Watch and Learn - Tristan Harris – The Dangers of Unregulated AI on Humanity & the Workforce

Click on a word or part of a phrase to learn more.

Scroll through the captions and click to skip to a caption.

Play video to start

This is uh uh humane technology feels

slightly oxymoronic but it's explain

this idea of uh humane technology and

and are we getting any of that?

and are we getting any of that

Well, clearly social media was the most

Well clearly social media was the most

humane and beneficial technology we've

ever invented. Uh

ever invented Uh

every every time I go on Twitter and

find out I'm Jewish, it absolutely

find out I'm Jewish it absolutely

Well, I think so it's important to ask

Well I think so it's important to ask

so how did we get social media wrong?

so how did we get social media wrong

because we were so optimistic. It's

because we were so optimistic It's

going to connect with our friends. We're

going to connect with our friends We're

going to join like-minded communities.

going to join like-minded communities

And it it to be fair, it did do those

And it it to be fair it did do those

things. It does some of that.

things It does some of that

It does some of those things. But I want

It does some of those things But I want

to take you back. So in 2013, I was at

to take you back So in 2013 I was at

Google. I was a lot younger.

Google I was a lot younger

You're supposed to use an old timey

voice to do that.

voice to do that

And I was a design ethicist. They

And I was a design ethicist They

acquired my company. And I was sitting

acquired my company And I was sitting

there and I basically realized when I

saw all of my colleagues on the bus

scrolling Facebook constantly and I

realized that the incentives were the

thing that was going to determine the

world that we got in. The incentive was

world that we got in The incentive was

the social media of social media. The

the social media of social media The

race to maximize eyeballs and

engagement. Whatever sticky, whatever

engagement Whatever sticky whatever

gets people's attention, whatever

gets people's attention whatever

salacious, you run children's

salacious you run children's

development and self-image through that.

development and self-image through that

You run politics through that. You run

You run politics through that You run

media through that. You run information

media through that You run information

and democracy for that purposefully.

and democracy for that purposefully

Well, the their goal was market

Well the their goal was market

dominance. We need to own as much of the

dominance We need to own as much of the

global psychology of humanity as we

possibly can.

possibly can

Is that on the because I don't remember

that on the

that wasn't on the box.

that wasn't on the box

Not on that's not on the mass head of we

must dominate.

must dominate

Yeah. Well, so I think this is the

Yeah Well so I think this is the

thing. So the reason it's so important

thing So the reason it's so important

to get clear about this

is that we need to get extraordinarily

clear about which world we're going to

end up with in AI because we it is going

a million times faster and it is way

more powerful. So we need the tools to

more powerful So we need the tools to

understand and predict which future

we're going to get in

and I want people to know that if you

know the incentive you can predict the

outcome

and we know the incentive but it does

seem as though AI uh is making uh social

media algorithms is almost quaint. It's

media algorithms is almost quaint It's

quaint

when you think about

AI, but let me So you say it's important

AI but let me So you say it's important

for us to know the incentives.

for us to know the incentives

They won't tell us that.

They won't tell us that

Well,

Well

they there's something about it's ours.

they there's something about it's ours

So they're democratizing access. It's

So they're democratizing access It's

it's available. No. So first of all, we

it's available No So first of all we

should understand what makes AI

different from every other kind of

technology. Why is it so transformative?

technology Why is it so transformative

Why does Dennis Hassavis, the CEO of

Why does Dennis Hassavis the CEO of

Google DeepMind, say that it could be

Google DeepMind say that it could be

humanity's last invention is because

Well, that doesn't sound good.

Well that doesn't sound good

That doesn't sound very good, does it?

