Digital Disruption

00;00;00;24 - 00;00;01;24

Hey everyone!

00;00;01;24 - 00;00;03;14

I'm super excited to be sitting down

00;00;03;14 - 00;00;07;10

with the authors of the icon, Doctor

Emily Bender and Doctor Alex Hannah.

00;00;07;26 - 00;00;11;06

Doctor Bender is a professor

at the University of Washington,

00;00;11;10 - 00;00;15;11

and Doctor Hannah is director of research

at the distributed AI Research Institute.

00;00;15;29 - 00;00;18;24

What's cool about these two

is that they're maybe the most vocal

00;00;18;24 - 00;00;19;28

critics of AI.

00;00;19;28 - 00;00;22;27

You'll hear anywhere

and think the whole thing is bullshit.

00;00;23;06 - 00;00;24;21

My words, not theirs.

00;00;24;21 - 00;00;27;24

What I want to ask them

is just how far the distrust goes.

00;00;28;04 - 00;00;30;29

What do they really think

this technology is good for?

00;00;30;29 - 00;00;34;16

And if it's as bad as they say,

what do we need to do to get the future

00;00;34;16 - 00;00;36;01

we actually want?

00;00;36;01 - 00;00;39;01

Let's find out.

00;00;39;25 - 00;00;40;24

I'm so excited to be

00;00;40;24 - 00;00;44;18

joined today by Doctor

Emily Bender and Doctor Alex Hannah.

00;00;45;02 - 00;00;47;07

Thanks so much to both of you

for joining today.

00;00;47;07 - 00;00;51;03

I wanted to start by just, you know,

asking a little bit about, you know,

00;00;51;03 - 00;00;54;11

the message that you two have been,

you know, sort of road showing these days.

00;00;54;11 - 00;00;56;12

You know,

you've got a very clear perspective

00;00;56;12 - 00;00;59;24

on, on AI, on,

on kind of the future of this technology.

00;01;00;07 - 00;01;03;13

And maybe for those who don't know,

could you lay that out for us?

00;01;04;01 - 00;01;04;22

Yeah.

00;01;04;22 - 00;01;09;28

So the book is called The Icon, and it is

what it says on the tin that AI is on.

00;01;10;03 - 00;01;14;28

First off, the Shin AI itself

is not a coherent set of technologies.

00;01;14;29 - 00;01;16;11

It is a marketing term

00;01;16;11 - 00;01;20;26

and has been from the beginning,

from the initial convening in 1956,

00;01;20;26 - 00;01;24;12

in which John McCarthy and Marvin Minsky

invited a bunch of folks

00;01;24;28 - 00;01;29;22

to Dartmouth College to have a discussion

around, quote unquote, thinking machines.

00;01;30;03 - 00;01;31;16

So that's one part of it.

00;01;31;16 - 00;01;35;12

The second part of it

is that the current era of er, AI,

00;01;35;13 - 00;01;38;15

the generative AI tools,

including large language models

00;01;38;15 - 00;01;42;11

and diffusion

models, really are premised on this idea

00;01;42;11 - 00;01;45;08

that there is a thinking behind mind

behind that.

00;01;45;08 - 00;01;48;01

That itself is a

can of a fly in the frame.

00;01;48;01 - 00;01;48;27

I'm so sorry.

00;01;50;20 - 00;01;51;19

And so then that

00;01;51;19 - 00;01;55;19

is, and we do that for one,

you know, a few different reasons.

00;01;55;19 - 00;02;00;13

One of them being that

there is a human desire to language

00;02;00;13 - 00;02;04;20

to a, to the synthetic, media,

00;02;04;20 - 00;02;08;27

but specifically the synthetic text

that is an output of these models.

00;02;09;14 - 00;02;12;04

And that leads us

to a whole bunch of different things.

00;02;12;04 - 00;02;13;16

If there's a,

00;02;13;16 - 00;02;17;18

potential mind behind these technologies,

then that means that they can be

00;02;17;18 - 00;02;23;10

a replacement for so many different types

of, things that require humans.

00;02;23;12 - 00;02;25;21

Things like white collar work,

00;02;25;21 - 00;02;29;11

social services,

medical services, teaching and the like.

00;02;29;25 - 00;02;33;21

I was just going to add quickly

to amplify the last point that because

00;02;33;21 - 00;02;37;05

especially with the large language models,

and let me back up one second and say,

00;02;37;05 - 00;02;40;25

I will never use the term artificial

intelligence to refer to technology,

00;02;41;01 - 00;02;44;04

because I think it is a misnomer,

and I think it just confuses things.

00;02;44;04 - 00;02;45;18

And so I will talk about automation

00;02;45;18 - 00;02;48;18

to talk about things in general

or name the specific technology.

00;02;48;21 - 00;02;51;17

So in the case of large language models,

especially when they are used

00;02;51;17 - 00;02;56;08

as synthetic text extruding machines,

we experience language.

00;02;56;09 - 00;02;58;20

And then we are very quick

to interpret that language.

00;02;58;20 - 00;03;02;01

And the way we interpret it involves

imagining a mind behind the text.

00;03;02;09 - 00;03;05;07

And we have these systems

that can output plausible looking text in

00;03;05;07 - 00;03;06;04

just about any topic.

00;03;06;04 - 00;03;09;29

And so it looks like we have nearly there

solutions to all kinds

00;03;09;29 - 00;03;12;29

of technological needs in society.

00;03;13;00 - 00;03;16;12

But it's all fake and we should not be

putting any credence into it.

00;03;17;21 - 00;03;19;03

I think that's so interesting.

00;03;19;03 - 00;03;21;14

And I'm I'm absolutely of the same mind,

by the way.

00;03;21;14 - 00;03;24;14

And I found myself laughing when I was,

when I was reading through your book,

00;03;25;11 - 00;03;28;23

you know, one of the first of all,

artificial intelligence.

00;03;28;23 - 00;03;29;22

I completely agree.

00;03;29;22 - 00;03;33;11

Like, first of all, I do have to

give credit because it is great marketing.

00;03;33;11 - 00;03;36;11

Like,

it just it's so evocative of something.

00;03;36;14 - 00;03;39;21

But you know, nobody can really seem

to define exactly what that is.

00;03;39;21 - 00;03;40;18

And of course, it,

00;03;40;18 - 00;03;43;22

you know, has all these ideas

and can be used for any purpose. But,

00;03;45;00 - 00;03;47;02

one of the things you do early on in

the book is you

00;03;47;02 - 00;03;50;11

kind of just pop that balloon by saying,

well, you know what?

00;03;50;11 - 00;03;53;03

If it wasn't called

artificial intelligence?

00;03;53;03 - 00;03;54;07

Can you share a little bit about,

00;03;55;08 - 00;03;56;24

you know, what what that sounds like.

00;03;56;24 - 00;04;00;13

And you know what

what why you encourage people to do that?

00;04;00;22 - 00;04;00;28

Yeah.

00;04;00;28 - 00;04;03;28

So we have a few fun alternatives

that we call on.

00;04;04;00 - 00;04;07;23

Early on in our podcast,

Alex coined Mathy Maths as a fun one.

00;04;08;08 - 00;04;12;00

There's also due to the Italian researcher

Stefano Torelli Salami,

00;04;12;04 - 00;04;14;10

which is an acronym

for Systematic Approaches

00;04;14;10 - 00;04;16;19

to Learning Algorithms

and Machine Inference.

00;04;16;19 - 00;04;19;12

And the funny thing about that is,

if you take the phrase

00;04;19;12 - 00;04;21;29

artificial intelligence

in a sentence like, you know, does

00;04;21;29 - 00;04;25;05

I understand

or can I help us make better decisions

00;04;25;05 - 00;04;27;18

and you replace it with mathy math

or salami?

00;04;27;18 - 00;04;29;27

It's immediately obvious

how ridiculous it is.

00;04;29;27 - 00;04;31;24

You know, does the salami understand?

00;04;31;24 - 00;04;34;01

Well,

the salami help us make better decisions.

00;04;34;01 - 00;04;36;07

It's. You know, it's absurd.

00;04;36;07 - 00;04;39;24

Just sort of putting that little flag in

there, I think is a really good reminder.

00;04;40;00 - 00;04;43;04

Now, I think that, that to me,

that just hits it

00;04;43;04 - 00;04;46;14

right on the head, and was exactly

what I was getting at there.

00;04;47;15 - 00;04;50;15

That there's so much people.

00;04;50;20 - 00;04;50;27

What?

00;04;50;27 - 00;04;54;06

Once you have the idea of a mind,

it just gets in your head, right?

00;04;54;06 - 00;04;58;02

And it doesn't actually gets the idea

that in some ways, like these algorithms

00;04;58;02 - 00;05;00;15

were describing, as you said,

they go back to the 50s, right?

00;05;00;15 - 00;05;04;00

As long as we've had computing,

we've had these notions.

00;05;04;00 - 00;05;07;00

And it seems like it seems like

it's really an extension of that.

00;05;07;12 - 00;05;08;26

I wanted to expand the question, though.

00;05;08;26 - 00;05;11;26

There's so much misinformation.

00;05;12;08 - 00;05;15;21

Dare I call it disinformation around

artificial intelligence

00;05;15;21 - 00;05;19;17

in this whole sphere

that benefits, you know,

00;05;19;17 - 00;05;22;24

big tech that benefits, you know,

specific people and organizations.

00;05;23;15 - 00;05;26;15

What do you think is the most dangerous

or the most dangerous

00;05;27;03 - 00;05;30;12

single myth or multiple myths

right now in this sphere?

00;05;31;21 - 00;05;32;08

Yeah, it's

00;05;32;08 - 00;05;36;08

hard to rank the myths behind all this

00;05;36;08 - 00;05;42;13

because all of this is so upsetting

and harmful.

00;05;43;04 - 00;05;47;18

I'd say that one of the things

that that I think underlies a lot of this

00;05;47;18 - 00;05;52;02

is that there's a notion of, let's say,

a singular type of intelligence.

00;05;52;16 - 00;05;56;08

And so that's something that's not really

well supported by science.

00;05;56;08 - 00;05;59;14

First off, that there's a notion

that intelligence can be

00;05;59;14 - 00;06;02;15

reduced to a single number.

00;06;02;27 - 00;06;06;18

And some of the stuff that has, you know,

some of the pseudoscience

00;06;06;18 - 00;06;09;24

that is supported, that is based

on a lot of eugenicist thought,

00;06;09;24 - 00;06;13;07

the idea that, like, you can have an IQ

test, the IQ test can be used

00;06;13;18 - 00;06;17;26

to rank order people and, but people and

00;06;19;04 - 00;06;22;20

and machines kind of in a line

and you can order them.

00;06;22;26 - 00;06;27;16

That's why you have people like, OpenAI

and I think

00;06;27;16 - 00;06;31;23

Amir Moratti had said that,

you know, we're going to develop GPT five.

00;06;31;24 - 00;06;37;13

It's one to have PhD level intelligence

or there's going to be, a street level,

00;06;38;26 - 00;06;39;28

agent that people

00;06;39;28 - 00;06;42;28

can, use for $20,000 a month.

00;06;43;06 - 00;06;44;11

And so it does a few things.

00;06;44;11 - 00;06;48;09

It reduces the notion of what

is what it means to be human to something

00;06;48;09 - 00;06;53;13

that is more computable or automatable,

which also has historical antecedents.

