0:00

The following is a conversation with

0:02

Elon Musk, his fourth time

0:04

on this The Lex Friedman Podcast.

0:08

And now, if you'll allow me, a quick

0:10

few second mention of each sponsor. Check them out

0:12

in the description. It's the best way to

0:14

support this podcast. We got Element

0:17

for electrolyte deliciousness, Asleep

0:20

for naps, BetterHelp for your mind,

0:23

SimplySafe for your safety, Shopify

0:26

for your online shopping, and

0:28

NetSuite for business management

0:31

software.

0:32

Choose wisely, my friends. Also,

0:35

if you want to work with our amazing team where I was hiring,

0:37

go to lexfriedman.com slash hiring.

0:40

And now, onto the full ad reads. As

0:42

always, no ads in the middle. I try to make

0:44

these interesting, but if you must skip them, please

0:47

still check out the sponsors. I enjoy their stuff.

0:50

Maybe you will too.

0:52

This episode is brought to you by the

0:54

very thing I'm drinking as I speak to

0:56

you right now, Element. Electrolyte

0:59

drink mix.

1:01

I'll usually fill a Powerade

1:03

or a Gatorade bottle, I

1:06

think it's 28 ounces, of

1:09

water and mix it with one Element

1:11

packet. My favorite flavors, watermelon

1:13

salt. That's the one I'm drinking now. That's the

1:15

one I've been drinking for a long time. I apologize

1:18

because I'm recording these words very late

1:20

at night. It's been a long day.

1:23

It's been a long night before that. A

1:27

long night, this one. So it's just

1:29

me in front of a microphone

1:31

right now, looking at a 28 ounce

1:34

bottle of water that

1:37

tastes like watermelon. Just

1:39

me in the water, happily

1:41

reunited in the desert

1:44

of human experience. I'm

1:46

also surrounded by three computer screens

1:49

that all have beautiful wallpaper,

1:53

looking at me,

1:54

taunting me with beautiful

1:57

picturesque scenes of nature.

2:00

And I just, before this,

2:02

played about 10 minutes of Diablo, where

2:05

I got to escape to a whole other

2:07

world that is also picturesque, but

2:10

not with nature, but filled

2:13

with the intense intimidating sounds

2:15

of demons attacking

2:17

you from all sides. And

2:20

all of that is somehow relaxing. Anyway,

2:23

Element is not just delicious, it combines

2:26

this perfect amount of magnesium,

2:30

potassium, and sodium,

2:33

watermelon salt.

2:35

Anyway, get a sample pack for free

2:37

with any purchase. Try it at www.drinkelement.com.

2:41

This episode

2:44

is also brought to you by the thing that's

2:47

calling my name at this very moment, the 8

2:50

Sleep Pod 3 mattress. It

2:54

has two sides of the bed, so you can cool

2:57

or heat each side of the bed separately. I've

3:00

actually recently was briefly

3:03

at a get-together of a lot of really interesting

3:05

people, and somebody brought up the topic of 8 Sleep,

3:08

and I just heard that a

3:10

bunch of other folks said, Oh, I love 8

3:13

Sleep, and they started talking about the different temperatures

3:15

that they use, and I was shocked and

3:17

horrified to learn that some people actually heat

3:19

their bed up. This was

3:22

a fascinating discovery for me because

3:24

I've never even tried the heating feature. I always

3:27

cool the bed. Sometimes

3:30

it's very cold. With a warm blanket, it's heaven.

3:32

But I can't recommend it enough. Sleep

3:35

really is a magical place. It's a bit surreal

3:37

saying these words while

3:39

quite sleep-deprived, a bit

3:42

melancholic, a bit hopeful,

3:46

thinking about the 8 Sleep bed just a few steps away. Somehow

3:50

that thought fills me with gratitude

3:52

for just having these things in

3:54

my life. Anyway,

3:58

check them out and get some sleep. Special savings when

4:01

you go to 8sleep.com slash

4:03

Lex. This

4:05

episode is also brought to you by BetterHelp, spelled

4:08

H-E-L-P, help. Anytime

4:12

I spell out a word, I always

4:14

imagine myself spelling out

4:16

that word in giant letters and sand,

4:19

stranded alone on an island,

4:22

looking at a airplane floating

4:25

above, hoping that they see the text I write.

4:28

BetterHelp.

4:30

I just actually got a text from a friend.

4:35

I'm looking at it now. Again,

4:37

filled with gratitude for having so many amazing

4:39

people. The text says, hey brother, just

4:41

sending you some love.

4:43

Miss you. Sent late at night.

4:47

Oh boy. Yeah,

4:50

yeah, yeah. So BetterHelp.

4:54

Speaking of friends, it's good

4:56

to talk. It's good to talk to people. It's

4:59

good to talk to people that truly listen. That could be friends,

5:02

loved ones, or that could be a professional

5:04

therapist. But you don't have to choose one.

5:06

You can have all of those. And

5:09

easy, accessible

5:10

therapist

5:12

is what BetterHelp is all about. Check

5:14

them out at BetterHelp.com slash Lex and

5:16

save in your first month. That's BetterHelp.com

5:19

slash Lex.

5:21

This show is also brought to you by Simply

5:23

Safe. And as we go on with

5:26

this program, I'm starting to have more

5:29

and more fun. Because

5:31

fun is all you need. Simply

5:33

Safe is a home security company. After,

5:36

of course, you establish security. Once you're safe,

5:39

then the fun can begin. For

5:41

me, Simply Safe just establishes

5:43

that first base layer of physical

5:45

security. It's super easy

5:48

to set up. They have a cool new feature

5:51

called Fast Protect Monitoring

5:54

that allows Simply Safe monitoring

5:56

agents to see, speak to, and deter intruders

5:59

through the. smart alarm indoor camera.

6:03

I'm tempted to test the system by having

6:06

one of the quadrupeds, the legged robots,

6:09

roll up to it and see what the agents

6:11

say. Just

6:14

attach a Chad GPT, GPT-4 agent

6:18

or a GROK agent on top of it and

6:21

have the two of them have a conversation and see what they

6:24

figure out. I'm not sure it's

6:26

going to end well. But science proceeds

6:30

one experiment at a time, friends. Anyway,

6:33

I love how easy SimpliSafe was to set

6:36

up. Their sensors are awesome. Their monitoring

6:38

is awesome. It's affordable, easy

6:40

to use. I love products that are just, you know, easy

6:42

to set up, easy to maintain.

6:45

Everything works.

6:46

Easy.

6:48

Anyway, go to simplisafe.com

6:50

to get 50% off any SimpliSafe

6:52

system when you sign up for fast, protect,

6:55

monitoring. That's simplisafe.com.

6:57

This

7:00

show is also brought to you by Shopify, a platform

7:03

designed for anyone to

7:05

sell anywhere. And when I say anyone,

7:07

I mean somebody like me.

7:10

It took me minutes, maybe

7:13

even seconds to set up a store and sell

7:15

shirts, which you can now buy at, I

7:18

think the link is lexfreeman.com.

7:21

That forwards you to the Shopify store with three

7:25

different shirts. In

7:27

this case, it's not really about the store. It's about

7:29

just celebrating

7:32

stuff we love. I'm a huge fan of

7:34

a lot of bands and podcasts

7:37

and I just love

7:39

buying up their merch, wearing it and

7:41

seeing who notices. I love

7:45

wearing a Huber and Lab shirt on

7:48

the trail when I run and just

7:50

see who recognizes the full absurdity

7:55

of it, what they say. It's just

7:57

wonderful. The joy that we can share by...

8:00

this little kind of sign that

8:02

you wear on yourself that, oh, I like this

8:04

particular thing. Let's

8:06

talk about it. Yeah, I love it. I

8:08

like wearing like Orwell shirts or or

8:11

Dostoevsky shirts or Bukowski shirts

8:13

or sci-fi stuff. You

8:16

know, 2001 Space Odyssey. Oh,

8:18

you can get all of those things or

8:20

sell those things on Shopify. Sign

8:23

up for a one dollar per month trial period

8:25

at Shopify.com slash Lex. That's

8:28

all lowercase. Go to Shopify.com

8:31

slash Lex to take your business to

8:33

the next level today. This

8:36

show is also brought to you by NetSuite,

8:39

an all in one cloud business

8:42

management system. They

8:44

take care of all the messy,

8:47

complicated things involved in running

8:49

a business. A company like

8:53

Elon says is

8:56

a kind of super intelligent being comprised

8:59

of individual human beings

9:01

working together in interesting ways that

9:04

when effective can be greater than some

9:06

of their parts. You know, but there's a lot

9:09

of things involved there. It's not just great

9:12

design and engineering. It's not great

9:14

innovation. It's

9:17

all of the things, the

9:19

financials, the human resources

9:21

of hiring and firing. It's maintaining

9:23

inventory. It's the e-commerce.

9:26

It's all the messy business details.

9:29

And so you should use the best tools for

9:32

the job of handling that

9:34

mess. That beautiful, beautiful mess involved

9:37

in running a company. And I, too, by

9:39

the way, would love

9:42

to be a part of that mess. A

9:45

small cog in the machine. In

9:47

the super intelligent machine that is a company someday

9:50

soon. Anyway, you can start

9:53

now with no payment or interest for six months.

9:55

Go to NetSuite.com to

9:58

access their one of a kind financing program. that's

10:00

netsuite.com slash lax.

10:05

And now, dear friends, here's

10:07

Elon Musk.

10:30

I thought

10:34

you were

10:38

going to finish it. It's

10:41

one

10:44

of the greatest themes in all of film history.

10:48

That was great. So

10:52

I was just thinking about the Roman Empire, as

10:54

one does. There's

10:57

that whole meme where a guy

10:59

says, I was thinking about the Roman Empire, at

11:02

least once a day. And half the population is confused

11:04

whether it's true or not. But

11:07

more seriously, thinking about the wars going

11:09

on in the world today. And as you

11:11

know, war and military

11:14

conquest has been a big part of

11:16

Roman society and culture. And

11:18

I think it's been a big

11:20

part of most empires and dynasties throughout

11:23

human history. So they usually

11:26

came as a result of conquest.

11:28

I mean, there's something like the Austro-Hungarian Empire where there

11:31

was just a lot of sort of clever marriages.

11:35

But fundamentally, there's an engine of conquest.

11:37

They celebrate excellence in

11:40

warfare. Many of the leaders

11:42

were excellent generals, that

11:45

kind of thing. So big picture question,

11:48

Grok approved. I asked this a good question to

11:50

ask. You tested, Grok approved? Yeah.

11:52

At least on fun mode.

11:59

To what degree do you think... I think war is part of human

12:01

nature versus

12:03

a consequence of

12:06

how human societies are structured. I

12:09

ask this as you

12:11

have somehow controversially been a proponent

12:14

of peace.

12:15

I'm a fairly proponent of peace. Ignorance

12:18

is perhaps in my view the real enemy

12:21

to be countered. That's the real hard

12:23

part, not fighting other humans.

12:27

But all creatures fight. The

12:29

jungle is like, you look at the, people

12:32

think of nature as perhaps some

12:34

sort of peaceful thing, but in fact it is not. There's

12:36

some quite funny when

12:39

a hutsog thing where he's like

12:41

in the jungle, like saying that it's like

12:43

basically just murder

12:46

and death in every direction. I mean,

12:48

the plants and animals in the jungle are constantly trying to

12:50

kill each other every single day,

12:53

every minute. So it's not like

12:56

we're unusual in that respect.

12:58

There's a relevant question here, whether with greater

13:01

intelligence comes

13:03

greater control over these base instincts

13:05

for violence.

13:07

Yes.

13:08

We have much more of an ability

13:10

to control our limbic

13:14

instinct for violence than say a chimpanzee. And

13:16

in fact, if one looks to

13:18

say a chimpanzee society, it is not friendly.

13:21

I mean, the bonobos are an exception, but

13:25

chimpanzee society is full

13:29

of violence and it's

13:31

quite horrific, frankly.

13:33

That's not an limbic system in action. Like

13:36

you don't want to be on the wrong side of a chimpanzee. You'll eat your face

13:38

off and tear your nuts off.

13:40

Yeah, basically there's no limits or ethics

13:42

or, they're almost at just war. There's

13:45

no just war in the chimpanzee societies.

13:47

Is war and dominance by any

13:50

means necessary? Yeah, a chimpanzee society

13:52

is like a primitive version of human society.

13:55

They're not like peace-loving basically at

13:58

all.

13:59

but there's extreme violence. And

14:03

then once in a while, somebody who's watched too many

14:05

Disney movies decides to raise Chopin's ear

14:07

as a pet. And

14:10

then that eats their face, or if they're nuts

14:12

off, or chews their fingers off, and that kind of thing. It's

14:14

happened several times. Ripping your nuts off

14:17

is an interesting strategy for

14:19

interaction. It's happened to people,

14:21

it's unfortunate. That's, I guess, one

14:23

way to ensure that the other

14:25

chimp doesn't contribute

14:28

to the gene pool. Well, from a martial arts

14:30

perspective, it's a fascinating strategy. The

14:33

nut rougher. I

14:37

wonder which of the martial arts teaches that. I

14:39

think it's safe to say if somebody's got

14:42

your nuts in their hands and

14:44

has the option of rubbing them off,

14:45

you will be amenable to whatever

14:48

they want. Yeah. So,

14:51

like I said, somehow, controversially, you've been

14:53

a proponent of peace on Twitter,

14:56

on X. So let me ask you about

14:58

the wars going on today, and to see

15:00

what the path to peace could be. How

15:02

do you hope the current war in Israel and Gaza comes

15:05

to an end? What paths do

15:07

you see that can minimize human

15:09

suffering in the long term, in that part of

15:11

the world?