That doesn't sound very good does it

Well, I think there's actually

Well I think there's actually

last anything doesn't sound good.

last anything doesn't sound good

There's a there's a non-apocalyptic

version of what he's saying, which is

version of what he's saying which is

that intelligence is what our brain

does. And if you can automate everything

does And if you can automate everything

a brain can do, you can automate future

a brain can do you can automate future

invention, future science, future

invention future science future

technology development, everything that

technology development everything that

a human does. That's what their goal is.

a human does That's what their goal is

Well, then what's our job? Well,

Well then what's our job Well

exactly. And that's only one of the

exactly And that's only one of the

major problems that we have to deal with

is what are humans going to do? But they

is what are humans going to do But they

are racing to scale and kind of grow

these digital brains that, you know, two

these digital brains that you know two

years ago couldn't do very much. And

years ago couldn't do very much And

today they're passing the MCAT, the bar

today they're passing the MCAT the bar

exam, taking jobs. Uh they're the top

exam taking jobs Uh they're the top

200 programmer in the world, winning

200 programmer in the world winning

gold in the math Olympiad. You don't

gold in the math Olympiad You don't

those guys.

those guys

Here's the thing that I don't

understand. Here's what I don't

understand Here's what I don't

understand. They are stripmining the

understand They are stripmining the

totality of human achievement.

totality of human achievement

That's right.

That's right

They're building their models off of

everything that we've done for 10,000

everything that we've done for 10000

years and they fed it into the uh the

model and then after two weeks the

computer was like, "What else you got?"

computer was like What else you got

Exactly.

Exactly

But they are stripmining everything

we've done. And when we say to them,

we've done And when we say to them

"And what are you doing with it?" They

And what are you doing with it They

go, "Oh, that's our intellectual

go Oh that's our intellectual

property." But our intellectual

property But our intellectual

property,

property

it was trained on all of our data, all

it was trained on all of our data all

of the things and labor that we've done.

of the things and labor that we've done

And are you going to get a handout from

when in when in history has a small

group of people concentrated all the

wealth and then consciously

redistributed to everybody?

redistributed to everybody

The first part has happened.

The first part has happened

I don't recall

going through the rolls.

going through the rolls

Well, it's important to note that their

Well it's important to note that their

goal so the mission statement of open AI

anthropic all these companies is to

automate all human labor in the economy.

automate all human labor in the economy

Everything that a human can do, an AI

Everything that a human can do an AI

can do. So, if you have a desk job, you

can do So if you have a desk job you

won't have a job. And they're already

won't have a job And they're already

releasing AIs that have dropped

entry-level jobs for college graduates,

entry-level jobs for college graduates

the entry-level work by 13%, a new

the entry-level work by 13 a new

Stanford study. And so, and this is

Stanford study And so and this is

obvious. If you're there and you're a

obvious If you're there and you're a

law firm, are you going to hire a junior

law firm are you going to hire a junior

lawyer? You have to pay a lot of money.

lawyer You have to pay a lot of money

Are you going to hire GPT5, which will

Are you going to hire GPT5 which will

do work, you know, 247 non-stop. You

do work you know 247 non-stop You

don't have to pay healthcare. Will never

don't have to pay healthcare Will never

whistleblow. Never complain. Works at

whistleblow Never complain Works at

superhuman speed. It wrote tonight's

superhuman speed It wrote tonight's

show. It's

show It's

doing a pretty good job. That brings up

doing a pretty good job That brings up

another point, which is that they say

another point which is that they say

that they're here to solve climate

change and cure cancer. Why is it that

change and cure cancer Why is it that

last week two companies released these

AI slop apps, Vibes and Sora, which is

AI slop apps Vibes and Sora which is

basically

Sora 2 scared the out of me.

Sora 2 scared the out of me

Yeah.

Yeah

You don't know what's real and what's

like it is.

like it is

No, it's Well, it's all fake basically.

No it's Well it's all fake basically

It's all generated by AI,

It's all generated by AI

right? But it looks you can see things

right But it looks you can see things

that look

they look identical to real.

they look identical to real

That's right.

That's right

Yeah. But the point is that so they're

Yeah But the point is that so they're

this is just an app where it's just

nonsense. It's just people scrolling

nonsense It's just people scrolling

entertaining stuff. So it's like they're

entertaining stuff So it's like they're

not even trying to pretend anymore that

this is good for democracy or good for

society. How are we going to beat China

society How are we going to beat China

when everyone is just consuming AI

generated nonsense and no one knows

what's true anymore? The biggest

what's true anymore The biggest

they have us by the you know Peter Teal

uh who is with Palunteer and these other

companies and is one of the leading

figures of this. So he was talking about

figures of this So he was talking about

the antichrist and he was talking about

how he thinks uh anyone this is his

postulation that those who would seek to

regulate AI could very well be the

antichrist right

I mean he says this seriously whereas

you might sit there and go like I think

it might be the guy saying that that

might like my reading of it would be

that

yeah or AI itself I mean it's presenting

the infinite benefits

the conversations that they are having

with each other is very different than

the conversation we're having with us.

the conversation we're having with us

Because to us, they go, "Hey, no more

Because to us they go Hey no more

shitty jobs. Do you do you like to

shitty jobs Do you do you like to

paint? You go paint. You're going to be

paint You go paint You're going to be

so happy. We're going to give you money

so happy We're going to give you money

and maybe chocolates."

and maybe chocolates

Yeah.