00;06;53;25 - 00;06;57;14

It also does this thing where

we're getting into this really dangerous

00;06;57;14 - 00;07;00;16

notion of what intelligence is

and what consciousness is,

00;07;00;23 - 00;07;04;02

as if one needs to be just smarter

to be conscious and therefore human.

00;07;04;17 - 00;07;07;17

And so that's a very dangerous line

to go down to.

00;07;08;00 - 00;07;09;09

And so,

00;07;09;09 - 00;07;14;03

you know, of the kind of rank ordering,

I mean, of kinds of applications,

00;07;14;04 - 00;07;17;14

the world,

it is probably not helpful to do that.

00;07;17;14 - 00;07;21;10

But that kind of foundation or myth around

AI itself is very dangerous.

00;07;22;12 - 00;07;25;12

And like Alex,

I have a hard time ranking these things

00;07;25;15 - 00;07;28;24

and I, you know,

want to add another piece of the puzzle,

00;07;28;24 - 00;07;31;04

which is another, another one of the myths

that I think is feeding into this

00;07;31;04 - 00;07;35;16

is this idea that if you take a system

that has unfathomably large data

00;07;35;16 - 00;07;39;07

inside of it,

it must therefore have an unbiased

00;07;39;07 - 00;07;42;13

sort of bird's eye view into

what's really going on in the world.

00;07;42;20 - 00;07;46;18

And that is definitely a kind of

wishful thinking, like we want there

00;07;46;18 - 00;07;50;27

to be some peak we could climb to from

which we could just understand everything.

00;07;51;05 - 00;07;53;23

But that's not how science works.

That's not how society works.

00;07;53;23 - 00;07;55;03

And it's not a path towards

00;07;55;03 - 00;07;58;09

building technologies that are functional,

let alone fair.

00;07;59;00 - 00;07;59;10

Right.

00;07;59;10 - 00;08;00;25

So, you know,

00;08;00;25 - 00;08;02;20

just just,

you know, following that thread,

00;08;02;20 - 00;08;05;09

it sounds like I mean, there's a few

things at play here. One of them is that

00;08;06;16 - 00;08;07;04

what we're

00;08;07;04 - 00;08;11;02

we have this kind of false foundation,

that intelligence itself

00;08;11;02 - 00;08;14;15

is anything more than this

kind of nebulous abstraction, right?

00;08;14;15 - 00;08;17;28

As soon as we say, oh, it's this,

we're already kind of kidding ourselves

00;08;17;28 - 00;08;20;28

and going down a dangerous road there.

00;08;21;12 - 00;08;23;13

And then it seems like there's almost.

00;08;23;13 - 00;08;26;28

And, you know, tell me if I'm, if I'm

misrepresenting you, but it seems like

00;08;27;11 - 00;08;32;16

there's this drive to pretend

we're creating something objective.

00;08;33;00 - 00;08;37;24

But this objective thing just also

so happens to look an awful lot like what?

00;08;37;24 - 00;08;42;06

Objectivity for those who don't have

a visual in heavy air quotes.

00;08;42;10 - 00;08;46;06

Looks like,

you know, from the from the office of,

00;08;46;06 - 00;08;49;06

you know, a California technology leader.

00;08;49;09 - 00;08;51;07

Is that fair? Yes.

00;08;51;07 - 00;08;54;11

And I mean, there's so much

in terms of scholarship that's showing

00;08;54;11 - 00;08;58;23

that the use of these systems

is anything but objective, right?

00;08;58;23 - 00;09;02;16

From the foundational

where from from Bellini and Gebru

00;09;02;23 - 00;09;06;13

to, you know, where from

that's like, island

00;09;06;24 - 00;09;11;26

can to my own work that's focused on the

data sets and their politics.

00;09;13;01 - 00;09;13;25

Really, any kind

00;09;13;25 - 00;09;17;09

of these systems

are making judgments of some kind,

00;09;17;09 - 00;09;22;00

and they are encoding

majoritarian notions of what it is to,

00;09;22;28 - 00;09;26;24

you know, have a face when it comes

to specifically the civil unions of Peru

00;09;27;02 - 00;09;29;27

or to what it looks like to be someone

00;09;29;27 - 00;09;32;27

in, a certain kind of job,

00;09;33;03 - 00;09;36;05

you know, which tends

to have higher status jobs that are,

00;09;36;05 - 00;09;39;05

that are whiter and Miller and so,

00;09;39;10 - 00;09;42;12

you know, there are those things

that could encode in that.

00;09;42;12 - 00;09;43;12

We have a

00;09;43;12 - 00;09;44;04

and we have a word

00;09;44;04 - 00;09;48;02

for what it means when we trust machines

to have an objective view.

00;09;48;02 - 00;09;50;04

And that's called automation bias.

00;09;50;04 - 00;09;53;06

The notion that, like,

we're going to see certain kinds of,

00;09;54;21 - 00;09;57;21

operations to something that's automated.

00;09;57;24 - 00;10;02;01

And so that objective view

is certainly not objective, is it is just,

00;10;02;18 - 00;10;07;04

reifying views

that are, that are, in the majority and,

00;10;07;11 - 00;10;10;11

and really to the detriment of people

who are not in that majority.

00;10;10;16 - 00;10;14;27

Along those lines, one of the themes or

sort of theses in your work seems to be,

00;10;15;11 - 00;10;19;22

you know, this notion that we can't

actually disentangle the technology

00;10;20;04 - 00;10;24;00

from the creators

and owners of the technology.

00;10;24;00 - 00;10;27;20

And I wanted to, you know, maybe ask you

to expand on that a little bit for people

00;10;27;20 - 00;10;31;20

who are, you know, just kind of learning

about this concept for the first time.

00;10;31;26 - 00;10;33;26

Yeah, absolutely. I think one of the,

00;10;35;02 - 00;10;36;17

parts of the con, in using

00;10;36;17 - 00;10;40;09

the phrase artificial intelligence is that

it's a way to displace accountability.

00;10;40;14 - 00;10;43;15

And if you look at the language

that's used around this, oftentimes

00;10;43;15 - 00;10;46;20

we put the so-called AI systems

in agent position.

00;10;46;29 - 00;10;49;28

We'll say, you know, ChatGPT

00;10;49;28 - 00;10;52;21

has, scraped the whole web, for example.

00;10;52;21 - 00;10;53;28

Chatbot didn't do that.

00;10;53;28 - 00;10;55;14

The engineers at OpenAI did.

00;10;55;14 - 00;10;58;16

And some other people

whose data, the data sets they repurposed.

00;10;58;19 - 00;10;59;14

Right.

00;10;59;14 - 00;11;00;27

And also, you can't scrape the whole web.

00;11;00;27 - 00;11;03;21

That's a whole separate conversation.

00;11;03;21 - 00;11;06;22

But to, to basically

always keep the people in the frame

00;11;06;22 - 00;11;09;01

and say, who built it? For what purpose

did they build it?

00;11;09;01 - 00;11;11;26

Who's using it? Who are they using it on?

00;11;11;26 - 00;11;16;05

And also whose labor was either

just flat out appropriated

00;11;16;05 - 00;11;18;17

or otherwise exploited

in the production of the systems,

00;11;18;17 - 00;11;21;09

really helped

keep the conversation grounded.

00;11;21;09 - 00;11;24;12

It's also helpful to note that in in,

00;11;24;18 - 00;11;29;08

I think so much of this conversation,

00;11;31;18 - 00;11;34;06

there's a notion that

00;11;34;06 - 00;11;37;19

these companies that are building

this are somehow

00;11;37;20 - 00;11;40;25

magnanimous, that they're doing this

for the benefit of all humanity.

00;11;41;08 - 00;11;44;16

This is something that's really well

underscored and sharing housework called

00;11;44;18 - 00;11;48;01

Empire of AI,

that focuses on the kind of drama,

00;11;49;03 - 00;11;51;29

empire of AI

that focuses specifically on OpenAI

00;11;51;29 - 00;11;56;03

and this sort of drama behind OpenAI,

and specifically on how

00;11;56;07 - 00;11;59;07

so many of the people behind

that technology are very fallible,

00;12;00;06 - 00;12;04;12

specifically around Tim Altman

and many that that surround him.

00;12;05;03 - 00;12;09;08

And so it's really, you know,

one thing that highlights in that Daniel

00;12;09;11 - 00;12;14;14

highlights is the way that you don't

come up with magnanimous technologies.

00;12;14;14 - 00;12;15;09

I mean, technologies

00;12;15;09 - 00;12;18;29

that actually work for people are ones

that are closer to those people

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

and the people themselves

that are built by those people,

00;12;22;06 - 00;12;25;06

or for those people, ideally by them.

00;12;25;08 - 00;12;30;00

And so, you know, like the idea

that there is no human in the frame is,

00;12;30;01 - 00;12;33;14

is a mechanism

not only to displace accountability,

00;12;33;14 - 00;12;36;14

but to displace

where the power really lies.

00;12;36;15 - 00;12;37;10

Right.

00;12;37;10 - 00;12;40;26

And one of the, you know, as I think about

that and I think about this notion of,

00;12;40;27 - 00;12;43;28

you know, benevolence,

the other sort of party line

00;12;43;28 - 00;12;46;02

that I've been hearing more and more

come out of that,

00;12;46;02 - 00;12;50;18

you know, from come out of those

organizations is that there's

00;12;51;27 - 00;12;53;19

the it's described as kind of a race.

00;12;53;19 - 00;12;55;10

And, oh, we have to keep doubling down.

00;12;55;10 - 00;12;58;29

We have to keep investing in this because,

you know, it's winner take all.

00;12;58;29 - 00;13;01;12

And like there's this conflicts here.

00;13;01;12 - 00;13;03;09

Right. That oh it's for everybody.

00;13;03;09 - 00;13;05;19

But we have to do it for everybody.

00;13;05;19 - 00;13;07;05

And and you know what.

00;13;07;05 - 00;13;10;08

What's your reaction to to

you know that narrative.

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

The the race framing is just ridiculous.

00;13;13;00 - 00;13;16;14

It's it's based on a misconception

of how science and technology work.

00;13;16;29 - 00;13;19;29

And this is some, some thinking

that I learned from Beth Singler,

00;13;20;13 - 00;13;23;03

who's looked into this in detail

says, look, you can look backwards

00;13;23;03 - 00;13;27;11

and trace the path of sort of what

built on what to get to where we are now.

00;13;27;11 - 00;13;29;21

And you can draw a straight line

if you want.

00;13;29;21 - 00;13;32;19

But looking to the future,

the future doesn't exist ahead of time.

00;13;32;19 - 00;13;34;25

And the people who say, you know,

00;13;34;25 - 00;13;38;06

artificial intelligence or artificial

general intelligence is at the end of this

00;13;38;06 - 00;13;40;09

path, and it's just a question of who

runs down it

00;13;40;09 - 00;13;44;09

the fastest, have completely misunderstood

how science works, right?

00;13;44;09 - 00;13;46;08

So first of all, there's

this an ill defined notion

00;13;46;08 - 00;13;47;08

that they're running towards,

00;13;47;08 - 00;13;50;06

but they are asserting that it's there

and it's definitely existing.

00;13;50;06 - 00;13;51;12

And we can get there.

00;13;51;12 - 00;13;55;05

And also asserting and that I think,

is a very Silicon Valley brain

00;13;55;05 - 00;13;56;11

way of thinking about things,

00;13;56;11 - 00;13;59;14

that the person who gets there first

is, as you said, winner takes all.

