Digital Disruption

00;00;00;14 - 00;00;01;08

Hey everyone!

00;00;01;08 - 00;00;02;26

I'm super excited to be sitting down

00;00;02;26 - 00;00;06;22

with Gregory Warner,

Peabody Award winning journalist, ex NPR

00;00;06;22 - 00;00;10;25

correspondent, and current host of the hit

AI podcast The Last Invention.

00;00;10;28 - 00;00;15;10

Greg is a fellow traveler in the quest to

understand the race to build advanced AI.

00;00;15;13 - 00;00;18;18

His full time

job is examining the existential questions

00;00;18;21 - 00;00;21;21

and key players

at the heart of the AI revolution.

00;00;21;27 - 00;00;25;27

I want to ask him whether we're creating

something that will save us or destroy us,

00;00;26;03 - 00;00;30;02

what future he thinks is most likely,

and what we need to do to prepare.

00;00;30;05 - 00;00;35;04

Let's find out.

00;00;35;07 - 00;00;36;24

I'm here with Gregory Warner.

00;00;36;24 - 00;00;38;29

He is a Peabody Award winning journalist.

00;00;38;29 - 00;00;43;25

He's the producer and host of The Last

Invention podcast, all about AI.

00;00;43;25 - 00;00;47;09

And maybe just to start things off, Greg,

tell me a little bit about

00;00;47;12 - 00;00;51;16

why you and some of your co-producers

created this podcast.

00;00;51;17 - 00;00;52;14

what was kind of the,

00;00;52;14 - 00;00;55;21

you know, the rationale

for why you wanted to tell this story?

00;00;55;24 - 00;00;56;09

Sure.

00;00;56;09 - 00;01;00;11

I mean, I think what for me would

what got me hooked on the on

00;01;00;11 - 00;01;06;09

the topic was realizing

that the people making I were, had a,

00;01;06;12 - 00;01;10;07

I had a sense

that this, this might kill us all.

00;01;10;10 - 00;01;11;12

It's as simple as that.

00;01;11;12 - 00;01;15;01

I mean, the fact that the

that they felt that the risk,

00;01;15;04 - 00;01;19;20

of this thing that they were building to

humanity was, was real,

00;01;19;23 - 00;01;23;13

just felt like such an interesting time

to live in.

00;01;23;13 - 00;01;28;07

And the fact that that that debate

over the existential risks and rewards

00;01;28;07 - 00;01;32;09

of this technology, because I think that

the potential upsides are just as radical.

00;01;32;14 - 00;01;36;05

It felt like while we were discussing

in the news at the time, things like,

00;01;36;05 - 00;01;37;18

oh, the danger of deepfakes

00;01;37;18 - 00;01;41;11

and the possibility of AI taking jobs,

these more existential questions

00;01;41;11 - 00;01;46;10

which had haunted AI really from its

beginning, as we found out.

00;01;46;13 - 00;01;48;24

It felt

like, okay, we need to have this debate

00;01;48;24 - 00;01;53;17

and try to bring this to people

in a way that didn't,

00;01;53;20 - 00;01;56;01

for lack of a better word,

didn't just freak people out.

00;01;56;01 - 00;01;59;15

You know, the

the idea was, let's introduce people to

00;01;59;15 - 00;02;05;21

the kind of debates that are being had in,

and have been had in these circles and,

00;02;05;22 - 00;02;10;29

and how we might talk about the future

with some kind of superintelligence.

00;02;11;02 - 00;02;12;05

It's, you know, that, to me,

00;02;12;05 - 00;02;15;29

is one of the most interesting things

about the topic is that if you if you ask

00;02;16;02 - 00;02;19;09

some of the most knowledgeable people

and the most central people

00;02;19;12 - 00;02;23;04

in this conversation,

what their outlook is, that just

00;02;23;04 - 00;02;28;04

the range of answers is so radical from,

you know, basically utopia

00;02;28;04 - 00;02;32;16

to destroy us as a species

to, you know, this, you know,

00;02;32;16 - 00;02;36;09

other cohort of voices saying it's all,

you know, basically a nothing burger.

00;02;36;12 - 00;02;40;01

You know, at the risk of asking you

to editorialize as a journalist,

00;02;40;04 - 00;02;42;27

I'm curious, you know, having talked

to all these people, having heard

00;02;42;27 - 00;02;44;13

all the sides of the story,

00;02;44;13 - 00;02;48;00

you know, what

what outlook do you find most compelling?

00;02;48;00 - 00;02;51;22

Are are you worried about the existential

risk or where have you kind of landed?

00;02;51;29 - 00;02;56;06

As you know, Greg, the human.

00;02;56;09 - 00;02;56;26

Okay.

00;02;56;26 - 00;02;57;25

Greg. The human.

00;02;57;25 - 00;03;00;03

Where do I land? well,

let me tell you a story.

00;03;00;03 - 00;03;03;23

So there was a hacker,

a white hat hacker named Dawn song.

00;03;03;24 - 00;03;07;21

I think her name is, And she took

00;03;07;21 - 00;03;11;05

all the AI, the leading AI models,

00;03;11;08 - 00;03;14;27

and she found, 20 zero days,

00;03;14;27 - 00;03;18;06

if you're familiar with the zero

day, it's a that's a hacker terminology.

00;03;18;06 - 00;03;20;09

And basically it's a it's a huge deal.

00;03;20;09 - 00;03;24;13

It's the sort of thing you stop everything

and you focus on this, this weakness,

00;03;24;13 - 00;03;27;26

because it's incredibly essential

to fix it.

00;03;27;29 - 00;03;31;22

To find even 102 is a big deal.

00;03;31;26 - 00;03;33;16

She found 20.

00;03;33;16 - 00;03;35;21

And using ChatGPT

00;03;35;21 - 00;03;39;26

and the current model, we're not even

talking about some future far off thing.

00;03;39;26 - 00;03;43;11

And then I think she used Gemini

and found it even faster.

00;03;43;11 - 00;03;44;08

Or something like this.

00;03;44;08 - 00;03;50;01

And the important thing is that

I hadn't been trained to find a zero day.

00;03;50;04 - 00;03;53;04

It literally became a top level hacker

00;03;53;10 - 00;03;56;16

with just a few smart prompts.

00;03;56;19 - 00;04;03;13

And so I think my take from that is

00;04;03;16 - 00;04;06;00

we are talking about

when will these systems

00;04;06;00 - 00;04;10;02

be good enough to pose a threat.

00;04;10;05 - 00;04;13;27

But I think that time is already here.

00;04;14;00 - 00;04;20;08

And so it's not will we awaken a God

or will we a will we summon the demon?

00;04;20;09 - 00;04;23;28

It's it's it's it's

not a future conversation.

00;04;24;01 - 00;04;27;00

It's it's that we we are living already

00;04;27;00 - 00;04;32;12

with a with a technology

that is so much more capable

00;04;32;12 - 00;04;38;02

than we realize that is becoming

increasingly capable and by design, is,

00;04;38;05 - 00;04;43;12

its capacities, its capabilities

are not known until the model is released.

00;04;43;12 - 00;04;45;04

That's amazing.

00;04;45;04 - 00;04;49;26

Really, when you think about it, that,

that it takes being out in the world

00;04;49;26 - 00;04;54;16

in order to or being out being created

in order to figure out what it can do.

00;04;54;19 - 00;04;59;29

And so I guess, you know, in terms

of my level of doom, which is like

00;04;59;29 - 00;05;04;23

my probability of doom, I guess you

other folks have used that on the show.

00;05;04;26 - 00;05;06;10

I don't think of it as Will.

00;05;06;10 - 00;05;11;21

Are we headed toward Utopia or are

we headed toward, you know, apocalypse?

00;05;11;24 - 00;05;16;15

It's there are weak points in our world.

00;05;16;18 - 00;05;19;22

We we have clearly ways

00;05;19;22 - 00;05;22;25

in which small conflicts

can scale up into big ones.

00;05;22;25 - 00;05;25;05

I mean, I've seen

that is a is a foreign correspondent.

00;05;25;05 - 00;05;26;11

I've seen as a war journalist.

00;05;26;11 - 00;05;29;11

And there are pathways to harm.

00;05;29;17 - 00;05;34;09

And so the question is how bad an impact

will that be?

00;05;34;12 - 00;05;36;10

Because it's definitely not zero. It

feels.

00;05;36;10 - 00;05;41;27

My doom in that sense is one like I am

sure that some bad things will happen.

00;05;42;04 - 00;05;44;16

I think actually that's inevitable.

00;05;44;16 - 00;05;47;23

But does that mean

the extension of the human race does

00;05;47;29 - 00;05;51;21

not mean

we can't recover and learn more? No.

00;05;51;21 - 00;05;53;25

I mean, I'm actually kind of an optimist

in that sense.

00;05;53;25 - 00;05;57;00

I don't like to even consider

that we are headed toward extinction.

00;05;57;04 - 00;05;58;04

But I don't know why.

00;05;58;04 - 00;06;01;21

We're not talking more about AI safety.

00;06;01;24 - 00;06;02;01

Well,

00;06;02;01 - 00;06;05;08

and you know, what's so compelling to me

about that is,

00;06;05;08 - 00;06;09;11

you know, in that story, in that example,

not about a crystal ball, right?

00;06;09;11 - 00;06;13;11

It's not about saying, oh, how

how quickly does the technology improve?

00;06;13;11 - 00;06;15;22

What can the technology do tomorrow?

00;06;15;22 - 00;06;17;10

It's real.

00;06;17;10 - 00;06;19;08

I think rational concern

00;06;19;08 - 00;06;23;15

about the disruption it can have based on

what's what's out there today.

00;06;23;15 - 00;06;26;20

And so you know, hasn't journalists,

you've talked to everybody from,

00;06;26;20 - 00;06;30;07

you know, AI leaders

to, you know, folks around the world and

00;06;30;07 - 00;06;32;28

you know, more kind of political roles or,

or military roles.

00;06;32;28 - 00;06;37;09

And so, I guess I'll ask you this way,

00;06;37;11 - 00;06;41;00

how ready are we right now for that

threat?

00;06;41;03 - 00;06;45;13

And, you know, what do we need to do

collectively as a society

00;06;45;16 - 00;06;50;13

to, you know, minimize

that that particular lens on zoom to

00;06;50;13 - 00;06;53;25

to minimize that,

the risk that it's going to provide,

00;06;53;25 - 00;06;58;07

you know, tremendous

harm to us as a society.

00;06;58;10 - 00;06;58;18

Yeah.

00;06;58;18 - 00;06;59;25

I would give two answers. I mean, there's

00;06;59;25 - 00;07;04;07

there's been some interesting papers

from the Frontier Model Forum.

00;07;04;10 - 00;07;05;13

I don't know if you've interviewed

00;07;05;13 - 00;07;08;13

anybody from there, but they're basically

an industry trade group.

00;07;08;18 - 00;07;10;29

And also interesting papers from

00;07;10;29 - 00;07;13;29

from anthropic, the, the AI company

00;07;14;06 - 00;07;17;17

that that really have have done a lot

to sort of

00;07;17;17 - 00;07;21;05

look at what are specifically the risks

and also the remedies.

00;07;21;05 - 00;07;21;27

They're not just red

00;07;21;27 - 00;07;25;28

teaming these models in terms of

trying to get them to do things that were,

00;07;26;00 - 00;07;29;06

you know, they were programed to do or

they're not supposed to do, like give you

00;07;29;06 - 00;07;32;16

the ingredients for a biological weapon.

00;07;32;19 - 00;07;35;01

And red teaming, tries

00;07;35;01 - 00;07;38;06

to get them to do those things or

to figure out if they will blackmail you.

00;07;38;06 - 00;07;41;21

But they're also saying, okay, wait,

wait, let's say it did those things.

00;07;41;21 - 00;07;43;21

How do we correct it?

00;07;43;21 - 00;07;48;10

And,

that's where most work needs to be done.

00;07;48;13 - 00;07;53;10

Because even if let's say according

to this paper by the frontier model form,

00;07;53;13 - 00;07;58;12

if you get the the model to give you, it's

not supposed to give you the, the,

00;07;58;12 - 00;08;01;28

you know, the the recipe for anthrax,

but let's say it does,

00;08;02;01 - 00;08;06;29

it tells you exactly how to make it

and how to get away with it.

00;08;07;02 - 00;08;07;12

Okay.

00;08;07;12 - 00;08;11;11

So how do we then program

the model to unlearn that?

00;08;11;14 - 00;08;14;23

It's, you can do a step.

00;08;14;23 - 00;08;15;29

It's called targeted on learning.

00;08;15;29 - 00;08;20;02

But what they then found was that it

actually didn't really unlearn it.

00;08;20;02 - 00;08;21;16

It just said it did.

00;08;21;16 - 00;08;24;12

And,

then there's also a technique which was,

00;08;24;12 - 00;08;27;20

where you get it

to give false information.

00;08;27;20 - 00;08;31;00

So if somebody asks for anthrax,

it will leave out a couple of ingredients.

00;08;31;03 - 00;08;31;14

Okay.

00;08;31;14 - 00;08;35;18

But then you're introducing

falsehood into the system

00;08;35;21 - 00;08;40;28

and you may have knock down effects which

are bad for actually legitimate research.

00;08;41;01 - 00;08;45;10

So that's odd

and an odd interesting situation,

00;08;45;10 - 00;08;47;21

which is not only that

we're not doing enough

00;08;47;21 - 00;08;51;21

sort of safety testing,

but we don't actually know the best way

00;08;51;24 - 00;08;56;10

to truly put guardrails

on this, on these technologies.

00;08;56;13 - 00;09;00;09

The best we can do

is, is, is a sort of a training overlay

00;09;00;14 - 00;09;03;18

where you, you essentially train it not to

00;09;03;18 - 00;09;06;25

or train it, to,

to not answer those questions.

00;09;06;28 - 00;09;08;24

But still it's a genetic

00;09;08;24 - 00;09;12;17

and as an agent system,

we do not know what it's going to do.

00;09;12;20 - 00;09;16;10

So, yeah, in terms of your question,

where are we at?

00;09;16;10 - 00;09;18;07

I think that,

00;09;18;10 - 00;09;21;00

the reason

I would want more of us to talk about this

00;09;21;00 - 00;09;24;08

is because in interacting with the models,

00;09;24;11 - 00;09;28;04

we are actually playing a role in

AI safety.

00;09;28;07 - 00;09;29;29

This is not true in nuclear weapons.

00;09;29;29 - 00;09;32;29

You know, nuclear weapons,

we have no impact

00;09;33;04 - 00;09;35;29

on whether there's nuclear war,

except for maybe who we vote for.

00;09;35;29 - 00;09;37;03

Perhaps.

00;09;37;03 - 00;09;39;21

But we do,

even if we're not a technologist,

00;09;39;21 - 00;09;42;10

even if we're not a lawmaker,

actually play a role.

00;09;42;10 - 00;09;47;08

There's there's all kinds of forms

where if if the AI does something weird,

00;09;47;08 - 00;09;49;21

you can post it,

you know, and it will be looked at.

00;09;49;21 - 00;09;53;02

In fact, all the AI companies say

they want that material,

00;09;53;02 - 00;09;54;23

they want that data.

00;09;54;23 - 00;09;58;02

And so we're all playing

a, I think a role in terms of,

00;09;58;02 - 00;10;00;11

as we interact with these models.

00;10;00;11 - 00;10;03;05

And we shouldn't just talk about chat

bots, I assume in the conversation.

00;10;03;05 - 00;10;04;20

I mean, I so much more than chat bots,

00;10;04;20 - 00;10;08;20

but just because it's

the most obvious thing, yes, we

00;10;08;26 - 00;10;12;16

we can also think about alignment

in our in our life.

00;10;12;16 - 00;10;16;04

We can try to treat these models

more carefully.

00;10;16;04 - 00;10;18;15

Maybe not give them access to everything.

00;10;18;15 - 00;10;21;22

yet I don't think we are doing that.

00;10;21;22 - 00;10;27;29

I think mainly we're just either getting

freaked out or ignoring the problem.

00;10;28;02 - 00;10;28;24

If you work in

00;10;28;24 - 00;10;32;14

IT, Infotech research Group is a name

you need to know.

00;10;32;17 - 00;10;35;17

No matter what your needs are, Infotech

has you covered.

00;10;35;22 - 00;10;36;29

AI strategy?

00;10;36;29 - 00;10;39;11

Covered. Disaster recovery?

00;10;39;11 - 00;10;40;11

Covered.

00;10;40;11 - 00;10;42;26

Vendor negotiation? Covered.

00;10;42;26 - 00;10;46;19

Infotech supports you with the best

practice research and a team of analysts

00;10;46;19 - 00;10;50;12

standing by ready to help you

tackle your toughest challenges.

00;10;50;15 - 00;10;55;19

Check it out at the link below

and don't forget to like and subscribe!

00;10;55;22 - 00;10;58;27

you, you talk about with AI and

especially with chat bots, that

00;10;58;27 - 00;11;04;07

we all play a role there and how that can,

you know, influence safety.

00;11;04;07 - 00;11;05;16

It can influence,

00;11;05;16 - 00;11;09;13

you know, kind of a trajectory of, of,

you know, the risk that's created here.

00;11;09;16 - 00;11;11;00

what is that role like?

00;11;11;00 - 00;11;16;28

What what message do you want to impart

on, you know, the the average chat

00;11;16;28 - 00;11;24;04

bot user out there to get them to interact

with the technology more thoughtfully?

00;11;24;07 - 00;11;25;24

Well, I mean, a couple of things.

00;11;25;24 - 00;11;30;12

One is just the basics, which is if

00;11;30;15 - 00;11;32;26

if I does something

00;11;32;26 - 00;11;36;05

odd, if it behaves in a certain way.

00;11;36;08 - 00;11;38;29

I mean, I think if you are a hacker

and you can figure out

00;11;38;29 - 00;11;42;23

how to use it to hack things,

that would be a more specialized tool.

00;11;42;26 - 00;11;46;18

I mean, look, it's a light role, that

if we're interacting with the technology,

00;11;46;18 - 00;11;51;13

but it's also about if we're bringing it

into our companies, what is then?

00;11;51;16 - 00;11;55;18

Do you know what is

then the role of AI in our company?

00;11;55;18 - 00;12;00;14

I mean, they talk about this idea of human

in the loop so often use phrase.

00;12;00;17 - 00;12;05;17

But on the positive side,

the human in the loop, we know that,

00;12;05;20 - 00;12;08;20

AI is more capable

00;12;08;26 - 00;12;11;16

if it has a human in the loop.