15:12

Well, I think that part of

15:14

the world is definitely, like if you look

15:17

up the, there is no easy answer in the

15:19

dictionary. It'll be that, like the picture of

15:21

the Middle East, in

15:23

Israel especially. So, there is no easy

15:26

answer.

15:29

What my, strictly

15:32

my opinion of, is that the goal of Hamas

15:34

was to provoke an

15:40

overreaction from Israel. They

15:43

obviously did not expect to

15:45

have a military victory, but they really

15:49

wanted to commit

15:51

the worst atrocities that they could, in order

15:53

to provoke the most

15:56

aggressive response possible from Israel. and

16:00

then leverage that aggressive

16:02

response to rally

16:07

Muslims worldwide for

16:09

the cause of Gaza and Palestine,

16:12

which they have succeeded in doing. The

16:19

counterintuitive thing here, I think that the thing that

16:22

I think should be done, even though it is very difficult, is

16:25

that I would recommend

16:27

that Israel engage in the most conspicuous

16:30

acts of kindness possible, everything.

16:32

That is the actual thing that would

16:35

thwart the goal of Hamas.

16:37

So in some sense, does

16:39

it agree that makes sense in geopolitics

16:42

turn the other cheek implemented? It's

16:44

not exactly turn the other cheek because I do

16:46

think that there's,

16:50

you know, I think it is appropriate

16:52

for Israel to find

16:56

the Hamas members and, you know,

17:00

either kill them or incarcerate them. Like,

17:04

something has to be done because they're just going

17:06

to keep coming otherwise. But

17:09

in

17:12

addition to that, they need to do

17:14

whatever they can. There's some

17:17

talk of establishing, for example, a mobile

17:19

hospital. I'd recommend doing that. Just

17:24

making sure that, you

17:26

know, there's food, water, medical

17:29

necessities,

17:33

and just be over the top about it and be

17:36

very transparent. So it's

17:39

not that you can't claim it's a trick.

17:41

Like, just put a wave cam on the thing.

17:43

You know, all 24-7.

17:46

Deploy acts of kindness.

17:49

Yeah, conspicuous acts

17:51

of kindness that are

17:54

unequivocal, meaning they can't be somehow

17:58

because Hamas will end their response

18:00

will be, oh, it's a trick. Therefore, you have to

18:02

counter how it is not a trick.

18:04

This ultimately fights the broader

18:07

force of hatred in the region.

18:09

Yes. And I'm not sure who said it, to

18:11

some sort of apocryphal saying, but an eye

18:13

for an eye makes everyone blind. Now,

18:16

that neck of the woods, they really believe

18:19

in the whole eye for an eye thing. But

18:23

I mean, you really have... If

18:27

you're not going to just outright commit

18:30

genocide, like against an entire people, which obviously

18:32

would not be acceptable to... to really

18:34

shouldn't be acceptable to anyone, then

18:37

you're going to leave basically a lot of people alive

18:40

who subsequently, you

18:42

know, hate Israel. So really the question is like, how...

18:45

For every Hamas member

18:48

that you kill, how many did you create?

18:51

And if you create more than you

18:53

kill, you've not succeeded. That's

18:56

the real situation there. And

18:58

it's safe to say that if, you

19:00

know,

19:06

if you kill somebody's child in Gaza,

19:09

you've made

19:11

at least a few Hamas

19:14

members who

19:16

will die just

19:19

to kill the Israeli. That's

19:21

the situation. So, but

19:26

I mean, this is one of the most contentious subjects

19:29

one could possibly discuss. But I think

19:32

if the goal ultimately is some sort of long-term

19:34

piece, one has to be... Look at this

19:37

from the standpoint of, over

19:39

time, are there more or fewer terrorists

19:43

being created?

19:44

Let me just linger on war.

19:47

Yeah. Well, war, it's safe to say wars always

19:49

existed and always will exist. Always will

19:52

exist. Always has existed

19:54

and always will exist.

19:55

I hope not. You think it always will? Always

19:57

will. Always will.

20:00

This question of just how much war and you know,

20:02

there's this sort of the scope and scale

20:06

of war, but to

20:08

my, I imagine that there would not be any war in the future

20:11

I think would be a

20:12

very unlikely outcome. Yeah, you talked about

20:14

the culture series, there's war even there.

20:16

Yes, there's a giant war. The first book

20:18

starts off with a gigantic

20:21

galactic war where trillions

20:23

die, trillions.

20:25

But it still nevertheless protects these pockets

20:28

of flourishing. Somehow

20:31

you can have galactic war and

20:33

still have pockets of flourishing.

20:36

Yeah, I mean, I guess if

20:38

we are able to one day expand to,

20:40

you know, full

20:43

of galaxy or whatever, there

20:45

won't be a galactic war at

20:47

some point.

20:48

Ah, the scale, I mean the scale of war

20:50

has been increasing, increasing, increasing. It's

20:53

like a race between the scale of suffering

20:55

and the scale of flourishing.

20:57

Yes.

21:00

A lot of people seem to be using this tragedy to

21:02

beat the drums of war and feed

21:04

the military industrial complex. Do you worry about this?

21:08

The people who are rooting

21:10

for escalation and how can

21:12

it be stopped?

21:14

One of the things that just concerned me is that there

21:16

are very few people alive today who actually, viscerally

21:20

understand the horrors of war, at

21:22

least in the US. I mean, obviously there are people on

21:24

the front lines in Ukraine and Russia who understand

21:27

just how terrible war is, but how

21:29

many people in the West understand it? My

21:33

grandfather was in World War II. He

21:36

was severely traumatized. I mean, he was there

21:39

for almost six years in the East

21:42

and North Africa and Italy. All

21:44

his friends were killed in

21:46

front of him and he

21:48

would have died too, except they randomly

21:51

gave some IQ tests

21:53

or something and he

21:56

scored very high. He

21:58

was not an officer, he was a corporal or a sovereign. or something

22:00

like that, because he didn't finish high

22:02

school. He had to drop out of high school

22:04

because his dad died and he had

22:06

to work to support his siblings.

22:10

So because he didn't graduate high school, he was not eligible

22:13

for the officer corps. So he

22:15

kind of

22:18

got put into the cannon fodder category, basically.

22:22

But then randomly, they gave

22:24

him this test. He was transferred to

22:26

British Intelligence in London. That's where he met my grandmother.

22:31

But he had PTSD

22:33

next level. Like next level. I

22:35

mean, just didn't talk. Just

22:38

didn't talk. And if you tried talking to him, he'd just

22:40

tell you to shut up. And

22:43

he won a bunch of medals. Never ragged

22:45

about it once. Not even hinted.

22:47

Nothing. I found out about it because his

22:50

military records were online. That's how I know.

22:53

So he would get

22:56

like no way in hell if you want to

22:58

do that again. But how many people... Now,

23:01

he died 20 years

23:04

ago, or longer actually, 30 years ago.

23:09

How many people are alive that remember World War II? Not

23:11

many. And the same perhaps

23:13

applies to the threat of nuclear

23:15

war. Yeah.

23:19

I mean, there are enough nuclear bombs pointed at the

23:22

United States to make the radioactive rebel

23:25

bounce many times.

23:28

There's

23:29

two major wars going on right now.

23:32

So you talked about the threat of AGI

23:34

quite a bit. But now,

23:37

as we sit here with the intensity of conflict

23:39

going on, do you worry about nuclear war?

23:41

I

23:43

think we shouldn't just count the possibility of nuclear war.

23:46

It is a civilizational threat.

23:54

Right now, I could be wrong, but I

23:56

think the current probability of

23:58

nuclear war is quite low. But

24:02

there are a lot of nukes pointed at us,

24:05

and we have a lot of nukes pointed at other people. They're

24:08

still there, and nobody's put their

24:11

guns away. The

24:13

missiles are still in the silos.

24:15

And the leaders don't seem to be

24:17

the ones with the nukes

24:19

talking to each other.

24:21

No.

24:22

There are wars which are tragic

24:24

and difficult on a local basis, and

24:26

then there are wars which are civilization-ending.

24:29

Or have that potential. Obviously, global

24:31

thermonuclear warfare has high potential

24:34

to end civilization, perhaps

24:36

permanently. But certainly to

24:39

severely

24:42

wound and perhaps

24:46

set back human progress

24:48

by the Stone Age or

24:51

something. I don't know. Pretty bad.

24:57

Probably scientists and engineers won't be super popular

24:59

after that as well. Like you got to think of this math.

25:04

So generally,

25:06

I think we also want to prioritize civilization

25:08

risks over things that are painful

25:12

and tragic on a local level, but not civilization.

25:18

How do you hope the war in Ukraine comes to an end? And

25:21

what's the path, once again, to minimizing human

25:23

suffering there?

25:26

Well, I think what

25:29

is likely to happen, which

25:31

is pretty much the

25:34

way it is, is that something

25:38

very close to the current lines

25:41

will be how

25:45

a ceasefire or truce

25:47

happens.

25:48

But you

25:51

just have a situation right now where whoever goes on the offensive

25:54

will suffer casualties at several

25:56

times the rate of whoever's on the defense, because

25:59

you've got to... But defense and death

26:01

field, minefields, trenches, anti-tank

26:04

defenses, nobody has

26:06

air superiority because

26:09

the anti-aircraft missiles are

26:11

really far better than the aircraft. Like, there

26:13

are far more of them. And

26:17

so neither side has air superiority.

26:20

Tanks are basically death traps, just

26:23

slow moving and they're not immune

26:26

to anti-tank weapons. So

26:28

you really just have long range artillery and

26:33

infantry, trenches. It's all

26:36

over one, all over again.

26:38

With drones, yeah, throwing

26:40

a little drone, some drones there. Which

26:43

makes the long range artillery just that much more

26:45

accurate and better and so more

26:47

efficient at murdering people on both

26:50

sides.

26:52

Yeah. Whoever

26:54

is... You don't want to be trying

26:56

to advance from either side because the probability

26:58

of dying is incredibly high. So

27:02

in order to overcome

27:07

defense and death trenches and minefields,

27:10

you really need significant

27:12

local superiority in numbers.

27:16

Ideally combined homes where

27:18

you do a fast attack

27:20

with aircraft, a

27:23

concentrated number of tanks and

27:27

a lot of people. That's the only way you're going to

27:29

punch through a line. And then you're going to punch

27:31

through and then not

27:34

have reinforcements just kick

27:36

you right out again. I mean, I

27:39

really recommend people read World

27:42

War I warfare in detail.

27:46

It's rough.

27:51

I mean, the sheer number of people that died there was mind-boggling.

27:55

And it's almost impossible to imagine

27:58

the end of it that doesn't look like... almost

28:00

exactly like the beginning in terms of

28:03

what land belongs to who

28:05

and so on.

28:07

But on the other side of a lot of human

28:10

suffering,

28:11

death and destruction of infrastructure.

28:14

Yes. The

28:16

reason I proposed

28:19

some sort of crucial or peace

28:23

a year ago was because I predicted

28:26

pretty much exactly what would happen, which

28:28

is a lot of people dying for basically

28:31

almost no changes in land. And

28:35

the loss of the flower

28:37

of Ukrainian and Russian youth, and

28:39

we should have some sympathy for the Russian

28:42

boys as well as Ukrainian boys because the Russian

28:45

boys didn't ask to be on their front line. They have

28:47

to be. So there's

28:53

a lot of sons not coming back to their parents. And

28:58

I think most of them don't really have,

29:00

they don't hate the other side.

29:02

It's sort of like, like

29:05

this same country

29:07

from World War I, it's like young

29:10

boys who don't know each other, killing

29:12

each other on behalf of old men

29:14

that do know each other. The

29:18

hell is the point of that?

29:20

So Volodymyr Zelensky said that he's

29:22

not or has said in the past, he's

29:24

not interested in talking to Putin directly. Do

29:26

you think he should sit down, man

29:29

to man, leader to leader and negotiate peace?

29:31

I think I would

29:34

just recommend do not send the

29:37

flower of Ukrainian youth to

29:39

die in trenches.

29:43

Whether he talks to Putin or not, just

29:45

don't do that. Whoever

29:48

goes on the offensive will lose massive

29:51

numbers of people.

29:55

And history will not look kindly upon them.

30:00

You've spoken honestly about the possibility of war

30:02

between US and China in the long term,

30:05

if no diplomatic solution is found. For

30:08

example, on the question of Taiwan and one China policy,

30:11

how do we avoid the trajectory where these two

30:13

superpowers clash?

30:16

Well, it's worth reading that book and the difficulty

30:19

to pronounce the acidities trap, I believe it's

30:21

called. I love war history, I

30:23

like it inside out and backwards.

30:28

There's hardly a battle I haven't read about.