00;13;59;14 - 00;14;00;29

And you'll sometimes hear people say,

00;14;00;29 - 00;14;05;12

well, we have to build the good

AI or the good AGI less evil.

00;14;05;14 - 00;14;08;13

You know, opposition builds the bad one

00;14;08;13 - 00;14;12;07

as if somehow building one technology

could prevent the building of another.

00;14;12;07 - 00;14;14;11

Like it

actually just doesn't make any sense.

00;14;14;11 - 00;14;18;09

And just to build on that,

I mean, specifically, the focus here is on

00;14;18;17 - 00;14;21;16

that the US has to build this

before China.

00;14;21;16 - 00;14;25;05

And so the boogeyman here

is, is the kind of sign of phobic idea

00;14;25;05 - 00;14;28;28

that the Chinese like AI or AGI

or whatever

00;14;28;28 - 00;14;32;29

it is, is going to be by virtue,

authoritarian.

00;14;33;06 - 00;14;37;06

Given that political environment,

they're going to say anything about what

00;14;37;09 - 00;14;41;07

Chinese builders are doing,

but as if what we're doing in the US

00;14;41;07 - 00;14;42;17

isn't authoritarian.

00;14;42;17 - 00;14;46;01

By being built by one company

that is commanding

00;14;46;01 - 00;14;49;18

so many resources,

that has so many implications with,

00;14;50;17 - 00;14;51;11

with,

00;14;51;11 - 00;14;55;05

national security organizations,

with with large energy companies.

00;14;55;12 - 00;14;58;12

I mean, these are things

which are centralization of power.

00;14;58;12 - 00;15;03;12

And that itself is not

as if it is a democratic version, the AI.

00;15;03;27 - 00;15;04;07

Right.

00;15;04;07 - 00;15;07;24

And so I think there's a real way

in which those even the claim around

00;15;08;07 - 00;15;11;05

the race dynamics also,

00;15;11;05 - 00;15;15;02

obscure what is happening

in terms of power centralization

00;15;15;12 - 00;15;18;12

and how that's undermining

democratic dynamics

00;15;18;13 - 00;15;21;13

in the US and in the West.

00;15;21;15 - 00;15;26;03

The centralization of power comment

is it's really interesting to me, right?

00;15;26;03 - 00;15;26;26

Because that's

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

that's another one of these tensions

in another one of these narratives about,

00;15;31;05 - 00;15;31;27

it's benevolent.

00;15;31;27 - 00;15;33;04

It's for everybody.

00;15;33;04 - 00;15;35;28

It's going to make everybody better.

00;15;35;28 - 00;15;38;00

Or except we own the platform. Right.

00;15;38;00 - 00;15;40;14

And by the way, we're charging you,

you know, 20 bucks a month

00;15;40;14 - 00;15;43;19

to have this new, better life,

which I find really interesting.

00;15;43;24 - 00;15;47;23

Is there is there merit in your minds

to, you know, there's a lot of talk about,

00;15;49;00 - 00;15;50;26

you know, maybe maybe,

00;15;50;26 - 00;15;55;10

you know, generative AI, maybe automation,

you know, some of these functions

00;15;55;18 - 00;16;00;22

might not be as good at professionals

at any given task,

00;16;01;03 - 00;16;05;07

but they they extend the entire market.

00;16;05;07 - 00;16;07;06

And market's

a bit of a dangerous word there,

00;16;07;06 - 00;16;10;15

but they extend the number of people

who services can be delivered to,

00;16;10;15 - 00;16;12;25

and they go

after the historically marginalized

00;16;12;25 - 00;16;16;25

in a way that is empowering

or makes their lives better.

00;16;16;28 - 00;16;18;23

Do you do you buy that?

00;16;18;23 - 00;16;21;23

Do you have optimism there or

do you think that's just part of the con?

00;16;21;24 - 00;16;23;05

It's really just part of the con.

00;16;23;05 - 00;16;26;17

The, the argument is always, well,

00;16;26;17 - 00;16;30;13

we don't have enough to provide

good education to everyone.

00;16;30;13 - 00;16;32;15

We don't have enough to provide

good health care to everyone.

00;16;32;15 - 00;16;34;09

And so these poor people are left out.

00;16;34;09 - 00;16;36;05

And so this is better than nothing.

00;16;36;05 - 00;16;37;29

And anytime you hear

this is better than nothing.

00;16;37;29 - 00;16;40;06

The question should always be

why was the alternative?

00;16;40;06 - 00;16;42;28

Nothing.

Because we have enormous resources.

00;16;42;28 - 00;16;46;00

If you look at the resources

that are being poured into these systems

00;16;46;05 - 00;16;49;04

and imagine instead

those resources were used for

00;16;49;04 - 00;16;52;03

shoring up education systems

and health care systems.

00;16;52;19 - 00;16;54;09

Imagine what we could do with that.

00;16;54;09 - 00;16;57;12

You know, sort of rich

and like rich investment

00;16;57;12 - 00;17;00;14

view of social services

rather than an austerity view.

00;17;00;24 - 00;17;04;06

And very much it's tied into

I mean, when you get behind this,

00;17;04;06 - 00;17;07;29

it's tied into this idea

that those of means are going to get

00;17;08;07 - 00;17;12;29

these bespoke human oriented services

that are by humans.

00;17;12;29 - 00;17;15;28

And we already see this

in other kind of technological domains.

00;17;16;10 - 00;17;20;08

Adrian Williams was on our podcast,

and she's a former charter school teacher.

00;17;20;08 - 00;17;23;29

And one of the things that she talks about

as being a former charter school teacher

00;17;23;29 - 00;17;27;06

is that when it comes to edtech, things

00;17;27;06 - 00;17;31;03

like Google Classroom and, classes, like,

00;17;32;09 - 00;17;32;26

that,

00;17;32;26 - 00;17;35;25

or content management systems

specifically in education,

00;17;35;25 - 00;17;40;08

they are disproportionately

offered in places where the students

00;17;40;08 - 00;17;44;01

are lower income, black and brown,

in lower income communities.

00;17;44;17 - 00;17;46;05

If you go to private schools,

00;17;46;05 - 00;17;50;02

there's very few screens or those screens

are used in a much different way.

00;17;50;02 - 00;17;53;23

They're not used for surveillance or use,

you know, they're pretty optional or,

00;17;53;23 - 00;17;56;23

you know, they're they're much more pen

and paper oriented.

00;17;57;06 - 00;17;59;19

And so we're seeing that dynamic.

00;17;59;19 - 00;18;04;10

And that's not actually providing

more opportunities

00;18;04;10 - 00;18;05;18

to people in those communities.

00;18;05;18 - 00;18;08;18

It is more heavily

surveilling those communities.

00;18;08;26 - 00;18;11;26

It is being used as a way to also,

00;18;13;06 - 00;18;17;08

say that there's being an offer of,

of that technology that they have access,

00;18;17;08 - 00;18;21;09

but that's not actually education

the same way where, you know,

00;18;21;09 - 00;18;24;09

one of the things that I boosters

are very excited about is,

00;18;25;04 - 00;18;28;16

following Ilya Sutskever

while he cheap and effective therapy.

00;18;29;07 - 00;18;34;08

Because people cannot afford therapists

that could speak to them specifically.

00;18;34;13 - 00;18;35;27

The people aren't getting therapy,

00;18;35;27 - 00;18;39;03

they're getting some kind of a machine

that might just be telling them

00;18;39;03 - 00;18;43;10

what they want to hear, or leading them

down to dangerous kind of, particular

00;18;43;10 - 00;18;47;27

behaviors, or maybe encouraging them

to engage in self-harm.

00;18;48;14 - 00;18;51;01

And so it's very what really gives away

the game

00;18;51;01 - 00;18;54;04

is this comment from Greg Corrado,

who is health of,

00;18;54;26 - 00;18;57;29

head of health

AI at Google Health stage of using term.

00;18;58;24 - 00;19;03;22

And he was debuting med palm to a Google

I o in a press junket

00;19;03;22 - 00;19;08;09

he had said to the Wall Street Journal,

you know, this thing is not something

00;19;08;09 - 00;19;10;22

I'd want in my own family's

medical journey,

00;19;10;22 - 00;19;13;08

but I'm very excited for it

to be available to everybody else.

00;19;14;26 - 00;19;16;29

Wow. Yeah.

00;19;16;29 - 00;19;19;08

What what

more do you need to say about a product

00;19;19;08 - 00;19;22;09

than, you know, I wouldn't

I wouldn't have my family use it.

00;19;22;09 - 00;19;23;25

Right.

00;19;23;25 - 00;19;27;00

It's like the, you know, the

the quotes from the, you know, heads of,

00;19;27;01 - 00;19;28;05

you know,

00;19;28;05 - 00;19;29;17

social media network saying, well,

00;19;29;17 - 00;19;31;14

I certainly wouldn't

want my kids on this, right.

00;19;31;14 - 00;19;34;14

It's like, okay, well that that's

that's the whole ballgame ban.

00;19;34;24 - 00;19;35;21

That's right. Yeah.

00;19;35;21 - 00;19;38;19

I think Sam Altman,

that one of the recent, I think the recent

00;19;38;19 - 00;19;41;19

Commerce meeting was asked by someone,

00;19;41;19 - 00;19;44;19

you know, you wouldn't want your,

00;19;44;19 - 00;19;48;13

your new kid to you know, be friends

with, with an AI agent.

00;19;48;13 - 00;19;51;13

And he was like, no way I would. So,

00;19;52;15 - 00;19;54;13

I think in a lot of the tension.

00;19;54;13 - 00;19;55;24

Very, very telling.

00;19;55;24 - 00;19;57;29

I just want to add that

I think that that shows that

00;19;57;29 - 00;20;00;11

that these folks don't see

the rest of the world is really people.

00;20;01;15 - 00;20;03;15

And then sort of

00;20;03;15 - 00;20;06;19

reveals the lie of saying, well, we're

doing this for the benefit of humanity,

00;20;06;22 - 00;20;08;20

but these services are not services

00;20;08;20 - 00;20;10;21

that we would consider

good enough for our families.

00;20;10;21 - 00;20;13;21

And so everybody else doesn't count

the way our families count.

00;20;14;08 - 00;20;14;18

Right?

00;20;14;18 - 00;20;18;13

There's almost like this,

like self deification or something, right?

00;20;18;13 - 00;20;21;11

Like there's we're

we're in some way outside humanity.

00;20;21;11 - 00;20;23;17

Humanity is this project.

00;20;23;17 - 00;20;24;25

And we are these,

00;20;24;25 - 00;20;28;14

you know, saviors who are going to come in

and tell you what's good for you.

00;20;29;01 - 00;20;30;20

Exactly.

00;20;30;20 - 00;20;34;14

So, you know, in a couple of minutes,

I want to come back to this, this notion

00;20;34;14 - 00;20;35;17

and what we do about it.

00;20;35;17 - 00;20;40;05

But just just before we get too deep into,

you know, what each of us can do.

00;20;40;22 - 00;20;44;09

I wanted to come back to something

you mentioned earlier, which is AGI.

00;20;45;01 - 00;20;47;10

And as we look at some of the,

00;20;47;10 - 00;20;49;26

you know, the big promises

00;20;49;26 - 00;20;53;19

or the big,

you know, stories around, you know,

00;20;53;19 - 00;20;56;23

technology right now, two of the big ones,

we talk about our AGI.

00;20;57;01 - 00;21;00;12

And on the opposite end of the spectrum,

this, you know, p doom

00;21;00;12 - 00;21;04;04

or probability of doom that it's going to,

you know, wipe out the human race.