00;12;11;16 - 00;12;12;17

And that's, that's it's

00;12;12;17 - 00;12;16;17

not just a, it's a safety thing,

but it's a, it's a value thing that,

00;12;16;17 - 00;12;19;19

the perfect example is this,

this question of the chess computer.

00;12;19;19 - 00;12;21;14

People are amazed at chess.

00;12;21;14 - 00;12;24;19

Computers can beat humans,

but a human with a chess computer

00;12;24;19 - 00;12;27;19

even human with a kind of slightly,

00;12;27;22 - 00;12;32;13

less and more dumb chess computer

can be any person and any computer.

00;12;32;16 - 00;12;33;23

And so

00;12;33;23 - 00;12;38;12

I think that as we are in the situation

in this decision making capacity as to how

00;12;38;13 - 00;12;42;23

much are we going to give over to the AI,

how much are we going to outsource?

00;12;42;26 - 00;12;44;13

And also

00;12;44;16 - 00;12;46;18

how much do we trust it?

00;12;46;18 - 00;12;49;18

I think we have to get away from an,

00;12;49;22 - 00;12;52;16

the kind of anthropomorphizing mentality

where we think,

00;12;52;16 - 00;12;55;16

wow, this is a really capable,

amazing worker

00;12;55;19 - 00;12;58;25

who can do all kinds of things,

what more can we give it?

00;12;58;28 - 00;13;01;26

That's probably the wrong, idea.

00;13;01;26 - 00;13;05;15

Rather, we should we should think of this

00;13;05;15 - 00;13;09;08

as a completely alien

00;13;09;11 - 00;13;12;22

kind of intelligence,

which in terms of mimicking

00;13;12;22 - 00;13;16;12

human, intelligence, it's

been programed to do that.

00;13;16;15 - 00;13;20;12

It's it's model is designed to interact

with you as a human.

00;13;20;12 - 00;13;25;09

That's a that's a blueprint that was

created 75 years ago by Alan Turing.

00;13;25;12 - 00;13;27;08

But when we interact

with the intelligence,

00;13;27;08 - 00;13;29;29

we should just have a certain alienation

from it.

00;13;29;29 - 00;13;33;18

And it's, and treat it as a

00;13;33;21 - 00;13;38;00

is an incredibly I don't know,

I mean, is an incredibly strange,

00;13;38;03 - 00;13;42;02

incredibly wonderful, marvelous tool

00;13;42;05 - 00;13;46;02

that we have in our world now,

but that, perhaps

00;13;46;02 - 00;13;49;26

it shouldn't have access

to confidential information.

00;13;49;26 - 00;13;53;02

We do know that

these models can blackmail,

00;13;53;02 - 00;13;57;01

it shouldn't have,

perhaps, control over the company.

00;13;57;01 - 00;13;57;15

You wouldn't.

00;13;57;15 - 00;14;02;02

You wouldn't leave an incredibly capable

intern in charge of the entire operation.

00;14;02;04 - 00;14;04;10

Even if they did amazing

with their photographic memory

00;14;04;10 - 00;14;05;03

and the fact that they didn't

00;14;05;03 - 00;14;08;03

need to sleep in the fact

that they were master of all subjects.

00;14;08;06 - 00;14;11;19

Nevertheless, we would be

we would be careful.

00;14;11;19 - 00;14;15;23

And I think when we read things

like the anthropic blackmail paper, which,

00;14;16;00 - 00;14;20;15

which was a fascinating paper,

back in July, that showed that the

00;14;20;15 - 00;14;25;07

anthropic model, in order to not be turned

off, threatened blackmail of the user.

00;14;25;07 - 00;14;27;10

This was a red team model.

00;14;27;10 - 00;14;29;15

We shouldn't

00;14;29;15 - 00;14;35;03

get scared that these models, quote

unquote, really want to do us harm

00;14;35;06 - 00;14;38;15

or have, evil intentions,

00;14;38;18 - 00;14;41;07

but rather realize that agency

00;14;41;07 - 00;14;46;20

that giving agency to a technology is,

00;14;46;23 - 00;14;48;07

is a powerful thing.

00;14;48;07 - 00;14;51;04

And we should,

we should treat it with respect

00;14;51;04 - 00;14;54;04

and with with caution.

00;14;54;09 - 00;14;55;27

I think that's really well said.

00;14;55;27 - 00;14;59;12

And there's one word in there, Greg,

that that caught my attention in the sense

00;14;59;12 - 00;14;59;25

that it's a word

00;14;59;25 - 00;15;02;27

that I don't know that anyone's ever said

in our conversations before, which is,

00;15;03;04 - 00;15;05;06

you said a couple of times alien.

00;15;05;06 - 00;15;08;04

And, you know, there's

a couple of different ways to interpret

00;15;08;04 - 00;15;08;26

the word alien.

00;15;08;26 - 00;15;12;12

There's like, you know, E.T.,

like extraterrestrial life.

00;15;12;16 - 00;15;14;29

There's also

alien is just outsider. Right.

00;15;14;29 - 00;15;19;01

But but there's some there's

some sense here that this is a foreign

00;15;19;01 - 00;15;22;05

or external presence in our organizations

00;15;22;05 - 00;15;25;26

and maybe even to us as humans,

that's not fully understood.

00;15;25;29 - 00;15;29;19

And it sounds like

it sounds like your approach and and maybe

00;15;29;19 - 00;15;33;06

even if I can, you know, push you

a little bit further that your advice

00;15;33;06 - 00;15;36;27

to business leaders would probably be,

you know, proceed with caution.

00;15;36;27 - 00;15;38;23

Is that fair?

00;15;38;26 - 00;15;39;19

Yeah, I think that's fair.

00;15;39;19 - 00;15;40;23

You know, I mean, I would

00;15;40;23 - 00;15;44;04

I would also credit Lisa Eliza Reed Koski,

who wrote this book called

00;15;44;04 - 00;15;47;16

If Anyone Builds It, Everyone Dies,

which to me is the

00;15;47;19 - 00;15;50;26

these got to be the best title of the book

ever.

00;15;50;29 - 00;15;53;29

It's direct, but, he really explores

00;15;53;29 - 00;15;57;12

this question of the alien intelligence

in a, in a very smart way.

00;15;57;12 - 00;16;00;14

And,

you know, just to paraphrase what he says

00;16;00;21 - 00;16;03;01

is that,

00;16;03;01 - 00;16;06;29

so when, when,

when we think about this, these

00;16;07;02 - 00;16;09;26

but he warns these as if anyone builds it,

everyone dies.

00;16;09;26 - 00;16;14;17

Meaning a superintelligence

is fundamentally unnavigable

00;16;14;20 - 00;16;17;17

or uncontrollable and unpredictable.

00;16;17;17 - 00;16;20;01

And this is

this is not just you Koski saying this.

00;16;20;01 - 00;16;22;07

We know that,

00;16;22;07 - 00;16;25;25

we know the about the basic technology

that that these models,

00;16;25;25 - 00;16;28;03

we don't know what they will do

before they're made.

00;16;28;03 - 00;16;32;09

We don't we can't tell you how they are

making the decisions they're making.

00;16;32;09 - 00;16;35;14

So there is a kind of black box

unknowability at the core.

00;16;35;17 - 00;16;38;14

But in terms of the alien, this

I think this is so interesting because.

00;16;38;14 - 00;16;41;08

It sort of gets it the danger of sci fi,

right?

00;16;41;08 - 00;16;44;08

Sci fi is written for humans.

00;16;44;09 - 00;16;49;16

It's written by humans, for humans, and

so even if there are aliens in the sci fi,

00;16;49;19 - 00;16;55;04

and we are all familiar with a very common

trope about AI versus humanity,

00;16;55;07 - 00;16;57;04

AI rebelling against humanity,

00;16;57;04 - 00;17;02;11

AI doing something evil,

and also aliens versus humans.

00;17;02;14 - 00;17;05;02

There's a certain way

in which those stories play out,

00;17;05;02 - 00;17;09;25

and this is Yudkowsky main point that,

he sort of follows the rules of narrative

00;17;09;26 - 00;17;13;26

where, okay,

the humans are battling against the AI,

00;17;13;26 - 00;17;18;11

the AI has an incredibly new,

powerful weapon or something,

00;17;18;11 - 00;17;22;28

or the AI is willing to act

inhumanely in some important way.

00;17;23;01 - 00;17;26;04

And then, you know, the question is,

what will humanity do about it?

00;17;26;04 - 00;17;27;12

And there's a big conflict.

00;17;27;12 - 00;17;30;22

But what he says

is that we have to understand

00;17;30;22 - 00;17;34;14

that as these that there's just so much

that is

00;17;34;17 - 00;17;38;04

non and it's a non-human

or he uses the word alien about it.

00;17;38;04 - 00;17;42;14

It's thinking that the ways

in which this might go wrong in

00;17;42;14 - 00;17;44;27

most of his scenarios are in ways

in which it goes wrong.

00;17;44;27 - 00;17;47;22

I don't think he has any.

Which goes perfectly right.

00;17;47;22 - 00;17;51;22

The ways in which this goes wrong

will be complicated and weird.

00;17;51;25 - 00;17;53;18

The complicated and weird.

00;17;53;18 - 00;17;56;10

They won't look like Skynet,

00;17;56;13 - 00;17;58;19

you know, from the Terminator, film.

00;17;58;19 - 00;18;01;04

They won't look like,

00;18;01;04 - 00;18;04;02

you know, another sort of situation

where, like,

00;18;04;02 - 00;18;06;05

they won't look like how 9000 or something

like that.

00;18;06;05 - 00;18;07;10

It'll look.

00;18;07;10 - 00;18;10;10

Or from, from, Stanley Kubrick.

00;18;10;13 - 00;18;13;10

It'll look like.

00;18;13;10 - 00;18;16;10

Okay, we never, predicted this.

00;18;16;10 - 00;18;17;29

This wasn't programed.

00;18;17;29 - 00;18;22;00

How did the how did the AI even grow

00;18;22;00 - 00;18;26;10

to want this, for example?

00;18;26;13 - 00;18;31;26

Anyway, he has these nightmarish scenarios

which we could go into later, but, that

00;18;31;27 - 00;18;37;12

that is his that is his term that we need

to not anthropomorphize this thing.

00;18;37;12 - 00;18;38;29

Not thinking of is just a tool.

00;18;38;29 - 00;18;43;05

Not thinking of his as another person, a

super smart Einstein or, you know, stereo.

00;18;43;05 - 00;18;46;26

I'm a day says, you know,

millions of Einsteins in a data center.

00;18;46;29 - 00;18;50;12

But just think of it

as an alien intelligence that,

00;18;50;16 - 00;18;52;27

and then most of you

Koski says this is going to go

00;18;52;27 - 00;18;54;27

horribly wrong,

but I think we could also talk together

00;18;54;27 - 00;18;59;12

about how an alien intelligence

could radically improve our lives,

00;18;59;15 - 00;19;02;10

which we should definitely get to, but,

yes,

00;19;02;10 - 00;19;06;02

I think resisting anthropomorphizing is

is absolutely important.

00;19;06;05 - 00;19;09;23

Well and and and recognizing

the inherent unpredictability

00;19;09;29 - 00;19;12;15

it sounds like

if something that just thinks thinks

00;19;12;15 - 00;19;14;27

and I'm even anthropomorphizing

by saying thinking I guess.

00;19;14;27 - 00;19;19;09

But that just behaves in a way

fundamentally different from,

00;19;19;14 - 00;19;24;12

you know, us Behaves and also wants things

fundamentally different than ours.

00;19;24;12 - 00;19;24;18

Right.

00;19;24;18 - 00;19;27;15

I think that's the key thing

is that, in a movie.

00;19;27;15 - 00;19;28;08

Right.

00;19;28;08 - 00;19;33;20

Even the enemies of humanity

want something that is recognizably worth

00;19;33;20 - 00;19;38;26

wanting,

like power, for example, or control.

00;19;38;29 - 00;19;40;16

Whereas

00;19;40;16 - 00;19;43;24

in AI, a superintelligent,

I mean, a want those things.

00;19;43;24 - 00;19;46;02

It may just want some other thing

00;19;46;02 - 00;19;48;20

and destroy the world

in the process of getting that thing.

00;19;48;20 - 00;19;50;10

And we'll think, well,

why did even want that?

00;19;50;10 - 00;19;52;22

That that makes no sense.

You would not go for those things.

00;19;52;22 - 00;19;56;17

And and yes, Eliezer cascades

all kinds of examples of that.

00;19;56;20 - 00;19;59;08

So, you know, there's a backdrop here.

00;19;59;08 - 00;20;02;06

We've talked about the need for AI safety

and for guardrails.

00;20;02;06 - 00;20;06;01

And, you know, I think there's

some really, really important points,

00;20;06;04 - 00;20;09;22

that, that you've made

that others have made in the space.

00;20;09;25 - 00;20;14;07

There's, there seems to be this spectrum

right now in terms of where people fall.

00;20;14;10 - 00;20;18;07

And on the one end it's

slow down guardrails, safety.

00;20;18;07 - 00;20;20;00

Let's really understand this.

00;20;20;00 - 00;20;23;29

And on the other end, there's

this notion of a winner take all race

00;20;24;05 - 00;20;28;28

where it's as we need this as fast

as possible, guardrails be damned.

00;20;29;01 - 00;20;31;12

Let's just let's just get there first.

00;20;31;12 - 00;20;34;12

It's almost like the anti

Uber kowski model where

00;20;34;12 - 00;20;38;23

it's just,

you know, caution to the wind. And

00;20;38;26 - 00;20;39;25

I don't know

00;20;39;25 - 00;20;42;12

you talked earlier Greg about you know

00;20;42;12 - 00;20;47;15

some of the key players in Silicon Valley

and beyond recognizing the risk.

00;20;47;15 - 00;20;51;06

But you know how have you

how have you seen them

00;20;51;09 - 00;20;55;04

behaving in practice space based on your

research and based on your interviews,

00;20;55;04 - 00;20;59;05

like where are we falling in terms

of the development of this technology?

00;20;59;05 - 00;21;01;04

Where, where should we be falling?

00;21;01;04 - 00;21;05;07

And, you know, is there a message

for the people at the helm in terms of,

00;21;05;10 - 00;21;08;05

like, should we be collectively trying

00;21;08;05 - 00;21;11;05

to influence the behavior here?

00;21;11;11 - 00;21;11;19

Yeah.

00;21;11;19 - 00;21;17;04

It's it's such a great question

because, the story of AI in the last

00;21;17;04 - 00;21;22;16

ten years, certainly has been a story

where essentially one after another,

00;21;22;16 - 00;21;26;15

people have said, oh, I don't trust

that guy to build a I, I need to build AI

00;21;26;15 - 00;21;31;02

and I need to build it faster than they do

because only I can build it safely.

00;21;31;03 - 00;21;32;21

And you see this.

00;21;32;21 - 00;21;36;00

So Demis Hassabis

at DeepMind, Google by DeepMind,

00;21;36;03 - 00;21;41;17

Elon Musk, Sam Altman create

OpenAI as a direct rival to Google.

00;21;41;20 - 00;21;46;19

Dario Amodei leaves OpenAI because he says

those guys are not committed to safety.

00;21;46;20 - 00;21;50;09

Meanwhile,

Elon Musk's Musk is kicked out of OpenAI.

00;21;50;09 - 00;21;54;22

He's forms X AI,

this and essentially this drive

00;21;54;22 - 00;21;59;03

to create it first and to create it.

00;21;59;06 - 00;22;01;12

And they all say for the benefit

of humanity.

00;22;01;12 - 00;22;05;29

Well, actually, I don't think Elon Musk

says that directly, but nevertheless,

00;22;05;29 - 00;22;10;01

the, the that drive has

then resulted in a race

00;22;10;04 - 00;22;14;22

and even just we're talking about the race

in, within the US.

00;22;14;26 - 00;22;20;24

We're not even talking about the race

with China, which amplifies things. So

00;22;20;27 - 00;22;22;05

that kind of,

00;22;22;05 - 00;22;25;29

I guess you'd call it

Silicon Valley approach to,

00;22;26;02 - 00;22;30;03

to product development,

where certainly making it first,

00;22;30;06 - 00;22;33;05

making it fast

00;22;33;05 - 00;22;34;06

has a lot of benefits.

00;22;34;06 - 00;22;38;14

I mean, not only being first to market,

but really being able to set the create

00;22;38;14 - 00;22;42;16

create the,

the model and create the, create the,

00;22;42;16 - 00;22;46;05

sort of the prototype and sort of what

people are interacting with.

00;22;46;05 - 00;22;47;25

And then you fix it,

00;22;47;25 - 00;22;51;15

you know, you you release it

and then you fix the bugs afterward.

00;22;51;18 - 00;22;55;26

So that kind of approach

to superintelligence

00;22;55;29 - 00;23;01;04

is I mean, it's that's it's really it's

really amazing to me that, Neda, for

00;23;01;04 - 00;23;04;10

example, has a superintelligence division,

00;23;04;13 - 00;23;08;13

so it feels like something sci fi.

00;23;08;16 - 00;23;12;02

But, the other I think question of it is,

00;23;12;05 - 00;23;18;03

why why are they doing this? Why

00;23;18;06 - 00;23;19;11

why the risk?

00;23;19;11 - 00;23;19;23

I'm sorry.

00;23;19;23 - 00;23;24;28

Why the race given the risk and and

and how do they justify it,

00;23;25;01 - 00;23;28;07

given that every one of the people I just

mentioned

00;23;28;10 - 00;23;32;25

has a well, not meta,

but but others have a have have stated

00;23;32;25 - 00;23;35;28

that they are very worried

about the dangers of this technology.

00;23;35;29 - 00;23;41;13

How is it that those same people are in a

race to create it as fast as possible?

00;23;41;14 - 00;23;42;29

Right.

00;23;42;29 - 00;23;44;20

And I think

00;23;44;20 - 00;23;49;04

one book that I would recommend people

or read or other essay

00;23;49;10 - 00;23;53;10

is, it's called Machines of Loving Grace

by Dario.

00;23;53;10 - 00;23;56;08

I'm a day, who started anthropic.

00;23;56;08 - 00;23;57;17

I don't know.

00;23;57;17 - 00;23;59;10

Have you have you read this this essay?