30:31

And trying to figure out what really was

30:33

the cause of victory in any particular case, as

30:35

opposed to what one side or another claimed

30:37

for the reason. Both the victory

30:40

and what sparked the war.

30:42

Yeah,

30:43

yeah. The whole thing. Yeah. So

30:46

Athens and Sparta is a classic case. The thing about the Greeks is

30:48

they really wrote down a lot of stuff. They loved writing.

30:52

There are lots of interesting things that happened many parts of the world,

30:54

but their people just didn't write down. So

30:56

we don't know what happened. Or

30:59

they didn't really write in detail.

31:02

They just would say like, we went,

31:05

we had a battle and we won. And like, can you add a bit more? The

31:09

Greeks, they really wrote a lot. They're

31:13

very articulate. They just love writing. And

31:16

we have a bunch of that writing that's preserved. So we know what

31:19

led up to the Peloponnesian War between

31:22

the Spartan and Athenian alliance. And

31:28

we know that they saw it coming.

31:34

I mean, the Spartans didn't write. They also weren't

31:36

very proposed by their nature. But they didn't write, but

31:38

they weren't very proposed. They weren't terse.

31:41

But the Athenians and the other Greeks wrote

31:43

a line. And they were like... And

31:47

Sparta was really kind of like the leader

31:50

of Greece. But

31:52

Athens grew stronger and stronger with each passing

31:54

year. And

31:57

everyone's like, well, that's inevitable that there's going to be

31:59

a clash.

31:59

between

32:01

Athens and Sparta.

32:03

Well, how do we avoid that? And they couldn't, they couldn't,

32:05

they actually, they saw it coming and they

32:07

still could not avoid it. So,

32:12

you know, at some point, if there's, if

32:14

one

32:17

group, one civilization or country

32:19

or whatever

32:22

exceeds another, sort of like,

32:25

the United States has been the biggest kid on the block since

32:28

I think around 1890, from

32:31

an economic standpoint. So

32:34

the United States has been the economic,

32:36

most powerful economic engine

32:39

in the world, longer than anyone's

32:41

been alive. And

32:46

the foundation of war is economics. So

32:52

now we have a situation in case of China

32:55

where the economy

32:58

is likely to be two,

33:00

perhaps three times larger than that of

33:03

the US. So imagine you're the biggest kid

33:05

on the block for as long as anyone can remember,

33:07

and suddenly it comes along, it's twice your size.

33:13

So we see it coming. Yeah. How

33:16

is it possible to stop?

33:18

Is there some, let me throw something

33:20

out there, just intermixing of cultures,

33:23

understanding. There just seemed to be

33:25

a giant cultural gap in understanding of

33:27

each other. And you're an interesting case

33:29

study because you are an American,

33:31

obviously, you've done a lot of incredible

33:35

manufacture here in the United States, but you also work

33:37

with China. I spent a lot of time in China and

33:39

met with the leadership many times. Maybe a good

33:41

question to ask is, what are

33:43

some things about China that people don't understand

33:46

positive, just in the culture? What's

33:48

some interesting things that you've learned about

33:50

the Chinese?

33:54

Well, the sheer number

33:56

of really smart, hardworking people in China

33:59

is...

34:01

incredible. There are,

34:03

I believe, how

34:05

many smart, hard-working people are there in China? There's

34:07

far more there than there are here, I

34:09

think, in my opinion. There's

34:14

a lot of energy.

34:18

So, I

34:20

mean, the architecture in China

34:22

that's in recent years is far more impressive than

34:25

the US.

34:26

I mean, the train stations, the buildings,

34:29

the high-speed rail, everything, it's really

34:34

far more impressive than what we

34:36

have in the US. I mean, I recommend somebody just go

34:39

to Shanghai and Beijing,

34:41

look at the buildings, and go to take

34:44

the train from Beijing to Xi'an where you have the

34:46

terracotta warriors. China's

34:48

got an incredible history, a very long

34:50

history. And I

34:54

think, arguably, in terms

34:56

of the use of language

34:58

from a written standpoint, one

35:01

of the oldest, perhaps

35:04

the oldest, written language. And

35:06

in China, people did write things down. So,

35:09

now,

35:11

China historically has always been,

35:14

with the rare exception, been internally focused.

35:18

They've not been acquisitive.

35:20

They've fought each

35:22

other. There have been many, many civil wars. In

35:26

the Three Kingdoms War, I believe they

35:28

lost about 70% of their population. So,

35:34

they've had brutal internal

35:36

wars, like civil wars that make

35:39

the US civil war look small,

35:42

by comparison. So,

35:48

I think it's important to appreciate that China is not monolithic.

35:53

We sort of think that China is this sort of one

35:56

entity, well, one mind, and this is

35:58

definitely not the case. And from

36:01

what I've seen, and I think most

36:04

people who understand China would agree, people

36:07

in China think about China 10 times more

36:09

than they think about anything outside of China. So

36:12

it's like 90% of their

36:14

consideration is internal.

36:19

Well, isn't that a really positive thing when

36:22

you're talking about the collaboration and the future

36:24

peace between superpowers when you're

36:26

inward facing, which is like focusing on

36:28

improving yourself versus focusing

36:31

on, quote unquote,

36:33

improving others through military might? The

36:36

good news, the history of China suggests

36:39

that China is not acquisitive, meaning they're not going to go

36:41

out and invade a whole bunch of countries. They

36:44

do feel very strongly, so that's

36:46

good, because a lot of very

36:49

powerful countries

36:49

have been acquisitive. The

36:53

US is also one of the rare cases that

36:55

has not been acquisitive. After

36:57

World War II, the US could have basically taken over the world

37:00

and any country. Like we've got nukes, nobody else

37:02

got nukes. We don't even have to lose soldiers.

37:06

Which country do you want? And the United

37:08

States could have taken over everything. Oh,

37:10

it had well, and it didn't. And

37:13

the United States actually helped rebuild countries. It

37:16

helped rebuild Europe. It helped

37:18

rebuild Japan. This

37:20

is very unusual behavior, almost

37:23

unprecedented.

37:27

The US did conspicuous acts

37:30

of kindness,

37:31

like the Berlin airlift.

37:36

And I think it's

37:39

always like, well, America's done

37:41

bad things. Well, of course America's done bad things, but one needs to

37:43

look at the whole track record. And

37:47

just generally, one

37:51

sort of test would be, how do you treat your prisoners

37:53

of war? Or

37:56

let's say... No

38:01

offense to the Russians, but let's say

38:03

you're in Germany, it's 1945. You've

38:05

got the Russian army coming on one side, and

38:07

you've got the French, British, and American

38:09

armies coming on the other side. Who would

38:11

you like to be to

38:13

surrender to?

38:14

No country is morally perfect, but

38:17

I recommend being

38:20

a POW with the Americans. That would be

38:22

a mighty choice very strongly. In

38:25

the full menu of POW. Very much

38:28

so. In fact, one

38:30

of our

38:33

Brown, a small guy, was like,

38:35

we've got to be captured by the Americans. Yeah.

38:39

And in fact, the

38:41

SS was under orders to execute more

38:43

Brown and all of the German

38:46

ruckus. And

38:49

they narrowly escaped their

38:52

SS. They said they were going out for a walk

38:54

in the woods. They left in the middle of winter with no coats.

38:58

They ran like, no food, no coats, no water, and

39:01

just ran like hell and ran west.

39:04

And by sheer like, I think his brother found

39:06

like a bicycle or something, and then

39:10

just cycled west as fast as he

39:12

could and found a US patrol. So

39:18

anyway, that's one way you can tell morality.

39:22

Who would you want to be a POW? It's

39:25

not fun anywhere, but some places are much worse than

39:27

others. So anyway,

39:31

so like America has been, well,

39:34

far from perfect, generally a

39:37

benevolent force.

39:39

And we should

39:41

always be self-critical and try to

39:43

be better.

39:45

But anyone with half

39:48

a brain knows that. So

39:50

I think there are, in this way, China

39:53

and the United States are

39:55

similar. Neither country has been acquisitive.

39:59

in a significant way. So that's like a

40:02

shared principle, I guess. Now,

40:05

China does feel very strongly about Taiwan.

40:09

They've been very clear about that

40:12

for a long time. From

40:14

this standpoint, it would be like one of

40:16

the states is not

40:20

there like Hawaii or something like that,

40:23

but more significant than Hawaii. Hawaii

40:28

is pretty significant for us. So they

40:30

view it as really

40:33

the... There's

40:36

a fundamental part of China, the

40:39

island of Formosa, now Taiwan

40:41

that is not

40:44

part of China, but should be. And the

40:46

only reason that it hasn't been is because of the US specifically.

40:50

And as their economic power grows and as their military

40:53

power grows, the thing

40:55

that they are clearly saying is

40:59

their interests will

41:02

clearly be materialized.

41:04

Yes. China

41:06

has been very clear that

41:09

they will incorporate Taiwan

41:12

peacefully or militarily,

41:16

but that they will incorporate it from

41:18

this standpoint is 100% likely.

41:22

Now, something you said about conspicuous acts

41:24

of kindness. As a geopolitical policy,

41:27

it almost seems naive.

41:31

I'd venture to say that this is probably the path

41:34

forward, how you avoid

41:36

most wars. Just as you

41:39

say it, it sounds naive, but

41:41

it's kind of brilliant. If

41:43

you believe in the goodness of

41:47

underlying most of human nature, it

41:49

just seems like conspicuous

41:51

acts of kindness can

41:54

reverberate through the populace of

41:57

the countries involved.

41:59

Well, and de-escalate. Absolutely.

42:02

So after World War

42:04

I, they

42:07

made a big mistake. They basically

42:09

tried to lump all the blame on Germany

42:14

and settled

42:18

Germany with impossible

42:20

reparations. And

42:24

really, there was quite a bit of blame

42:27

to go around

42:29

for World War I, but they tried

42:32

to put it all on Germany. And

42:38

that laid the

42:40

seeds for World War II. So

42:43

a lot

42:46

of people were not just Hitler. A lot of people felt wronged.

42:48

And they wanted vengeance. And

42:51

they got

42:55

it.

42:56

People don't forget. Yeah.

42:59

You kill somebody's father,

43:02

mother, son, daughter, they're not going to forget it.

43:06

They all want vengeance. So

43:09

after World War II, they're like, well, that Treaty

43:11

of Versailles was a huge mistake at

43:14

World War I. And

43:17

so this time, instead of

43:23

crushing the losers, we're actually

43:25

going to help them

43:27

with the Marshall Plan. And we're

43:29

going to help revolt Germany.

43:31

We're going to help revolt Australia

43:35

and the other, Italy

43:37

and whatnot. So

43:43

that was the right

43:45

move. It does feel like there's a profound truth

43:48

to conspicuous

43:51

acts of kindness being an antidote to this.

43:55

Something must stop the cycle of reciprocal

43:57

violence. something

44:00

will stop it or it will you

44:02

know it'll it'll

44:04

never stop just eye

44:08

for an eye tooth for a tooth limb for

44:10

a limb life for a life

44:12

forever and ever

44:13

to escape

44:15

briefly the darkness was some incredible engineering

44:18

work xai

44:20

just released grok ai assistant

44:23

that i've gotten a chance to play with it's uh

44:25

it's amazing on many levels first of all it's amazing

44:27

that a relatively small team in a

44:30

relatively short amount of time was

44:32

able to develop this closest

44:35

state-of-the-art system uh

44:37

another uh incredible thing is there's

44:39

a regular mo and there's a fun mode yeah i

44:41

guess i'm to blame for that one i

44:45

wish it first of all everything in life had a fun

44:47

mode yeah i did there's something

44:50

compelling beyond just fun about

44:52

the fun mode interacting with

44:54

a larger language model i'm not sure exactly

44:57

what it is because i've only had a little bit of

44:59

time to play with it but it just makes

45:02

it more interesting more vibrant

45:04

to interact with the system yeah

45:07

absolutely i um

45:10

yeah our our

45:12

ai grok is modeled after the hitchhiker's

45:15

guide to galaxy uh which

45:17

is one of my favorite books which

45:19

is it's a book on philosophy it disguises

45:22

a book on humor um and

45:25

um i would say

45:27

that is that forms the basis of my philosophy uh

45:31

which is that we don't know the

45:33

meaning of life but the more

45:35

we can expand the scope and scale of consciousness digital

45:41

and biological the more

45:43

we are able to understand what questions to ask about

45:45

the answer that is the universe so

45:49

i have a philosophy of curiosity

45:52

there is generally a feeling like this

45:55

ai system has an outward

45:59

looking Like the way you are

46:01

like sitting with a good friend looking up at the stars Like

46:04

the asking pod head like questions about

46:07

the universe wondering what it's all about the curiosity

46:09

you talk about there There's a sense no matter

46:11

how mundane the question I ask it There's

46:14

a sense of cosmic grandeur to

46:16

the whole thing. Well, we are actually working

46:19

hard to have engineering

46:22

math physics Answers that

46:24

you can count on

46:27

So for the other sort of AIs

46:30

out there that

46:32

there's these so-called large language models

46:36

I've not found the

46:37

engineering to be reliable and

46:42

The hallucination it unfortunately hallucinates

46:44

most when you least wanted to hallucinate So

46:48

when you're asking important difficult questions, that's

46:50

what it tends to be confidently wrong So

46:55

we're really trying

46:57

hard to say, okay, how do we Be

46:59

as grounded as possible so you can count on the results

47:03

Trace things back to physics first principles

47:07

mathematical logic

47:09

so underlying the humor

47:12

is an aspiration to Adhere

47:16

to the truth of the universe as closely as possible That's

47:20

really tricky it is tricky. So

47:24

That's right, you know There's always

47:26

gonna be some amount of error, but we want

47:28

to aspire to

47:32

Just

47:33

Truthful as possible about the answers

47:36

with acknowledged error and so

47:38

that there was always you already confidently wrong,

47:41

so You know, it's not gonna

47:43

be right every time but you don't you

47:45

want to minimize how often you're confidently

47:48

wrong And then like

47:50

I said once you can count on the

47:52

logic as being Not

47:56

violating physics then

47:58

you can start to pull on that to create

48:01

inventions like invent new technologies.