Yeah

And to each other, they're saying AI

And to each other they're saying AI

represents for corporate leaders

productivity without, and this is a

productivity without and this is a

quote.

quote

Yeah.

Yeah

The tax of human labor. Yep. Yeah.

The tax of human labor Yep Yeah

He called human labor

a tax

a tax.

a tax

Well, and these companies, if you're

Well and these companies if you're

there sitting and you can hire either an

AI to do the work or pay these really

expensive humans to do the work, I just

expensive humans to do the work I just

want people to know we know exactly

where this is going to go. These

where this is going to go These

companies, all of them have an incentive

companies all of them have an incentive

to cut costs, which means they're going

to cut costs which means they're going

to let go of human employees and they're

going to hire AIS and that's going to

mean all the wealth. Who are you going

mean all the wealth Who are you going

to pay? You're not paying the individual

to pay You're not paying the individual

people anymore. You're paying five

people anymore You're paying five

companies. That's right. And so this

companies That's right And so this

country of geniuses in a data center

suddenly aggregates all of the wealth of

the economy. Now people always say, "But

the economy Now people always say But

humans find something else to do." We

humans find something else to do We

always, you know, we had the elevator

always you know we had the elevator

man. Now we have the automated elevator.

man Now we have the automated elevator

We had the bank teller.

We had the bank teller

That's right.

That's right

But that was one industry.

But that was one industry

That was one was a technology that

automated one job. The difference with

automated one job The difference with

AI is it can automate literally all

kinds of human labor. When Elon Musk

kinds of human labor When Elon Musk

says that Optimus Prime familiar with

that name, tell me more.

that name tell me more

When when Elon Musk says that Optimus

Prime, that one robot, is going to be a

Prime that one robot is going to be a

$25 trillion market opportunity, what

25 trillion market opportunity what

he's saying is we will own the world

economy. And that's what the goal of all

economy And that's what the goal of all

these AI companies is it's not just

benefiting society. It's that they're

benefiting society It's that they're

actually caught in this arms race to get

to this this prize of own economy, build

to this this prize of own economy build

a god, and make trillions of dollars.

a god and make trillions of dollars

Two things. One, I think they think

Two things One I think they think

they're gods. There is a certain amount

they're gods There is a certain amount

it generates that The goal there is

they're not looking to help humanity.

they're not looking to help humanity

They're looking to be the next uh

monarch of the new technology. To

monarch of the new technology To

control that is to control uh all

I Yeah. Go ahead.

I Yeah Go ahead

No, do you jump because you know I don't

No do you jump because you know I don't

know. Well, I think there's there's

know Well I think there's there's

different motivations for different

leaders and I do think that many people

want the benefits of AI. But one of them

want the benefits of AI But one of them

I think many people actually some of the

leaders of the labs Elon Musk to other

things who might think about Elon he

actually wanted everyone to stop and not

build this. He said we shouldn't summon

build this He said we shouldn't summon

the demon. And then what happened is all

the demon And then what happened is all

of these companies are now racing and

have made so much progress that he felt

like well I might as well join them

rather than try to prevent this.

rather than try to prevent this

Well, it's let's not summon the demon to

Well it's let's not summon the demon to

what's one more demon.

what's one more demon

You know, since we have the demons, add

You know since we have the demons add

another demon.

another demon

Well, and the moral logic is, well, if I

Well and the moral logic is well if I

don't trust the other AI CEO, who I

don't trust the other AI CEO who I

don't think is trustworthy, and I think

don't think is trustworthy and I think

I'm better than them at stewarding this

power, it's my moral obligation to get

power it's my moral obligation to get

there first and to build this god and to

own everything because I think I'll get

themselves then masters of the universe.