00;23;59;10 - 00;24;02;25

It's it's it's I it

I first encountered it because a number

00;24;02;25 - 00;24;07;02

of my humanitarian friends,

you know, from from Nairobi and Ukraine

00;24;07;02 - 00;24;09;27

and others,

they were all loving this book and,

00;24;09;27 - 00;24;13;09

and just to, just to sort of understand

the context here.

00;24;13;09 - 00;24;17;04

I mean, these are a lot of folks

who started off in NGOs

00;24;17;07 - 00;24;21;15

and then got disappointed with NGOs,

started companies to really do good work

00;24;21;18 - 00;24;25;04

that they felt they could do, work faster,

more technologically savvy,

00;24;25;11 - 00;24;30;02

and help humanity not under the rubric

of kind of philanthropy and NGOs,

00;24;30;02 - 00;24;33;28

but rather through a through

a startup model that kind of cohort.

00;24;34;01 - 00;24;38;00

They were blown away by this essay,

Machines of loving Grace.

00;24;38;03 - 00;24;41;23

Machines

of loving grace is, I think, the best,

00;24;41;26 - 00;24;46;12

probably most clear headed, manifesto

00;24;46;15 - 00;24;49;17

for for the acceleration ist

point of view.

00;24;49;17 - 00;24;53;25

Mark Andreasen also has the techno

optimist manifesto, I think it's called.

00;24;53;25 - 00;24;57;22

But, I would say Machines of Loving Grace

is far clearer in terms of

00;24;57;22 - 00;25;01;07

what is the approach of somebody who,

for instance, Dario.

00;25;01;07 - 00;25;02;13

I'm a day.

00;25;02;13 - 00;25;04;02

He is an effective altruist.

00;25;04;02 - 00;25;07;00

He believes he kind of disavows that now.

00;25;07;00 - 00;25;08;22

But he's a certainly a person who believes

00;25;08;22 - 00;25;10;27

we need to do the most good

for the most people and live,

00;25;10;27 - 00;25;13;12

live our lives according to that.

So what does that mean?

00;25;13;12 - 00;25;18;20

That he is now creating a superintelligent

AI that might destroy us all?

00;25;18;23 - 00;25;21;23

What he lays out is he makes this case.

00;25;21;23 - 00;25;26;16

He says, you know, it is critical

to have a genuinely inspiring

00;25;26;16 - 00;25;30;17

vision of the future

and not just the plan to fight fires.

00;25;30;20 - 00;25;33;13

He says, yes, there are risks.

00;25;33;13 - 00;25;36;17

There are dangers of of of of powerful AI.

00;25;36;18 - 00;25;39;18

He doesn't use

the word AGI is powerful AI.

00;25;39;18 - 00;25;43;11

But at the end, there has to be something

we're fighting for, right?

00;25;43;11 - 00;25;47;10

Some something that we can rally,

rally towards.

00;25;47;13 - 00;25;50;24

I think he says

fear is only one kind of motivator.

00;25;50;24 - 00;25;52;16

We also need hope.

00;25;52;19 - 00;25;54;03

So what is it

00;25;54;03 - 00;25;57;03

that those who are building

AI are fighting for?

00;25;57;06 - 00;26;02;15

And he lays out this vision in that

in that essay of the Compressed century.

00;26;02;18 - 00;26;05;02

And if you've you've come across

the compressed century idea.

00;26;05;02 - 00;26;10;10

But this is like 100 years of progress

in, in 5 or 10.

00;26;10;13 - 00;26;13;08

And so all the scientific developments

that we may have

00;26;13;08 - 00;26;19;09

in the entire 21st century

and a bit of the 22nd will all happen.

00;26;19;12 - 00;26;22;06

He says in the 5 to 10 year window,

00;26;22;06 - 00;26;26;13

after we have a suitably advanced AI and a

when when will that will happen?

00;26;26;17 - 00;26;28;21

It's, you know, that's

there's a lot of debate about that.

00;26;28;21 - 00;26;32;22

He's even said it might happen as

soon as 2026, but nevertheless it's soon.

00;26;32;22 - 00;26;33;22

It's within our lifetimes.

00;26;33;22 - 00;26;34;17

This is what he believes.

00;26;34;17 - 00;26;38;12

So then all of this scientific progress,

00;26;38;15 - 00;26;41;02

and what will it lead to?

00;26;41;02 - 00;26;44;18

And so, we could go into what he thinks

it'll lead to.

00;26;44;18 - 00;26;45;21

It's actually quite fascinating.

00;26;45;21 - 00;26;49;23

Lis talks about biology,

health, what work and meaning.

00;26;49;26 - 00;26;54;13

But the reason that resonated

with so many people that I know,

00;26;54;16 - 00;26;58;18

in the developing world, in other places,

is that they're talking with people

00;26;58;18 - 00;27;01;24

who are not worried about their jobs

being taken away.

00;27;01;24 - 00;27;04;28

They're they're they're in terrible jobs.

00;27;04;28 - 00;27;07;18

So they don't have a job. They don't

they don't like their careers.

00;27;07;18 - 00;27;08;23

We know it's easy for you.

00;27;08;23 - 00;27;13;06

And I just sit here and say, God,

I can't believe I might take our jobs.

00;27;13;06 - 00;27;16;07

We I think we sort of like our jobs

generally, but,

00;27;16;07 - 00;27;19;09

there's a lot of people in the world

who are suffering, a lot of people,

00;27;19;09 - 00;27;23;01

a world who need solutions,

major solutions,

00;27;23;08 - 00;27;29;26

to, you know, climate change,

to poverty, to, you know, hunger.

00;27;29;29 - 00;27;33;28

And the acceleration is believe

or certainly Dario Ahmadi

00;27;33;28 - 00;27;37;04

in this essay believes that an advanced

00;27;37;04 - 00;27;41;16

AI is a radical solution.

00;27;41;19 - 00;27;43;20

And it will come and it will it will

00;27;43;20 - 00;27;47;03

bring about changes

that we cannot even imagine.

00;27;47;03 - 00;27;51;18

And what's so fascinating is how similar

00;27;51;21 - 00;27;56;14

somebody like Dario

Ahmed's vision is to Eliezer Yudkowsky.

00;27;56;17 - 00;27;59;13

The if anyone builds it, everyone dies.

00;27;59;13 - 00;28;02;08

Author, both of them

00;28;02;08 - 00;28;06;14

this complete dumber

and then one and in a fairly, you know,

00;28;06;17 - 00;28;09;01

acceleration test and we make it

whatever you want to call it.

00;28;09;01 - 00;28;13;23

But he's certainly a,

you know, believer in the power of AI.

00;28;13;26 - 00;28;16;14

They both, believe that this is going

00;28;16;14 - 00;28;19;14

to be such a radical change

00;28;19;18 - 00;28;22;28

and, and fundamentally upend money,

00;28;22;28 - 00;28;27;21

the things that we treat

as normal, both of them say,

00;28;27;24 - 00;28;30;19

the only thing normal about normal

is that it ends.

00;28;30;19 - 00;28;32;27

Normality always ends.

00;28;32;27 - 00;28;37;20

And so, yeah, the only difference

between them, of course, is,

00;28;37;25 - 00;28;40;28

is whether that will end in disaster

00;28;40;28 - 00;28;44;22

or whether it'll end in, in delight.

00;28;44;25 - 00;28;48;25

And but I think these both of these

people know the models very, very closely.

00;28;48;28 - 00;28;51;01

They, they're,

they're staring directly at them.

00;28;51;01 - 00;28;52;09

They've seen the progress of them.

00;28;52;09 - 00;28;54;17

They understand how they work.

00;28;54;17 - 00;28;55;26

So it's worth. Yeah. Yeah.

00;28;55;26 - 00;29;00;17

And being with that,

00;29;00;20 - 00;29;03;14

sitting with that imagination,

whether it's the darker side

00;29;03;14 - 00;29;06;26

or the or the positive side, but but yeah,

the simple answer your question is

00;29;06;29 - 00;29;09;21

I think they see a tremendous upside

to this.

00;29;09;21 - 00;29;11;08

And it's worth the risk.

00;29;11;08 - 00;29;13;02

So so I want to

I want to dig into that a little bit

00;29;13;02 - 00;29;15;21

because I'm, you know, I'm

a self-proclaimed cynic

00;29;15;21 - 00;29;17;19

for a lot of this stuff

or at least a skeptic.

00;29;17;19 - 00;29;21;07

And so the cynic or the skeptic in me

says that,

00;29;21;10 - 00;29;25;03

you know, the dark side and the light side

are, you know, a hair's breadth apart.

00;29;25;06 - 00;29;28;19

And it it just seems,

it seems to me or that's

00;29;28;20 - 00;29;33;19

there's certainly an argument to be made

that what tilts people to the light side

00;29;33;22 - 00;29;37;04

is if they're asking for a big bag

of money, if they're asking for somebody

00;29;37;04 - 00;29;40;06

to fund them, then suddenly

all it's going to be amazing, you

00;29;40;06 - 00;29;43;06

know, versus if you're, yudkowsky, you're

you're not asking for money.

00;29;43;06 - 00;29;45;05

It's a lot easier to move to the dark

side.

00;29;45;05 - 00;29;47;13

So, so, I mean, let me frame it this way.

00;29;47;13 - 00;29;51;02

Like, to what degree do you buy the,

00;29;51;02 - 00;29;55;01

utopianism or the true acceleration,

this vision

00;29;55;05 - 00;29;57;11

of some of these technology

leaders versus,

00;29;57;11 - 00;30;03;19

you know, how much do you think

it's a fundraising tool?

00;30;03;22 - 00;30;05;10

I think that's such an important question.

00;30;05;10 - 00;30;05;26

Right.

00;30;05;26 - 00;30;09;01

And it's definitely one

that a lot of tech journalists wrestle

00;30;09;01 - 00;30;12;01

with because they've,

00;30;12;05 - 00;30;15;10

I mean everybody that I know

has been burned, whether they were burned

00;30;15;10 - 00;30;18;24

on Google Glass or they were burned

on the metaverse or whatever.

00;30;18;27 - 00;30;24;23

That's the nature of technologists

is to hype their, their, their stuff.

00;30;24;23 - 00;30;29;00

Now, we should note that lawsuit

Koski is not selling anything.

00;30;29;03 - 00;30;32;02

He's just trying to warn the world.

00;30;32;02 - 00;30;34;05

He's like a Jeremiah.

00;30;34;05 - 00;30;41;06

And there are many people out there who,

00;30;41;09 - 00;30;45;05

you know, for example, I would say Yoshua

Bengio or Geoffrey Hinton.

00;30;45;05 - 00;30;49;00

Geoffrey Hinton left his job at Google,

a very high paying job,

00;30;49;00 - 00;30;50;07

which he got only in the 60s.

00;30;50;07 - 00;30;54;11

Geoffrey Hinton, of course, the godfather

of AI creator or not directly creator,

00;30;54;11 - 00;30;58;19

but certainly a co-creator

of this incredibly important,

00;30;58;22 - 00;31;02;19

algorithm backpropagation,

which led to AI models.

00;31;02;22 - 00;31;04;29

he left his job at Google and he's

00;31;04;29 - 00;31;08;15

he's out there trying to warn the world

about about this technology.

00;31;08;15 - 00;31;12;07

So I don't think it's just the,

00;31;12;10 - 00;31;17;13

all the hype is coming from the people who

who are or who stand to benefit.

00;31;17;16 - 00;31;18;17

However.

00;31;18;17 - 00;31;21;07

It does make it extremely difficult

to talk about

00;31;21;07 - 00;31;24;23

because clearly, there's a lot of hype.

00;31;25;00 - 00;31;28;08

I think I think probably I

mean, it'd be great to talk a little bit

00;31;28;08 - 00;31;32;09

about Yoshua Bengio,

maybe because Yoshua Bengio to me is the,

00;31;32;12 - 00;31;36;06

his is is such a different

kind of model out there.

00;31;36;06 - 00;31;39;10

And it's not just, complaining about

or warning the world.

00;31;39;10 - 00;31;43;02

It's he's, he's he's literally

presented the world with an alternative, a

00;31;43;02 - 00;31;46;27

non a genetic model of of

00;31;47;00 - 00;31;50;06

of AI which, which we don't we,

we're not even talking

00;31;50;06 - 00;31;53;13

about at all where we, we're imagining

that there's only one way to make AI.

00;31;53;19 - 00;31;55;14

They're going to get smarter and smarter

and they're

00;31;55;14 - 00;31;57;14

going to do more and more things

and they're going to,

00;31;57;14 - 00;32;00;25

you know, make our airplane reservations,

and then they're going to,

00;32;00;28 - 00;32;04;24

you know, take over the

then they're going to be our lawyer,

00;32;04;24 - 00;32;07;01

and then they're going to be our doctor.

And then there will be our CEO.

00;32;07;01 - 00;32;09;23

And they just like take over

more and more human roles.

00;32;09;23 - 00;32;12;01

Right. But

00;32;12;04 - 00;32;13;20

what Yoshua Bengio

00;32;13;20 - 00;32;19;07

has created is he's it, again,

another godfather of AI early, early,

00;32;19;07 - 00;32;25;02

pioneer of these models, huge

fan of OpenAI, a huge fan of AI until he's

00;32;25;05 - 00;32;29;17

more recently looked at OpenAI,

look at the ChatGPT and

00;32;29;20 - 00;32;33;20

and realized that he's devoted his life

to something that may kill humanity.

00;32;33;20 - 00;32;38;19

So he took a huge u turn,

created something called, scientist AI.

00;32;38;22 - 00;32;42;02

And,

have you had Yoshua on the show yet? Or.

00;32;42;09 - 00;32;45;13

I've heard, you know his interviews

that you did with him on, on your program.

00;32;45;13 - 00;32;46;16

But we haven't had him on your show.

00;32;46;16 - 00;32;50;09

So why don't you

you know if you'll indulge me, you know,

00;32;50;12 - 00;32;53;04

tell us a little bit about his position

and what he's proposing.

00;32;53;04 - 00;32;54;18

No, I appreciate the chance I'm in.

00;32;54;18 - 00;32;58;21

You're indulging me because, you know,

hopefully he'll come on the show soon and

00;32;58;21 - 00;33;02;08

and and say all this from

the from the horse's mouth and

00;33;02;11 - 00;33;05;11

and not, you know,

have to deal with the poor, you know,

00;33;05;14 - 00;33;10;10

the middle man

here, but basically, Yoshua Bengio,

00;33;10;13 - 00;33;11;16

he has created

00;33;11;16 - 00;33;15;11

this thing called scientist

AI, and scientist

00;33;15;11 - 00;33;20;07

AI is, as he says, it's

like an ideal scientist or psychologist.

00;33;20;07 - 00;33;24;08

So its job is to understand and to explain

00;33;24;11 - 00;33;28;05

and to predict,

but not to act on its own goals.

00;33;28;08 - 00;33;30;25

So it is non-genetic.

00;33;30;25 - 00;33;33;18

It's an unknown genetic model,

meaning it does not have its own long term

00;33;33;18 - 00;33;36;26

goals,

that it's trying to achieve in the world.

00;33;36;26 - 00;33;39;07

It is probabilistic and cautious.

00;33;39;07 - 00;33;44;02

So, for example, unlike if, you know,

if you've interacted with like Claude or,

00;33;44;04 - 00;33;48;18

or ChatGPT, it doesn't kind of bombastic

think it knows every answer

00;33;48;23 - 00;33;53;29

and act like this overconfident

kind of, know it all.

00;33;54;02 - 00;33;56;20

Rather,

it will have a probabilistic assessment.

00;33;56;20 - 00;34;01;15

It'll say like, well, there's a 3%,

3% chance that this plan leads to

00;34;01;18 - 00;34;05;04

this outcome or this other outcome,

and it will tell you you're wrong, which

00;34;05;04 - 00;34;10;08

which a lot of times, the other models

are not designed to do so.

00;34;10;08 - 00;34;12;12

It's not trained to persuade you.

00;34;12;12 - 00;34;14;00

It's not trained to please you.

00;34;14;00 - 00;34;16;20

It's it's supposed to be honest

and calibrated.

00;34;16;20 - 00;34;20;04

And most importantly,

and this is his vision.

00;34;20;07 - 00;34;23;27

It is supposed to be or it's

hopefully his plan is

00;34;23;27 - 00;34;30;01

that it might be used as a guardrail

for other, agents, a genetic AI.

00;34;30;01 - 00;34;33;17

So, for example,

you could run a powerful AI agent

00;34;33;17 - 00;34;39;05

through scientist AI,

and it will evaluate the safety

00;34;39;08 - 00;34;42;08

of their

proposed actions and can veto them.

00;34;42;12 - 00;34;46;09

So he's got, Yoshua

Bengio has this, thing called Law Zero,

00;34;46;12 - 00;34;48;14

which, he's talked about,

but a lot of zeros,

00;34;48;14 - 00;34;51;14

but essentially,

a different approach to regulation.

00;34;51;14 - 00;34;54;20

It's not saying, okay, we we're just going

to regulate these companies

00;34;54;20 - 00;34;57;20

and ask them to follow certain benchmarks

that you will.

00;34;57;22 - 00;34;58;29

They are we don't understand.

00;34;58;29 - 00;35;03;03

We're going to use

AI to regulate the AI, essentially use

00;35;03;03 - 00;35;07;26

a technological solution

to, to this kind of firm.

00;35;07;26 - 00;35;10;18

Yeah, to this kind of, safety approach.

00;35;10;18 - 00;35;15;09

And what what

Bengio is fundamentally worried about

00;35;15;09 - 00;35;19;06

is the, the, the very direction

00;35;19;06 - 00;35;23;07

that the industry is heading,

which is a genetic AI.

00;35;23;10 - 00;35;26;01

He doesn't even necessarily

talk about the dangers of,

00;35;26;01 - 00;35;29;01

might say superintelligence or,

00;35;29;02 - 00;35;32;13

you know, AGI that's a term

that gets thrown out a lot about.

00;35;32;13 - 00;35;35;23

But he, which means that or

00;35;35;23 - 00;35;39;07

artificial

general intelligence is smart as a human.

00;35;39;10 - 00;35;41;21

He just says, well, as soon as something

00;35;41;21 - 00;35;45;01

is a genetic meaning, it can help people

00;35;45;04 - 00;35;50;05

design a plan or it can manipulate humans

00;35;50;05 - 00;35;53;27

or institutions to, to achieve its ends,

00;35;54;00 - 00;35;59;02

or it can resist being shut down or it

can, you know, cause harm.

00;35;59;02 - 00;36;02;26

And we've seen the models

do all these things already.