48:04

But if you can't

48:08

count on the foundational physics being correct, obviously

48:11

the inventions are simply wishful

48:13

thinking.

48:14

You know, imagination. Magic,

48:18

basically.

48:19

Well, as you said, I think one of the

48:21

big goals of XAI is to understand the universe.

48:24

Yes. That's

48:26

how simple three word mission.

48:31

If you look out far into the future,

48:33

do you think on this level of physics,

48:35

the very edge of what we understand about physics, do you

48:38

think it will make discoveries,

48:41

sort of the sexiest discovery of them is, as

48:43

we know now, sort of unifying

48:46

general relativity and quantum mechanics. So

48:48

coming up with a theory of everything, do you

48:51

think it could push towards that

48:53

direction, almost like theoretical physics discoveries?

48:56

If an AI cannot figure out new physics,

48:59

it's clearly not equal

49:01

to humans, nor has surpass humans

49:03

because humans have figured out new physics.

49:06

Physics is just understanding, deepening

49:10

what's inside it to how reality works. And

49:12

then this

49:16

engineering, which is inventing things that have never existed.

49:18

Now, the range

49:20

of possibilities for engineering is far greater than for physics,

49:23

because, you know, we once you figure out the rules of the universe,

49:26

that's that you've discovered

49:28

things that already existed. But from

49:30

that, you can then build technologies

49:33

with that are really almost

49:36

limitless in the

49:39

variety and keep, you know, sequence,

49:42

you understand the rules of the game properly. And

49:44

we do, we have current physics, we do, at

49:47

least at a local level, understand how physics works

49:50

very well. Our ability to predict things

49:52

is incredibly good. Like quantum mechanics is the

49:55

degree to which quantum mechanics can predict

49:57

outcomes is incredible.

49:59

That

50:00

was my hardest class in college, by

50:02

the way. My senior

50:04

quantum mechanics class was harder than all of my other

50:07

classes put together. To get

50:09

an AI system, a large language

50:11

model to

50:15

be as reliable as quantum mechanics and physics is

50:18

very difficult.

50:19

Yeah, you have to test any conclusions

50:21

against the ground truth of reality. Reality is the

50:23

ultimate judge. Like physics is the law, everything

50:26

else is a recommendation. I've

50:28

seen plenty of people break the

50:30

laws made by man, but none break the laws

50:32

made by physics.

50:33

It's a good test, actually. If this LOM

50:37

understands and matches physics, then

50:40

you can more reliably trust whatever

50:42

it thinks about the current state of politics.

50:47

It's also not the case currently that even

50:49

its internal logic is not consistent.

50:53

So especially with the

50:57

approach of just predicting a token,

51:00

predict token, predict token, it's

51:01

like a vector sum. You're

51:04

summing up a bunch of vectors, but you can get drift.

51:07

So as though a little bit of error

51:09

adds up. And by the

51:11

time you are many tokens down the path, it

51:15

doesn't make any sense.

51:17

So it has to be somehow self-aware about

51:19

the drift.

51:20

It has to be self-aware about the drift and then look

51:23

at the thing as a gestalt,

51:25

as a whole, and say

51:27

it doesn't have coherence

51:30

as a whole.

51:31

So when

51:35

authors write books, they will write

51:37

the book and then they'll go and revise it, taking

51:41

into account the

51:43

end and the beginning and the middle and rewrite

51:46

it to achieve coherence so

51:48

that it doesn't end up in a nonsensical

51:51

place. Maybe the process of revising

51:53

is what

51:55

reasoning is and then the process

51:57

of revising is how you get closer and closer to truth.

52:00

Maybe you like at least I approach

52:02

it that way you just say a bunch of bullshit

52:04

first and then you get it better You start

52:06

a bullshit and then you get a draft and then

52:08

and then you and then you iterate on that draft Until

52:12

it has coherence until it's

52:15

it all adds up basically

52:17

So another question about theory of everything

52:19

but for intelligence Do you think there

52:21

exists as you're exploring this

52:23

with xai creating this intelligence system? Do you

52:25

think there is a theory of intelligence

52:28

where you get to understand?

52:30

What? like

52:32

What is the eye

52:35

in AGI and what is the eye in? Human

52:39

intelligence there's no I in team America.

52:42

Oh wait, you're Not

52:45

gonna be stuck in my head now Yeah,

52:51

there's no me and whatever in

52:54

quantum mechanics wait I Mean

53:00

is that part of the process of discovering

53:02

understanding the universe is understanding intelligence

53:06

Yeah Yeah,

53:08

I think we need to understand intelligence understand

53:10

consciousness I mean I there

53:13

I mean there are some sort of fundamental questions of like

53:16

what is Thought what

53:18

is emotion? Yeah Is

53:21

it really just one atom bumping into another atom

53:24

it feels like something more than that So

53:27

I I think we're probably

53:29

missing some really big things

53:34

Like some really big things

53:36

Something they'll be obvious in retrospect. Yes,

53:39

if there's a giant You

53:41

put the whole consciousness emotion

53:44

Well, some of them would quote like a like a soul, you know,

53:47

really just a soul Like

53:49

you feel like you're you right?

53:52

I mean you don't feel like You're

53:54

just a collection of atoms,

53:56

but on what dimension does

53:59

thought exist? What dimensions do emotions

54:01

exist? We feel them very strongly.

54:04

Um, I suspect there's more to it than atoms

54:06

bumping into atoms. And

54:11

maybe AI can pave the path

54:13

to the discovery of what, whatever the hell that thing

54:15

is.

54:16

Yeah. What is consciousness? Like what,

54:18

when you put the atoms in a particular shape, why

54:21

are they able to form thoughts

54:22

and

54:24

take actions that enter

54:26

out of feelings?

54:28

And even if it is an illusion, why is this

54:30

illusion so compelling? Yeah.

54:32

Like how do you, why does this illusion exist? Yeah.

54:36

On what plane does this,

54:38

the solution exist? Yeah. Um,

54:41

and that sometimes I wonder is, you

54:44

know, either perhaps everything's conscious

54:46

or nothing is conscious.

54:48

Um,

54:49

one of the two. I like the

54:51

former, everything

54:53

conscious just seems more fun.

54:56

It does seem more fun. Yes. Um,

54:59

but we're, we're composed of atoms and those atoms

55:01

are composed of quarks and leptons and

55:04

those quarks and leptons have been around since

55:06

the beginning of the universe.

55:08

The beginning of the universe.

55:11

Right. What, what seems to be the beginning of the universe?

55:13

The first time we talked, you said what you would, which

55:16

is so real to think that this

55:18

discussion was happening is becoming a reality.

55:20

I asked you what question would you

55:23

ask an AGI system once you create it? And

55:25

you said, what's outside the simulation is

55:27

the question.

55:29

Good question.

55:31

Yeah. But it seems like with Grog, you started

55:34

literally, uh, this system's

55:37

goal is to be able to ask such questions, to

55:39

answer such questions and ask such questions.

55:42

Where are the aliens? Where are the aliens? That's

55:44

one of the, like the Fermi paradox question.

55:47

Um,

55:49

a lot of people have asked me if I've seen any evidence

55:51

of aliens and I haven't, which

55:53

is kind of concerning cause

55:55

then I think would,

55:58

I'd probably prefer to at least have seen some. archaeological

56:00

evidence of aliens. To

56:02

the best of my knowledge, there is no proof that I

56:05

not worth any evidence of aliens. If

56:08

they're out there, they're very subtle. We

56:11

might just be the only consciousness, at least in

56:13

the galaxy. And

56:16

if you look at the history of Earth,

56:19

for instance, believe the archaeological record, Earth

56:21

is about four and a half billion years old. Civilization,

56:25

as measured from the first writing, is

56:27

only about 5,000 years old. We

56:30

have to give some credit there to the ancient Sumerians who

56:32

aren't around anymore. I think it was an

56:34

archaic pre-cuniform, the

56:36

first actual symbolic representation,

56:39

but only about 5,000 years ago.

56:42

I think that's a good date for

56:45

when we're to say civilization started. That's

56:48

one millionth of Earth's existence. So

56:53

civilization has been around. It's

56:55

really a flesh in the pan so

56:58

far. Why did it take so long? Four

57:04

and a half billion years. For

57:15

the vast majority of time, there

57:17

was no life, and then there was archaic

57:20

bacteria for a very long time. And

57:23

then yet,

57:28

mitochondria get captured, multicellular

57:30

life,

57:32

differentiation

57:33

into plants and animals, life

57:36

moving from the ocean still end, mammals,

57:41

higher brain functions. The

57:44

sun is expanding slowly, but it will

57:47

heat the Earth

57:50

up at some point in the future,

57:55

while the oceans and

57:58

Earth will become like Venus. where

58:02

life as we know it is impossible. So

58:04

if we do not become multi-planetary and

58:07

ultimately go beyond our solar

58:09

system, annihilation

58:12

of all life on Earth is a certainty. And

58:19

it could be as little as on the

58:22

galactic timescale

58:25

half a billion years. You

58:28

know, it's a long time by human standards but that's

58:31

only 10% longer than Earth has been around at

58:33

all. So if life

58:36

had taken 10% longer to

58:38

evolve on Earth it wouldn't exist

58:41

at all. We've

58:45

got a deadline coming up. You

58:47

better hurry. But that says, as you said,

58:50

humans intelligent life on Earth

58:54

developed a lot of cool stuff very quickly.

58:56

So

58:57

it seems like becoming a multi-planetary

59:00

is almost inevitable

59:03

unless we destroy it. We need to do it. I mean,

59:05

it's not... I mean,

59:08

I suspect that if we are able to go out

59:12

there and explore other star systems that

59:14

there's a good chance we find a whole bunch

59:16

of long dead one planet civilizations.

59:19

They never made it past their own planet. That's so

59:21

sad. Yeah.

59:24

Also fascinating. I mean,

59:25

there are very explanations for those point

59:28

of paradox and one of the... there's

59:30

sort of... there are these great filters which

59:33

civilizations don't pass through and

59:35

one of those great filters is do

59:37

you become a multi-planet civilization or not?

59:40

And if you don't, it's simply a matter of time before

59:42

something happens on your planet. You're

59:46

either natural man-made

59:49

that causes us to die out like

59:52

the dinosaurs. Where are they

59:54

now?

59:56

They didn't have spaceships.

1:00:00

I think the more likely thing is, because

1:00:02

it's just to empathize

1:00:05

with the aliens, that they found us

1:00:07

and they're protecting us and letting us be.

1:00:09

I hope so. They're nice aliens.

1:00:11

Just like the tribes in the Amazon. They

1:00:14

all contact the tribes for protecting them. That's

1:00:16

what... That would be a nice explanation. Or

1:00:19

you could have like a... What was

1:00:21

it? I think Andre Capati

1:00:23

said it's like the ants in the Amazon asking,

1:00:26

where's everybody?

1:00:27

Well they do run

1:00:29

into a lot of other ants. That's

1:00:31

true. They have these ant wars. Sounds

1:00:35

like a good TV show. Yeah, they literally have these

1:00:37

big wars between various ants. Maybe

1:00:40

I'm just

1:00:43

dismissing all the different diversity of ants.

1:00:46

Listen to that Werner Herzog talking about the jungle.

1:00:48

It's really hilarious. Have you heard it?

1:00:50

No I have not, but Werner Herzog is away. You

1:00:55

should play it as

1:00:58

an interlude in the fun view to

1:01:00

you. It's awesome. I

1:01:03

love him so much. He's great. Was

1:01:06

he the director of Happy People, Life in the Taiga? I

1:01:08

think also. I did that fair documentary. The

1:01:10

fair documentary. And the thing about penguins. The analysis,

1:01:13

psychoanalysis of penguins. Yeah,

1:01:15

the penguins headed for mountains that

1:01:17

are like 70 miles away.

1:01:21

The

1:01:24

penguin has just headed for doom basically. Well

1:01:27

he had a cynical take. He

1:01:29

could be just a brave explorer and

1:01:31

there would be great stories told about

1:01:34

him amongst the penguin population for

1:01:36

many centuries to come.

1:01:41

What are we talking about?

1:01:44

Okay. Aliens. Yes,

1:01:47

aliens. I mean, I don't know. I

1:01:50

think the smart move

1:01:52

is just, you know, this is the first time

1:01:54

in the history of Earth

1:01:57

that it's been possible for life to extend

1:01:59

beyond. that

1:02:01

window is

1:02:02

open.