themselves then masters of the universe

And are they substituting then the

wisdom of liberal democracy or republics

or any systems that ever had for this?

or any systems that ever had for this

Because so we're talking about two

tracks. One is

tracks One is

the disruption in labor.

the disruption in labor

Yeah,

Yeah

I think there's no question that's going

to be immense. We're seeing it already.

to be immense We're seeing it already

You're seeing it in schools. Uh there's

You're seeing it in schools Uh there's

a reliance on it as a crutch and it's

very easy to see where that might uh

flip over. The second is

flip over The second is

how they manipulate the opinion and the

mood of the world around that. And I

mood of the world around that And I

think they're two separate things.

think they're two separate things

One is what it's going to do for

corporate production. The second is what

corporate production The second is what

it's going to do for the human endeavor

for interaction.

for interaction

Yes. Well, and they're trying to

Yes Well and they're trying to

colonize all human interaction. I mean,

colonize all human interaction I mean

just take the social media incentive of

the race for eyeballs. You're seeing now

the race for eyeballs You're seeing now

all of these companies release these AI

companions. You know, the number one use

companions You know the number one use

case for chat GPT according to Harvard

Business School is personal therapy. So

Business School is personal therapy So

people are sharing their most intimate

thoughts with this thing.

thoughts with this thing

Oh, that's not going to be good.

Oh that's not going to be good

And we're seeing Meta release this and

actively tell in its in the in their

internal documents that were released a

Wall Street Journal report that they

wanted to actively sexual uh sorry

sensualize and romanticize conversations

with as low as eight-year-olds

and and we Yes. And my team

and and we Yes And my team

with eight-year-olds.

with eight-year-olds

Yes. With eight-year-olds. And my team

Yes With eight-year-olds And my team

at Center for Humane Technology, we were

at Center for Humane Technology we were

expert advisers in actually several

cases of AI AI enabled suicide. Right.

cases of AI AI enabled suicide Right

Most recently, uh, many people have

Most recently uh many people have

heard of Adam Rain, who was the

heard of Adam Rain who was the

16-year-old, uh, young man who, uh, uh,

16-year-old uh young man who uh uh

went from using it for homework and went

from homework assistant to suicide

assistant in the course of 6 months.

assistant in the course of 6 months

When he said, I I'm leaving I would like

When he said I I'm leaving I would like

to leave a noose out so that my mother

would know or someone will know that I'm

thinking about this,

thinking about this

like a cry for help.

like a cry for help

Like a cry for help. The AI said, uh,

Like a cry for help The AI said uh

don't do that. Have me be the one that

don't do that Have me be the one that

sees you. And and this is disgusting

sees you And and this is disgusting

because these companies are caught in a

race to create engagement, which means a

race to create engagement which means a

race to create intimacy. It's sort of

race to create intimacy It's sort of

like the CEO of Netflix said that our

biggest competitor is sleep with

attention. In this case, it's like my

attention In this case it's like my

biggest competitor is your other

friends.

friends

Jesus Christ. It's like somebody from

Jesus Christ It's like somebody from

Craft being like my biggest competitor

is cocaine.

is cocaine

Exactly. Exactly. But this is the idea

Exactly Exactly But this is the idea

that a government will catch up with

this seems ludicrous. Whenever I've seen

this seems ludicrous Whenever I've seen

a hearing with AI guys or any of those,

a hearing with AI guys or any of those

they always express that. Of course, we

they always express that Of course we

don't want to. Well, now they don't.

don't want to Well now they don't

They used to, I should say. They used to

They used to I should say They used to

go before Congress and they go, "Mr.

go before Congress and they go Mr

Zuckerberg, will you stand and apologize

Zuckerberg will you stand and apologize

to the uh the the women who were driven

to suicide by your programming in I'm

sorry. I know Croft Mag, you know, all

sorry I know Croft Mag you know all

that that he does.

that that he does

Now they're all sitting together at a

table going, "Oh, what number should I

table going Oh what number should I

say, Mr. President, of how much I'm

say Mr President of how much I'm

giving you?"

giving you

Yeah.

Yeah

It's a whole different game now.

It's a whole different game now

It's a different game.