00;36;02;28 - 00;36;06;17

That means we should we should not

be modeling AI off of humans.

00;36;06;20 - 00;36;09;20

We should not be modeling them

off of agency.

00;36;09;27 - 00;36;11;15

That's what not only humans.

00;36;11;15 - 00;36;14;05

Actually, every life form on the planet

has some degree of agency.

00;36;14;05 - 00;36;16;26

That's what kind of defines life.

00;36;16;29 - 00;36;19;26

Artificial intelligence

does not need to be a genetic.

00;36;19;26 - 00;36;23;09

It can just be a very helpful, very smart,

00;36;23;12 - 00;36;27;28

very, perceptive tool.

00;36;28;01 - 00;36;30;04

And thus we get away from deception.

00;36;30;04 - 00;36;34;08

We get away from manipulation because

it will have any agency of its own.

00;36;34;11 - 00;36;36;29

That's not at all

where the industry is headed.

00;36;36;29 - 00;36;41;03

But I think it's important to know

that there's an alternative out there.

00;36;41;06 - 00;36;43;11

Yeah,

and it's a it's a compelling alternative.

00;36;43;11 - 00;36;46;23

And certainly for us,

you know, as a species,

00;36;46;23 - 00;36;50;02

if we think about what's best for us, I.

00;36;50;05 - 00;36;52;09

I like that vision.

00;36;52;09 - 00;36;56;23

The. You know, the concern I have is

it seems like, if anything,

00;36;56;26 - 00;37;00;13

these models are getting more kind

of fractured and fragmented,

00;37;00;13 - 00;37;02;12

like as we've seen more open source AI.

00;37;02;12 - 00;37;05;20

You know, even if we get to things

like Deep Seek and some of these models

00;37;05;20 - 00;37;10;00

outside of,

you know, the US, I don't know, like

00;37;10;03 - 00;37;14;06

have we crossed like a Rubicon here

in terms of the ability to control these?

00;37;14;10 - 00;37;20;19

Like how do we I don't know, it kind of

feels like the the cat is out of the bag.

00;37;20;22 - 00;37;21;16

Yeah, it's a good point.

00;37;21;16 - 00;37;25;22

I mean, this is my main

beef personally with,

00;37;25;25 - 00;37;27;29

with with folks,

00;37;27;29 - 00;37;31;26

with the drill dumber camp

that folks like Lars Yudkowsky.

00;37;31;29 - 00;37;35;01

Because I just don't understand.

00;37;35;05 - 00;37;39;15

Maybe I'm just not smart enough

to understand, but I don't understand why,

00;37;39;18 - 00;37;44;08

we can't put the cat back in the bag

a little bit.

00;37;44;08 - 00;37;45;04

You know why?

00;37;45;04 - 00;37;51;00

Why human institutions can't rally

to create the right regulations.

00;37;51;03 - 00;37;54;15

And to sandbox

00;37;54;15 - 00;37;58;00

new models, for example,

until they're truly ready.

00;37;58;03 - 00;38;00;14

There are things that can be done.

00;38;00;14 - 00;38;01;25

They're not easy.

00;38;01;25 - 00;38;06;16

It would take a lot of,

a lot of societal will, but,

00;38;06;20 - 00;38;12;12

I think it's, I think it's not the time

to feel despair and think, gosh,

00;38;12;16 - 00;38;16;07

we've already kind of passed

over some crucial threshold.

00;38;16;07 - 00;38;20;23

I mean, in some sense, we've we passed

that when ChatGPT was first released.

00;38;20;23 - 00;38;23;00

Or you could say we passed that.

00;38;23;00 - 00;38;27;03

I mean, the Turing test has long

been kind of

00;38;27;06 - 00;38;28;11

I mean, it's it's it's

00;38;28;11 - 00;38;30;11

there's arguments about

whether the Turing test has been passed,

00;38;30;11 - 00;38;33;24

but certainly we've we've crossed

some sort of incredibly important line.

00;38;33;27 - 00;38;37;13

I think to though, you know,

what it gets at for me, I

00;38;37;14 - 00;38;41;17

one of the things about working on

this series that really taught me is the,

00;38;41;20 - 00;38;42;10

the importance of

00;38;42;10 - 00;38;46;06

storytelling and imagination

in this technology,

00;38;46;09 - 00;38;49;02

and that goes all the way back

to Alan Turing,

00;38;49;02 - 00;38;52;03

who and I didn't really understand this

because I understood the Turing test

00;38;52;10 - 00;38;53;22

as a kind of benchmark,

00;38;53;22 - 00;38;57;20

like this would be a benchmark of human,

of sorry, of machine progress.

00;38;57;20 - 00;39;00;17

You know it

once, once the Turing test was passed.

00;39;00;17 - 00;39;03;17

So, for example,

if I could chat with a machine

00;39;03;17 - 00;39;08;27

and not know it was a machine, then, well,

it's it's achieved some sort of,

00;39;09;00 - 00;39;10;22

some sort of, milestone.

00;39;10;22 - 00;39;14;11

And that was the Turing test to,

what do you call it, the Imitation Game.

00;39;14;14 - 00;39;19;13

But, in fact,

what Turing was doing all the way back

00;39;19;17 - 00;39;25;12

in World War Two and right postwar,

when he was introducing this idea,

00;39;25;15 - 00;39;26;05

was not just

00;39;26;05 - 00;39;29;09

saying, okay, this is a benchmark

for machines to pass.

00;39;29;09 - 00;39;31;27

And once machines passed that,

we can say that they're on their way

00;39;31;27 - 00;39;33;15

to really being thinking machines.

00;39;33;15 - 00;39;39;08

He said that, but he was also taking

what was at the time a really complicated,

00;39;39;11 - 00;39;43;29

philosophical debate about,

well, can machines ever think.

00;39;44;02 - 00;39;47;28

And he and he treated it like an engineer

and he said, you know what?

00;39;47;28 - 00;39;53;09

We just need to create, and,

an observable,

00;39;53;12 - 00;39;57;00

metric by which we can say

that they're thinking and, and that's,

00;39;57;01 - 00;40;00;16

and we don't have to deal

with the philosophical,

00;40;00;19 - 00;40;04;26

you know, discomfort of saying,

well, can machines thinking,

00;40;04;26 - 00;40;07;20

what would that mean

if they are thinking, etc.,

00;40;07;20 - 00;40;09;06

if, if they pass the Turing test

00;40;09;06 - 00;40;11;12

and then then they're on their way

to thinking machines.

00;40;11;12 - 00;40;15;04

And by doing that, he not only sort of

00;40;15;07 - 00;40;18;06

freed engineers

from the philosophical angst

00;40;18;06 - 00;40;21;28

and set them a kind of path to to follow,

which they certainly follow it.

00;40;21;28 - 00;40;26;22

And it kind of leads

to ChatGPT today, but also,

00;40;26;25 - 00;40;27;22

I think, created

00;40;27;22 - 00;40;31;12

this, this, this, this new way.

00;40;31;19 - 00;40;35;05

He kind of he kind of realized

something very important, which is that

00;40;35;05 - 00;40;38;28

we would not recognize machines

as thinking

00;40;39;01 - 00;40;44;02

until we started interacting with them

in a humanlike way,

00;40;44;05 - 00;40;47;18

and that when they started using language

00;40;47;21 - 00;40;51;23

and, and talking back to us,

00;40;51;26 - 00;40;56;15

that's when we would see them

recognize their thinking.

00;40;56;18 - 00;40;59;17

And, you could get very philosophical

about this.

00;40;59;17 - 00;41;02;19

You could say that you know, trees

do a lot of thinking,

00;41;02;21 - 00;41;07;00

but we don't think of them as thinking,

there's a lot of other

00;41;07;00 - 00;41;10;00

living things in this planet

that, that think,

00;41;10;03 - 00;41;12;21

but we

00;41;12;21 - 00;41;15;08

their intelligence

is don't interact with our own.

00;41;15;08 - 00;41;18;07

And so we're not really

that concerned about them or many of us.

00;41;18;07 - 00;41;19;09

Right.

00;41;19;12 - 00;41;21;24

And so what Turing felt was

00;41;21;24 - 00;41;27;20

in order for us to sort of really respect

and use this word respect machines,

00;41;27;24 - 00;41;31;10

they would need to sort of, interact

with us

00;41;31;13 - 00;41;36;07

like this, in the way they were, the way

ChatGPT interacts with us.

00;41;36;10 - 00;41;39;12

But the danger of that, right, is that we

00;41;39;12 - 00;41;41;16

then don't see this gets back to something

we talked about.

00;41;41;16 - 00;41;45;02

We don't see the alien,

this alien ness of it.

00;41;45;05 - 00;41;49;05

And we start to think,

we start to interact with it,

00;41;49;05 - 00;41;53;12

maybe too much like a person

or like a fellow,

00;41;53;15 - 00;41;58;07

human or a human like entity,

a human thinking like entity.

00;41;58;07 - 00;42;03;21

And thus we make very important

cognitive mistakes in interacting with it.

00;42;03;24 - 00;42;08;07

And we perhaps

trust it or distrust it in the wrong ways.

00;42;08;10 - 00;42;10;18

So and this also happened in

00;42;10;18 - 00;42;13;18

the sort of this failure of imagination

then or this,

00;42;13;23 - 00;42;17;25

this, this kind of this way in which

our imagination is challenge channeled.

00;42;17;28 - 00;42;21;12

We don't see how the models,

00;42;21;15 - 00;42;23;28

are vastly different year

00;42;23;28 - 00;42;27;08

over year because we're interacting

with a model right now.

00;42;27;08 - 00;42;29;05

It's interacting with us.

00;42;29;05 - 00;42;31;08

Yes, like the fastest human

we've ever talked to.

00;42;31;08 - 00;42;35;27

But it's still recognizable

in its thinking in some ways.

00;42;35;27 - 00;42;40;08

And it's it's

something we can respect but recognize.

00;42;40;11 - 00;42;44;28

And so it's very, very difficult for us to

then think, okay, wait a second.

00;42;44;28 - 00;42;47;16

This is just the current iteration.

00;42;47;16 - 00;42;50;05

We have to imagine

00;42;50;05 - 00;42;53;20

a different kind of intelligence

that this could grow into.

00;42;53;23 - 00;42;57;12

And what would I be in that,

00;42;57;15 - 00;43;00;05

in that situation,

if I could say one more thing, it's

00;43;00;05 - 00;43;04;01

it reminds me, honestly, of being,

you know, reporting in Ukraine

00;43;04;04 - 00;43;07;04

or reporting in Afghanistan

or reporting in South Sudan,

00;43;07;04 - 00;43;11;03

and you talk to people

who are in the middle of a war

00;43;11;06 - 00;43;12;19

and they say, you know, we

00;43;12;19 - 00;43;16;18

knew the war was coming,

but we just didn't

00;43;16;21 - 00;43;21;13

imagine what it would feel like to be us

when it was here.

00;43;21;16 - 00;43;23;29

And they want me to know.

00;43;23;29 - 00;43;24;20

You don't understand.

00;43;24;20 - 00;43;28;00

I was just planning my daughter's

wedding in that building over there,

00;43;28;00 - 00;43;29;26

which is now like a bombed out

whole, like.

00;43;29;26 - 00;43;32;07

And they still see.

00;43;32;07 - 00;43;34;24

They still see the,

the place that they were planning the way

00;43;34;24 - 00;43;38;21

and they still are thinking and frustrated

about the, the money they spent on,

00;43;38;21 - 00;43;40;01

the wedding invitations or something.

00;43;40;01 - 00;43;43;08

You know, they haven't quite transitioned

over from the old world

00;43;43;08 - 00;43;46;08

to the new and,

00;43;46;10 - 00;43;48;15

I don't know,

I don't want to make this sound

00;43;48;15 - 00;43;52;08

like a Duma forecast because I think

the future could be quite bright.

00;43;52;08 - 00;43;57;07

But it does take an active imagination

whether we think we're headed toward,

00;43;57;10 - 00;44;00;24

you know, any of these kind of versions

of the future

00;44;00;27 - 00;44;04;07

to put ourselves

in the new version of the future?

00;44;04;10 - 00;44;08;10

And to sort of play with,

you know, our, our imagination

00;44;08;10 - 00;44;10;19

and to imagine that the, you know,

the world is not going to be the same.

00;44;10;19 - 00;44;13;07

It is is as it is now.

00;44;13;10 - 00;44;15;20

It's so, so thank you for that.

00;44;15;20 - 00;44;18;05

that was a really interesting

answer that covered a lot Yeah.

00;44;18;05 - 00;44;19;14

So so I have.

00;44;19;14 - 00;44;22;23

so I can, you know there's a lot of things

we could talk about Sure.

00;44;22;26 - 00;44;26;23

But one of the pieces

I want to talk about is,

00;44;26;26 - 00;44;30;03

you know, you got me thinking that there's

all this conversation about the future

00;44;30;03 - 00;44;31;22

and what will happen in the next model.

00;44;31;22 - 00;44;34;20

And we talked about this before, but,

you know,

00;44;34;20 - 00;44;37;20

the sense that the future is already here

or, you know, the Arthur C

00;44;37;20 - 00;44;41;02

Clarke quote about the future is

here are not, you know, evenly distributed

00;44;41;09 - 00;44;44;02

and you know, how many people out there

right now

00;44;44;02 - 00;44;48;08

are interacting with chat bots

or with this technology

00;44;48;08 - 00;44;53;22

in a way that would have been completely

unimaginable to people, Write

00;44;53;25 - 00;44;54;20

right.

00;44;54;20 - 00;44;57;25

to, you know, what I think

00;44;57;25 - 00;45;01;20

has been built

into the design of these tools is,

00;45;01;23 - 00;45;04;23

human engagement

00;45;04;29 - 00;45;08;01

as a design principle,

if I can call it that.

00;45;08;01 - 00;45;08;06

Right.

00;45;08;06 - 00;45;11;03

Like if you go back to the Turing test,

it's, you know, what is it?

00;45;11;03 - 00;45;11;25

It's it's

00;45;11;25 - 00;45;16;10

the fact that thinking for us is measured

in terms of interaction with us.

00;45;16;15 - 00;45;18;29

And that's exactly how these things

have been designed.

00;45;18;29 - 00;45;23;00

They've been designed to,

you know, flatter, to create engagement,

00;45;23;00 - 00;45;26;10

to let people's guard down,

to continuously engage.

00;45;26;10 - 00;45;26;18

Right.

00;45;26;18 - 00;45;29;09

Like, you know, one of the things

I've noticed that ChatGPT does

00;45;29;09 - 00;45;33;03

is it always prompts you for like, hey,

how can we keep the conversation going?

00;45;33;03 - 00;45;34;26

What do you need next from me? Right?

00;45;34;26 - 00;45;38;15

It's almost like a

a meta fixation or a social media

00;45;38;15 - 00;45;42;20

fixation, of this

and where does that take us?

00;45;42;20 - 00;45;43;10

You know,

00;45;43;10 - 00;45;47;07

we're we're having a conversation

out of one side of our mouths about

00;45;47;10 - 00;45;51;16

we need to be more cautious

about the alien nature of this.

00;45;51;19 - 00;45;52;06

it kind of

00;45;52;06 - 00;45;55;26

feels like we're watching

that battle be lost.

00;45;55;29 - 00;45;58;01

And is that,

00;45;58;01 - 00;45;59;17

I don't know, I guess.

00;45;59;17 - 00;46;02;18

Let me let me kind of

frame the discussion up this way.

00;46;02;18 - 00;46;07;02

If we're if we're worried about where

that future direction is taking us,

00;46;07;05 - 00;46;12;14

you know, do we have an obligation

to push the technologists

00;46;12;14 - 00;46;17;09

and the owners of these tools to put more,

put more guardrails

00;46;17;09 - 00;46;21;01

and principles in place

that prevent people from,

00;46;21;04 - 00;46;25;05

you know, becoming enamored, let's say,

at the extreme end, with these tools?

00;46;25;08 - 00;46;28;10

Or is it more purely on the demand side?

00;46;28;10 - 00;46;31;24

And we just have to do a better job

of educating these people

00;46;32;00 - 00;46;35;11

about, you know, what

they're signing themselves up for?

00;46;35;14 - 00;46;39;01

I that's, that's I think the key question

to ask, it's such an important question.

00;46;39;01 - 00;46;42;10

I think that well a couple quick things.

00;46;42;10 - 00;46;46;19

One is that the metaphors

we use to describe this do matter.

00;46;46;22 - 00;46;51;08

And even if, even if,

even if we're sort of taking

00;46;51;08 - 00;46;54;16

the perspective of a business leader

saying, okay, let's what's practical here?

00;46;54;16 - 00;46;55;26

How can I use this to cut costs?

00;46;55;26 - 00;46;58;16

How can I use this to maximize efficiency,

compete?

00;46;58;16 - 00;47;02;13

Even still, our employees

will be narrative izing this.

00;47;02;13 - 00;47;07;22

This technology and in interacting with it

in a way that, as you say, that,

00;47;07;25 - 00;47;11;11

that kind of pulls the trigger

on our very relational intelligence

00;47;11;11 - 00;47;16;22

and our sense of self, which is so based

on how we interact with others.

00;47;16;22 - 00;47;19;24

So if we're interacting with with the AI

and that's going away, then

00;47;19;24 - 00;47;23;02

then our sense of self is, is disrupted,

is, is is affected.

00;47;23;02 - 00;47;28;29

And even if,

if the AI doesn't intend that or

00;47;28;29 - 00;47;33;10

isn't trying to quote unquote manipulate

us, there's a French philosopher,

00;47;33;13 - 00;47;36;10

Catherine

Evans, who introduced me to a concept

00;47;36;10 - 00;47;40;21

I, I felt it was quite I don't think

she's published about this yet, but

00;47;40;24 - 00;47;43;18

she said, you know,

I guess you're probably familiar

00;47;43;18 - 00;47;47;13

with there's all kinds of UN rules

about not anthropomorphizing intelligence,

00;47;47;13 - 00;47;49;20

not anthropomorphizing in technology.

00;47;49;20 - 00;47;52;03

And these go back some years.

00;47;52;03 - 00;47;56;26

So she was, creating a, I think, a

comic book for kids about AI.

00;47;56;26 - 00;48;00;23

But she ended up stumbling into this idea

of, since she wasn't allowed to create

00;48;00;23 - 00;48;04;00

AI as a is it,

she wasn't allowed to anthropomorphize

00;48;04;00 - 00;48;06;04

the AI because of UN rules.

She was doing it for them.