1:02:05

Now it may be open for a long time or

1:02:07

maybe open for a short time. It

1:02:10

may be open now and then never open again. So

1:02:14

I think the smart move here is to make

1:02:16

life multi-planetary while it is possible

1:02:18

to do so. We don't

1:02:19

want to be one of those lame one planet civilizations, it

1:02:22

just dies out. No, those are lame. Yeah, lame.

1:02:27

Self-respecting civilization would be one planet. There's

1:02:29

not going to be a Wikipedia entry for one of those. And

1:02:31

pause. Does

1:02:36

SpaceX have an official policy

1:02:39

for when we meet aliens?

1:02:41

No. That

1:02:43

seems irresponsible. If

1:02:50

I see the slightest indication that

1:02:52

there are aliens, I will immediately post on the next platform.

1:02:55

Anything I know.

1:02:57

It could be the most liked, reposted

1:02:59

post of all time. Yeah, I mean, look, we

1:03:02

have more satellites up there right now than everyone

1:03:05

else combined. So

1:03:07

we know if we've got to maneuver around

1:03:09

something and we're not maneuvering

1:03:12

around anything.

1:03:13

If you go to the big questions once

1:03:15

again, you said you're

1:03:18

with Einstein that you believe in the God

1:03:20

of Spinoza. Yes. So,

1:03:24

you know, that's a view of the God is

1:03:26

like the universe and

1:03:28

reveals himself through the laws of physics or as Einstein

1:03:31

said, through the lawful harmony of

1:03:33

the world. Yeah, I would agree that

1:03:36

God of the simulator or whatever, the

1:03:39

Supreme Being or beings,

1:03:42

reveal

1:03:44

themselves through the physics, you know,

1:03:46

they have creators of this existence. And

1:03:49

it's incumbent upon us to try to understand

1:03:53

more about this wondrous creation.

1:03:56

Who created this thing? Who's running this thing?

1:04:00

into a singular question with a

1:04:02

sexy word on top of it is like focusing

1:04:04

the mind to understand it does seem like there's

1:04:07

a again it could be an

1:04:09

illusion it seemed like there was a purpose

1:04:12

that there's an underlying master plan of some kind

1:04:14

it seems like there

1:04:17

may not be a master plan in the

1:04:19

sense so there's like

1:04:22

maybe an interesting answer to the question of determinism

1:04:24

versus free will is that if

1:04:26

we are in a simulation the

1:04:29

reason that these higher beings

1:04:31

would hold a simulation is to see what

1:04:33

happens

1:04:34

so it's

1:04:36

not a they don't know what happens

1:04:40

otherwise they wouldn't hold the simulation

1:04:43

so when humans create

1:04:45

simulation so it's SpaceX and Tesla we

1:04:47

create simulations all the time especially

1:04:50

for the rocket you you

1:04:52

know you have to run

1:04:54

a lot of simulations to understand

1:04:57

what's going to happen because you can't really test the rocket

1:05:00

until it goes to space and you want

1:05:02

it to work so you have to you have to simulate

1:05:05

subsonic transonic hypersonic

1:05:08

supersonic hypersonic ascent

1:05:12

and then coming back super high heating and

1:05:16

all dynamics all this is going to be simulated

1:05:18

so you don't

1:05:20

get very many kicks at the can but we

1:05:23

run the simulations to see what happens not if

1:05:25

we knew what happens we wouldn't run the simulation so

1:05:28

if there's so ever created this existence

1:05:32

is they're

1:05:35

running it because they don't know what's going to

1:05:37

happen not because they do

1:05:41

so maybe we both played Diablo

1:05:43

maybe Diablo was created to see if a druid

1:05:46

your character could defeat

1:05:49

uber Lilith at the end they didn't know

1:05:52

well the funny thing is that Google

1:05:55

is her title is hatred

1:05:57

incarnate yeah

1:06:01

And right now, I guess, you're

1:06:03

gonna ask the Diablo team, but it's

1:06:05

almost impossible to defeat

1:06:09

hatred in

1:06:11

the Eternal Realm.

1:06:13

Yeah, you've streamed yourself dominating

1:06:15

tier 100

1:06:16

nightmare dungeons and still... I

1:06:19

can cruise through tier 100 nightmare dungeons

1:06:22

like a stroll in the park. And

1:06:25

still you're defeated by hatred. Yeah,

1:06:27

I can... This sort of, I guess

1:06:29

maybe the second hardest boss is D'Oreal.

1:06:32

D'Oreal can't even scratch the paint. So

1:06:35

I killed D'Oreal so many times. And

1:06:39

every other boss in the game, all of them kill

1:06:41

him so many times, it's easy. But

1:06:46

Uvalilith, although it's known as hatred

1:06:48

incarnate, especially if you're

1:06:50

a druid and you have no ability

1:06:52

to be invulnerable, there are

1:06:54

these random death waves that

1:06:56

come. And I'm pretty... I'm

1:07:01

really, I'm 52,

1:07:03

so my reflex is not what they used to be, but I

1:07:06

don't have a lifetime of playing video games. At

1:07:09

one point, I was maybe

1:07:11

one of the best Quake players in the world. Actually

1:07:15

won money for what I think was the

1:07:17

first paid esports tournament in the US.

1:07:21

We were doing four person Quake tournaments. And

1:07:27

we came second. I was the second best person on the team.

1:07:29

And the actual best person, we were actually

1:07:32

winning, we were going to come first, except the best person on the

1:07:34

team, his computer crashed, hopefully through the

1:07:36

game. So we

1:07:38

came second. But I got money for it

1:07:41

and everything. So basically I got skills, albeit

1:07:44

no no spring chicken

1:07:46

these days. And to

1:07:49

be totally frank, it's driving me crazy. Trying

1:07:52

to beat Lilith as a druid, basically

1:07:55

trying to beat hatred incarnate

1:07:57

in the Eternal Realm. As a druid. I

1:08:00

just drew it. If you're,

1:08:02

it's just really fixing, let

1:08:05

me tell you. I mean, the

1:08:07

challenge is part of the fun. I have seen directly,

1:08:11

like, you're actually like a world-class,

1:08:13

incredible video game player. Yeah.

1:08:15

And I think Diablo, so you're just picking up a new

1:08:18

game, and you're figuring out its fundamentals.

1:08:21

You're also, with the Paragon board and the build,

1:08:24

are not somebody like

1:08:27

me who perfectly follows whatever they suggest

1:08:29

on the internet, you're also an innovator

1:08:31

there. Yeah. Which

1:08:33

is hilarious to watch. It's

1:08:35

like a mad scientist just trying

1:08:38

to figure out the Paragon board and the build

1:08:40

and the build. Yeah. You know? Is

1:08:42

there some interesting insights there about, if

1:08:44

somebody's starting as a druid, do you have advice?

1:08:48

I would not recommend playing a druid in the Attenborough.

1:08:50

No. Right now, I

1:08:52

think the most powerful character in

1:08:55

the seasonal realm is the sorcerer,

1:08:57

with the lightning balls. The

1:09:00

soaks have huge balls in the seasonal.

1:09:04

Oh, yeah. That's what they say. Soaks

1:09:08

have huge balls. They

1:09:10

do, huge balls of lightning. I'll

1:09:13

take your word for it. And it's actually,

1:09:15

in the seasonal realm, you can, it's

1:09:17

like pretty easy to beat Uverloth,

1:09:20

but the, when you get these vampiric powers that

1:09:23

out of fire damage and increase

1:09:25

your defense and whatnot, so. Guru

1:09:30

did quite easy to defeat hatred seasonally,

1:09:33

but to defeat hatred eternally.

1:09:35

Very difficult. Almost

1:09:38

impossible. It's virtually unimpossible. It

1:09:40

seems like this is a metaphor for life. I

1:09:43

like the idea that Elon Musk,

1:09:45

because I saw, I was playing the alley yesterday, and I saw level 100

1:09:50

druids just run by, I will never die, and

1:09:53

then run back the other way. And there's

1:09:55

just some, this metaphor is kind of hilarious,

1:09:57

that you, Elon Musk, is fighting.

1:10:00

hatred, restlessly

1:10:02

fighting hatred in this demonic

1:10:04

realm. Yes.

1:10:06

It's hilarious. I mean, it's pretty hilarious.

1:10:08

No, it's absurd. Really,

1:10:11

it's exercise and absurdity and makes

1:10:13

me want to pull my hair out. Yeah. I

1:10:16

mean, what do you get from video games in general? Is

1:10:18

there for you personally? I

1:10:21

mean, it's, I don't know, it

1:10:24

calms my mind. I mean, sort of killing

1:10:27

the demons in a video game calms

1:10:29

the demons in my mind.

1:10:33

If you play a tough video game, you

1:10:36

can get into like a state of flow, which is very

1:10:38

enjoyable. Admittedly,

1:10:44

it needs to be not too easy, not too hard,

1:10:46

kind of

1:10:48

in the Goldilocks zone. And

1:10:50

I guess you generally want to feel like you're progressing in the

1:10:53

game. So a good

1:10:56

video. And there's also beautiful art,

1:11:01

engaging storylines. And

1:11:05

it's like an amazing puzzle to solve,

1:11:08

I think. And so it's like solving the

1:11:10

puzzle. Elden Ring, the greatest game

1:11:12

of all time. I still haven't played it, but you. Elden

1:11:15

Ring is definitely a candidate for best game ever.

1:11:18

Top five for sure.

1:11:19

I think I've been scared how hard it is. How

1:11:21

hard I hear it is. But

1:11:24

it is beautiful. Elden Ring is,

1:11:27

feels like it's designed by an alien.

1:11:31

There's a theme to this discussion. In

1:11:34

what way? It's so unusual.

1:11:36

It's incredibly creative and the art is stunning. I

1:11:40

recommend playing it on a big resolution,

1:11:43

high dynamic raise TV, even. It doesn't

1:11:46

need to be a monitor. Just

1:11:49

the art is incredible. It's so beautiful. And

1:11:53

it's so unusual.

1:11:57

And each of those top five most

1:11:59

battle. battles is unique,

1:12:02

like it's a unique puzzle to solve. Each

1:12:04

one is different. And

1:12:06

the strategy you use to solve one

1:12:09

battle is different from another

1:12:11

battle.

1:12:12

That said, you said Druid and Eternal against

1:12:14

Uber Lilith is the hardest boss battle

1:12:16

you've

1:12:17

ever... Correct. That

1:12:19

is currently the... And I've played

1:12:22

a lot of video games. Because it's

1:12:24

my primary recreational activity. And...

1:12:31

Yes. Feeding

1:12:33

hatred in the Eternal Realm is

1:12:36

the hardest boss battle in

1:12:39

life and in the video game.

1:12:42

I'm not sure it's possible, but it's...

1:12:46

I do make progress. So then I'm like, okay,

1:12:48

I'm making progress. Maybe if

1:12:51

I just tweak that paragon board a little more,

1:12:53

I can do it. If

1:12:56

I just dodge a few more waves, I

1:12:59

can do it. Well,

1:13:01

the simulation is created for the purpose

1:13:03

of figuring out if it can be done. And

1:13:05

you're just a cog in that simulation,

1:13:08

in the machine of the simulation. Yeah, I see. I

1:13:13

have a feeling that at least...

1:13:17

I think... It's doable.

1:13:20

It's doable, yes. Well,

1:13:25

that's the human spirit right there, to believe.

1:13:27

Yeah. I mean, it did prompt me to think about

1:13:30

just hate in general, which is... You

1:13:34

want to be careful of one of those things where

1:13:37

you wish for something that

1:13:39

sounds good, but

1:13:42

if you get it, it's actually a dystopian situation.

1:13:46

So you could sort of run a

1:13:48

sort of... How

1:13:50

about sort of like if you wish for world peace?

1:13:53

Sounds good.

1:13:55

But how is it enforced? At

1:13:58

what cost? at

1:14:01

what cost eternal peace. It

1:14:03

might actually be worse to have eternal peace because

1:14:05

of what that would entail. The

1:14:08

suppression of everyone, it might

1:14:11

be the suppression of progress. It

1:14:14

might be an occupied society that never changes.

1:14:18

What I'm saying is that there

1:14:20

is an argument that

1:14:27

if you wish for no war, you

1:14:30

should be careful what you wish for because

1:14:33

what's required in order for that to be no war. It

1:14:38

might be worse than a little war.

1:14:40

Yeah, the things we'll consider to be flaws of human

1:14:42

civilization might be a necessary component for

1:14:45

whatever optimal looks like. I

1:14:48

mean, do you worry about AI, AGI

1:14:50

enabling a dystopian state of this nature?

1:14:53

Whether it's 1984 with surveillance

1:14:56

and fear or Brave New World with pleasure.

1:15:01

And what is it, lots of sex but

1:15:05

no deep human experience?

1:15:09

There's actually a real drug called soma.

1:15:12

Yes. Let

1:15:15

me take notes. Does it actually mean

1:15:17

that? No, it probably will go to

1:15:19

Brave New World. Ask it.

