory is correct, then

52:56

by some miracle, this

52:59

technological marvel that you have

53:01

produced will lead

53:04

to a new flowering of the humanities. Do

53:07

you go for that, John? Do

53:12

I go for it? I would go for it if

53:15

it were going to happen. Do I put that in

53:17

a slightly sloppy way? I

53:21

think it would be wonderful. It's something to hope

53:24

for. So

53:26

far – now I'm going to be the

53:28

cynic – so

53:31

far what I see in students is more

53:33

and more focus – for

53:36

Stanford students – more

53:38

and more focus on

53:40

technology. Computer science is still the

53:42

biggest major at this university. And

53:47

we have tried at HAI. We

53:49

have actually started

53:51

a program called Embedded Ethics,

53:54

where the CS at the end of

53:56

ethics is capitalized,

53:58

so it's confusing. computer science. That

54:02

will catch the kids' attention. No,

54:04

we don't have to catch their attention. What

54:07

we do is virtually

54:10

all of the courses in

54:12

computer science, the introductory courses, have

54:16

ethics components built in. So

54:19

a problem set, a

54:21

problem set this week, and that will

54:23

have a whole bunch of very

54:26

difficult problems,

54:30

computer science problems, and then it will have

54:32

a very difficult ethical challenge.

54:35

It will say, here's the situation.

54:37

You are programming a computer,

54:40

programming an AI system, and

54:43

here's the dilemma. Now

54:46

discuss. What are you going to do? So

54:49

we're trying to bring—this is

54:52

what Fei-Fei wanted. We're trying to

54:54

bring— This is new. —ethics

54:56

within the last couple of years,

54:59

two, three years. We're

55:01

trying to bring the attention to

55:03

ethics into the computer science

55:05

curriculum. And

55:08

partly that's because they're not—students

55:12

tend to follow the path

55:14

of least resistance. Well, they also— Let's put

55:16

it again. I'm saying things crudely

55:18

again and again, but someone must say it. They

55:20

follow the money. So as

55:23

long as this valley that surrounds

55:25

us rewards brilliant young

55:27

kids from Stanford with CS

55:29

degrees as richly as it

55:31

does, and it is amazingly

55:33

richly, they'll go get CS

55:35

degrees, right? Well,

55:37

I do think it's a little

55:39

crude. I

55:44

think money is one

55:46

surrogate measure of also

55:50

what is advancing in our

55:52

time. You know, the technology

55:54

right now truly is

55:56

one of the biggest drivers of the

55:58

changes of our— of our

56:01

civilization. When you're talking about what does

56:03

this generation of students talk about, I

56:05

was just thinking that 400 years ago,

56:07

you know, when the scientific

56:10

revolution was happening, what is in

56:12

the dorms? Of course it's all

56:14

young men in Cambridge

56:16

or Oxford, but that must also be

56:19

a very exciting and interesting time. Of

56:21

course there was an Internet and social

56:23

media to propel the travel

56:25

of the knowledge, but imagine there

56:28

was, you know, the

56:31

blossoming of discovery and of

56:33

our understanding of the physical

56:35

world. Right now we're

56:37

in that kind of great era

56:40

of technological blossoming. It's

56:42

a digital revolution. So the

56:45

conversations in the dorm, I think, is

56:48

a blend of the meaning of who

56:50

we are as humans as well as

56:52

our relationship to these technology we're building.

56:55

And so it's

56:58

a properly taught

57:01

technology can

57:04

subsume or embed philosophy,

57:08

literature. Of course, can inspire.

57:10

And also think about it,

57:12

what follows scientific revolution is

57:14

a great period of change

57:16

of political, socio-economical change. And

57:19

we're seeing that. All for the better. Right.

57:22

And I'm not saying it's necessarily

57:25

for the better, but we

57:27

are seeing, we're having even

57:29

peaked the digital revolution, but

57:32

we're already seeing the political,

57:34

socio-economic changes. So this is

57:36

again back to Stanford High

57:38

when we founded it five years ago. We

57:41

believe all this is happening

57:44

and this is an institute

57:46

where these kind of conversations,

57:48

ideas, debates should

57:50

be taking place. Education programs should

57:53

be happening. And that's part of

57:55

the reason we did this.

58:00

Let me tell you, as you pointed

58:02

out, I just finished teaching a course

58:04

called Philosophy of Artificial Intelligence. About which

58:06

I found out too late, I would

58:08

have asked permission to audit your course,

58:10

John. No, now you're too old. And

58:15

about half of the students were computer

58:17

science students, who were planned to be

58:19

computer science majors. Another

58:22

quarter planned to be

58:24

symbolic systems majors, which

58:26

is a major that is

58:29

related to computer science. And then

58:32

there was a smattering of others. And

58:36

these were people, every one of

58:38

them, at the end of the

58:40

course, and I'm not saying this to brag,

58:42

every one of them said, this is the

58:44

best course we've ever taken. And

58:47

why did they say that? It

58:49

inspired, it made them think.

58:53

It gave them a framework for

58:55

thinking, a framework for trying

58:57

to address some of these problems, some

58:59

of the worries that you've brought out today.

59:01

And how do we

59:04

think about them, and how do we

59:06

not just become panicked because

59:10

of some science fiction movie that

59:12

we've seen, or because we

59:14

read Ray Kurzweil. Maybe

59:18

it's just as well I didn't take the course.

59:20

I'm sure you're going to give me a C-minus at best.

59:24

Great inflation. So

59:27

it's clear that these

59:29

kids, the students, are

59:34

looking for the

59:38

opening to think these things

59:41

and to understand how to

59:43

address ethical questions, how

59:46

to address hard philosophical

59:48

questions. And

59:52

that's what they got out of the course. And

59:55

that's a way of looking for meaning in

59:57

this time. Yes, it is. Dr.

1:00:00

Fei-Fei Li and Dr. John

1:00:03

Echamendi, both of the Stanford

1:00:05

Institute for Human-Centered Artificial Intelligence.

1:00:08

Thank you. Thank you, Peter. Thank you, Peter. For

1:00:11

Uncommon Knowledge and the Hoover Institution and Fox

1:00:13

Nation, I'm Peter Robinson.