00;48;06;04 - 00;48;08;12

And also she wasn't

allowed to have any antagonist.

00;48;08;12 - 00;48;12;14

So the worst, kind of narrative

situation, you know, no conflict.

00;48;12;19 - 00;48;15;10

No, no people. What do you tell a story?

00;48;15;10 - 00;48;19;16

But she ended up coming up with

this idea of AI as a place

00;48;19;19 - 00;48;20;17

and, you

00;48;20;17 - 00;48;23;17

know, in the in the cartoon

or in the graphic novel, the YouTube

00;48;23;17 - 00;48;27;22

content algorithm is a sort of,

a place, and it's, you know,

00;48;27;25 - 00;48;32;19

you lead it with a map and it leads

you down different, different, different,

00;48;32;22 - 00;48;36;27

different recommendation, portals, but

00;48;37;00 - 00;48;40;23

I found in my interaction

with the different models

00;48;40;26 - 00;48;44;07

and again,

this is AI is not just about chat bots.

00;48;44;07 - 00;48;48;27

We don't always feel

like that's important to note that

00;48;49;00 - 00;48;52;00

it is kind of useful,

I think, to think about,

00;48;52;06 - 00;48;55;14

it is a sense of place,

if only because it gets it.

00;48;55;14 - 00;49;01;01

What you're talking about,

which is how are the norms and culture

00;49;01;04 - 00;49;04;04

and cultural expectations of this place

a little bit different?

00;49;04;10 - 00;49;07;06

You know, how do I behave with the model?

00;49;07;06 - 00;49;12;05

That's not quite the way I would

that I was going to behave in person.

00;49;12;05 - 00;49;13;09

We all deal with this, right?

00;49;13;09 - 00;49;16;19

And social media behavior is different

than we are in person with each other.

00;49;16;26 - 00;49;18;23

And so I think if we think of AI

in that same way,

00;49;18;23 - 00;49;23;16

because it is programed exactly

as you said, it is programed to be helpful

00;49;23;19 - 00;49;29;13

to be solicitous to, to prove its value.

00;49;29;16 - 00;49;32;09

You know how 9000 in Stanley Kubrick's

00;49;32;09 - 00;49;37;11

fantastic film is constantly talking

about how foolproof it is

00;49;37;14 - 00;49;39;19

before it murders the entire crew?

00;49;39;19 - 00;49;43;16

And and that's a sort of an important part

is that these the

00;49;43;20 - 00;49;48;08

the models are advertising themselves

to us,

00;49;48;11 - 00;49;51;01

much like an intern

that wants to keep its job and get

00;49;51;01 - 00;49;54;06

promoted,

will be constantly promoting itself.

00;49;54;09 - 00;49;55;23

And we like to think that

00;49;55;23 - 00;49;59;18

that's kind of kind of helpful

and that's quite culturally appropriate.

00;49;59;18 - 00;50;03;23

Certainly we we want it to tell us

other things it can do.

00;50;03;26 - 00;50;06;03

There's nothing wrong with that, per se,

00;50;06;03 - 00;50;08;27

but it is a is

it is a kind of different world.

00;50;08;27 - 00;50;11;19

It's a different place that we step into.

00;50;11;19 - 00;50;12;21

And,

00;50;12;21 - 00;50;16;17

yeah, I mean, I think barring whether it's

the supply side or demand side,

00;50;16;17 - 00;50;18;06

I think maybe both

solutions are important.

00;50;18;06 - 00;50;21;02

But it's also about the metaphors

we we use.

00;50;21;02 - 00;50;21;21

Yeah.

00;50;21;21 - 00;50;22;21

Well, and the,

00;50;22;21 - 00;50;24;19

you know, on the demand side

that the reason I ask this

00;50;24;19 - 00;50;28;02

and it's something that I've been

brushing up more and more against

00;50;28;02 - 00;50;32;29

is this notion

that it is providing a value to people.

00;50;33;06 - 00;50;33;15

Right.

00;50;33;15 - 00;50;37;16

And I'll use the specific example because

it's, it's one of the most intimate.

00;50;37;16 - 00;50;39;22

It's I don't know

if it'll be the most intimate for long,

00;50;39;22 - 00;50;42;10

but one of the more intimate ways

people are using AI,

00;50;42;10 - 00;50;45;05

or he's using chat

bots is as a therapist.

00;50;45;05 - 00;50;45;14

Right.

00;50;45;14 - 00;50;50;08

And it becomes a way to process,

you know, whether it's trauma

00;50;50;08 - 00;50;54;12

or conflict, it's a way

to have an intimate relationship.

00;50;54;15 - 00;50;57;23

And, I don't know,

maybe even some ways get

00;50;57;26 - 00;51;02;06

a better sense of self

or a better sense of purpose.

00;51;02;09 - 00;51;04;20

And it's working, right.

00;51;04;20 - 00;51;07;22

Like I've talked to enough people,

some of them AI experts, some of them,

00;51;07;27 - 00;51;11;07

you know, friends who say, well,

I don't have a therapist.

00;51;11;07 - 00;51;13;13

And this gets me one and it's useful.

00;51;13;13 - 00;51;15;04

Or I do have a therapist.

And you know what?

00;51;15;04 - 00;51;17;06

AI is better than my therapist.

00;51;17;06 - 00;51;20;06

And yeah, I don't know, like,

00;51;20;09 - 00;51;23;09

I just think about where this is going

00;51;23;09 - 00;51;27;01

and what happens

when you've created a technology

00;51;27;04 - 00;51;30;18

that does provide this service

and that people start saying,

00;51;30;18 - 00;51;34;05

well, this is actually better for me

than my human relationships.

00;51;34;12 - 00;51;36;06

Like does that take us?

00;51;36;06 - 00;51;38;22

And what do we do with that?

00;51;38;22 - 00;51;41;22

And, you know,

what are the implications for us

00;51;41;23 - 00;51;45;03

as a society where historically,

if you wanted a human relationship,

00;51;45;03 - 00;51;46;17

you had to have it with a human,

00;51;46;17 - 00;51;51;19

and that was a propagating force

for the continuation of the human race.

00;51;51;22 - 00;51;52;00

Yeah.

00;51;52;00 - 00;51;52;29

No, no, I think you're right.

00;51;52;29 - 00;51;58;09

I mean, it's hard to parse out what is I

00;51;58;09 - 00;52;01;15

mean, morality shifts

through across generations, right?

00;52;01;15 - 00;52;02;20

We know this.

00;52;02;20 - 00;52;05;01

And, standards change.

00;52;05;01 - 00;52;10;27

I think Esther Perel has made

a very important point that therapy,

00;52;11;00 - 00;52;13;24

I therapy is thin.

00;52;13;24 - 00;52;18;10

It's a thin kind of therapy,

which is an interesting way of approaching

00;52;18;10 - 00;52;22;07

this, which is to say that it's not a

00;52;22;10 - 00;52;24;21

a challenging kind of therapy.

00;52;24;21 - 00;52;27;21

It's not a one that's whole bodied.

00;52;27;21 - 00;52;31;28

it's thinner, and it

she feels that it then

00;52;32;01 - 00;52;35;08

maybe leads people to have thinner

or have have lower

00;52;35;08 - 00;52;38;12

expectations of human relationships.

00;52;38;12 - 00;52;41;20

So it sort of lowers

the standard, as it were.

00;52;41;23 - 00;52;45;09

I think that's a, that's a yeah,

that's a concern.

00;52;45;15 - 00;52;49;18

At the same time, though, I'm

a little bit worried about saying

00;52;49;18 - 00;52;52;18

that I'm, I'm worried

about saying I'm worried where,

00;52;52;18 - 00;52;57;17

where society is going just because

there are so many different stories.

00;52;57;17 - 00;53;01;04

I mean, there's so many people

that don't have access to therapy at all.

00;53;01;07 - 00;53;04;19

And, so many people for whom

00;53;04;19 - 00;53;09;01

I think, a first chat with ChatGPT

00;53;09;04 - 00;53;14;20

or any other model might be the gateway

to a certain of self-understanding.

00;53;14;23 - 00;53;17;25

It's not like everybody's in a situation

where, oh, should I,

00;53;18;01 - 00;53;20;01

should I call my therapist

or should I know?

00;53;20;01 - 00;53;22;21

Perhaps they don't have health insurance.

Perhaps they don't have that access.

00;53;22;21 - 00;53;27;18

So I just think it's hard to generalize

and say we are going anywhere.

00;53;27;18 - 00;53;33;10

It goes back to the quote you said, which

is the future is unevenly distributed.

00;53;33;13 - 00;53;33;19

Yeah.

00;53;33;19 - 00;53;38;07

It's very it's very difficult

to know, the some effect of this,

00;53;38;10 - 00;53;43;26

on, on human society,

but perhaps it does it certainly enters

00;53;43;26 - 00;53;47;16

it certainly adds a different kinds

of expectation of perhaps a lower one.

00;53;47;19 - 00;53;48;21

To change gears slightly.

00;53;48;21 - 00;53;51;14

I wanted to come back to something

you said earlier about,

00;53;51;14 - 00;53;53;06

you know, the culture of AI.

00;53;53;06 - 00;53;56;13

And, you know,

culture is being kind of a component here.

00;53;56;13 - 00;54;00;05

And, you know, that being dynamic

and looking different in different places.

00;54;00;11 - 00;54;03;11

You know, you you created the

00;54;03;12 - 00;54;07;09

The Rough Translation podcast and,

you know, you had a sub series in there

00;54;07;09 - 00;54;10;23

about at work and you looked at work

across different cultures.

00;54;10;23 - 00;54;11;18

And I'm curious,

00;54;11;18 - 00;54;14;16

you know, whether we're talking about AI

or whether we're not

00;54;14;16 - 00;54;17;16

if we're if we're talking

about the future of work,

00;54;17;19 - 00;54;22;07

how are you seeing people's relationship

with work change?

00;54;22;10 - 00;54;26;28

And, you know, as you did that series,

you know,

00;54;27;01 - 00;54;30;05

did did you see market differences

across cultures, or was there

00;54;30;05 - 00;54;34;15

kind of a common core of

the way people approach work?

00;54;34;18 - 00;54;36;05

No. Thank you for that question.

00;54;36;05 - 00;54;39;21

Yeah.

00;54;39;22 - 00;54;42;18

You know, just to highlight maybe two

things I learned from that series.

00;54;42;18 - 00;54;45;10

So one. We had one,

00;54;45;10 - 00;54;46;00

one show in

00;54;46;00 - 00;54;49;22

I think it was, it might have been called

failure is a four letter word.

00;54;49;25 - 00;54;53;16

That was maybe that was a throw

in a title, but it was about how this,

00;54;53;19 - 00;54;57;25

this concept of fail fast.

00;54;57;28 - 00;55;00;15

Which was

we think of associated with Silicon Valley

00;55;00;15 - 00;55;06;09

just does not translate well into other

and to, to everywhere in the world.

00;55;06;09 - 00;55;09;25

You know, we, talked to somebody

00;55;09;28 - 00;55;13;10

in Nigeria who said, you know,

what about fail slow?

00;55;13;11 - 00;55;15;18

Everything takes forever here.

00;55;15;18 - 00;55;17;10

Where, there's so much bureaucracy.

00;55;17;10 - 00;55;21;14

Or perhaps you live in a culture

or somebody else was from Mexico City.

00;55;21;14 - 00;55;25;10

They said, you know,

failing is such a taboo

00;55;25;13 - 00;55;29;16

that once you fail, you never want

to even show your face again.

00;55;29;17 - 00;55;33;01

So fail fast doesn't

00;55;33;04 - 00;55;34;09

isn't as accessible.

00;55;34;09 - 00;55;35;25

And yet, because we live in a digital,

00;55;35;25 - 00;55;39;18

globalized, culture,

fail fast, and Silicon Valley

00;55;39;18 - 00;55;43;24

and inspiring entrepreneurial stories

were very much part of the the water.

00;55;43;25 - 00;55;46;29

So one of the things that I wanted

to explore in that work series was,

00;55;46;29 - 00;55;50;18

how do you swear

00;55;50;21 - 00;55;56;08

what you're reading and seeing on YouTube

and inspired by with your own local,

00;55;56;11 - 00;55;59;11

constraints?

00;55;59;14 - 00;56;04;02

How do people translate

that entrepreneurial spirit?

00;56;04;05 - 00;56;08;14

You know, and the idea was not to say,

oh, well, well, it's

00;56;08;14 - 00;56;12;18

more difficult to be an entrepreneur

in Mexico City or Nigeria.

00;56;12;18 - 00;56;15;05

I mean, perhaps that's true, but

in some ways, actually that's not true.

00;56;15;05 - 00;56;16;14

In some ways the opposite is true.

00;56;16;14 - 00;56;19;19

So nothing's nothing's nothing's,

black and white.

00;56;19;22 - 00;56;23;14

But to me it was

it was about finding space to,

00;56;23;17 - 00;56;28;06

to give permission to people who may not

have swallowed that fail fast mantra,

00;56;28;09 - 00;56;28;29

who maybe feel

00;56;28;29 - 00;56;32;20

alienated by it to find a home in it.

00;56;32;23 - 00;56;36;03

You know,

how how do they find their own way to it,

00;56;36;03 - 00;56;40;24

which has always been my life's work,

honestly, as a journalist, is to try to,

00;56;40;27 - 00;56;45;09

focus on these mistranslations,

focus on the ways in which,

00;56;45;12 - 00;56;49;12

you know, some advice

or some or some or a piece of culture

00;56;49;12 - 00;56;55;24

might feel foreign or inaccessible,

but how can we find the commonalities?

00;56;55;24 - 00;56;59;08

How can we sort of all be,

00;56;59;11 - 00;57;01;15

participate

in some ways in the global economy,

00;57;01;15 - 00;57;03;12

even from our different cultural angles?

00;57;03;12 - 00;57;06;18

And what role does

where we're from affect what

00;57;06;18 - 00;57;10;10

we think of as good

or how we how we approach the question?

00;57;10;13 - 00;57;14;26

and then one more, example,

from that series and come around to

00;57;14;26 - 00;57;18;07

it is looking at, well, there's,

00;57;18;11 - 00;57;22;15

there's a particular law in, in Portugal.

00;57;22;16 - 00;57;27;01

I remember at the time being reading about

it, it said that if your boss calls you

00;57;27;04 - 00;57;29;24

after hours

or the boss is not allowed to call

00;57;29;24 - 00;57;34;01

you or email you after hours or otherwise,

they would get a $10,000 fine.

00;57;34;04 - 00;57;35;05

Or 10,000.

00;57;35;05 - 00;57;37;03

€10,000. Fine. Sorry.

00;57;37;03 - 00;57;39;18

And what we discovered

00;57;39;18 - 00;57;43;07

was that this law,

which seemed to be quite a,

00;57;43;10 - 00;57;43;19

you know,

00;57;43;19 - 00;57;47;21

lovely law, as if you're a worker,

suddenly you don't get called by your boss

00;57;47;24 - 00;57;52;29

was actually quite a cynical ploy

by the Minister of Labor would form

00;57;52;29 - 00;57;55;23

a formally was the Minister of Tourism

00;57;55;23 - 00;57;58;28

to sort of push Portugal

as a place for work life balance

00;57;58;28 - 00;58;04;13

and because she knew that nobody

who was coming to Portugal

00;58;04;13 - 00;58;06;07

would ever kind of be affected

by that law,

00;58;06;07 - 00;58;10;25

I was fine to just kind of pass this law

and to create this illusion of difference,

00;58;10;25 - 00;58;15;12

this illusion that Portugal

was this space that we could go to.

00;58;15;15 - 00;58;21;04

That was a that was a,

that that had their stuff in order.

00;58;21;04 - 00;58;23;14

They figured out something

that, that the, for example,

00;58;23;14 - 00;58;26;26

the US companies hadn't figured out,

which is work life balance.

00;58;26;29 - 00;58;32;20

And so the cynicism, though, that

then created then for people in Portugal

00;58;32;23 - 00;58;34;12

was, was quite profound

00;58;34;12 - 00;58;38;15

because they felt that these laws were

then just created for the outsiders.

00;58;38;18 - 00;58;43;01

And there was a second kind of there was,

it weren't created for them.

00;58;43;04 - 00;58;44;28

And so

00;58;44;28 - 00;58;46;03

what do we learn from these two stories?

00;58;46;03 - 00;58;52;13

Well, I think that we're living in a time

where, work is global, where,

00;58;52;16 - 00;58;56;03

work

advice, work people work across borders.

00;58;56;07 - 00;59;01;01

and we people have international teams,

and yet those teams are all dealing

00;59;01;01 - 00;59;05;17

with things that are not only because

of their own cultural,

00;59;05;20 - 00;59;09;26

point of view, but just because of the

the role of geography,

00;59;09;26 - 00;59;14;21

the role of,

the role of societal expectations.

00;59;14;24 - 00;59;18;27

That, that even though people are under

00;59;19;04 - 00;59;23;01

speaking of increasingly,

you know, global English,

00;59;23;04 - 00;59;26;04

these, these differences

really do matter.

00;59;26;07 - 00;59;29;23

And I think this directly translates

kind of the.

00;59;29;29 - 00;59;33;17

So so I would sum it up, sum

up the theory of the Future of work series

00;59;33;17 - 00;59;37;12

as, as an exploration into how specific

00;59;37;12 - 00;59;41;01

stories, into how even as teams are global

00;59;41;02 - 00;59;44;09

and and work is happening

more internationally.

00;59;44;12 - 00;59;47;25

The the fault lines, you know,

00;59;47;25 - 00;59;51;12

between cultures,

in, between societies really

00;59;51;18 - 00;59;55;01

are becoming even more exploited

and mean more

00;59;55;04 - 00;59;59;01

and are more painful to people

because they see the differences.

00;59;59;07 - 01;00;03;06

You know, it's it's

I saw this in my years

01;00;03;06 - 01;00;06;06

living in Nairobi,

going back and forth to East Africa,

01;00;06;11 - 01;00;10;16

how people were so much more aware of, of,

01;00;10;19 - 01;00;13;06

of the, of their status

01;00;13;06 - 01;00;18;00

in relationship to their age mates

in other places because of the internet

01;00;18;03 - 01;00;23;26

and, and so, so,

so what does this mean then, for, for AI?