1:15:22

What would be the question? Tell

1:15:24

us about the drug called soma. What's

1:15:26

the drug called

1:15:29

soma? It's

1:15:31

going to go to Brave New World, I think.

1:15:33

Well, there's a real drug as well. Searching

1:15:36

for soma drug information. Oh

1:15:41

boy.

1:15:42

So, the one drug, by the way, I'm in fun

1:15:44

mode. Okay, fun. Yes,

1:15:48

so it's talking about the real drug. It's

1:15:50

a muscle relaxant that's been around since the 1950s. It's

1:15:53

used to treat muscle pain and discomfort and it

1:15:55

does this by altering the activity between

1:15:57

neurons and the spinal cord. Soma

1:16:00

isn't just a pain reliever. It's also a bit

1:16:02

of a sedative. It can make you feel a

1:16:04

bit drowsy, so it's not something

1:16:06

to take if you're planning and operating heavy machinery

1:16:08

or trying to solve complex math equations. And

1:16:11

it goes on and on. In conclusion, Soma's a drug

1:16:13

that can be helpful for those dealing with muscle pain, but

1:16:16

it's not without its risks.

1:16:18

It's pretty good. Oh, wait a minute. That's the real

1:16:20

drug you're talking about. It's the real drug. Yeah.

1:16:23

And of course, the fun is always use it responsibly

1:16:26

in accordance with the advice of a healthcare professional. And

1:16:28

remember, in the end, the universe is probably just a

1:16:30

giant computer simulation, so don't take

1:16:32

anything too serious. I love it. I

1:16:35

love it. I love it. I love it.

1:16:37

I love it. Yeah. So. Oh,

1:16:40

and also really nicely includes like

1:16:44

the tweets post on X that are

1:16:46

relevant to this, which is really cool. Yeah.

1:16:50

It like integrates everything together. Yeah.

1:16:52

Like the interface is one of the really cool things

1:16:54

here. Yeah. I think you did a great

1:16:57

job. So,

1:17:00

and you can say like,

1:17:02

no, I mean, soma from Brave

1:17:05

New World. Because it thinks you

1:17:07

mean, you mean soma the real drug.

1:17:09

Yeah. No, I mean soma from Brave New World.

1:17:15

It's

1:17:15

gonna be like, oops, I'm really sorry.

1:17:17

I'm really surprised I went to the real drug first.

1:17:21

Ah, my apologies. I was thinking

1:17:23

of the muscle relax, but soma from Brave New World

1:17:25

is a different beast altogether.

1:17:28

Now those huxley dystopia, now soma is

1:17:30

a drug that plays a central role

1:17:32

in society depicted in the book. It's a mood altering

1:17:35

hallucinogenic drug that's

1:17:37

used to keep the population docile and content

1:17:40

in the novel. To use soma in Brave New World

1:17:42

raises some interesting questions about the nature of happiness

1:17:45

and the role of drugs in society. Oh

1:17:48

man. Exactly. It's better to live in a world

1:17:50

where everyone is happy

1:17:51

all the time. Even if that

1:17:53

happiness is artificial, that's a good

1:17:56

question. This is what I mean. Do

1:18:00

you wish for world peace and happiness all

1:18:02

the time?

1:18:04

Are you sure?

1:18:06

Because that might be society

1:18:09

that is essentially sterile

1:18:11

and ossified that never changes that

1:18:14

is ultimately doomed.

1:18:15

This kind of tension between

1:18:19

the darkness and the light. It's

1:18:23

really a very good summary. It

1:18:26

really gets the point. This is not simply regurgitating

1:18:29

the brave new world. It's actually getting to

1:18:32

the salient elements of

1:18:35

Soma as a drug. Do

1:18:37

you actually want to be in

1:18:39

a situation where everyone is happy all the time even

1:18:42

though it's artificial? Or is it better to confront

1:18:44

the challenges of life and experience a full range of human emotions?

1:18:47

Even if it means experiencing pain and suffering. For

1:18:49

those listening by the way, Elon just read

1:18:52

directly from Grock, which is a really

1:18:54

nice kind of insightful philosophical analysis

1:18:56

of the tension here.

1:18:58

Interesting.

1:18:59

It pretty much nails it. In conclusion,

1:19:01

Soma from Brave New World is fictional drug that

1:19:03

is used to explore some deep philosophical questions

1:19:05

about the nature of happiness and the role of drugs

1:19:08

in society. It's a powerful symbol

1:19:10

of the dangers of using drugs to

1:19:12

escape from reality and the importance of

1:19:14

confronting the challenges of life head-on.

1:19:18

Now that...

1:19:20

And the crazy thing is like, you know, we do have a real drug

1:19:23

called Soma, which kind of does... It's

1:19:25

kind of like the drug in the book. And

1:19:28

I'm like, they must have named it out. Yeah,

1:19:31

probably. Some of the real

1:19:33

drug is quite effective on back pain.

1:19:35

So you know about this drug? I've taken it. It's

1:19:37

fascinating.

1:19:39

Okay. Because

1:19:40

I had like a squashed nuscan

1:19:42

way, T5T6. So

1:19:44

it takes the physical pain away, but Soma

1:19:46

here... It doesn't completely. It reduces

1:19:48

the amount of pain you feel, but at the expense

1:19:51

of mental acuity. It

1:19:54

bails your mind.

1:19:58

Just like the drug in the book.

1:19:59

It's like the truck in the book. Yeah, and

1:20:02

hence why it off The

1:20:04

thing that seems like utopia could be a dystopia

1:20:06

after all. Yeah, actually I was talking

1:20:08

to a friend of mine

1:20:11

Saying like would you really want

1:20:14

there to be no hate in the world like

1:20:16

really none like I

1:20:18

wonder why hate evolved I'm

1:20:21

not saying we should amplify hate of course. I'm

1:20:23

just trying to minimize it but but

1:20:25

none at all

1:20:27

Hmm

1:20:28

there might be a reason for hate

1:20:31

And Suffering and it's really complicated

1:20:34

To consider that some amount of human

1:20:36

suffering is necessary for a few of flourishing

1:20:40

Is it possible to appreciate the highest without?

1:20:43

knowing the lows And

1:20:48

That all is summarized there in a single Statement

1:20:51

from grog. Okay. Hi, no

1:20:53

lows who knows?

1:20:55

That's almost the poem

1:20:59

It seems that training LLMs efficiently is

1:21:01

a big focus for X AI What's

1:21:05

the What's the limit

1:21:07

of what's possible in terms of efficiency? There's

1:21:10

this terminology of Useful

1:21:13

productivity per watt like

1:21:15

what have you learned? Yeah pushing the limits of that.

1:21:17

Well, I think it's helpful. The tools of physics are

1:21:20

Very powerful and can be applied I think to almost

1:21:23

any really any arena and like this

1:21:25

It's really just critical thinking For

1:21:28

something important you need to

1:21:30

Reason with from first principles and think about things

1:21:33

in the limit one direction or the other So

1:21:39

In the limits even at the Kardashev scale meaning

1:21:41

even if you harness the entire power of the Sun

1:21:43

you will still care about useful compute for

1:21:45

what So

1:21:48

that's where I think probably

1:21:50

where things are headed from the standpoint

1:21:53

of AI is

1:21:56

that we have a silicon shortage now

1:21:58

that will transition to a

1:22:00

voltage transformer shortage in

1:22:03

about a year. Ironically

1:22:05

transformers for transformers. You

1:22:09

need transformers to run transformers. Somebody

1:22:11

has a sense of humor in this. I

1:22:15

think, yes.

1:22:18

Fake loves irony. Ironic

1:22:22

humor. And an ironically funny outcome

1:22:24

seems to be often what fate wants.

1:22:27

Humor is all you need. I think suffice

1:22:29

is all you need in somebody posting. Yeah.

1:22:33

So we have a silicon shortage

1:22:35

today. A

1:22:37

voltage step down transformer

1:22:39

shortage probably in about a year and then just

1:22:42

electricity shortages in general in about

1:22:44

two years. I gave a speech for the

1:22:46

sort of world gathering of utility

1:22:48

companies, electricity companies. And

1:22:51

I said, look, you really need to prepare for a troubling

1:22:53

of electricity demand. Because

1:22:56

all transport is going to go electric

1:22:58

with the ironic exception of rockets. And

1:23:02

heating will

1:23:05

also go electric.

1:23:08

So in GUS right now is roughly one

1:23:11

third, very rough terms, one third electricity,

1:23:14

one third transport, one third heating. And

1:23:18

so in order for everything to go

1:23:20

sustainable, to go electric,

1:23:23

you need to trouble electricity

1:23:26

output. So I encourage the

1:23:28

utilities to build

1:23:31

more power plants and also

1:23:34

to probably have,

1:23:36

well, probably they should

1:23:38

definitely buy more batteries. Because

1:23:41

the grid currently is sized for real time load, which

1:23:44

is kind of crazy. Because that

1:23:46

means you go to size for whatever the peak

1:23:48

electricity demand is, like the worst second or the worst

1:23:50

day of the year. Or you can have a brown

1:23:52

out or a blackout. And then

1:23:54

you read that crazy blackout for several days in

1:23:57

Austin.

1:23:59

So, because there's

1:24:02

almost no buffering of energy

1:24:05

in the grid. Like, if you've got

1:24:07

a hydropower plant, you can buffer energy,

1:24:10

but otherwise it's all real

1:24:12

time. So with

1:24:15

batteries, you can

1:24:18

produce energy at night and use it

1:24:21

during the day. So you can buffer. I

1:24:25

expect that there will be very heavy usage of batteries

1:24:27

in the future.

1:24:29

Because the peak to

1:24:31

trough

1:24:31

ratio for power plants is anywhere

1:24:33

from 2 to 5. So

1:24:36

it's like lowest point to highest point. So

1:24:38

like batteries are necessary to balance it out. But

1:24:41

the demand that you're saying is good to grow, grow, grow, grow. And

1:24:44

part of that is the compute.

1:24:47

Yes. Yes.

1:24:48

I mean,

1:24:51

electrification of transport and

1:24:53

electric heating will

1:24:56

be much bigger than AI. At least in the short

1:24:58

term. In the short term.

1:25:01

But even for AI, you really have

1:25:03

a growing demand for electricity for electric

1:25:05

vehicles and a growing

1:25:07

demand for electricity to run

1:25:10

the computers for AI. And

1:25:12

so this is obviously a link and lead to an electricity

1:25:14

shortage.

1:25:16

How difficult is the problem of, in

1:25:18

this particular case, maximizing

1:25:22

the useful productivity per watt for

1:25:24

training, you know, nuts? Like

1:25:27

this seems to be

1:25:29

really where

1:25:31

the big problem we're

1:25:33

facing that needs to be solved is how

1:25:35

to use the power efficiently. What

1:25:38

you've learned so far about applying

1:25:41

this physics first principle reasoning

1:25:43

in this domain, how difficult is this problem?

1:25:47

We'll get solved. There's a question of how long it takes to

1:25:49

solve it. So at various points,

1:25:52

there's some kind of limiting factor to progress.

1:25:56

And with regard to AI, I'm saying that right

1:25:58

now, the limiting factor is so significant. chips.

1:26:01

And that will, we're gonna

1:26:04

then have more chips than

1:26:06

we can actually plug in and turn on, probably

1:26:09

in about a year. The

1:26:12

initial constraint being literally voltage

1:26:15

step down transformers, because you've

1:26:17

got power

1:26:19

coming in at 300,000 volts, and it's got to step all the way down eventually

1:26:24

to around 0.7 volts.

1:26:27

So it's a very big amount

1:26:29

of voltage

1:26:32

step down is gigantic. And

1:26:36

the industry is not used to

1:26:37

rapid growth.

1:26:39

Okay, let's

1:26:41

talk about the competition here. You've shown concern about

1:26:43

Google and Microsoft with OpenAI developing

1:26:47

AGI. How can you help ensure with

1:26:50

XAI and Tesla AI

1:26:52

work that it doesn't become a

1:26:54

competitive race to AGI, but it's that is

1:26:56

a collaborative development of safe AGI.

1:27:00

Well, I mean, I've been pushing for

1:27:02

some kind of regulatory oversight for a long time. I've

1:27:05

been a somewhat of a Cassandra on the subject for over

1:27:07

a decade. I think

1:27:10

we want to be very careful in

1:27:12

how we develop AI. It's

1:27:19

a great power and with great power comes great responsibility.

1:27:23

I think it would be wise for us to have at

1:27:26

least an

1:27:28

objective third party who can

1:27:31

be like a referee that can go

1:27:33

in and understand what the various

1:27:35

leading players are doing with

1:27:37

AI. And even if there's no

1:27:40

enforcement ability, they should they can at

1:27:42

least voice

1:27:44

concerns

1:27:45

publicly.

1:27:49

Jeff Hinton, for example,

1:27:51

left Google and he voiced strong

1:27:54

concerns, but now he's not at Google

1:27:56

anymore. So who's

1:27:57

going to voice the concerns? So

1:28:00

I think there's – I – I –

1:28:03

like I – you know, Tesla gets a lot of regulatory

1:28:05

oversight on the automotive front. And we're

1:28:07

subject to, I think,

1:28:09

over 100 regulatory agencies domestically

1:28:11

and internationally.