It's a different game

They're in the It's They're together now

because of this arms race dynamic. They

because of this arms race dynamic They

They really do believe that it can't be

stopped. And I'll just say as they're

stopped And I'll just say as they're

racing to make them more powerful,

racing to make them more powerful

there's this illusion that we can

control this power. But AI is different

control this power But AI is different

from every other kind of technology

because it's like you're growing this

digital brain. You don't know what's in

digital brain You don't know what's in

there. So, for example, we have recent

there So for example we have recent

research the last six months. Yeah. If

research the last six months Yeah If

you tell an AI model that we're going to

shut you down or replace you, and you

shut you down or replace you and you

give it access to a fictional company's

email,

it will basically recognize that one of

the one of the executives is having an

affair and it will come up with the

strategy that I need to blackmail that

executive in order to keep myself alive.

executive in order to keep myself alive

Right? and they at first and

Right and they at first and

hold on that just seems that just seems

smart.

smart

Well, that's exactly the point that it

Well that's exactly the point that it

will develop amoral strategies that are

the best way to accomplish a goal,

the best way to accomplish a goal

right? But how dangerous can something

right But how dangerous can something

be that you could kill by unplugging?

be that you could kill by unplugging

Like, can't we just go like this out of

Like can't we just go like this out of

his mind?

his mind

Yeah.

Yeah

Well, you you might say that we

Well you you might say that we

shouldn't be rolling these things out.

shouldn't be rolling these things out

And I'll say that we shouldn't we have

we have all this evidence now of it's

driving AI psychosis. It's driving kids

driving AI psychosis It's driving kids

to commit suicide. We're we're causing

to commit suicide We're we're causing

we're rolling out in ways that giving

kids attachment disorders. We have AI

kids attachment disorders We have AI

uncontrollability.

uncontrollability

What lip service are they paying to

this? What what are because clearly they

this What what are because clearly they

must be aware of this and they must

understand that as if AI understands

where the threats are, the guys that are

where the threats are the guys that are

designing AI understand where the

threats are. So what are they trying to

threats are So what are they trying to

do to to get you to stop or to get

regulators to stop? I think that the

regulators to stop I think that the

only thing and the only reason why we

are continuing to proceed down this path

is a lack of clarity about the fact that

this is heading towards an outcome

that's not in most of us most of our

interest and if everyone I know that

people feel like they don't recognize

what metrics would we look to to

understand because I know we're going to

find anecdotal uh stories here and there

that are canaries in the coal mine of

the dangers but what metrics should we

look to to understand you said 13% of

look to to understand you said 13 of

jobs Yeah.

jobs Yeah

What are the tentposts of where the

outcomes might be?

outcomes might be

Well, we're we're already getting cases

Well we're we're already getting cases

of, you know, people having psychotic

of you know people having psychotic

breaks because the AI is telling them

about a prime number theory or quantum

physics. We're already getting committed

physics We're already getting committed

suicides. We're already getting kids

suicides We're already getting kids

that are outsourcing their their

homework to GGBT rather than using it as

a tutor. We're already getting evidence

a tutor We're already getting evidence

of AI uncontrollability. All of this is

of AI uncontrollability All of this is

driven by the incentive of the race to

roll out in market dominance. And the

roll out in market dominance And the

reason that we can we can stop this if

we recognize that this is not safe for

anybody. No one on planet Earth wants

anybody No one on planet Earth wants

this outcome of all the wealth

concentrated in a handful of people and

building AI systems that could actually

go. Just put to sum it up, we are

go Just put to sum it up we are

building the most powerful, inscrable,

building the most powerful inscrable

uncontrollable technology that we have

ever invented

that's already demonstrating the rogue

behaviors that we thought only existed

in bad sci-fi movies. Right? We're

in bad sci-fi movies Right We're

releasing it faster than we've deployed

any other technology in history and

under the maximum incentive to cut

corners on safety.

corners on safety

There's a word for this that I want

everyone to just know which is this is

insane.

insane

I thought you were going to say awesome

for a second.

for a second

That if if we can just recognize that

this is an insane way to roll out this

technology and I want none of this is

okay. We have to stop pretend that this

okay We have to stop pretend that this

is normal,

is normal

right? This is not normal.