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

What what you know,

00;21;05;01 - 00;21;08;12

what's your outlook around,

you know, both of these in the next.

00;21;08;12 - 00;21;08;21

Yeah.

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

You know, in

any sort of reasonable time horizon.

00;21;11;26 - 00;21;13;29

I mean any time horizon at all. Right.

00;21;13;29 - 00;21;15;25

So it's it's all fiction.

00;21;15;25 - 00;21;17;23

And one of the things

it's very frustrating to me

00;21;17;23 - 00;21;20;18

is the way that the dreamers,

the people who think this is

00;21;20;18 - 00;21;22;19

going to be the end of humanity

and the boosters,

00;21;22;19 - 00;21;24;28

the ones who think it's going to solve

all of our problems,

00;21;24;28 - 00;21;27;29

present themselves

as like two ends of a spectrum.

00;21;28;07 - 00;21;31;00

And the media picks

this up and sort of amplifies

00;21;31;00 - 00;21;34;05

it and gives the idea that there's

there's just one spectrum,

00;21;34;05 - 00;21;37;08

so you're either fully doomed

or fully booster or somewhere in between.

00;21;37;16 - 00;21;39;20

But in fact, it's

two sides of the same coin.

00;21;39;20 - 00;21;41;26

So the Duma is, say,

00;21;41;26 - 00;21;45;14

artificial intelligence or artificial

general intelligence is a thing.

00;21;45;20 - 00;21;48;11

It's imminent. It's inevitable.

00;21;48;11 - 00;21;49;17

And it's going to kill us all.

00;21;49;17 - 00;21;52;22

And the boosters say AI

slash AGI is a thing.

00;21;52;27 - 00;21;55;27

It's inevitable, it's imminent, and it's

going to solve all of our problems.

00;21;55;28 - 00;21;57;19

And putting it that way,

I hope it makes it very clear

00;21;57;19 - 00;22;00;13

that these are actually the same position,

and there's no daylight between them.

00;22;00;13 - 00;22;04;06

And it's just once

you've gone down this fantasy path

00;22;04;10 - 00;22;06;03

which turn you take at the end.

00;22;06;03 - 00;22;06;12

Yeah.

00;22;06;12 - 00;22;09;24

So and

and just to be put a finer point on it to

00;22;10;06 - 00;22;13;22

there's a notion of doom

and there's a notion of hope.

00;22;13;22 - 00;22;17;03

And I think

to what we initially want to say

00;22;17;03 - 00;22;21;17

is we reject this probability

framing at all as if one can one

00;22;21;17 - 00;22;25;26

imagine this as the question

of completely fake probabilities.

00;22;26;20 - 00;22;29;10

And the thing that really sets me off,

especially

00;22;29;10 - 00;22;32;10

about, folks like the rationalists.

00;22;32;17 - 00;22;33;13

So if it's like,

00;22;37;00 - 00;22;38;29

Daniel, Carter,

00;22;38;29 - 00;22;42;26

that's who I completely

just completely mangled his name.

00;22;43;03 - 00;22;46;10

But the one of the authors of the ER,

2027 document

00;22;46;22 - 00;22;51;22

and others, of his colleagues

is that they're putting completely fake

00;22;51;22 - 00;22;55;03

probability leads to fake events.

00;22;56;20 - 00;22;58;23

And it's really in for folks

00;22;58;23 - 00;23;01;27

that really envision themselves

to be very empirically minded.

00;23;01;27 - 00;23;06;15

It's very it's just very much

just made up kinds of time horizons.

00;23;06;15 - 00;23;08;24

And it's it's really a work.

00;23;08;24 - 00;23;12;09

And Emily and I are both social scientists

and are both empiricist.

00;23;12;09 - 00;23;15;09

So that's actually very weird to see.

00;23;16;02 - 00;23;17;05

So that's one thing about that.

00;23;17;05 - 00;23;21;22

And then AGI itself serves

as this very nebulous concept as well,

00;23;21;22 - 00;23;24;22

this kind of idea

that there is some kind of a,

00;23;25;09 - 00;23;26;15

intelligence that.

00;23;26;15 - 00;23;30;21

So it's going to be,

very capable at many different tasks.

00;23;31;25 - 00;23;33;14

And we're

00;23;33;14 - 00;23;36;14

not really sure how well defined that is.

00;23;36;14 - 00;23;38;05

I mean, it doesn't seem

really well-defined at all.

00;23;38;05 - 00;23;41;09

We just, recently, published

00;23;41;09 - 00;23;46;00

a, a piece at Tech

Policy Press called The Myth of AGI

00;23;46;07 - 00;23;49;13

that was focusing

specifically on the idea on

00;23;50;18 - 00;23;53;18

the idea that this notion is nebulous,

00;23;54;00 - 00;23;56;00

and that it has a very particular

00;23;56;00 - 00;23;59;00

view of the world is wildly, scoped.

00;23;59;07 - 00;24;01;29

And somehow there's supposed to be,

00;24;01;29 - 00;24;06;13

you know, OpenAI or anthropic

being the, the particular organization

00;24;06;13 - 00;24;10;02

that is going

to, ensure that we receive this.

00;24;10;22 - 00;24;14;05

And so that itself is,

you know, like very,

00;24;15;09 - 00;24;17;01

very wrongheaded.

00;24;17;01 - 00;24;18;13

Right.

00;24;18;13 - 00;24;20;06

So I'm just no, it's it's interesting.

00;24;20;06 - 00;24;23;27

And I'm just I'm just reflecting on that,

a little bit because if you

00;24;24;23 - 00;24;28;29

if you take all that and sort of

synthesize it, that most of what

00;24;28;29 - 00;24;32;16

we're hearing in this space is,

you know, as you described a con, right?

00;24;32;16 - 00;24;33;18

It's marketing.

00;24;33;18 - 00;24;34;19

It's nonsense.

00;24;34;19 - 00;24;37;25

I mean, I'm not going to ask you

if you think this is a bubble.

00;24;37;25 - 00;24;40;19

I think that the writing

is on the wall there.

00;24;40;19 - 00;24;45;00

But I did want to ask you, you know,

do you think this bubble is going to burst

00;24;45;07 - 00;24;48;01

and is it already on a trajectory

to burst?

00;24;48;01 - 00;24;51;15

Do we, as you know, individuals

00;24;51;15 - 00;24;55;16

and maybe as leaders need to do something

differently for it to burst?

00;24;55;16 - 00;24;58;16

And you know, assuming it does, what

do we think that's going to look like?

00;24;58;23 - 00;25;00;20

Yeah, I mean it's going to burst.

00;25;00;20 - 00;25;05;11

And I think it's not a question of of

if it's a question of when.

00;25;05;19 - 00;25;08;11

And it's not a question of

00;25;08;11 - 00;25;11;04

and moreover, it's

a question of how. Right.

00;25;11;04 - 00;25;14;19

And so there's a few different ways

this bubble burst, a burst big.

00;25;15;02 - 00;25;20;08

And we got sort of the inklings of that

when we saw the freak out around Deep Sea,

00;25;20;19 - 00;25;24;11

in which Nvidia lost

a whole bunch of value in its stock.

00;25;25;13 - 00;25;25;22

And they

00;25;25;22 - 00;25;28;22

said, oh, maybe we don't need

these massive data centers to do so.

00;25;28;22 - 00;25;31;15

But the bigger you know, the bigger worry

00;25;31;15 - 00;25;34;15

here is not that there's going to be

this is really more efficient models.

00;25;34;20 - 00;25;39;17

It's that it's not going to solve

productivity or solve wages

00;25;39;25 - 00;25;42;25

in the way that the boosters

think it will.

00;25;43;05 - 00;25;46;05

And so a lot of,

you know, a lot of people have already

00;25;46;12 - 00;25;49;20

said, well, there's a huge revenue bubble

that needs to be,

00;25;50;14 - 00;25;53;12

you know, that needs to be,

00;25;53;12 - 00;25;55;09

sorry, there's a huge debt bubble

00;25;55;09 - 00;25;58;11

in terms of all the, all the,

00;25;59;10 - 00;26;03;12

the GPUs and the data centers and all that

infrastructure needs to get paid for.

00;26;04;09 - 00;26;08;14

And at some point they're

the facade is going to come.

00;26;08;14 - 00;26;12;16

And is that going to be something like

the first AI winter in which, you know,

00;26;12;16 - 00;26;16;18

there was the, the Lighthouse report

in which they said, okay,

00;26;16;27 - 00;26;20;08

this isn't this isn't panning out,

no more government funding,

00;26;20;19 - 00;26;23;17

or is it going to be something

that looks a lot more like Uber,

00;26;23;17 - 00;26;26;17

where more money gets thrown at it

00;26;26;19 - 00;26;30;28

over and over and over again,

until finally a profit gets turned?

00;26;32;03 - 00;26;34;18

The thing is, so much

00;26;34;18 - 00;26;37;21

money is going to get thrown on it,

and it's not going to turn a profit.

00;26;37;25 - 00;26;40;25

The revenue margins have been so low

00;26;41;01 - 00;26;44;07

and the investment has been so high.

00;26;44;17 - 00;26;47;17

So, so either going to happen

or it's going to happen quickly.

00;26;48;07 - 00;26;51;07

To your question of what

we should be doing about it, I think that,

00;26;51;16 - 00;26;53;01

certainly we should be resisting the hype.

00;26;53;01 - 00;26;54;06

And that's part of our goal

00;26;54;06 - 00;26;58;07

in writing this book, is to help people

articulate their objections to the hype.

00;26;58;07 - 00;27;02;22

But I think also it's really important

to not let systems that we rely on

00;27;02;22 - 00;27;06;13

get rebuilt around the false promises

of artificial intelligence,

00;27;06;21 - 00;27;10;02

because it's going to be harming us

while the bubble is still going.

00;27;10;02 - 00;27;12;01

And instead of getting, you know,

00;27;13;06 - 00;27;16;20

actual thoughtful medical notes

at the end of a doctor's appointment,

00;27;16;20 - 00;27;19;28

we get the output of a synthetic text

extruding machine with some errors

00;27;19;28 - 00;27;23;22

and the doctor saying, well,

I it's not my fault if it's wrong.

00;27;24;03 - 00;27;24;24

The system did it.

00;27;24;24 - 00;27;27;20

And, you know, medical

providers are under a lot of stress.

00;27;27;20 - 00;27;30;27

And in many cases this would happen

because their employer would say, well,

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

you got to see three extra people in a day

00;27;32;19 - 00;27;34;28

now because you're not spending time

doing the clinical notes or whatever.

00;27;34;28 - 00;27;37;28

So it can be harmful

while it's still going on.

00;27;38;00 - 00;27;41;06

But if you think ahead to

when the systems fall apart

00;27;41;06 - 00;27;44;07

and the vendors aren't there anymore,

so what?

00;27;44;08 - 00;27;46;05

How much has things been restructured?

00;27;46;05 - 00;27;48;25

How much have people's jobs been changed?

00;27;48;25 - 00;27;53;11

Either people being asked to do more

in the same amount of time, or people's

00;27;53;11 - 00;27;56;20

jobs being turned from stable jobs

into very casualised

00;27;56;20 - 00;28;00;08

gig work jobs,

because the AI supposedly could do it.

00;28;00;08 - 00;28;02;23

And so the more we can resist that

restructuring,

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

the better off we're going to be,

00;28;04;00 - 00;28;08;02

sort of regardless of when the the scales

finally fall from people's eyes.