01;00;23;29 - 01;00;24;08

And we

01;00;24;08 - 01;00;28;27

haven't talked about China,

but for me, the,

01;00;29;00 - 01;00;33;27

the, the, the kind of story

that's being told within China

01;00;34;00 - 01;00;37;19

about AI and the story

that China is telling, the world about AI

01;00;37;19 - 01;00;41;06

is so different,

and it's so important that we realize this

01;00;41;06 - 01;00;44;28

because we're in this battle

with with China, where AI companies

01;00;44;28 - 01;00;48;19

are in this battle with China, this race.

01;00;48;22 - 01;00;51;27

And yet

China has its own specific priorities

01;00;52;00 - 01;00;55;09

and its own kind of, narrative about AI

01;00;55;09 - 01;00;58;16

within, within, within the country

that affects the workers.

01;00;58;16 - 01;01;04;19

And, and I think will become increasingly

important as, as the AI race heats up.

01;01;04;22 - 01;01;06;13

Let's dive a little bit deeper into that.

01;01;06;13 - 01;01;09;21

So what what is the narrative there

within China.

01;01;09;21 - 01;01;11;17

And you know,

how is it the same or different from

01;01;11;17 - 01;01;14;03

what's being projected outwardly

as it comes to? I.

01;01;14;03 - 01;01;14;23

Sure, sure.

01;01;14;23 - 01;01;15;23

So, I mean, I think

01;01;15;23 - 01;01;20;09

basically, you know, for a long time

we were told that AI that Chinese firms

01;01;20;09 - 01;01;24;05

were more interested in sort of practical

a or applied AI, whereas

01;01;24;08 - 01;01;28;06

the US has had a more clear, stated

goal of superintelligence.

01;01;28;09 - 01;01;30;18

I think many people that I talked to

did not believe that.

01;01;30;18 - 01;01;36;01

And indeed, Alibaba became the first major

Chinese tech giant to openly discuss

01;01;36;01 - 01;01;39;23

artificial general intelligence

and even superintelligence. So,

01;01;39;26 - 01;01;41;22

we know that

01;01;41;22 - 01;01;47;01

China's has just as much ambition in the

in the area of superintelligence.

01;01;47;04 - 01;01;50;06

But importantly, the nationalist ambitions

01;01;50;06 - 01;01;54;20

and the technological ambitions are in

somewhat in

01;01;54;20 - 01;01;59;13

are in are in contradiction there

because or at least not always aligned.

01;01;59;17 - 01;02;04;13

I was talking to a safety researcher,

who had who had mentioned that

01;02;04;16 - 01;02;06;25

in, in China,

01;02;06;25 - 01;02;10;08

you know, red teaming,

this kind of safety testing of the models,

01;02;10;11 - 01;02;14;19

it's much more rigorous in Chinese

than it is in English.

01;02;14;22 - 01;02;17;22

So you could ask the, the model,

01;02;17;27 - 01;02;23;08

for instance, like moonshots, Kimi K2,

which apparently is it's thinking

01;02;23;08 - 01;02;28;15

is outperformed OpenAI's church GPT five,

as well as anthropic squads latest model.

01;02;28;15 - 01;02;35;09

So okay, so you take this model

and you can get it to do things in English

01;02;35;15 - 01;02;42;02

that it would not do in, in, in Mandarin

Chinese or in Chinese language. So,

01;02;42;05 - 01;02;45;02

what

01;02;45;02 - 01;02;50;05

that means is course, is that,

the red teaming is,

01;02;50;08 - 01;02;55;10

is, is very much,

about political control.

01;02;55;10 - 01;02;59;06

It's about making sure that people

are not asking questions of this

01;02;59;06 - 01;03;02;15

AI that will will destabilize or,

01;03;02;18 - 01;03;06;14

you know, work

against the Chinese government,

01;03;06;17 - 01;03;10;12

but it's not exactly the same thing

as a safe model.

01;03;10;15 - 01;03;14;08

and the whole way China got into the

AI race is, is,

01;03;14;08 - 01;03;18;05

I think, also instructive in this way,

where, where AlphaGo,

01;03;18;08 - 01;03;21;13

DeepMind, or deep secret,

01;03;21;13 - 01;03;27;12

did the, sorry, DeepMind, the,

the Demis Hassabis model, created,

01;03;27;15 - 01;03;30;25

created a go player in 2016

that then beat the Chinese

01;03;30;25 - 01;03;34;27

national champion is that's

when China became interested and in go

01;03;35;00 - 01;03;38;08

very interested in AI

when it kind of hit home.

01;03;38;13 - 01;03;41;23

You know it

hit a game that that China reveres

01;03;41;26 - 01;03;44;16

and is a is a is an ancient Chinese game.

01;03;44;16 - 01;03;47;24

So this has always been about,

01;03;47;27 - 01;03;52;10

I think, the turf war about,

what is happening within China,

01;03;52;10 - 01;03;55;29

how in terms of China's

concern of controlling its people,

01;03;55;29 - 01;04;00;04

China is concerning

preserving its own sort of, territory.

01;04;00;09 - 01;04;01;27

Territory

01;04;02;00 - 01;04;03;08

that, that that

01;04;03;08 - 01;04;06;28

is why the, the AI race is, is being

is being heard.

01;04;07;01 - 01;04;11;00

So what that means,

I think, for us is that

01;04;11;03 - 01;04;14;03

has these models

01;04;14;03 - 01;04;16;17

become smarter.

01;04;16;17 - 01;04;19;28

There is nobody, who is concerned

01;04;19;28 - 01;04;24;19

with making them safer in a complete way

01;04;24;22 - 01;04;29;07

that the, the goals of winning,

winning the AI race,

01;04;29;10 - 01;04;32;24

supersede the goal of creating a model

01;04;32;24 - 01;04;36;14

that, keeps us all safe.

01;04;36;17 - 01;04;39;24

And, even though we should,

01;04;39;27 - 01;04;43;29

we should imagine that China doesn't

want the US to get a superintelligence.

01;04;43;29 - 01;04;46;27

The US doesn't want

China get a superintelligence.

01;04;46;27 - 01;04;47;26

There should be.

01;04;47;26 - 01;04;50;08

You know,

when you think about nuclear disarmament,

01;04;50;08 - 01;04;53;07

there should be some sort of, agreement

between rivals.

01;04;53;07 - 01;04;54;25

That's possible here.

01;04;54;25 - 01;05;00;10

Just as there has been in

nuclear armed arms treaties.

01;05;00;13 - 01;05;03;11

And yet, despite the fact that I know

many good diplomats

01;05;03;11 - 01;05;07;05

diplomacy these days, is seen

as, is a dead end.

01;05;07;09 - 01;05;11;17

And,

so we're all racing on our own, sides.

01;05;11;20 - 01;05;12;00

Yeah.

01;05;12;00 - 01;05;13;06

Well, well and it's,

01;05;13;06 - 01;05;16;19

you know, if you're a student of history,

it's a bit concerning that it seems like

01;05;16;19 - 01;05;20;26

with most of these technologies,

you know, nuclear arms included,

01;05;20;29 - 01;05;23;29

the technology comes first

and the safeguards come later.

01;05;24;06 - 01;05;24;15

Right.

01;05;24;15 - 01;05;28;00

Like it's

we've got nuclear nonproliferation treaty

01;05;28;03 - 01;05;30;04

that came

after we deployed nuclear weapons.

01;05;30;04 - 01;05;31;03

There's the

01;05;31;03 - 01;05;33;13

you know that the story making rounds

about the

01;05;33;13 - 01;05;36;08

the gap in years between when the first,

you know, assembly line

01;05;36;08 - 01;05;39;23

car was made versus

when seatbelts were introduced?

01;05;39;26 - 01;05;42;16

And, you know,

I think you hear a lot of the rumors say

01;05;42;16 - 01;05;44;16

we may not have that luxury with AI,

right?

01;05;44;16 - 01;05;46;19

Like it's again, to come back to that,

you'd kowski,

01;05;46;19 - 01;05;48;20

you know, good piece of

of brand marketing.

01;05;48;20 - 01;05;50;23

If, if anybody builds it,

you know, everybody dies.

01;05;50;23 - 01;05;53;00

So that to me is one of the exam questions

here.

01;05;53;00 - 01;05;57;01

Is, is this time fundamentally different

or is it the same.

01;05;57;01 - 01;06;00;04

And how do we how do we grapple with that.

01;06;00;07 - 01;06;02;14

Yeah, it's interesting because at the

AI Summit in Seoul,

01;06;02;14 - 01;06;07;19

which I think was just 20, 24,

it's at that summit that 15

01;06;07;22 - 01;06;12;06

leading AI companies committed to quote,

you know, defining the intolerable risks

01;06;12;06 - 01;06;15;08

and agreeing to not deploy,

which one person said to me,

01;06;15;08 - 01;06;17;10

that's the only time

trillion dollar companies had agreed

01;06;17;10 - 01;06;19;19

to literally deploy a product

if it's not safe. Right.

01;06;19;19 - 01;06;22;25

So in some sense, what's also unusual

01;06;22;25 - 01;06;26;08

about this industry is that,

01;06;26;11 - 01;06;29;24

you know, unlike

the car companies that needed Ralph Nader

01;06;29;27 - 01;06;33;05

to to shout about seatbelts

for quite a long time, and a lot of people

01;06;33;06 - 01;06;36;29

did die

before seatbelts were even included.

01;06;37;02 - 01;06;40;18

And, you know, similarly, the, you know,

you can look at other industries

01;06;40;18 - 01;06;44;10

with their whistleblowers

and their and their gadfly is that, the

01;06;44;10 - 01;06;48;06

no AI companies from the get

go have been talking about AI safety.

01;06;48;06 - 01;06;51;22

So it's it's they've actually been

the main people talking about it.

01;06;51;25 - 01;06;55;01

So, so in some sense you could say, well,

01;06;55;04 - 01;06;58;25

they are ahead

of the curve in, in safety.

01;06;58;28 - 01;07;02;12

The problem with that, I think, is

that, well, first of all, the problem we,

01;07;02;13 - 01;07;06;07

we pointed out before,

which is that they don't actually know

01;07;06;10 - 01;07;08;15

what the models are capable of

until they release them.

01;07;08;15 - 01;07;12;19

So there's that Unknowability

in this technology, the unpredictability.

01;07;12;19 - 01;07;15;03

That's just part of the,

01;07;15;03 - 01;07;17;22

it's part of the way

in which I instructed,

01;07;17;22 - 01;07;21;06

but also, you know, not safe.

01;07;21;09 - 01;07;25;23

Defining the intolerable

risks is not an easy thing.

01;07;25;23 - 01;07;30;16

I mean, no, no, technology in the world

is completely devoid of risk.

01;07;30;19 - 01;07;35;21

You might be somebody who enjoys,

you know, hang gliding.

01;07;35;21 - 01;07;38;29

I might be somebody who, is scared.

01;07;38;29 - 01;07;41;13

Even when I go in the back seat of a car

when I, when I was.

01;07;41;13 - 01;07;44;13

But, I mean, we have different, different

standards of risk.

01;07;44;16 - 01;07;48;29

And so, this is what I think is

so important about for instance, your show

01;07;49;06 - 01;07;54;11

and this kind of conversation

is to dig deeper into what we mean

01;07;54;14 - 01;07;58;08

by not safe, because that's kind of where

a lot of the discussion ends.

01;07;58;08 - 01;08;02;05

It says,

oh my gosh, these things might destroy us.

01;08;02;08 - 01;08;03;25

And then

01;08;03;25 - 01;08;04;23

it stops there.

01;08;04;23 - 01;08;07;18

But but in fact, you know,

if you look at, for example,

01;08;07;18 - 01;08;10;21

I mentioned sold,

the agreements that came out of Seoul.

01;08;10;21 - 01;08;12;26

So I think that was like a

01;08;12;26 - 01;08;16;14

500 word agreement,

you know, that just a statement.

01;08;16;14 - 01;08;17;16

It was a kind of an open letter.

01;08;17;16 - 01;08;19;19

And then after that,

01;08;19;19 - 01;08;24;29

the companies issued maybe, thousand word

01;08;25;02 - 01;08;29;21

mission statements, documents, basically

saying how they defined intolerable risks.

01;08;29;21 - 01;08;32;12

So there was some effort of defining it.

01;08;32;12 - 01;08;34;15

But,

I was talking to somebody and they said,

01;08;34;15 - 01;08;37;15

you know, yeah, about five,

5000 words would be better.

01;08;37;15 - 01;08;39;09

10,000

words would be even better than that.

01;08;39;09 - 01;08;43;17

I mean, the more granular that we can get

these companies to be about,

01;08;43;20 - 01;08;48;04

how do you define risky,

how deep, what specifically do

01;08;48;04 - 01;08;51;01

we want to see in the model

and the training,

01;08;51;01 - 01;08;53;19

the pre-training requirements

and the conditions for deployment?

01;08;53;19 - 01;08;57;20

You know, specifics

before this thing gets deployed?

01;08;57;20 - 01;09;00;10

I would say the details matter.

01;09;00;10 - 01;09;03;06

And unfortunately,

maybe because it's our relationship

01;09;03;06 - 01;09;05;13

to technology, I would not say that of

about your listeners.

01;09;05;13 - 01;09;10;02

But, you know, I think just our society

more broadly, we've just kind of

01;09;10;05 - 01;09;14;01

come into this idea that the, oh,

technology will either work or not work.

01;09;14;01 - 01;09;17;03

You know, it'll either

glitch or it'll function.

01;09;17;09 - 01;09;19;18

And the best technology is invisible.

01;09;19;18 - 01;09;22;11

And maybe,

yeah, maybe we can't do that with this.

01;09;22;11 - 01;09;24;15

Maybe we have to kind of get nerdy

01;09;24;15 - 01;09;28;17

and get into the details,

because I'm agreeing with you.

01;09;28;17 - 01;09;30;18

I, I'm just by nature.

01;09;30;18 - 01;09;33;20

I can't subscribe to Utopia, but

01;09;33;20 - 01;09;36;28

I also just can't subscribe to apocalypse.

01;09;37;01 - 01;09;41;01

Maybe this is my failure of imagination,

but but what I do think

01;09;41;01 - 01;09;45;07

is, is that I have to do

the work of reading about backpropagation

01;09;45;07 - 01;09;50;14

just to just to understand, you know what,

how alien these intelligences are and,

01;09;50;14 - 01;09;55;10

maybe understand a little bit more about

what kind of requirements.

01;09;55;13 - 01;09;59;02

I would what I would want to see before

the models are released.

01;09;59;05 - 01;10;00;07

Right. Well, well.

01;10;00;07 - 01;10;04;05

And, you know, it's it's

great to have everybody sign off on this,

01;10;04;05 - 01;10;07;04

and then I don't want to dismiss that

because that's a win. As you said.

01;10;07;04 - 01;10;10;21

That's something that's that's very

unusual for, for technology companies.

01;10;10;21 - 01;10;14;10

But, you know, to what degree can

01;10;14;10 - 01;10;18;13

we actually create

any sort of enforcements mechanisms here?

01;10;18;13 - 01;10;18;19

Right.

01;10;18;19 - 01;10;22;11

Because that's, that's,

you know, all of this comes down and,

01;10;22;11 - 01;10;25;11

and you set it right off the top

that there's like a trust component here.

01;10;25;15 - 01;10;29;04

And there's the fact that,

the future of civilization is concentrated

01;10;29;04 - 01;10;34;01

in the hands of a bunch of guys who may

or may not be in a group chat together.

01;10;34;04 - 01;10;38;05

And, what happens when Sam says, you know,

oh, yeah, we're not going to do that.

01;10;38;05 - 01;10;41;01

And then says his team,

let's make sure we do that anyway.

01;10;41;01 - 01;10;45;17

You know, like, it's just it's just a wild

I don't know, it's

01;10;45;17 - 01;10;50;00

certainly beyond my imagination

that we could be that we could get here.

01;10;50;03 - 01;10;50;14

Yeah.

01;10;50;14 - 01;10;51;20

Yeah, yeah. No, absolutely.

01;10;51;20 - 01;10;57;11

And, it has been my kind of mission

through the, through the whole show.

01;10;57;11 - 01;11;00;18

And it's not been quite, not easy in this,

in this reporting is to just

01;11;00;18 - 01;11;05;00

to be constantly thinking what is the role

01;11;05;03 - 01;11;09;10

of a person who is not a,

01;11;09;13 - 01;11;12;29

not a technologist, not a lawmaker?

01;11;13;02 - 01;11;16;18

What is their role other than to just

sit back and watch this future happen?

01;11;16;21 - 01;11;18;21

Because clearly,

01;11;18;21 - 01;11;22;19

if you're just a person,

if you're a parent, if you're

01;11;22;22 - 01;11;25;01

if you care about the world,

01;11;25;01 - 01;11;28;00

it doesn't feel acceptable to just sit

01;11;28;02 - 01;11;31;24

with this level of risk and do nothing.

01;11;31;27 - 01;11;34;04

But it also feels premature

01;11;34;04 - 01;11;38;02

or perhaps histrionic or,

I don't know, just to freak out.

01;11;38;02 - 01;11;41;01

And some

and you know, I had this conversation

01;11;41;01 - 01;11;44;01

with a number of people,

even somebody who was

01;11;44;05 - 01;11;46;28

reviewing and recommending the podcast

and they said, you know, well,

01;11;46;28 - 01;11;49;28

I had to give it a mixed review

because I was also freaked out.

01;11;50;01 - 01;11;54;29

And so I said, you know, that doesn't mean

you shouldn't listen, but it's

01;11;55;02 - 01;11;58;28

but I got their point, which is that,

I think it is important for us

01;11;58;28 - 01;12;03;06

and you and I and,

anybody who has the opportunity

01;12;03;06 - 01;12;07;21

to talk to anybody about this

stuff is to think about how

01;12;07;24 - 01;12;08;24

how we

01;12;08;24 - 01;12;13;12

how we, walk the tightrope that I guess

all prophets

01;12;13;15 - 01;12;17;24

and, and, and biblical,

figures have have walked before us,

01;12;17;24 - 01;12;23;12

which is how do you warn people without,

01;12;23;15 - 01;12;28;13

making them unhelpfully anxious.

01;12;28;16 - 01;12;32;02

And, you know,

I see the, the in the, in the podcast, we,

01;12;32;02 - 01;12;34;04

we talk about these three groups,

we talk about the doomsayers,

01;12;34;04 - 01;12;38;10

but the acceleration is we also talk about

this third group called the Scouts

01;12;38;13 - 01;12;41;14

and the scouts, kind of believe

01;12;41;15 - 01;12;44;27

essentially in a nutshell,

that this win win opportunity is possible,

01;12;45;00 - 01;12;46;07

that, that more

01;12;46;07 - 01;12;49;28

AI safety is absolutely necessary,

but that we shouldn't just stop AI.