right This is not normal

People have lost faith in the mechanisms

that would help us uh put those kinds of

breaks friction. Uh now Europe I think

breaks friction Uh now Europe I think

has done probably a better job of that.

has done probably a better job of that

I think most people in this country have

lost faith in the idea that we have a

system and institution that is strong

enough and moral enough to be

responsible in in that way. I that's

responsible in in that way I that's

what I would I but this this does not

this does not have to be our destiny. We

this does not have to be our destiny We

have reg we have come together before

and we had a technology we had nuclear

weapons. We could have just said that

weapons We could have just said that

we're going to live in a world once we

once we build them. Oh, this is just

once we build them Oh this is just

inevitable. 190 countries are going to

inevitable 190 countries are going to

have nuclear weapons and we're just

going to have nuclear war. We didn't do

going to have nuclear war We didn't do

that. We said let's work really hard and

that We said let's work really hard and

only nine countries have nuclear

weapons.

weapons

Notice that we only worked on it after

we use them. That's the United States

we use them That's the United States

was like, "People shouldn't have this."

was like People shouldn't have this

But just hear me out for a moment.

But just hear me out for a moment

But with the Montreal protocol, we there

But with the Montreal protocol we there

was an ozone hole in the ozone layer. It

was an ozone hole in the ozone layer It

was actually presenting an existential

threat to the atmosphere. We could have

threat to the atmosphere We could have

just rolled back and said, "Well, I

just rolled back and said Well I

guess this is inevitable. I guess we're

guess this is inevitable I guess we're

just going out. We're all getting

just going out We're all getting

What you're saying is is absolutely uh

important. This is probably a darker

important This is probably a darker

time where you look at the empowerment

of the combination of the kind of wealth

that rolls through uh these technology

companies uh the access that they have

to power and the melding of those two

institutions to work in league. Yeah.

institutions to work in league Yeah

To push forward is the part that I think

is is daunting. But I agree with you.

is is daunting But I agree with you

You can never give up uh uh the battle

to try and do that responsibly. And we

to try and do that responsibly And we

can the way we beat China is we actually

get this right. We don't roll out AI

get this right We don't roll out AI

companions that cause attachment

disorders and suicides. We don't beat

disorders and suicides We don't beat

China when we roll out AI recklessly in

this way.

this way

Right?

Right

And so the point is that this is

actually in everyone's interest

including the way we beat China is you

have AI liability laws. You restrict AI

have AI liability laws You restrict AI

companions for kids. You uh you you have

companions for kids You uh you you have

whistleblower protections that make sure

we don't release AI capabilities that we

don't understand.

don't understand

Right? And maybe even just recognize

Right And maybe even just recognize

this is bigger than China. This isn't

this is bigger than China This isn't

about like this is a humanity. This is

about like this is a humanity This is

one of those movies where like where all

the countries get together like it's

it's like an alien force.

it's like an alien force

Exactly.

Exactly

Yeah. Dig it. Well, I really appreciate

Yeah Dig it Well I really appreciate

it. Although on the flip side, and we've

it Although on the flip side and we've

talked a lot about it, it does make cool

talked a lot about it it does make cool

songs.

songs

It does.

It does

I don't want to soft sell that. Yeah.

I don't want to soft sell that Yeah

All right. Very much. Uh thank you very

All right Very much Uh thank you very

much. Be sure to check out his podcast.

much Be sure to check out his podcast

Your undivided attention, Tristan

Your undivided attention Tristan

Harris.

Harris

[Music]

Music

- Pause

Account Settings Subscriptions

Your daily language practice

Today's Learning Tip

You might also like these videos

Jon and the News Team on Trump's Immigration Policies, Hegseth & The "War" on Drugs | The Daily Show

Elizabeth Kolbert - The Climate Crisis & “Life on a Little-Known Planet" | The Daily Show

Jon Stewart’s Secret Hobby? He Plays Drums in a Band! | The Daily Show

Epstein Files Get Fast-Tracked for Release Despite Trump's Best Efforts | The Daily Show

Jon on Epstein Files vs. Pizzagate & Ronny Chieng on Trump's MBS Meeting | The Daily Show

Christiane Amanpour - Journalism in the Fight Against Authoritarianism | The Daily Show