00;28;08;02 - 00;28;10;12

And we're not forced to have this

everywhere.

00;28;10;12 - 00;28;14;10

Moreover, and one thing

we mentioned in the book and Emily took up

00;28;14;19 - 00;28;18;27

a piece of that is we have a piece called

That the grimy residue of the AI bubble

00;28;19;08 - 00;28;24;07

and the job loss is one thing,

but we can think of two other things

00;28;24;07 - 00;28;28;06

that are going to be left over if we don't

try to do some mitigation right now.

00;28;28;06 - 00;28;31;05

One of them is the environmental damage

that's already been done,

00;28;31;15 - 00;28;34;17

both in terms of the carbon

that's been put into the atmosphere,

00;28;34;25 - 00;28;37;25

the water that's been used for data center

cooling.

00;28;37;27 - 00;28;40;27

Also the types of,

00;28;40;28 - 00;28;44;16

externalities

in terms of air pollution and

00;28;45;17 - 00;28;46;18

forever chemicals that are

00;28;46;18 - 00;28;49;23

put into the ground from semiconductor

or construction.

00;28;50;14 - 00;28;51;29

So those are hard to reverse.

00;28;51;29 - 00;28;56;18

So we're already on track to blow past

the goals of the Paris climate agreement.

00;28;57;04 - 00;29;00;19

And then the kind of spills

and the information ecosystem,

00;29;00;19 - 00;29;03;18

the idea that we already has

so much synthetic text out there,

00;29;04;13 - 00;29;06;10

it's going to be hard to suss out,

you know,

00;29;06;10 - 00;29;09;23

what is synthetic text

and what is not synthetic text.

00;29;10;05 - 00;29;14;22

And we already see that babble

being done, in places like Wikipedia,

00;29;14;22 - 00;29;19;14

which is trying to fight the onslaughts

of the Lem generated and therefore,

00;29;20;18 - 00;29;23;08

not very trustworthy outputs

00;29;23;08 - 00;29;26;10

of synthetic media, text, media machines.

00;29;27;02 - 00;29;30;12

So that's one part

thinking more of what we can do about it.

00;29;30;12 - 00;29;32;19

Well, one thing is trying to

00;29;32;19 - 00;29;35;29

be more cool on investment,

especially in data centers.

00;29;36;20 - 00;29;39;27

That is harming, communities

in the here and now.

00;29;40;05 - 00;29;44;16

And whether it's the data centers,

operating in southwest Memphis

00;29;44;29 - 00;29;49;03

that is a living box town,

predominantly black and poor neighborhood.

00;29;49;28 - 00;29;54;22

Or whether it is, areas

like Northern Virginia and London County

00;29;55;04 - 00;29;59;11

or the outskirts of Atlanta, where more

and more data centers are being built,

00;30;00;01 - 00;30;05;00

specifically relying on fossil

fuel powered, power plants.

00;30;05;17 - 00;30;05;28

Yeah.

00;30;05;28 - 00;30;09;09

I'm, I'm again,

just just kind of reflecting on that.

00;30;09;09 - 00;30;10;11

Alex, the

00;30;10;11 - 00;30;12;05

I have the same concern

about the environment,

00;30;12;05 - 00;30;14;19

and I'm just I'm tying

it back to your comment about Uber,

00;30;14;19 - 00;30;18;10

because the thing that the thing that

really gets me concerned about this is,

00;30;18;18 - 00;30;22;06

you know, Uber is kind of throwing good

money after bad for a long time, right?

00;30;22;06 - 00;30;23;16

It's the it's the old joke of,

00;30;23;16 - 00;30;25;22

you know, we're losing money

on Interac every interaction.

00;30;25;22 - 00;30;27;23

But we'll make it up at scale. Right.

00;30;27;23 - 00;30;31;11

But the the thing that worries me

about this, when you combine it with that

00;30;31;11 - 00;30;34;18

kind of arms race narrative, is it

feels like it's not even linear.

00;30;34;18 - 00;30;36;12

It's like exponential, right?

00;30;36;12 - 00;30;38;18

It's like everybody's saying like,

00;30;38;18 - 00;30;41;03

we need more and more data, more

and more data centers.

00;30;41;03 - 00;30;45;04

Like, I can very easily picture

kind of an asymptotic curve

00;30;45;04 - 00;30;48;11

where while we still haven't found

a solution, it's still not profitable.

00;30;48;11 - 00;30;49;05

And so.

00;30;49;05 - 00;30;54;03

Well, let's just try throwing ten times as

much energy or 100 times as much energy.

00;30;54;13 - 00;30;58;13

And, you know, either

this whole thing collapses is fake

00;30;58;13 - 00;31;00;07

or we do an awful lot of damage.

00;31;01;21 - 00;31;04;12

You know, in the interim there. So,

00;31;04;12 - 00;31;06;17

you know that that's certainly

a piece that resonates with me.

00;31;06;17 - 00;31;08;07

And, you know, Emily,

I was certainly thinking about,

00;31;08;07 - 00;31;14;03

you know, your comments on automation

and the need for us to be more thoughtful

00;31;14;03 - 00;31;17;19

about, you know, where we automate

and how we make sure we're not just

00;31;17;19 - 00;31;21;07

replacing a solid foundation

with a flimsy foundation.

00;31;21;10 - 00;31;25;20

One of the things about these ever larger

data centers, data sets,

00;31;26;00 - 00;31;29;06

and so on, is that it becomes a metric

00;31;29;06 - 00;31;32;21

that the people who are spending

the money can say, look, we made it bigger

00;31;32;28 - 00;31;37;10

because what they're trying to build

is actually not well formed.

00;31;37;10 - 00;31;38;13

It's not it's not well conceived.

00;31;38;13 - 00;31;40;03

And so you can't evaluate

how close you are

00;31;40;03 - 00;31;41;21

to building the thing

you're trying to build.

00;31;41;21 - 00;31;43;15

And so they've got something

they can measure instead.

00;31;43;15 - 00;31;46;16

And that thing is in fact

environmentally quite ruinous and built on

00;31;46;16 - 00;31;47;19

stolen labor and so on.

00;31;48;20 - 00;31;48;29

Right.

00;31;48;29 - 00;31;51;15

It's that more parameters,

more computations.

00;31;51;15 - 00;31;54;15

And, you know, that must be good

because it's because it's bigger.

00;31;55;04 - 00;31;57;12

I want it to come back to the,

00;31;57;12 - 00;32;00;27

you know, and put a finer point

on sort of the, the adoption piece.

00;32;00;27 - 00;32;04;25

And I'm thinking specifically for,

you know, leaders of industry for,

00;32;04;25 - 00;32;07;27

for organizational leaders

who are, you know, very much,

00;32;07;27 - 00;32;10;11

you know, everywhere you turn, I'm

sure you are exposed to it too.

00;32;10;11 - 00;32;13;10

I this I that if you don't do

it, you're behind the curve.

00;32;13;10 - 00;32;16;04

What do you have any specific guidance?

00;32;16;04 - 00;32;19;20

You know, for, for for these people

in terms of what they can do to,

00;32;20;14 - 00;32;23;13

I don't know, maybe be more responsible

here?

00;32;23;13 - 00;32;27;06

Does that mean just fully saying,

you know, no automation, none of this?

00;32;27;06 - 00;32;29;01

Or is there an approach they can take

00;32;29;01 - 00;32;32;14

that's just going to yield better results

and protect them from some of this?

00;32;32;25 - 00;32;33;15

You know.

00;32;33;15 - 00;32;35;19

There's certainly a time

and a place for automation,

00;32;35;19 - 00;32;37;05

but you want to automate something

00;32;37;05 - 00;32;40;06

when you can very specifically say

what it is that you're automating

00;32;40;15 - 00;32;44;07

and do you have very good reason

to believe that the information needed for

00;32;44;07 - 00;32;45;18

the output is in the input?

00;32;45;18 - 00;32;50;03

When you can evaluate how well it works

in your current use case, when you, have

00;32;50;09 - 00;32;55;04

sufficient recourse for things for someone

who's been harmed by the automation.

00;32;55;04 - 00;32;57;27

Because one of the things about automation

is that it scales whatever you're doing

00;32;57;27 - 00;33;00;27

and if it's getting it wrong in that run,

getting it wrong is harmful to people,

00;33;00;27 - 00;33;02;20

and you do it

more and more and faster and faster.

00;33;02;20 - 00;33;04;21

You've got to be prepared

to make things right.

00;33;06;05 - 00;33;08;14

Or, you know, decide no, that's too

harmful.

00;33;08;14 - 00;33;09;27

It's not the kind of thing

that can be made right.

00;33;09;27 - 00;33;11;07

We're not going to go down that path.

00;33;11;07 - 00;33;13;28

So, you know, advice to leaders

who are making decisions here.

00;33;13;28 - 00;33;16;28

I would say, first of all, you know,

think about values.

00;33;17;18 - 00;33;20;03

And I know that

for many people in business,

00;33;20;03 - 00;33;22;01

the only value that matters is stakeholder

value.

00;33;22;01 - 00;33;24;10

And the way to get stakeholder value

right now is a promise AI.

00;33;24;10 - 00;33;27;27

So setting that aside to say, okay,

well what are our other values?

00;33;28;15 - 00;33;33;06

And to what extent does

this automation actually speak to them.

00;33;33;06 - 00;33;38;06

And especially thinking about, you know,

what could go wrong when you put

00;33;38;15 - 00;33;40;11

if you're talking about something

like ChatGPT,

00;33;40;11 - 00;33;44;08

a synthetic text extruding machine,

into the middle of a sensitive process,

00;33;44;15 - 00;33;48;21

you know, how might that impact

what, you know, your company's reputation?

00;33;49;09 - 00;33;52;09

Or what it

is that you say that you stand for?

00;33;52;11 - 00;33;55;25

And think about sort of like the,

the long term durability of the systems

00;33;55;25 - 00;33;56;23

you're putting in place.

00;33;56;23 - 00;33;59;11

Is this still going to work with

this still work?

00;33;59;11 - 00;34;03;01

If OpenAI went belly up and didn't

have access to electricity anymore,

00;34;03;08 - 00;34;05;21

but this still work. You know, if,

00;34;06;20 - 00;34;07;17

it turned out that

00;34;07;17 - 00;34;10;17

large language models aren't all that

and so on.

00;34;10;19 - 00;34;11;17

Just a few things.

00;34;11;17 - 00;34;13;26

I mean, there's just a few data points

I want to add to that.

00;34;13;26 - 00;34;17;25

So there was a survey

that was done by an org planning

00;34;17;25 - 00;34;20;24

platform in the UK called Org View.

00;34;21;01 - 00;34;25;05

They found that, 55% of companies

that replace workers with,

00;34;25;05 - 00;34;28;05

I regret the decision.

00;34;28;18 - 00;34;31;18

So already

have some buyer's remorse happening.

00;34;31;19 - 00;34;34;28

And I'm assuming that there's a notion

that there would be,

00;34;36;17 - 00;34;40;09

workers that would be able to that

that still were there

00;34;40;09 - 00;34;44;08

could, using AI tools

to shore up the difference.

00;34;44;08 - 00;34;47;08

But that is just that is noxious.