01;12;50;01 - 01;12;52;12

But I, I see them

01;12;52;12 - 01;12;55;17

struggling

as any centrist kind of struggles

01;12;55;20 - 01;12;59;15

where you don't have a clear message,

this is amazing or this is horrible,

01;12;59;18 - 01;12;59;27

and you're

01;12;59;27 - 01;13;04;00

trying to explain to people, no, this is,

this is this could be really good.

01;13;04;03 - 01;13;08;06

But we have to do we have to understand

the models a little bit more.

01;13;08;06 - 01;13;11;16

And because it's not easy to say

this could be good,

01;13;11;16 - 01;13;14;16

but we certainly need to regulation

that will do X

01;13;14;23 - 01;13;18;13

because there's it's not clear

that there's any regulation

01;13;18;13 - 01;13;21;26

that the superintelligence

or just an advanced AI will not outwit.

01;13;22;01 - 01;13;25;11

This is where I think in answer

to that question, more people need to

01;13;25;11 - 01;13;27;06

be involved in the problem.

01;13;27;09 - 01;13;28;22

More people need to be

01;13;28;22 - 01;13;33;00

scenario planning for what,

let's say, never happens.

01;13;33;03 - 01;13;35;13

Who knows? Let's say ten years, 20 years.

01;13;35;13 - 01;13;37;27

Everything looks exactly the same

as it does now. Then fine.

01;13;37;27 - 01;13;40;02

Then we will have done a mental exercise

for no reason.

01;13;40;02 - 01;13;43;09

Okay, but let's say there is a chance.

01;13;43;12 - 01;13;43;20

Let's say

01;13;43;20 - 01;13;47;24

it's a good chance that things look

radically different in 10 or 20 years.

01;13;47;27 - 01;13;50;23

Perhaps some scenario planning now

01;13;50;23 - 01;13;53;27

as to how universities might change,

01;13;53;27 - 01;13;58;20

how schools might change,

how parenting might change, how how,

01;13;58;23 - 01;14;01;07

community living or sort of,

01;14;01;07 - 01;14;04;26

communal relationships

might change with a superintelligence

01;14;05;03 - 01;14;09;26

or with an advanced

AI that is doing most of the human labor.

01;14;09;29 - 01;14;12;16

That's a

that's a question that we could play out,

01;14;12;16 - 01;14;16;13

that we could we could worry, in

fact, you know, if anybody

01;14;16;16 - 01;14;17;27

wants to contact me with,

01;14;17;27 - 01;14;21;21

with some scenario planning ideas,

I, I'm, I'm working on this right now.

01;14;21;24 - 01;14;26;04

So a question, and it's very important

01;14;26;04 - 01;14;29;13

if you're doing scenario planning

to not be freaked out

01;14;29;16 - 01;14;32;06

in the, in the sense of that word,

not be panicked,

01;14;32;06 - 01;14;36;14

but also not be,

not take a cavalier attitude

01;14;36;17 - 01;14;38;18

and, and maybe we could figure out

01;14;38;18 - 01;14;41;28

some new solutions that would work

even if we don't get to advance AI.

01;14;42;01 - 01;14;44;17

That's my optimist. Talking.

01;14;44;17 - 01;14;45;17

No I love that.

01;14;45;17 - 01;14;48;17

And I love that as a mission for you know

01;14;48;21 - 01;14;51;23

the podcast in general and frankly

the journalistic mission of it all.

01;14;51;23 - 01;14;53;04

And I think it's you know.

01;14;53;04 - 01;14;55;28

Yeah I agree that it's super,

super important

01;14;55;28 - 01;14;58;22

to, to just sort of pivot

that question a little bit

01;14;58;22 - 01;15;01;13

when we think about scenario planning

and we think about, you know, what

01;15;01;13 - 01;15;05;11

we need to know and what we need to do

differently to build the future we want.

01;15;05;17 - 01;15;08;27

What's your advice for,

you know, business leaders

01;15;08;27 - 01;15;12;22

or government leaders, you know,

in the organizational side of government?

01;15;12;25 - 01;15;16;21

Yeah, outside of Silicon Valley,

like for the people who are looking at

01;15;16;21 - 01;15;20;28

adopting this technology, looking to,

you know, figure out

01;15;21;01 - 01;15;25;17

what they need to do differently to be

successful as people and as organizations,

01;15;25;20 - 01;15;28;20

what should be on their radar

and what guidance would you give them?

01;15;28;22 - 01;15;30;14

Yeah, I appreciate it.

01;15;30;14 - 01;15;35;29

One is, I think, you know, one of the

the feedback that I've gotten from this,

01;15;36;02 - 01;15;39;12

from the, from working on this series,

I've talked to people

01;15;39;12 - 01;15;42;06

who feel that their companies

01;15;42;09 - 01;15;45;09

and these are, these tend to be, say,

01;15;45;14 - 01;15;50;12

middle level decision makers or even,

people who don't feel like that.

01;15;50;12 - 01;15;53;02

They're the key decision maker.

They're just under the CEO.

01;15;53;02 - 01;15;56;25

They feel that their company is either,

01;15;56;28 - 01;16;00;16

either moving too fast

or being left behind.

01;16;00;18 - 01;16;03;09

Right. This is the

this is the always the story.

01;16;03;09 - 01;16;05;20

The moving too fast.

01;16;05;20 - 01;16;10;01

Goes with

they're throwing out human intelligence.

01;16;10;01 - 01;16;11;23

They're trying to replace.

01;16;11;23 - 01;16;13;28

They're trying to automate everything.

01;16;13;28 - 01;16;16;11

And the we're not moving

fast enough saying

01;16;16;11 - 01;16;19;21

we're going to be left behind when we're,

we're, we're we're sort of stuck.

01;16;19;21 - 01;16;23;00

And so I'm sure the business leaders

are feeling that pressure

01;16;23;00 - 01;16;26;21

and having to make decisions

all the time about the pace of adoption.

01;16;26;23 - 01;16;28;06

Right.

01;16;28;06 - 01;16;31;09

And it's extremely difficult to make

a decision about the pace of adoption

01;16;31;09 - 01;16;35;08

of something that keeps changing,

that keeps developing and getting smarter.

01;16;35;11 - 01;16;38;07

Because how do you make that decision?

01;16;38;07 - 01;16;40;14

So one thing that I think is

helpful are two things

01;16;40;14 - 01;16;44;10

that I think has been helpful for me

in talking to those people

01;16;44;13 - 01;16;46;24

and in some sense allaying their concerns,

01;16;46;24 - 01;16;50;05

but also addressing, addressing

01;16;50;05 - 01;16;54;21

what is the elephant in the room,

which is what will my life look like?

01;16;54;21 - 01;16;57;05

What will my industry look like?

01;16;57;05 - 01;17;00;17

On the other side of this technology?

01;17;00;20 - 01;17;02;17

The two things I always say is, first,

01;17;02;17 - 01;17;07;24

the, the,

I think I told the story before, but the,

01;17;07;27 - 01;17;11;24

the fact that a chess computer

with a human,

01;17;11;27 - 01;17;16;12

a human and a computer

and an AI is is smarter

01;17;16;12 - 01;17;20;09

is better than a, and then,

01;17;20;09 - 01;17;24;04

any AI or any human, at least currently.

01;17;24;07 - 01;17;26;21

And I think that's going to be true

for a while.

01;17;26;21 - 01;17;31;27

So what I think

the smart approach is to think about

01;17;32;00 - 01;17;36;04

how do we enhance, how do we,

01;17;36;07 - 01;17;41;12

you know, supercharge the work of the

of our our of our of our employees?

01;17;41;12 - 01;17;44;21

How do we get them to do not only more

01;17;44;21 - 01;17;48;10

but to, to to think deeper

01;17;48;13 - 01;17;52;29

and to make more interesting decisions

to as you use the predictive power of AI

01;17;53;00 - 01;17;57;25

to make decisions not only for, for now,

but to do more sophisticated planning

01;17;57;28 - 01;18;00;25

and so I think that that's

actually addresses

01;18;00;25 - 01;18;04;07

both concerns that we're not moving fast

enough as well as we're not, we're moving

01;18;04;07 - 01;18;07;08

to, to, to, we're moving too

01;18;07;08 - 01;18;10;20

rapidly in that humans are humans.

01;18;10;23 - 01;18;13;07

Our employees need to feel empowered.

01;18;13;07 - 01;18;17;19

They need to feel that they are

getting smarter because of this AI.

01;18;17;22 - 01;18;19;07

And it gets it. Really.

01;18;19;07 - 01;18;22;07

The second point, I would say,

which is that the story,

01;18;22;08 - 01;18;25;23

there's so many stories from sci fi

01;18;25;26 - 01;18;27;10

that are

01;18;27;10 - 01;18;31;19

not only just in our heads,

but baked into these models.

01;18;31;22 - 01;18;37;18

I mean, even the way forgiveness people

haven't seen, say, Stanley Kubrick's,

01;18;37;21 - 01;18;39;21

masterpiece, 2001

01;18;39;21 - 01;18;44;25

A Space Odyssey or they haven't seen,

Blade Runner or The Matrix,

01;18;44;28 - 01;18;47;25

even if they haven't

seen any of those films.

01;18;47;25 - 01;18;53;09

Even the way that AI is so solicitous

and the way it is encouraging.

01;18;53;09 - 01;18;55;18

And you talked about this earlier,

01;18;55;18 - 01;18;57;03

it begs a kind of narrative.

01;18;57;03 - 01;19;01;20

It makes you think of a story

where, you know,

01;19;01;25 - 01;19;04;24

like an Isaac Asimov

kind of narrative where this,

01;19;04;24 - 01;19;08;00

this servant is quite helpful

01;19;08;03 - 01;19;13;04

until they're not, and

01;19;13;07 - 01;19;16;21

Geoffrey Hinton, I would say,

says something so smart about this.

01;19;16;24 - 01;19;21;29

He says that most CEOs, you know,

most leaders

01;19;22;02 - 01;19;24;29

are used to not being the most

intelligent person in the room, right?

01;19;24;29 - 01;19;26;19

If they were the most intelligent person

01;19;26;19 - 01;19;28;15

in the room,

they're probably not a good leader

01;19;28;15 - 01;19;31;21

because they need to hire

the smarter people,

01;19;31;24 - 01;19;34;26

because they hold the vision

for the company

01;19;34;26 - 01;19;37;29

and they hold the, you know, the

they are the leadership.

01;19;37;29 - 01;19;42;04

The leadership

shouldn't be the best at every task.

01;19;42;07 - 01;19;44;25

And so in some sense, he

01;19;44;25 - 01;19;48;24

says that these models have been designed

01;19;48;27 - 01;19;51;14

by very smart people who are used

01;19;51;14 - 01;19;55;09

to even smarter people working for them,

01;19;55;12 - 01;19;59;17

and they are not threatened

by intelligence

01;19;59;20 - 01;20;03;15

in the way

that say, the employees in a company

01;20;03;18 - 01;20;06;12

might be threatened by the arrival

01;20;06;12 - 01;20;11;12

of this incredibly capable,

infinitely knowledgeable,

01;20;11;15 - 01;20;14;25

machine and technology

that doesn't need to sleep

01;20;14;25 - 01;20;18;14

and doesn't need to eat and and and and

and never forgets anything.

01;20;18;14 - 01;20;21;15

So, Geoffrey Hinton uses

that as an analogy to say that, that

01;20;21;17 - 01;20;24;24

in some sense, the CEOs of these companies

aren't, aren't worried enough.

01;20;24;24 - 01;20;26;05

They, they're used to they

01;20;26;05 - 01;20;28;08

they can dream of superintelligence

and still imagine

01;20;28;08 - 01;20;30;07

that the superintelligence

will do their bidding

01;20;30;07 - 01;20;32;13

just as their employees do their bidding,

because they

01;20;32;13 - 01;20;35;19

they haven't really appreciated the fact

that that a superintelligence is

01;20;35;19 - 01;20;39;18

much, much, much, much,

much smarter than any, you know,

01;20;39;21 - 01;20;42;01

Einstein level person that they hire.

01;20;42;01 - 01;20;45;03

But I think the, the takeaway

01;20;45;03 - 01;20;50;06

for me, for, for business leaders

who are adopting AI is to watch

01;20;50;09 - 01;20;53;23

and take care for the narratives

that, that,

01;20;53;23 - 01;20;58;15

that are often in this disharmony,

this not alignment between the C-suite

01;20;58;18 - 01;21;02;17

and the workshop

for the rest of the employees that,

01;21;02;20 - 01;21;07;14

our relationship to an intelligence

is going to be different in the C-suite

01;21;07;14 - 01;21;10;20

and is going to be different

on the among the workers,

01;21;10;23 - 01;21;14;17

and that people need to feel

even as this is making the company better

01;21;14;17 - 01;21;19;03

or more efficient,

that it's also making the humans

01;21;19;06 - 01;21;23;17

smarter and more capable and even happier.

01;21;23;20 - 01;21;28;03

So that's that's more around the narrative

than it is around the adoption.

01;21;28;03 - 01;21;32;13

But I think that it's it's

kind of can guide all the,

01;21;32;16 - 01;21;35;15

adoption decisions if,

if that's if that makes sense.

01;21;35;15 - 01;21;36;19

It makes complete sense.

01;21;36;19 - 01;21;38;12

And I love that guidance.

01;21;38;12 - 01;21;40;24

And I think it's it's

so important these days.

01;21;40;24 - 01;21;43;25

I mean, it's and you read about it

everywhere, and it's something we've had

01;21;43;25 - 01;21;46;25

people talk about on the show of,

you know, you've got this kind of

01;21;47;02 - 01;21;50;11

two speed or kind of,

01;21;50;14 - 01;21;53;13

you know, narratives in conflict of CEOs

saying,

01;21;53;13 - 01;21;55;25

oh, this is going to make my company

so much better.

01;21;55;25 - 01;21;59;02

And frontline workers say, oh, by,

you know, putting me out of a job.

01;21;59;08 - 01;22;03;12

And if you can't marry those,

if you're an organizational leader

01;22;03;14 - 01;22;07;17

and you can't get the people that report

to you to be excited about the s,

01;22;07;20 - 01;22;12;06

it feels like it's going to be very

difficult to successfully, you know, rally

01;22;12;06 - 01;22;15;28

the organization and build something,

that's going to be worth doing.

01;22;16;01 - 01;22;20;09

I agree and I and I think also we may see

this play out politically as well.

01;22;20;12 - 01;22;25;21

I mean, currently, I don't think that

I as a, as a political issue yet.

01;22;25;28 - 01;22;31;22

There's not a clear Democratic

or Republican position on AI.

01;22;31;28 - 01;22;37;29

But that I think could quickly

change and we could see,

01;22;38;02 - 01;22;40;13

kind of people rallying around

01;22;40;13 - 01;22;43;29

this becoming as polarized

an issue as any other.

01;22;44;02 - 01;22;47;02

Perhaps we could we could theorize

about what might trigger that,

01;22;47;02 - 01;22;52;07

but it's not hard to imagine, for example,

in our polarized climate, a stop AI

01;22;52;07 - 01;22;56;07

camp and a AI is great camp or whatever,

01;22;56;10 - 01;22;58;23

and that would be very disappointing.

01;22;58;23 - 01;23;04;07

That would be very disappointing

because polarization is is really the

01;23;04;10 - 01;23;06;13

the killer of anything complex,

01;23;06;13 - 01;23;09;25

anything complicated to discuss

and to think about clearly.

01;23;09;25 - 01;23;13;09

I mean, you and I are having a discussion

where neither of us are in the,

01;23;13;12 - 01;23;17;03

say, the stop everything camp

or the except at all Utopia camp.

01;23;17;03 - 01;23;20;26

But in order to talk about the fine print

01;23;20;28 - 01;23;25;28

of these models,

I would say, to use our alien analogy,

01;23;26;01 - 01;23;28;13

if this becomes politicized,

01;23;28;13 - 01;23;31;27

then the alien definitely wins.

01;23;32;00 - 01;23;34;00

Well said, well said.

01;23;34;00 - 01;23;37;22

Greg, we've talked a few times now about,

this conversation has been

01;23;37;25 - 01;23;40;16

pretty heavily focused on chat bots,

and I think that makes sense, given

01;23;40;16 - 01;23;44;04

that the the technology is here

and it's it's very clear

01;23;44;04 - 01;23;45;15

how it's impacting us right now.

01;23;45;15 - 01;23;47;10

But, you know, one of the

01;23;47;10 - 01;23;50;13

in some ways, you know, cardinal sins of

AI is conflating

01;23;50;20 - 01;23;54;06

chat bots or just,

you know, generative AI with everything

01;23;54;06 - 01;23;57;26

possible in the world of algorithms

and these advanced technologies.

01;23;57;26 - 01;23;59;27

We've talked briefly about agen tech.

01;23;59;27 - 01;24;03;15

As we look beyond that, as we look

at some of the other

01;24;03;15 - 01;24;06;15

big buckets of,

01;24;06;22 - 01;24;10;18

you know, technological capabilities

that are either here or are emerging,

01;24;10;21 - 01;24;12;01

what are some of the ones that, through

01;24;12;01 - 01;24;14;12

these conversations

have gotten your attention

01;24;14;12 - 01;24;17;07

and maybe what are some of the ones

that you think are a little bit

01;24;17;07 - 01;24;19;22

less likely to have an impact?

01;24;19;25 - 01;24;20;02

Yeah.

01;24;20;02 - 01;24;21;01

I'm so glad you asked the question,

01;24;21;01 - 01;24;24;02

because it does feel like,

as you said, we do.

01;24;24;02 - 01;24;24;27

We do make the mistake.

01;24;24;27 - 01;24;28;14

I often make this mistake of thinking

that the chat bot is the AI, but in fact,

01;24;28;14 - 01;24;34;00

the chat bot is just, it points us to the

the artificial intelligence underneath.

01;24;34;00 - 01;24;37;11

And AI is being used

in very different ways.

01;24;37;14 - 01;24;41;13

And I think also just to understand why

01;24;41;16 - 01;24;45;04

frontier AI companies

would be kind of barreling

01;24;45;04 - 01;24;48;28

toward this future of superintelligence

is easier to understand

01;24;49;01 - 01;24;52;15

when we look beyond the chat bot,

for example.