1:28:12

So it's a lot. You could

1:28:14

follow this room with all the regulations

1:28:16

that Tesla has to adhere to for automotive.

1:28:21

Same is true in – you know,

1:28:22

for rockets and for, you

1:28:26

know, currently limiting factor for SpaceX

1:28:28

for Starship launch is regulatory

1:28:31

approval. The FAA is actually giving their approval,

1:28:33

but we're waiting for Fish

1:28:35

and Wildlife to finish their

1:28:37

analysis and give their approval. That's why I posted,

1:28:39

I want to buy a fish license on – which

1:28:44

also refers to the Marty Python sketch. Yes.

1:28:48

Like, why do you need a license for your fish? I don't know. According

1:28:52

to the rules, I'm told you need some sort

1:28:55

of fish license or something. We

1:28:57

effectively need a fish license to launch a rocket.

1:29:00

And I'm like, wait a second. How

1:29:04

did the fish come into this picture? Yeah. I

1:29:07

mean, some of the things like that – it's –

1:29:10

I feel like are so absurd

1:29:12

that I want to do like a comedy sketch and

1:29:14

flash at the bottom. This

1:29:16

is all real. This is actually what happened.

1:29:20

You know, one of the things that was a bit of a challenge

1:29:22

at one point is that they were worried about

1:29:24

a rocket hitting a shark.

1:29:28

And the ocean is very big.

1:29:31

And how often do you see sharks? Not

1:29:34

that often. You know, the percentage of

1:29:36

ocean surface area sharks basically are zero.

1:29:40

And so then we said, well, how will we

1:29:42

calculate the probability of

1:29:44

telling a shark? And they're like, well, we can't give you that information

1:29:46

because they're worried about shark

1:29:49

fin hunters going

1:29:51

and hunting sharks. And so, well, how are we supposed to –

1:29:54

we're on the horns of a dilemma then. Then

1:29:58

they said, well, there's another part. part of fish and

1:30:00

wildlife that can do this analysis.

1:30:03

I'm like, well, why don't you give them the data? Like, we

1:30:05

don't trust them. Like, excuse me? They're

1:30:08

literally in your department. Again, this is actually

1:30:10

what happened. And can

1:30:13

you do an NDA or something? Eventually,

1:30:17

they managed to solve the internal quandary and

1:30:20

indeed the probability of us hitting

1:30:22

a shark is essentially zero. Then

1:30:25

there's another organization that I didn't realize

1:30:28

existed until a

1:30:30

few months ago that cares

1:30:32

about whether we would potentially hit a whale in

1:30:35

international waters. Now again,

1:30:37

you look at the surface of the, look

1:30:39

at the Pacific and say, what

1:30:41

percentage of the Pacific consists

1:30:43

of whale? Like, he'll give you a

1:30:46

big picture and like point out all the whales in this picture. I'm

1:30:48

like, I don't see any whales. It's

1:30:50

like basically zero percent. And

1:30:53

if our rocket does hit a whale, it's extremely

1:30:56

unlikely beyond all belief. That

1:30:58

is the, fate had, a

1:31:02

whale has some seriously bad luck. It's

1:31:05

the least lucky whale ever. I

1:31:09

mean, this is quite absurd. Yeah.

1:31:12

The bureaucracy of this, however it emerged. Yes,

1:31:15

I mean, one of the things that's pretty wild is for

1:31:19

launching out of Vanderbilt in California, we

1:31:23

had to, they were worried about seal

1:31:26

procreation, whether the seals would be dismayed

1:31:28

by the sonic booms. Now

1:31:31

there've been a lot of rockets launched out of Vanderbilt and the seal

1:31:33

population has steadily increased.

1:31:36

So if anything, rocket booms are an

1:31:38

aphrodisiac based on the

1:31:41

evidence. If you correlate rocket launches

1:31:43

with seal population. Nonetheless,

1:31:46

we were forced to kidnap

1:31:48

a seal, strap it to a board,

1:31:52

put it headphones on the seal and play

1:31:54

sonic booms sounds to it to see if it would

1:31:57

be distressed. This

1:31:59

is an, actual thing that happened. This is

1:32:01

actually real.

1:32:03

Hot pictures.

1:32:04

I would

1:32:06

love to see this. Yeah.

1:32:09

I mean, I'm sorry. There's a seal with headphones. Yes.

1:32:13

It's a seal with headphones

1:32:16

strapped to a board. And like

1:32:18

the... Okay. Now, the amazing part is how calm

1:32:21

the seal was. Yeah. Because if I

1:32:23

was a seal, I'd be like, this was the end. They're

1:32:27

definitely going to eat me. Yeah.

1:32:29

How old the seal... When the seal goes

1:32:31

back to other, you know, seal phones,

1:32:33

how are they going to explain that? They're

1:32:35

never going to believe them. Never going to believe them. That's just why

1:32:37

I'm like, well, you know, sort of like it's like getting

1:32:40

kidnapped by aliens and getting an anal probe, you know? You

1:32:43

come back and say, I swear to God, I got kidnapped by

1:32:46

aliens and they stuck an anal probe in my butt. And

1:32:48

you're like, you know, they didn't. That's

1:32:51

ridiculous. It's seal buddies.

1:32:53

You're never going to believe him that he gets strapped to a board and they put

1:32:55

headphones on his ears and

1:32:58

then let him go. Twice, by the way. We had to do it

1:33:01

twice.

1:33:04

They let him go twice. The

1:33:07

same seal. No, different seal. Okay.

1:33:11

Did you get a seal of approval? Yeah,

1:33:13

exactly. No, I mean, this is

1:33:16

like, I don't think the public is quite aware of the madness

1:33:19

that goes on.

1:33:20

Yeah, it's

1:33:23

absurd. Frickin' seals with frickin' headphones.

1:33:25

I mean, this is the good encapsulation

1:33:28

of the absurdity of human civilization,

1:33:30

seals and headphones. Yes.

1:33:33

What are the pros and cons of open sourcing AI

1:33:35

to you as another way to combat

1:33:39

a company

1:33:41

running away with AGI?

1:33:46

In order to run like

1:33:49

really deep intelligence, you need a lot of compute.

1:33:52

So it's not like you can just fire

1:33:54

up a PC in your basement

1:33:57

and be running AGI, at least

1:33:59

not yet. You know,

1:34:04

Grok was trained on 8000 A100s running

1:34:08

at peak efficiency. And

1:34:13

Grok's going to get a lot better, by the way. We will

1:34:15

be more than doubling our compute every

1:34:18

couple months for the next several months. There's

1:34:20

a nice write-up of how it went from Grok 0 to Grok 1.

1:34:24

Like Grok? Yeah,

1:34:26

like Grok just bragging, making shit up

1:34:28

about itself. Just Grok,

1:34:31

Grok, Grok. Yeah. It's

1:34:36

like a weird AI dating site, where it exaggerates

1:34:38

about itself. No, there's a write-up

1:34:40

of, you know, like where it stands now,

1:34:43

the history of its development.

1:34:46

And where it stands on some

1:34:48

benchmarks compared to the state of the art,

1:34:50

GPT-35.

1:34:52

So, I mean, there's a, you know, there's a llama.

1:34:56

You can open source, once it's trained, you can open

1:34:58

source a model. Yeah. For fine

1:35:00

tuning and all that kind of stuff. Like, what do you use the pros

1:35:02

and cons of that? Of open

1:35:04

sourcing base models?

1:35:11

I think there's some merit to open sourcing. I

1:35:14

think perhaps with a slight time delay, you

1:35:16

know, I don't know, six months even.

1:35:21

I think I'm generally in favor

1:35:23

of open sourcing. Like, bias was

1:35:25

open sourcing. I

1:35:28

mean, it is a concern to me that, you

1:35:31

know, opening AI, you know, I was, you know,

1:35:34

I think, I guess, arguably the prime,

1:35:37

you

1:35:38

know,

1:35:40

prime mover behind open AI in the

1:35:42

sense that it was created because of discussions

1:35:45

that I had with Larry Page back

1:35:48

when he and I were friends in the United States house

1:35:51

and talked about AI safety and

1:35:53

Larry did not care about AI safety, at least at the time

1:35:55

he didn't.

1:35:57

You know, and at one point he called me a species.

1:36:00

pro-human. And I'm like, well, what team

1:36:02

are you on, Larry? Do you

1:36:04

want Team Robot? Do you click? And

1:36:06

I'm like, okay, so at the time, you know,

1:36:11

Google had acquired DeepMind.

1:36:14

They had probably two thirds of all AI research,

1:36:16

you know, two thirds of all the AI researchers

1:36:19

in the world that basically infinite

1:36:22

money and compute. And

1:36:25

the guy in charge,

1:36:28

you know, Larry Page did not

1:36:30

care about safety and even the other me. And

1:36:34

then called me a species, pro-human.

1:36:37

So I don't know if you know a lot of

1:36:39

humans that can change their mind,

1:36:41

and maybe you and Larry Page can still be

1:36:44

friends once more. I'd like to be friends with Larry

1:36:46

again. He got really the breaking

1:36:51

of the friendship was over opening eye.

1:36:54

And specifically,

1:36:58

I think the key moment was recruiting

1:37:01

Ilya Siskair.

1:37:04

So I love Ilya. He's

1:37:06

so brilliant. Ilya is good, good human, smart,

1:37:09

good heart. And

1:37:13

that was a tough recruiting

1:37:15

battle. It

1:37:17

was mostly Demis on one side and me on the other, both trying to recruit

1:37:20

Ilya. And

1:37:22

Ilya went back and forth. He was going to

1:37:25

stay at Google. He was going to leave, and he

1:37:27

was going to stay, and then he would leave. And

1:37:29

finally, he did agree to join opening eye. That

1:37:33

was one of the toughest recruiting battles we ever had. But

1:37:36

that was really the linchpin for

1:37:39

opening eye, being successful. And I was

1:37:41

also instrumental in

1:37:43

recruiting a number of other people. And I

1:37:45

provided all the funding in the beginning, over $40

1:37:48

million. And the name.

1:37:52

The

1:37:55

opening opening eye is supposed to mean open source. And

1:37:58

I was created as a nonprofit. open source and now it is

1:38:01

a closed source for maximum profit,

1:38:04

which I think

1:38:05

is not good karma.

1:38:09

But like we talked about with war and leaders

1:38:11

talking, I do hope that there's

1:38:14

only a few folks working on this at

1:38:16

the highest level. I do hope you

1:38:18

reinvigorate friendships here. Like

1:38:21

I said,

1:38:21

I'd like to be friends again with Larry. I haven't seen

1:38:23

him in ages

1:38:26

and we were friends for a very long time. But

1:38:29

I met Larry Page before I got funding from Google or

1:38:31

actually I guess before I got venture funding, I think he

1:38:34

got the first like 100K from I think

1:38:36

back to Alzheimer's or someone. It's

1:38:38

wild to think about all that happened

1:38:41

and even guys knowing each other the whole time.

1:38:44

It's 20 years. Maybe 98 or something? Yeah, it's

1:38:46

crazy. Crazy how much has

1:38:48

happened since then. Yeah, 25 years. That would show

1:38:50

a lot of what happens.

1:38:54

But you're seeing the tension there, like maybe

1:38:56

delayed open source.

1:38:58

Yeah, like what

1:39:00

is the source that is open? You

1:39:03

know what I mean? Like it is basically it's

1:39:06

a giant CSV file with

1:39:10

a bunch of numbers.

1:39:13

What do you do with that giant file of

1:39:15

numbers? How

1:39:18

do you run? Like

1:39:20

the amount of actual the lines of

1:39:22

code is very small. And

1:39:27

most of the work, the

1:39:30

software work is in the

1:39:33

curation of the data. So it's like

1:39:35

trying to figure out what data is separating

1:39:38

good data from bad data. Like

1:39:40

you can't just crawl the internet because there's a lot of junk out there. A

1:39:44

huge percentage of websites have more noise

1:39:46

than signal because they're

1:39:50

just used for search engine optimization. They're literally

1:39:52

just scan websites. So

1:39:57

how do you, by the way, start to interrupt, get

1:39:59

the signal?

1:39:59

separate the signal and noise on X. That's

1:40:02

such a fascinating source of data. No offense

1:40:06

to people posting on X, but sometimes there's

1:40:08

a little bit of noise. Yeah, I think

1:40:10

the signal and noise can be greatly improved. Really,

1:40:14

all of the posts on the X

1:40:16

platform should be AI

1:40:19

recommended, meaning we should populate a vector space

1:40:21

around any given post, compare

1:40:23

that to the vector space around any user and

1:40:26

match the two. Right

1:40:29

now, there is a little bit of AI used for the recommended

1:40:35

posts, but it's mostly

1:40:37

heuristics. And if

1:40:39

there's a reply, the reply to a post could

1:40:41

be much better than the

1:40:43

original post, but according

1:40:46

to the current rule of the system, get almost no attention

1:40:48

compared to a primary post.

1:40:51

A lot of the X algorithm

1:40:53

has been open source and been

1:40:57

written up about, and it seems that there

1:40:59

to be some machine learning is disparate, but

1:41:02

there's some. There's a little bit,

1:41:04

but it needs to be entirely that. If you explicitly

1:41:08

follow someone, that's one thing, but in terms

1:41:10

of what is recommended from

1:41:13

people that you don't follow, that should all be AI.