00;34;48;04 - 00;34;52;23

The largest study that's been done or

one of the largest studies, I'm assuming,

00;34;53;17 - 00;34;57;20

but I don't know if there's been a larger

study that's been, done

00;34;57;20 - 00;35;00;23

in, in, in Denmark, of

00;35;00;23 - 00;35;04;13

25,000 workers across 7000 organizations,

00;35;05;06 - 00;35;09;22

suggested that there were very modest

productivity gains,

00;35;10;17 - 00;35;14;12

by workers

using artificial intelligence tools,

00;35;15;08 - 00;35;17;16

at the rate of 3%.

00;35;17;16 - 00;35;21;27

But those gains were offset

by the new labor

00;35;21;27 - 00;35;24;10

displacing tasks

that they had to deal with.

00;35;24;10 - 00;35;28;11

There was also no increase in earnings,

for those workers.

00;35;28;27 - 00;35;33;16

And so these are not proving to be

the end of the

00;35;33;16 - 00;35;37;21

the kind of huge

an amazing kind of productivity,

00;35;37;21 - 00;35;41;02

productivity gains that the companies

are making, making them out to be.

00;35;41;20 - 00;35;45;14

So the message to business leaders,

I would say, is really invest

00;35;45;14 - 00;35;49;11

in your people and really thinking about

what are your people doing?

00;35;49;17 - 00;35;50;25

How can you better support

00;35;50;25 - 00;35;53;25

people who are already doing

the work that you need to do?

00;35;53;28 - 00;35;58;25

If people are having an issue

with being productive

00;35;58;28 - 00;36;02;16

and answering what are ways in which

they can be supported organizationally

00;36;02;29 - 00;36;06;10

and thinking about this kind of

with an organizational sociologist

00;36;06;10 - 00;36;09;10

hat on and thinking about what are ways

in which

00;36;09;12 - 00;36;12;25

those, you know,

a product can be made better by that.

00;36;13;25 - 00;36;14;24

The reason why

00;36;14;24 - 00;36;19;09

large language models are very attractive

is because if their values,

00;36;19;09 - 00;36;22;18

as Emily suggests, are very much only for

00;36;23;11 - 00;36;26;11

or shareholders,

00;36;26;11 - 00;36;29;18

shareholders often like to see layoffs,

because that means you can,

00;36;30;08 - 00;36;33;08

be making, more profit per headcount.

00;36;33;15 - 00;36;38;01

But if your product is not,

keeping up the task and staying

00;36;38;01 - 00;36;41;24

as high quality as was then, that's

going to be a huge issue if you're not,

00;36;42;27 - 00;36;46;04

and so in some cases, there's

some really interesting cases

00;36;46;04 - 00;36;49;29

in which some companies are leaning into

not using AI and saying that,

00;36;49;29 - 00;36;52;29

no, we're actually giving you

a high touch experience

00;36;53;00 - 00;36;57;17

and that is very important to signify,

because I think

00;36;57;17 - 00;37;01;15

a lot of people are culturally saying

that AI is shoddy and jank.

00;37;01;15 - 00;37;04;15

And what Tressie

McMillan Cotner said it is bit

00;37;04;20 - 00;37;07;09

the outputs are not really verifiable.

00;37;08;29 - 00;37;10;21

Or if they

00;37;10;21 - 00;37;14;05

if you can verify them,

they are very labor intensive to do so.

00;37;14;27 - 00;37;20;08

The outputs of images are very shoddy

and take a lot of labor to correct,

00;37;20;08 - 00;37;23;08

especially down in the,

in the supply chain.

00;37;23;29 - 00;37;26;19

And this is overall

of that value proposition.

00;37;26;19 - 00;37;29;20

So there's when you paint

that picture, Alex,

00;37;30;00 - 00;37;32;02

you know, I was thinking back

to what you were talking about

00;37;32;02 - 00;37;35;15

earlier in EdTech

and our conversation earlier about

00;37;36;10 - 00;37;39;22

you know, this sense

that this technology is going

00;37;39;22 - 00;37;43;26

to open up these whole new audiences

for what's going on here.

00;37;43;26 - 00;37;47;01

And you also mentioned and I've seen

first hand, by the way, that,

00;37;47;24 - 00;37;49;24

oh, you know, there's companies

not using AI.

00;37;49;24 - 00;37;52;09

And you get this, this hands on approach.

00;37;52;09 - 00;37;56;18

You know, my my concern when I read

all of this together is that sure,

00;37;56;19 - 00;38;00;03

there are companies that will say, yes,

you get there's no AI, it's human only.

00;38;00;22 - 00;38;04;08

And but of course they'll look at that

with, you know, money bags

00;38;04;08 - 00;38;07;18

in their eyes and say, well,

now of course there's this hefty premium

00;38;07;18 - 00;38;11;02

for anything with no AI or for human,

you know, for human touch.

00;38;11;11 - 00;38;14;27

And so we end up with this world

that's, you know, even more extreme where.

00;38;15;04 - 00;38;20;08

Yeah, it's it's AI slop for the masses

that we know is mid or low quality.

00;38;20;14 - 00;38;24;00

And then you have to pay a premium for,

you know,

00;38;24;19 - 00;38;25;24

that the quality of service

00;38;25;24 - 00;38;28;24

that you're either getting today

or certainly were getting a few years ago,

00;38;29;05 - 00;38;29;11

yeah.

00;38;29;11 - 00;38;33;07

We are talking about edtech

and this is, you know, as I look across

00;38;33;07 - 00;38;37;10

the spectrum of not just corporations

but also social services,

00;38;37;10 - 00;38;40;10

it feels like there's

this like this erosion

00;38;40;28 - 00;38;44;07

of quality or, you know, some people

call it like an intuitive ification.

00;38;45;08 - 00;38;47;24

And so, you know, to what degree are

the two of

00;38;47;24 - 00;38;51;20

you seeing this in your research across

some of these different sectors?

00;38;51;20 - 00;38;55;00

And, you know, in your minds,

is there a way that we can reverse this?

00;38;56;21 - 00;38;58;15

Yeah, I mean, it's a good question.

00;38;58;15 - 00;39;01;25

And I think we're seeing a little bit of,

of kind of a rush.

00;39;02;08 - 00;39;05;14

You know, there's definitely this insured

application process this term that,

00;39;06;03 - 00;39;08;29

or a doctor coined.

00;39;08;29 - 00;39;10;24

The thing about it

00;39;10;24 - 00;39;13;25

is that, thinking about what we should do,

the reverse.

00;39;13;25 - 00;39;16;24

I mean, one thing is that

00;39;16;24 - 00;39;21;02

what we hope and I think one of

the hopes of the book is to suggest that

00;39;22;00 - 00;39;24;10

the kind of Lem output

00;39;24;10 - 00;39;27;15

is just not up to snuff for any critical

task.

00;39;27;15 - 00;39;28;08

Right.

00;39;28;08 - 00;39;31;25

And so, you know,

if there's synthetic output,

00;39;31;25 - 00;39;35;27

it is just seen as either

scammy or spammy.

00;39;36;07 - 00;39;40;11

And because Manila Times

where it is, it's as well

00;39;40;21 - 00;39;43;21

is in scams and in spam.

00;39;43;28 - 00;39;49;10

And that's because it's, it's

sort of this thing where output is

00;39;49;17 - 00;39;54;24

where you don't really care

about the quality of the synthetic text.

00;39;56;09 - 00;39;57;03

And you don't

00;39;57;03 - 00;40;00;09

care whether any of these things within

that text

00;40;00;17 - 00;40;04;03

has any verifiable

truth, has any truth claims in it.

00;40;04;09 - 00;40;05;22

You don't really care

about the truth value.

00;40;05;22 - 00;40;09;19

So there's a paper called

ChatGPT is bullshit, right?

00;40;09;19 - 00;40;13;05

And it uses the Henry Frankfurt

definition of bullshit.

00;40;13;05 - 00;40;17;02

The idea that the bullshitter doesn't care

about the truth values of their claims,

00;40;17;02 - 00;40;19;04

they're just trying to reach their goal.

00;40;19;04 - 00;40;20;00

They're there.

00;40;20;00 - 00;40;23;05

Therefore, like Trump

is like that bullshit or par excellence

00;40;23;05 - 00;40;26;08

because it's just like,

get to get from get, get the deal.

00;40;26;08 - 00;40;27;07

Right.

00;40;27;07 - 00;40;32;29

And so to some degree, there is, you know,

like, we can see that any kind of

00;40;32;29 - 00;40;38;11

synthetic output is just that kind of,

you know, detritus is this kind of waste.

00;40;38;25 - 00;40;42;13

And so I don't know if humans

as being the value differentiator.

00;40;42;13 - 00;40;45;14

It's more like, well,

in the best of cases,

00;40;45;18 - 00;40;49;07

anybody that's doing anything

that's worth its salt should be,

00;40;50;05 - 00;40;53;20

you know, done with humans

in the loop in a meaningful way.

00;40;54;17 - 00;40;58;07

I just want to add there that, you know,

you said we might end up in a situation

00;40;58;07 - 00;41;01;16

where the current status quo

becomes the luxury tier,

00;41;01;16 - 00;41;04;22

where you actually have people involved,

and then everybody else gets

00;41;04;22 - 00;41;05;15

pushed down to this.

00;41;05;15 - 00;41;06;06

Well, you've got to deal

00;41;06;06 - 00;41;10;08

with the crappy synthetic system,

and there's no magic bullet here.

00;41;10;08 - 00;41;12;05

We just have to resist it

at every turn. Right?

00;41;12;05 - 00;41;15;01

If this is coming into your school system,

say no. Right?

00;41;15;01 - 00;41;17;10

If this is coming into your workplace,

say no.

00;41;17;10 - 00;41;20;09

And that's part of what we're trying to do

in this book, is to empower people

00;41;20;13 - 00;41;24;20

to use no at every turn

and to recognize it when it's happening.

00;41;25;11 - 00;41;29;05

And also, you know, the

the no can be firm, it can be angry,

00;41;29;05 - 00;41;30;28

but it can also be humorous.

00;41;30;28 - 00;41;32;04

And this is where we recommend

00;41;32;04 - 00;41;36;01

ridiculous practices and saying, this is

this is low value, it's fake, it's bad.

00;41;36;08 - 00;41;39;14

It's, it's,

you know, amid all of these things,

00;41;40;06 - 00;41;43;13

because sometimes the people

who are making the decisions,

00;41;43;24 - 00;41;47;08

it's not moneybags they have in their eyes

with stars and in particular

00;41;47;12 - 00;41;50;10

the sparkle emoji that got appropriated,

00;41;50;10 - 00;41;53;10

by the, the tech companies for this stuff.

00;41;53;14 - 00;41;55;20

It looks like magic.

And so I want the magic.

00;41;55;20 - 00;41;58;20

And so to empower people

to educate those around us

00;41;58;20 - 00;42;01;22

so that we collectively make better

decisions is really important.

00;42;02;05 - 00;42;04;15

It makes it makes complete sense to me.

00;42;04;15 - 00;42;06;29

I wanted to ask the two of you

a slightly different question.

00;42;06;29 - 00;42;11;07

So one of the things I normally ask, you

know, guests that I speak to here is I,

00;42;11;07 - 00;42;12;28

you know,

I ask them what they think is bullshit.

00;42;12;28 - 00;42;15;28

And I'm not going to ask the two of you

that because I think we've spent

00;42;16;01 - 00;42;18;27

we spent quite enough time talking about,

you know, what is bullshit.

00;42;18;27 - 00;42;19;21

And I you know,

00;42;19;21 - 00;42;23;11

I know we've got some, some strong and,

you know, well supported views here.