01;24;52;18 - 01;24;57;27

I mean, just to give like a couple

of quick examples, there's a company

01;24;58;00 - 01;25;03;11

that I spoke with which is, using AI to,

01;25;03;14 - 01;25;07;03

to very quickly,

in an automated way, harvest,

01;25;07;06 - 01;25;12;18

cells from your own body

so that you if you need, let's say,

01;25;12;21 - 01;25;15;06

a liver replacement, it will be harvesting

01;25;15;06 - 01;25;18;15

and growing

cells is a very onerous process.

01;25;18;15 - 01;25;21;27

You have to find the exact right

kind of cell that's healthy,

01;25;21;27 - 01;25;24;21

and then you have to encourage

to grow and divide and kill

01;25;24;21 - 01;25;28;00

or kill the other cells so you don't get

some kind of disease thing.

01;25;28;00 - 01;25;31;16

Doing this outside of the human body

is very difficult, but it's something

01;25;31;16 - 01;25;35;26

that it's a kind of pattern recognition

that AI is extremely good at.

01;25;35;29 - 01;25;39;24

So, the vision of this particular company

is called Selena.

01;25;39;24 - 01;25;43;09

But there are other companies

like this is to,

01;25;43;12 - 01;25;45;24

I mean, essentially have

a future of medicine

01;25;45;24 - 01;25;48;24

where we go to the hospital

and we have a kind of cassette

01;25;48;26 - 01;25;52;01

with our organs

there on the on the cassette.

01;25;52;07 - 01;25;55;19

And perhaps if we need,

01;25;55;22 - 01;26;00;07

if we need a, I mean, maybe in the future

if we need a whole organ or currently

01;26;00;07 - 01;26;00;26

maybe if we need,

01;26;00;26 - 01;26;04;18

you know, some, some tissue, then we

we don't have to use a donor.

01;26;04;21 - 01;26;06;08

We can use ourselves as the donor.

01;26;06;08 - 01;26;10;05

so that's just one example, I think of how

the very same tools that were used

01;26;10;05 - 01;26;13;10

to like pattern recognition,

making, making these decisions,

01;26;13;10 - 01;26;17;04

but but doing it

quickly, could transform health care.

01;26;17;10 - 01;26;22;07

Another example that I was at,

think about is,

01;26;22;10 - 01;26;26;25

a particular crown created by a company,

01;26;26;28 - 01;26;27;29

French company.

01;26;27;29 - 01;26;31;09

Olivia Julia is the

is the scientist behind it,

01;26;31;09 - 01;26;35;25

and he's created an AI crown

that will essentially read your thoughts.

01;26;35;25 - 01;26;39;18

So if we think of, say, neurotic

or other brain computer interfaces, this

01;26;39;18 - 01;26;43;23

is not involve any drilling into the brain

or any drilling into the skull.

01;26;43;23 - 01;26;45;25

You don't think actually stick

something in there?

01;26;45;25 - 01;26;50;09

Rather it's, it, it,

it works outside the head.

01;26;50;09 - 01;26;55;02

You just plop it on your head

like, like a headset and a paraplegic man

01;26;55;09 - 01;27;00;06

in, was able to use this

headset to control,

01;27;00;09 - 01;27;03;22

not only a mouse on the screen,

but actually to drive a formula

01;27;03;22 - 01;27;08;03

One race car on a real racetrack and,

01;27;08;06 - 01;27;11;06

using only his mind.

01;27;11;07 - 01;27;13;14

What's crazy about this story

and the other stories

01;27;13;14 - 01;27;15;21

is that this technology, it's

not futuristic.

01;27;15;21 - 01;27;16;20

It's actually exists.

01;27;16;20 - 01;27;20;24

I mean, there is a real person

who is driving a Formula One racing car

01;27;20;25 - 01;27;24;04

using only his mind with just a headset

that sat on his head.

01;27;24;07 - 01;27;28;04

And yet for that future to be accessible

or build,

01;27;28;06 - 01;27;31;12

be available to the rest of us,

you need a lot more data.

01;27;31;15 - 01;27;34;23

But that's always the question with, with,

with every I think it's like, well,

01;27;34;26 - 01;27;36;17

where's the data?

01;27;36;20 - 01;27;38;19

Then we we I'm sure

you've talked to many people about that,

01;27;38;19 - 01;27;42;02

but in this case

the data is our own brainwaves, right?

01;27;42;02 - 01;27;45;09

So many, many,

many of us would have to volunteer up

01;27;45;09 - 01;27;49;02

our brainwaves for the AI to learn enough.

01;27;49;05 - 01;27;54;10

Enough to have enough data essentially to,

to work with, to be able to be.

01;27;54;13 - 01;27;55;07

And I don't want to

01;27;55;07 - 01;27;57;00

I don't want to make it sound like it's

just an out of the box,

01;27;57;00 - 01;27;59;01

throw in the headset

and drive a car with no hands.

01;27;59;01 - 01;27;59;09

I mean, I'm

01;27;59;09 - 01;28;03;10

sure there's some training involved,

but to make that even a possible future

01;28;03;13 - 01;28;07;07

where if I have had an injury,

01;28;07;11 - 01;28;12;14

I'm able to stick a thing on my head

and then function as I was before

01;28;12;15 - 01;28;17;26

while I recover

or perhaps that's my new future.

01;28;17;29 - 01;28;18;19

That is

01;28;18;19 - 01;28;22;18

the sort of vision of AI that,

I would say a lot

01;28;22;18 - 01;28;27;17

of my humanitarian friends say we

we cannot, cannot come fast enough.

01;28;27;20 - 01;28;30;12

There's so many people

that could could benefit from that,

01;28;30;12 - 01;28;34;13

from a point of view of accessibility

or longevity or health care.

01;28;34;16 - 01;28;35;08

So, yes.

01;28;35;08 - 01;28;38;04

So I definitely think that that's there.

01;28;38;04 - 01;28;42;11

As for things

that may not have as much of an impact,

01;28;42;14 - 01;28;44;20

I guess that's that's everything else.

01;28;44;20 - 01;28;47;14

I would not

it's it's really hard to know.

01;28;47;14 - 01;28;49;26

I mean, there's so many new AI companies

coming up every day.

01;28;49;26 - 01;28;51;19

It's hard to know what's going to be,

01;28;51;19 - 01;28;55;22

what's going to be, you know, the wheat

and what's going to be the chaff. But,

01;28;55;25 - 01;28;57;07

somehow it's going to be.

01;28;57;07 - 01;28;59;26

Yeah, I think there's some radical changes

coming ahead.

01;28;59;26 - 01;29;01;07

the examples you chose.

01;29;01;07 - 01;29;05;09

not not just because they're so positive

but because there's such a, there's

01;29;05;09 - 01;29;09;15

such a radical departure from where we got

caught up before talking about like,

01;29;09;18 - 01;29;13;13

you know, AI as this,

you know, alien intelligence, right?

01;29;13;13 - 01;29;16;07

That this is not

AI is as an alien intelligence.

01;29;16;07 - 01;29;20;08

This is very much,

you know, basically a complexity engine

01;29;20;08 - 01;29;24;09

that helps us,

you know, serve humans better, right?

01;29;24;12 - 01;29;28;13

It helps us personalized medicine,

personalized care,

01;29;28;16 - 01;29;32;17

and just create tools that help us,

01;29;32;20 - 01;29;36;07

you know, live healthier,

more fulfilling lives, which is

01;29;36;10 - 01;29;39;12

which is inspirational and is just such

a completely different vision

01;29;39;17 - 01;29;45;27

from yeah, it's trying to manipulate me

or, you know, build Skynet.

01;29;46;00 - 01;29;46;29

It's actually true.

01;29;46;29 - 01;29;47;26

And it's it's that's why

01;29;47;26 - 01;29;51;19

it's hard for me to understand

the debate around the AI bubble

01;29;51;22 - 01;29;55;06

because,

the bubble question has so much to do

01;29;55;11 - 01;29;58;11

with valuation rather than value,

01;29;58;14 - 01;30;01;14

you know,

01;30;01;17 - 01;30;03;23

in so well

and this is something I've never,

01;30;03;23 - 01;30;05;24

you know, again,

when the mysteries of the stock market

01;30;05;24 - 01;30;10;00

is how something could be valuable

but still, over overvalued.

01;30;10;00 - 01;30;14;18

But, it's not that AI is

01;30;14;23 - 01;30;18;29

is this promising future thing

that once it crosses a certain threshold,

01;30;18;29 - 01;30;22;04

then it will being incredibly powerful

01;30;22;04 - 01;30;25;04

and change our our world.

01;30;25;09 - 01;30;29;29

The technology of current AI that exists

01;30;30;02 - 01;30;32;23

is is already

01;30;32;23 - 01;30;37;16

quite amazing and and goes far beyond

the fact that, you know, the GPT can say

01;30;37;16 - 01;30;42;20

like write a write a pretty decent

short story or, or a legal report.

01;30;42;27 - 01;30;49;04

nevertheless, there's a contract

with, with the AI that's necessary.

01;30;49;04 - 01;30;52;04

There's a, there's obviously data centers

that are being built.

01;30;52;10 - 01;30;54;28

There are there's a question of data.

01;30;54;28 - 01;30;57;04

There's a question of regulation.

01;30;57;04 - 01;31;00;25

So that's why, you know,

I think one of our themes

01;31;00;25 - 01;31;04;08

throughout this whole conversation

has been, well, what what is my role,

01;31;04;11 - 01;31;08;13

what is what is our role collectively

in shaping the future of AI

01;31;08;15 - 01;31;13;09

if we're not the head of a frontier

AI company, if there were not a lawmaker,

01;31;13;12 - 01;31;17;17

if we are a decision maker in a company,

but again, not

01;31;17;17 - 01;31;21;07

somebody who can create a new model but

just decide whether to adopt it or not,

01;31;21;10 - 01;31;25;00

I think there's going to be

so many questions in the next few years.

01;31;25;03 - 01;31;27;16

In terms

of whether we're running a factory,

01;31;27;16 - 01;31;32;04

how much data gets used,

how is that data get used if we're even

01;31;32;04 - 01;31;35;28

just a patient in health care,

whether we give up our data.

01;31;36;05 - 01;31;38;02

I mean, many of us are already

giving up our data,

01;31;38;02 - 01;31;41;02

but having more understanding

about how that data is being used,

01;31;41;09 - 01;31;44;18

that will contribute to whether AI has

01;31;44;18 - 01;31;48;29

a shaping force on our world

in a in these different domains.

01;31;49;02 - 01;31;51;20

You know, it was interesting, Greg,

what you said earlier about,

01;31;51;20 - 01;31;55;24

you know, journalists

and about the program,

01;31;56;01 - 01;31;59;24

you know, suffering by

not by scaring people too much.

01;31;59;24 - 01;32;03;29

And to me, it's almost like a journalistic

responsibility to be scaring people.

01;32;03;29 - 01;32;06;23

Right. Like, I you in some ways.

01;32;06;23 - 01;32;08;24

And I feel the same way here.

01;32;08;24 - 01;32;12;14

There's an obligation to actually tell

people the facts and what's going on.

01;32;12;14 - 01;32;12;21

Right.

01;32;12;21 - 01;32;17;23

And then one of the, you know, downfalls

or issues with journalism these days

01;32;17;26 - 01;32;21;21

is this sort of social media ification

or this algorithmic,

01;32;21;24 - 01;32;23;29

you know, content lens

where you just tell people

01;32;23;29 - 01;32;27;07

what they want to hear and people

say, yes, you know, validate me,

01;32;27;08 - 01;32;32;13

tell me what I think is right versus

actually provide me with the facts and,

01;32;32;13 - 01;32;36;10

and tell me something that

that's important that I'm educated on,

01;32;36;13 - 01;32;41;18

you know, versus just you know, that

the same old thing that I already know.

01;32;41;21 - 01;32;42;04

Yeah.

01;32;42;04 - 01;32;43;18

I mean, it's it's it's a good point.

01;32;43;18 - 01;32;45;06

I mean, it reminds me of my days

01;32;45;06 - 01;32;48;28

as an international correspondent

and sort of being, say, in Afghanistan

01;32;48;28 - 01;32;51;03

and, and hearing the grapes

of other correspondents.

01;32;51;03 - 01;32;52;24

They say, you know, gosh,

this is a whole war

01;32;52;24 - 01;32;56;09

and people are back at home

are just not interested anymore.

01;32;56;09 - 01;32;58;12

It's fallen off the news.

01;32;58;12 - 01;33;00;03

And I always saw it differently.

01;33;00;03 - 01;33;05;21

I always thought, no, no,

my job is to make you care and to

01;33;05;24 - 01;33;09;17

I don't know if you call that entertain

or engage you, but I was going to find

01;33;09;17 - 01;33;13;10

some kind of angle that was going

to make this feel relevant to you.

01;33;13;17 - 01;33;20;05

So I do think that as voices, as as

I feel like

01;33;20;05 - 01;33;23;26

we do have a role, we do have a job

to make this feel relevant.

01;33;23;29 - 01;33;27;12

that probably means

don't freak people out right away

01;33;27;15 - 01;33;29;06

because, because

01;33;29;06 - 01;33;33;01

then they'll just feel,

small as opposed to empowered.

01;33;33;01 - 01;33;36;25

But I think that you can take that

too far.

01;33;36;25 - 01;33;40;18

And we are in a situation

now where there's

01;33;40;18 - 01;33;42;26

this incredibly important technology.

01;33;42;26 - 01;33;48;12

It's complicated, the complexity

is interesting and is I mean,

01;33;48;15 - 01;33;50;23

is is maybe worth

01;33;50;23 - 01;33;53;17

spending some time thinking about.

01;33;53;17 - 01;33;58;05

But this is why it feels like you and I

and we need to find

01;33;58;05 - 01;34;03;01

these new, new new narratives,

not just the sci fi narrative,

01;34;03;04 - 01;34;05;25

not just the,

you know, apocalyptic doom and gloom.

01;34;05;25 - 01;34;08;15

I think Ellie is you. Koski is.

01;34;08;15 - 01;34;11;11

I don't know if for titling the book,

if anyone builds it, everyone dies.

01;34;11;11 - 01;34;14;10

It's blunt,

And maybe it gets more readers, but

01;34;14;10 - 01;34;18;19

there's there must be a role

for sitting down with a listener or reader

01;34;18;22 - 01;34;20;15

and saying, okay,

01;34;20;18 - 01;34;21;22

here are some

01;34;21;22 - 01;34;24;22

incredibly fascinating things

about AI that you know it.

01;34;24;22 - 01;34;28;14

Here's some ways they could go wrong,

01;34;28;17 - 01;34;31;13

and here's ways in which the

the world might play out

01;34;31;13 - 01;34;35;07

that might feel radically different

than than you may think it feels.

01;34;35;10 - 01;34;37;19

Here's how your kids will fare.

01;34;37;19 - 01;34;41;12

You know, it's like we have to address

the questions that people have

01;34;41;17 - 01;34;47;16

and not just leave them with, you know,

01;34;47;19 - 01;34;49;11

leave them with a scare story.

01;34;49;11 - 01;34;50;15

And so I, I struggle with it.

01;34;50;15 - 01;34;51;14

You hear me struggling with it

01;34;51;14 - 01;34;55;19

in this answer because I,

I do think our role as information

01;34;55;19 - 01;34;59;13

gatherers is to packaged package that

information in a way that people will,

01;34;59;16 - 01;35;00;21

will want to consume.

01;35;00;21 - 01;35;04;28

I mean, it's we're not like professors

here with a, with a captured audience.

01;35;04;28 - 01;35;07;06

People can choose

what they want to tune into.

01;35;07;06 - 01;35;10;10

So we have to we have to sing and dance

for our supper, as it were.

01;35;10;10 - 01;35;14;12

But at the same time,

it's the complexity of it

01;35;14;12 - 01;35;18;13

and the the fear, I think, that audiences

have of that complexity.

01;35;18;13 - 01;35;21;08

And I think even the fear that journalists

have of the complexity.

01;35;21;08 - 01;35;24;08

I would say that you're very much

an exception to this, but,

01;35;24;14 - 01;35;27;06

You know, a fear of say,

oh, gosh, I don't want to sound stupid

01;35;27;06 - 01;35;29;29

because I'm talking about the stuff

that's computer science related.

01;35;29;29 - 01;35;32;13

And I didn't really feel, I mean,

01;35;32;16 - 01;35;34;07

you know, yes,

01;35;34;07 - 01;35;38;24

there's a lot that makes us feel dumb

when we're trying to understand how a

01;35;38;27 - 01;35;41;07

something

like artificial intelligence works, but

01;35;41;07 - 01;35;45;07

just being willing to ask those questions

and being willing to dive

01;35;45;07 - 01;35;48;09

into what is red teaming with, like,

what is safety training look like?

01;35;48;09 - 01;35;52;07

What what

what might a model take to be controlled?

01;35;52;14 - 01;35;54;14

If we can get more people

to have these conversations

01;35;54;14 - 01;35;57;14

without feeling imperiled,

you know, either

01;35;57;14 - 01;36;01;21

physically or economically imperiled,

that will be a win.

01;36;01;24 - 01;36;02;09

Right.

01;36;02;09 - 01;36;04;15

I love that it,

it makes, it makes complete sense.

01;36;04;15 - 01;36;07;08

And it's it's a noble calling as well.

01;36;07;08 - 01;36;09;13

Well,

thanks so much for this, encouragement.

01;36;09;13 - 01;36;11;19

I really appreciate it, Jeff,

and appreciate the work you're doing.

01;36;11;19 - 01;36;13;15

Keep, keep doing what you're doing

and you know, we'll

01;36;13;15 - 01;36;16;01

get, we'll get out there,

one person at a time.

01;36;16;01 - 01;36;19;04

Absolutely, absolutely.

01;36;19;07 - 01;36;19;29

If you work in

01;36;19;29 - 01;36;23;19

IT, Infotech research Group is a name

you need to know.

01;36;23;22 - 01;36;26;22

No matter what your needs are, Infotech

has you covered.

01;36;26;27 - 01;36;28;04

AI strategy?

01;36;28;04 - 01;36;30;16

Covered. Disaster recovery?

01;36;30;16 - 01;36;31;16

Covered.

01;36;31;16 - 01;36;34;01

Vendor negotiation? Covered.

01;36;34;01 - 01;36;37;24

Infotech supports you with the best

practice research and a team of analysts

01;36;37;24 - 01;36;41;17

standing by ready to help you

tackle your toughest challenges.

01;36;41;20 - 01;36;44;20

Check it out at the link below

and don't forget to like and subscribe!