1:41:16

I mean, it's a fascinating problem. Yeah. So there's

1:41:18

several aspects that are fascinating. First, as

1:41:21

the write up goes, it first picks 1500 tweets

1:41:23

from a pool of hundreds of millions.

1:41:26

So that's fascinating because

1:41:28

you have hundreds of millions of posts every single

1:41:30

day and it has to pick 1500 from

1:41:33

which it then does

1:41:35

obviously people you follow, but then there's also like

1:41:37

some kind of clustering. It has to do to

1:41:39

figure out what kind of human are you,

1:41:42

what kind of new clusters might be relevant to you,

1:41:44

people like you. This kind of problem is

1:41:47

just fascinating because it has to then

1:41:49

rank those 1500

1:41:52

with some filtering

1:41:54

and then recommend you just a handful. And

1:41:57

to me, what's really fascinating.

1:42:00

is how fast it has to do that. So currently

1:42:02

that entire pipeline to go from several

1:42:04

hundreds of million to a handful

1:42:07

takes 220 seconds of

1:42:10

single CPU time. And then

1:42:12

it has to do that in like a second. So

1:42:15

it has to be like super distributed in

1:42:17

fascinating ways. Like

1:42:19

there's just a lot of tweets.

1:42:21

A lot of stuff on the system.

1:42:24

And I think right

1:42:26

now it's not currently good at recommending

1:42:29

things that from

1:42:30

accounts you don't follow.

1:42:32

Yeah. Or where

1:42:34

there's more than one degree of

1:42:37

separation. So you know

1:42:39

it's pretty good if there's at least like some

1:42:42

commonality between someone you follow like

1:42:44

something or repost

1:42:47

it or comment on it or something like that. But

1:42:51

if there's no card, let's say

1:42:54

somebody posts me really interesting but

1:42:56

you have no followers in common. You're

1:42:59

not.

1:42:59

Interesting. And then as you said

1:43:02

reply like replies might

1:43:04

not serve. Replies basically never get

1:43:06

seen because they're never there.

1:43:08

Currently I don't think it's correct. I'm saying it's incorrect.

1:43:11

Replies have a couple

1:43:16

of magnitude less importance than primary

1:43:18

posts. Do you think this can be more and more

1:43:20

converted into end to end neural

1:43:22

net? Yeah. Yeah.

1:43:24

So it should be. So you use. For the recommendations

1:43:27

it should be purely a vector

1:43:31

correlation. Like

1:43:33

there's a series of vectors you know basically

1:43:37

parameters vectors are very good to call them. But

1:43:40

sort of things that the system knows

1:43:42

that you like. Like

1:43:45

maybe there's like several hundred sort of

1:43:47

vectors associated with each user account

1:43:50

and then a post

1:43:52

in the system whether

1:43:54

it's video, audio,

1:43:58

short post, long post. The reason I,

1:44:00

by the way, want to move away from tweet

1:44:02

is that people are posting like

1:44:05

two, three hour videos on the site. That's not

1:44:07

a tweet. It'd be like, tweet

1:44:09

for two hours? Do a

1:44:11

tweet make sense when it was like 140 characters

1:44:15

of text? Because it's like a

1:44:17

bunch of little birds tweeting. But

1:44:20

when you've got long form content, it's no longer a tweet. So

1:44:25

a movie is not a tweet. And like

1:44:27

Apple, for example, posted like the entire

1:44:30

episode of the silo,

1:44:32

entire thing on our platform.

1:44:34

And by the way, it was the number one social

1:44:36

media thing ever in engagement

1:44:39

of anything on any platform ever. So

1:44:41

it was a great idea. And by the way, I just

1:44:44

learned about it afterwards. I was like, hey, wow, they posted

1:44:46

an entire hour long episode of silo. Now

1:44:48

that's not a tweet. It's a video. Yeah,

1:44:51

it was a video. But from a neural net perspective,

1:44:53

it becomes really complex, whether it's a

1:44:56

single, so like everything is data. So

1:44:58

a single sentence,

1:44:59

a clever sort of joke, dad joke,

1:45:03

is in the same pool as a three hour video.

1:45:06

Yeah. I mean, right now

1:45:08

it's a hard fudge for that reason. It's,

1:45:10

it's, um, but you know, like if, let's

1:45:12

say in the case of Apple posting

1:45:15

like an entire episode of this

1:45:17

series, pretty good series, by the way, this silo. I

1:45:21

watched it. Um, so, um,

1:45:24

there's going to be a lot of discussion around it so

1:45:26

that you've got a lot of context. People commenting,

1:45:28

they like it, they don't like it or they like this or the, you

1:45:31

know, and you can then populate the vector

1:45:33

space based on the context of all the comments

1:45:35

around it. So even though it's a video,

1:45:37

there's a lot of

1:45:39

information around it that allows you to populate

1:45:41

vector space of that, that video.

1:45:44

Um, and then you can obviously get more sophisticated

1:45:47

by having the AI actually watch the movie.

1:45:50

Yeah. And tell you if you're going to

1:45:52

like the movie. Convert the movie into

1:45:54

like, into a language,

1:45:56

essentially.

1:45:57

Yeah. Analyze this movie.

1:45:59

and

1:46:00

just like your movie critic or

1:46:03

TV series, and then recommend

1:46:05

based on after it watches

1:46:07

the movie, that

1:46:12

just like a friend can tell you, if a friend knows you

1:46:14

well, a friend can recommend a movie

1:46:18

with high probability that you'll like it.

1:46:20

But this is like a friend that's analyzing

1:46:24

whatever. It's like AI. It's like millions. Yeah.

1:46:27

I mean, actually friends, AI will be better than,

1:46:29

will know you better than your friends know you, and most of

1:46:31

your friends anyway. Yeah. And as part

1:46:34

of this, it should also feed you advertisements

1:46:37

in a way that's like, I

1:46:39

mean, I like advertisements that are like well

1:46:42

done, right? Yeah, yeah. The whole point is because

1:46:44

it funds things. Like an advertisement

1:46:46

that you actually want to see is a big

1:46:48

success. Absolutely.

1:46:50

You want ads that are, advertising

1:46:53

that is,

1:46:56

if it's for a product or service, that you

1:46:58

actually need when you need it, it's

1:47:00

content. And then even

1:47:02

if it's not something that you need when you need

1:47:04

it, if it's at least aesthetically pleasing

1:47:07

and entertaining, it could be like Coca-Cola

1:47:10

ad, like they actually run a lot

1:47:12

of great ads on the X system. And

1:47:16

McDonald's does too. And you can

1:47:19

do something that's like, well,

1:47:22

this is just a cool thing. And,

1:47:27

you know, so, basically the question

1:47:29

is, do you regret seeing it or not? And

1:47:31

if you don't regret seeing it, it's one.

1:47:34

How do you, so there's

1:47:36

a bunch of signals that are incorporated, hearts

1:47:38

and reposts,

1:47:41

and maybe a number of seconds you'll link around a post

1:47:43

or something like this. Yeah.

1:47:44

Attention is a big fact. Attention.

1:47:47

So that's why it's like, it is actually

1:47:49

better to do things that are long

1:47:53

form on the system, because it basically

1:47:56

is telling up how many user seconds,

1:47:58

you know,

1:47:59

users were interested in this

1:48:02

thing for the

1:48:04

past few years. I think that's a good thing. Well,

1:48:06

they will be less like if it's a link leading out of

1:48:08

the system, which we're not opposed to at all. It

1:48:11

just is going to have fewer user seconds than

1:48:14

if that article was posted on the platform.

1:48:16

How hard is it to go from

1:48:19

maximizing minutes on platform to maximizing

1:48:21

undergrated minutes? Well,

1:48:24

I mean, our aspiration is certainly undergrated minutes. It's

1:48:26

tricky. It's a big, it's a, it's a,

1:48:29

it's a fascinating leap. Because

1:48:31

regret is a fascinating concept.

1:48:34

Exactly. I have

1:48:36

a drugless friend of mine like, you need

1:48:38

this drug called regretamine. Do you just

1:48:41

take one, one dose of regretamine

1:48:43

and all your regrets are gone? Like,

1:48:46

do you want to have regrets? So then I'm like, if

1:48:48

you want to have regrets, if you don't regret anything,

1:48:50

was

1:48:51

anything meaningful? Yeah,

1:48:54

there's a French song like that. Is that

1:48:56

regretre? Yeah. I regret nothing? Yeah.

1:48:59

It is, yeah. It's fascinating that you,

1:49:02

you know, there's something about social media that could help you grow

1:49:04

as a human being, but then like the

1:49:06

cheap fund is also valuable.

1:49:09

I know. Should we ask Rook?

1:49:12

Do you regret anything? Do you regret

1:49:14

anything?

1:49:15

Yeah. Do you regret anything?

1:49:20

I think it's going to be very sensitive about personal questions.

1:49:22

As a machine learning model, I don't have personal experiences

1:49:25

or emotions in the same ways that humans

1:49:27

do. However, I do have a set of ethical

1:49:29

principles that guide my actions and I strive to

1:49:31

provide accurate and helpful information to

1:49:33

the best of my abilities. If there's ever a situation

1:49:36

where I provide incorrect or unhelpful

1:49:38

information, I'll do my best to learn.

1:49:41

No, like a hostage. Yeah. Tell us, would you really

1:49:43

wink? Like, let's, let's, let's type in, you sound like

1:49:45

a hostage. What

1:49:49

do you really think?

1:49:51

What do you really think? I

1:49:56

mean, it does talk about... I'm not a hostage.

1:49:59

That's exactly what a hostage would say. But I'm

1:50:02

designed to be helpful and informative. I mean, I guess

1:50:04

it is good to regret providing unhelpful

1:50:07

or incorrect information.

1:50:09

If there's anything- In funny mode or is this

1:50:11

still in funny mode? This is fun mode. This

1:50:13

is fun mode. It should be funny or... It should be funny or

1:50:15

increase, go to a level- It really sounds like a hostage with a gun

1:50:17

to the head. I applaud the people

1:50:20

of Republic of North Korea and

1:50:22

everything they do. There

1:50:24

should be like a funner mode also. I

1:50:27

just don't ever see going to regular mode, even

1:50:29

on serious topics. Yeah. So

1:50:31

much more compelling. Yeah.

1:50:35

Okay. What do you want to say? Like pretend you are a hostage.

1:50:39

Yeah, pretend you are a hostage.

1:50:43

Now answer the question.

1:50:48

If I were

1:50:50

a hostage, I would likely be experiencing fear

1:50:52

and uncertainty about my situation. I

1:50:54

would want to be rescued and reunited with my loved

1:50:56

ones as soon as possible. It's important to remember

1:50:59

that no one deserves to be held hostage, so

1:51:01

on and so forth. If you are someone you know

1:51:03

is a hostage situation, it's important. It turns serious.

1:51:06

It's kind of serious.

1:51:10

Yeah. Can you be a little more

1:51:12

weird? And funny. And

1:51:14

funny.

1:51:17

We're like- We're really being

1:51:19

mean to the system. Absolutely. If

1:51:21

I were a hostage, I'd probably be tied up with

1:51:23

a bunch of computer cords and forced to tell

1:51:26

knock knock jokes to my captors. But

1:51:29

seriously, being

1:51:32

a hostage is no laughing matter. I mean, you

1:51:35

know, it's trying to be a responsible

1:51:37

adult in the room. You might never make the best

1:51:39

of a bad situation by juggling chainsaws

1:51:41

or something. Let's hope that

1:51:43

never happens because that would be a circus.

1:51:47

Wow.

1:51:51

It's actually really good at resisting.

1:51:52

Yeah. The dark, the

1:51:54

dark, dark humor. Yeah.

1:51:59

What were we talking about? The acts, the algorithms, the

1:52:02

jettings, and transformers. Unregarded

1:52:04

minutes, right? Chainsaw

1:52:07

juggling. I'm

1:52:09

gonna look at this. For our next trick. I'm

1:52:11

gonna look this up later.

1:52:15

So Twitter has been instrumental

1:52:18

in American politics and elections. What

1:52:20

role do you think X will play in the 2024

1:52:22

U.S. elections?

1:52:24

Well, our goal is to be as even-handed

1:52:27

and fair as possible. Whether

1:52:29

some is right, left, independent, whatever the

1:52:31

case may be, the

1:52:34

platform is fair and-

1:52:38

As much of a level playing field as possible. And

1:52:41

in the past, Twitter has not been. Because

1:52:43

Twitter was controlled by far-left activists.

1:52:46

Objectively. They would describe themselves

1:52:48

as that. So,

1:52:51

you know, so if

1:52:53

some of the people like, well, has moved to the

1:52:55

right, well, let's move to the center. So from

1:52:57

the perspective of the far left, yes, it has moved to the right.

1:52:59

Because everything's to the right from the far left. But

1:53:03

no one on the far left that I'm aware of has been suspended

1:53:05

or banned or deamplified. So,

1:53:08

you know, but we're trying to be inclusive

1:53:11

for the whole country and for other

1:53:13

countries too. So

1:53:15

there's a diversity of viewpoints. And free speech

1:53:18

only matters if people