00;42;23;19 - 00;42;27;07

I wanted to flip the question around

and ask, you know, in this sphere,

00;42;27;29 - 00;42;29;17

what isn't bullshit?

00;42;29;17 - 00;42;31;12

What what are you excited about?

00;42;31;12 - 00;42;35;26

What's a good use of leverage,

generative AI or some of these,

00;42;36;13 - 00;42;40;12

you know, newer or modern technologies

and, and, you know, is it really a case of

00;42;40;16 - 00;42;41;23

say no to everything?

00;42;41;23 - 00;42;44;19

Just shut it off,

you know, full, you know,

00;42;44;19 - 00;42;47;28

you know, hands of work

over your eyes and ears or are there,

00;42;47;28 - 00;42;51;25

you know, very specific, very targeted

use cases where, you know,

00;42;52;07 - 00;42;55;07

there's potential for,

you know, excitement and value.

00;42;55;20 - 00;43;00;19

So I'm very excited

about the uses of language technology

00;43;00;27 - 00;43;03;22

that are for community empowerment.

00;43;03;22 - 00;43;08;07

And so and I'm being very specific here,

I'm not saying LMS and I'm not saying

00;43;08;07 - 00;43;13;18

diffusion models in, in and

or generative AI, but the cases in which

00;43;13;27 - 00;43;18;08

there are things that empower communities

to do things which served community.

00;43;18;08 - 00;43;22;02

So an example that we talk about in

the book is the example take the media,

00;43;22;25 - 00;43;27;03

in which they provide machine translation

and automatic speech

00;43;27;03 - 00;43;31;04

recognition tools for the Terraria

Maori language.

00;43;31;27 - 00;43;35;01

So focusing on

and the thing that's exciting about that

00;43;35;09 - 00;43;38;07

is that they have,

the people in the community

00;43;38;07 - 00;43;41;29

have control over

which data gets used for training models.

00;43;42;08 - 00;43;45;02

They ask their community elders on

00;43;45;02 - 00;43;48;02

what, what data that could be used.

00;43;48;09 - 00;43;50;14

Certain data can't be used.

00;43;50;14 - 00;43;53;17

And those are tools that are in which data

00;43;53;17 - 00;43;56;29

is owned by the community

as well as the computing power itself.

00;43;57;10 - 00;43;59;27

So this is kind of like the anti open AI.

00;43;59;27 - 00;44;04;00

The idea instead of building this big

everything machine that every language

00;44;04;00 - 00;44;09;05

everywhere, you have a very narrowly

scoped task, that works

00;44;09;05 - 00;44;14;05

for the specific community and building

more technical medias is amazing.

00;44;14;05 - 00;44;18;03

And it's one thing that they're trying

to do in powering a net worth

00;44;18;03 - 00;44;22;18

of individuals, through a federation

that we called the freaking Federation.

00;44;23;08 - 00;44;24;29

That is just doing off the ground now.

00;44;26;04 - 00;44;27;27

And those things were really

00;44;27;27 - 00;44;31;26

focused

on building the imagination of folks stuff

00;44;32;02 - 00;44;35;11

and basically doing it in a way

that is not environmentally ruinous,

00;44;35;16 - 00;44;38;16

that is, does not rely on data, but

00;44;38;24 - 00;44;41;24

and is really providing

for a specific need.

00;44;42;00 - 00;44;45;25

So I would like to add there, I'm

a technologist, just like Alex.

00;44;46;05 - 00;44;49;01

I ran a professional master's program

in computational linguistics

00;44;49;01 - 00;44;51;16

training people

how to build language technologies.

00;44;51;16 - 00;44;55;02

So I definitely think there are good use

cases for things like language technology.

00;44;55;02 - 00;44;58;02

And the take media example is wonderful.

00;44;58;16 - 00;45;02;12

But I see no beneficial

use case of synthetic text.

00;45;03;06 - 00;45;05;13

And I actually look into this

from a research perspective.

00;45;05;13 - 00;45;08;21

I have a talk called ChatGPT one

if ever synthetic text

00;45;08;21 - 00;45;12;03

safe, desirable and appropriate,

or those adjectives in some order.

00;45;12;03 - 00;45;13;29

I don't remember the exact title.

00;45;13;29 - 00;45;16;25

And basically it has to be a situation

where, first of all,

00;45;16;25 - 00;45;21;05

you have created the synthetic text

extruding machine ethically.

00;45;21;05 - 00;45;23;21

So without environmental ruin,

00;45;23;21 - 00;45;27;07

without labor exploitation,

without data theft, we don't have that.

00;45;27;07 - 00;45;30;02

But assuming that we did,

you would still need to meet for criteria.

00;45;30;02 - 00;45;31;04

So it has to be a situation

00;45;31;04 - 00;45;34;22

where you either don't care

about the veracity of the output,

00;45;35;03 - 00;45;38;01

or it's one where you can check it

more efficiently

00;45;38;01 - 00;45;40;05

than just writing the thing

in the first place yourself.

00;45;40;05 - 00;45;42;24

It has to be a situation

where you don't care about originality,

00;45;42;24 - 00;45;44;15

because this way,

the way the systems are set up,

00;45;44;15 - 00;45;47;27

you're not linked back to the source

where an idea came from.

00;45;48;15 - 00;45;51;24

And then thirdly, it has to be a situation

where you can effectively

00;45;51;24 - 00;45;54;28

and efficiently identify and mitigate

any of the biases that are coming out.

00;45;55;10 - 00;45;58;09

And I tried to find something

that would fit those categories.

00;45;58;09 - 00;46;02;18

And I don't so certainly language

technology is useful.

00;46;02;21 - 00;46;06;06

Other kinds of well scoped technology

where it makes sense to go from X

00;46;06;06 - 00;46;10;05

input to Y output, and you've evaluated it

in your local situation.

00;46;10;05 - 00;46;10;23

Great.

00;46;10;23 - 00;46;14;08

But you know,

the giant random eight ball y.

00;46;15;09 - 00;46;16;25

But it's interesting, right?

00;46;16;25 - 00;46;20;10

And the y question

is, is very interesting to me.

00;46;20;10 - 00;46;20;19

Right.

00;46;20;19 - 00;46;23;18

Because I think most of most of what

you said there is not controversial.

00;46;23;18 - 00;46;23;25

Right.

00;46;23;25 - 00;46;27;11

It's it's,

you know, hugely energy intensive.

00;46;27;20 - 00;46;32;08

It's probably a lower quality

than what people would come up with

00;46;32;08 - 00;46;33;21

if it weren't synthetic.

00;46;33;21 - 00;46;36;13

And yet, you know,

you have to juxtapose that with the fact

00;46;36;13 - 00;46;40;18

that there is an awful lot of adoption

and people seem to be getting,

00;46;40;25 - 00;46;44;11

you know, value out of that and how,

you know, maybe just what that says

00;46;44;11 - 00;46;48;25

about us as people that were willing

to settle for, you know, something worse

00;46;48;25 - 00;46;52;00

because it's because it's easier

and not worry about the,

00;46;52;22 - 00;46;55;16

you know, any

any of the details behind the scenes.

00;46;55;16 - 00;46;56;00

So it's.

00;46;56;00 - 00;46;58;12

Yeah, I'm

just still kind of thinking about that.

00;46;58;12 - 00;47;01;10

I mean, I, I'm every time

someone says, well, I'm using it

00;47;01;10 - 00;47;04;27

because, you know, I don't have time

or I'm using it because it's easier.

00;47;05;04 - 00;47;07;23

I think it's always worth asking. Well,

why don't you have time?

00;47;07;23 - 00;47;11;06

And if it's easier, what are you

what's the opportunity cost there?

00;47;11;06 - 00;47;14;22

What are you missing out on for

not actually connecting with a person?

00;47;14;22 - 00;47;17;22

To have a conversation

or thinking through something yourself.

00;47;17;24 - 00;47;20;24

And, you know, oftentimes

the source of the problem isn't

00;47;20;29 - 00;47;22;17

the person themselves

who made that decision,

00;47;22;17 - 00;47;25;14

but the the structure

is that put them into the corner

00;47;25;14 - 00;47;27;29

where it felt like

this was the best way out.

00;47;27;29 - 00;47;31;03

But what I got to tell, deans

of universities

00;47;31;03 - 00;47;33;07

from the west coast of the US and Canada,

00;47;33;07 - 00;47;38;19

last year was the only use of ChatGPT

for a university is as a contrast eye test

00;47;38;19 - 00;47;43;09

to see where resources are missing,

where students and staff turn to this,

00;47;43;09 - 00;47;47;02

it means something is lacking

in terms of what

00;47;47;02 - 00;47;50;20

they would need to actually fully engage

in the educational project,

00;47;50;27 - 00;47;54;00

and that information is a value

to administrators.

00;47;54;00 - 00;47;56;12

But like, that's it. That's the end of it.

00;47;56;12 - 00;48;00;18

I will also say the cases in which people

say that, well, I find this very helpful.

00;48;00;18 - 00;48;02;15

I would also note that

00;48;02;15 - 00;48;06;28

I think that's a pretty even term

in terms of the people who are using it.

00;48;06;28 - 00;48;10;09

It's kind of a limited number of workers

who are using it.

00;48;10;09 - 00;48;14;06

So in this survey,

Pew did they found that 17% of

00;48;14;06 - 00;48;17;28

workers

were using Lims at least some of the time.

00;48;19;01 - 00;48;22;01

And I think 1% were using them

all the time.

00;48;22;11 - 00;48;26;08

We're finding cases

which I think there's a mismatch reality

00;48;26;08 - 00;48;29;29

in which especially business leaders

find them to be more useful

00;48;29;29 - 00;48;32;29

than many other people who are more junior

00;48;33;00 - 00;48;36;00

on in in the, job ladder.

00;48;36;16 - 00;48;40;21

There was another story that came out

by Eno Schreiber in the New York Times

00;48;40;21 - 00;48;45;04

that talked about the uses of Lims

at Amazon, effectively, how

00;48;45;14 - 00;48;50;29

the deployment of LMS has become

almost mandatory, and Amazon in the,

00;48;51;14 - 00;48;54;12

in the, in the programing work

00;48;54;12 - 00;48;59;20

and how even though more senior devs

appreciated that, more junior devs

00;48;59;20 - 00;49;03;14

were really being forced to use it,

and their work was looking

00;49;03;14 - 00;49;06;03

much more like factory work

than the kind of creative work

00;49;06;03 - 00;49;10;01

that often goes with with software

engineering and programing.

00;49;10;18 - 00;49;15;01

And so I think that kind of notion

that it is kind of useful is

00;49;15;15 - 00;49;18;07

happens, is probably happening only

00;49;18;07 - 00;49;21;23

for a very narrow

set of workers and people.

00;49;22;15 - 00;49;26;23

Maybe it's happening more for students

because they're being constrained and push

00;49;26;23 - 00;49;31;05

for time, to not engage in classes

to the bane,

00;49;32;01 - 00;49;34;07

of, of their instructors.

00;49;34;07 - 00;49;38;01

And we've talked to a lot of instructors

and our instructors ourselves.

00;49;38;18 - 00;49;41;18

And so that itself, I think is, is,

00;49;41;28 - 00;49;44;28

is another area

where a lot of this is happening.

00;49;45;02 - 00;49;46;23

But I think that narrative

00;49;46;23 - 00;49;50;19

and what's happening across most workers,

there's a big mismatch there.

00;49;51;04 - 00;49;54;14

So just just following that thread

for a minute, Alex, on the worker side,

00;49;54;29 - 00;49;58;23

one of the again,

one of the narratives we hear is that,

00;49;59;08 - 00;50;03;00

you know, there's kind of two sides

to adopting these technologies.

00;50;03;00 - 00;50;05;11

There's,

you know, organizations or enterprises

00;50;05;11 - 00;50;08;11

that can adopt them to try and drive

productivity.

00;50;08;15 - 00;50;09;16

Organizations

00;50;09;16 - 00;50;13;15

and maybe that displaces workers, maybe

that's the do more with less mandate.

00;50;13;21 - 00;50;17;14

But then there's this other narrative

of as an individual worker,

00;50;17;27 - 00;50;22;28

you should adopt some of these,

you know, synthetic language tools

00;50;23;04 - 00;50;26;17

or some of these automation tools

because it's empowering for you.

00;50;26;23 - 00;50;30;06

It helps you take back,

you know, a modicum of control

00;50;30;06 - 00;50;33;09

from your employer

by making you more efficient.

00;50;33;09 - 00;50;34;06

Maybe it takes less

00;50;34;06 - 00;50;37;23

time to do what you're doing,

or you can do a better quality work.

00;50;38;11 - 00;50;40;18

You know, in the same amount of time

that you're doing worse quality work.

00;50;40;18 - 00;50;42;29

Now, do you think there's merit

to that argument,

00;50;42;29 - 00;50;44;21

or do you think

that's still worth resisting?

00;50;45;21 - 00;50;48;14

Yeah, I don't

think it's there's really merit to that.

00;50;48;14 - 00;50;53;02

I mean, I think the cases in which work

these tools have been deployed

00;50;53;02 - 00;50;58;21

have been in cases

which there may be some marginal gains,

00;50;58;21 - 00;51;03;12

but you have to check that

the output of this very meticulously.

00;51;03;24 - 00;51;07;05

There's been many, many different cases

in which people, like

00;51;07;05 - 00;51;10;05

lawyers, have been using tools like this

in legal briefs.

00;51;10;15 - 00;51;13;26

And those legal briefs have made

00;51;13;27 - 00;51;16;27

have a lot of made up case law. And,

00;51;17;06 - 00;51;20;12

there's cases

in which journalists have used them,

00;51;20;12 - 00;51;24;03

and it is made up a bunch of books

that don't exist.

00;51;24;03 - 00;51;25;11

And these are

00;51;25;11 - 00;51;28;11

I think there was a case with the Chicago

Sun-Times, which I think they had,

00;51;29;15 - 00;51;30;24

some of that was in their pipeline,

00;51;30;24 - 00;51;33;24

and nothing was a journalist

at such angles to Chicago Sun-Times.

00;51;34;04 - 00;51;36;01

But they had made a list of fake.

00;51;36;01 - 00;51;38;22

But so there and these are kind

of coming up over and over.

00;51;38;22 - 00;51;44;20

So if you have cases in which you're

you're using these tools with the intent

00;51;44;20 - 00;51;48;07

of being more productive, often

it is doing the opposite.

00;51;48;07 - 00;51;51;08

It's slowing workers down,

making it less productive.

00;51;51;16 - 00;51;55;02

There's also the case

that it makes you less collaborative

00;51;55;02 - 00;51;58;02

because you sort of produce something,

but you are not.

00;51;58;06 - 00;52;02;15

Then passing something on to a coworker

to use it

00;52;02;15 - 00;52;05;26

in any kind of verifiable,

verifiable, or useful way.

00;52;06;04 - 00;52;09;07

There's a good example where,

in talking with animators,

00;52;09;07 - 00;52;12;07

as I was talking with somebody

from the Innovation Guild

00;52;12;08 - 00;52;15;26

who reported that they were under pressure

to use,

00;52;16;16 - 00;52;20;26

Midjourney to produce like an image

and then fix all the artifacts

00;52;20;26 - 00;52;24;13

that come out in the image

and what you get at the bid journey

00;52;24;13 - 00;52;28;12

or stable diffusion or whatever is

you get a PNG or a Jpeg file

00;52;29;00 - 00;52;31;28

and then you might have

to fix the artifacts that come out of it.

00;52;31;28 - 00;52;34;28

But really, what you want to work with

is something like an illustrator file

00;52;35;04 - 00;52;37;18

that has many different layers

to it. Right?

00;52;37;18 - 00;52;39;03

And that's not what it produces.

00;52;39;03 - 00;52;41;01

And so if you're actually

going to fix the artifacts,

00;52;41;01 - 00;52;44;20

you actually have to reproduce the image

with all its, all its layers.

00;52;45;05 - 00;52;48;05

And so that's not actually helpful

as a tool of collaboration.

00;52;48;14 - 00;52;52;05

It's actually breaking

the collaborative pipeline there.

00;52;52;21 - 00;52;53;10

And so

00;52;54;16 - 00;52;57;08

there may be solo, folks

00;52;57;08 - 00;53;01;03

and, you know, individual contributors

that may find them very useful.

00;53;01;10 - 00;53;03;11

But once you get into a larger

organization

00;53;03;11 - 00;53;05;24

and what you have to actually

verified information,

00;53;05;24 - 00;53;07;07

it falls apart pretty quickly.

00;53;07;07 - 00;53;08;27

I would just add briefly to that.

00;53;08;27 - 00;53;11;23

I would ask workers

who are using something

00;53;11;23 - 00;53;14;24

because they feel like it helps them

speed things up to think about

00;53;14;24 - 00;53;17;23

how long they actually get

to maintain the benefits of that

00;53;17;28 - 00;53;20;26

before they're just asked to do more

in the same amount of time.

00;53;20;26 - 00;53;22;14

I think that these kinds of

00;53;22;14 - 00;53;25;24

so-called efficiencies generally

are not going to accrue to the workers.

00;53;26;03 - 00;53;27;09

Right, right.

00;53;27;09 - 00;53;29;27

That the and there's

and there's certainly a risk of that.

00;53;29;27 - 00;53;33;15

So, you know, just kind of tying

a bunch of this conversation together.

00;53;34;17 - 00;53;36;06

We're seeing all these trends

00;53;36;06 - 00;53;40;09

happening right now around the hype,

around adoption, around pushback.

00;53;40;17 - 00;53;43;17

When you look out, you know, over the,

you know,

00;53;43;28 - 00;53;46;19

your outlook for the years to come.

00;53;46;19 - 00;53;50;09

To what degree are you kind of optimistic

or pessimistic about the trajectory

00;53;50;09 - 00;53;56;03

we're on right now and our ability to,

you know, get to, you know, bend the curve

00;53;56;03 - 00;54;00;04

in a way that's actually net positive for,

you know, us as people.

00;54;00;17 - 00;54;02;10

So I am an optimist at heart.

00;54;02;10 - 00;54;04;01

I also don't make predictions.

00;54;04;01 - 00;54;05;05

So I can tell you sort of what

00;54;05;05 - 00;54;11;10

what gives me hope though is you know,

watching people, standing up and saying no

00;54;11;10 - 00;54;14;10

and watching people

adopting ridiculous practices.

00;54;15;03 - 00;54;16;29

Also,

00;54;16;29 - 00;54;20;06

taking a page from Karen

Howe's amazing book Empire of I, she,

00;54;20;06 - 00;54;23;28

she tells amazing

stories of people in, Chile

00;54;23;28 - 00;54;27;23

and Uruguay who organize

to resist the imposition of data centers.

00;54;28;00 - 00;54;32;17

And she makes the point that,

you know, people who have had more power

00;54;32;18 - 00;54;36;18

taken from them nonetheless

maintain agency and nonetheless push back.

00;54;36;26 - 00;54;41;04

And I think that, it is really important

to resist narratives of inevitability.

00;54;41;24 - 00;54;44;28

Even the ones that say, oh, well,

you know, it's here to stay.

00;54;44;28 - 00;54;46;11

So we have to learn to live with it.

00;54;46;11 - 00;54;48;22

That is still a narrative of inevitability

00;54;48;22 - 00;54;51;11

and therefore

still a bid to steal our agency.

00;54;51;11 - 00;54;54;01

But we all have agency

and we can continue to claim it.

00;54;54;01 - 00;54;56;18

I'm less of an

optimist. I think the Leah's,

00;54;58;03 - 00;54;59;21

the pessimistic part of me is just

00;54;59;21 - 00;55;03;01

that there's more and more investment

that's being put into this.

00;55;03;01 - 00;55;03;28

I mean, we have one of,

00;55;03;28 - 00;55;07;26

I think, the largest tech investment round

we've seen with open AI,

00;55;08;04 - 00;55;12;26

an investment round led by SoftBank

that was to the tune of $40 billion.

00;55;13;10 - 00;55;15;26

And so more more money. Sorry after that.

00;55;15;26 - 00;55;19;05

But I think, you know, an optimistic

reading of that

00;55;19;05 - 00;55;22;05

is that this is kind of the last gasp.

00;55;22;05 - 00;55;25;12

It is

we are tossing this much money at it.

00;55;25;22 - 00;55;27;28

This is the big bet.

00;55;27;28 - 00;55;31;15

If you don't come out of this,

this is you know, this is

00;55;31;15 - 00;55;33;02

this is going to be your last chance.

00;55;33;02 - 00;55;37;03

And it's not like

Vassar is known for good investments.

00;55;38;16 - 00;55;40;18

And we work is indicative of that.

00;55;41;25 - 00;55;42;08

And so

00;55;42;08 - 00;55;46;01

it might be the case that we're seeing

a lot of that, that bubble,

00;55;46;18 - 00;55;49;24

really reaching its, its peak in size.

00;55;50;13 - 00;55;53;21

The things that are optimistic

are the kind of ways in which workers,

00;55;53;29 - 00;55;55;19

in particular are pushing back.

00;55;55;19 - 00;55;59;14

We're seeing efforts from the Writers

Guild of America, of course,

00;55;59;14 - 00;56;03;05

that have strong protections

around generative AI in their workplace.

00;56;03;14 - 00;56;06;13

We've seen some of that work

for public workers in Pennsylvania

00;56;06;13 - 00;56;09;25

and from SEIU, the Writers Guild,

00;56;10;07 - 00;56;15;03

and the Authors Guild, also, or, sorry,

the Authors Guild in the animators film

00;56;15;09 - 00;56;18;15

as also being, organizations,

00;56;19;04 - 00;56;21;25

taking a strong line of that,

00;56;21;25 - 00;56;24;01

line against generative AI.

00;56;24;01 - 00;56;26;21

And that really does, give me hope.

00;56;26;21 - 00;56;29;21

And I think we're seeing

a really nice confluence here of people

00;56;29;24 - 00;56;32;24

trying to understand

what's behind all this.

00;56;32;29 - 00;56;36;03

Is this all hype

and what can we do about it?

00;56;36;03 - 00;56;39;21

And I think to that end,

I think our book is is very helpful,

00;56;39;21 - 00;56;42;21

and I hope it's a tool for folks

seeking that out.

00;56;42;25 - 00;56;43;20

Amazing.

00;56;43;20 - 00;56;46;00

I appreciate the thorough answer,

and I wanted to say a big

00;56;46;00 - 00;56;48;12

thank you to each of you, Alex

and Emily for joining today.

00;56;48;12 - 00;56;50;08

I thought this was a really fascinating

conversation.

00;56;50;08 - 00;56;52;10

And I appreciate

your time. Thank you so much.

00;56;53;10 - 00;56;54;19

This is great to talk to you today, Jeff.