Latest / Radio Davos / AI and inequality: Nobel-winning economist Daron Acemoglu on how to get AI right

Transcript

AI and inequality: Nobel-winning economist Daron Acemoglu on how to get AI right · Radio Davos

0:00
AI is going to remake every aspect of our lives. And not just our lives.
0:04
The lives of 6 billion people who are outside of the United States and China,
0:08
they have no voice, and we have a voice.
0:10
We still live in a democracy.
0:12
It is our responsibility to have a say about how our future
0:15
is being remade, even if it's being remade by two or three companies.
0:19
It doesn't mean that their CEOs have the right to make every decision.
0:23
I think the democratic input is a very important part of that.
0:29
Welcome to Radio Davos.
0:31
This is the podcast that looks at the biggest challenges and how we might solve them.
0:36
I'm Linda Lacina stepping in for Robin Pomeroy
0:38
this week as we had the chance to snag a very special guest
0:41
in our World Economic Forum's New York City offices.
0:45
Our guest is a renowned economist, an MIT professor,
0:49
and a 2024 Nobel Prize winner.
0:52
He is also the world's foremost expert on AI's impact
0:56
on the economy, which is one of the most critical issues of our time.
0:59
He has a new book out.
1:01
It is What Happened to Liberal Democracy. And a lot of very smart
1:05
people are calling it a “tour de force,” a “must read,”
1:08
and even “the most optimistic thing that they've read in a very long time.”
1:11
He is going to explain the crisis as he sees it, crisis
1:15
that is widening the economic divide, what's needed to fix it, and also the role
1:19
that AI, especially pro-worker AI, could play in solving this issue.
1:24
Welcome to Daron Acemoglu.
1:26
Thank you, Linda. It's my pleasure to be with you.
1:28
We are very, very honored to have you.
1:30
Why don't we get started, dig right in.
1:32
What is this crisis?
1:33
What is the biggest crisis that is facing the economies and humanity
1:37
right now as you see it?
1:39
I believe that liberal democracy
1:43
is one of humanity's greatest achievements.
1:46
If you look at our past,
1:49
the million-year revolution,
1:53
the existence in fighting bands,
1:57
or very unfree societies,
2:01
people at that point
2:04
would have had a very hard time imagining
2:07
that we could build a complex society with very different people.
2:12
But all cohabiting, participating in politics mostly peacefully,
2:17
and using state institutions
2:20
for creating also shared prosperity,
2:24
meaning economic growth from which many people,
2:27
most demographic groups, are also taking their fair share.
2:32
That is just an incredible achievement.
2:34
But the more immediate challenge is that I think that achievement is now in danger.
2:39
And we need to understand why liberal democracy is in crisis today
2:43
and how we can rebuild it because liberal ideas.
2:47
And liberal democratic institutions, I am convinced
2:50
have a special role to play in our future as well.
2:54
And we can make them better so that they can play that role.
2:57
And just to get our listeners and our watchers sort of all on the same page,
3:02
can you define liberal democracy
3:04
as you see it and as you wrote about it in the book?
3:05
Liberalism, at its heart, is about individual freedoms,
3:09
freedom of expression, association, religion,
3:14
but also
3:15
about turning those into an actual behaviour
3:20
in which people can realize their aspirations.
3:23
And that requires state institutions
3:27
that provide opportunities, education, infrastructure,
3:31
insurance against very tough things happening to people such as disability
3:37
or inability to earn a living from their families,
3:42
but all the time
3:44
consistent with liberal principles.
3:47
And liberal democracy is the most successful
3:51
institutionalization of those liberal ideas because it brings
3:56
elements of liberal thought
3:58
and principles together with democratic participation.
4:02
If we're not able to
4:04
get liberal democracy out of crisis, what happens? Today,
4:08
liberal democracy is viewed
4:10
as the establishment.
4:14
And anti-establishment feelings are everywhere.
4:17
And not only those movements
4:21
that challenge liberal democracy are illiberal.
4:24
They're not democratic.
4:26
They don't want to provide voice and ability to self-government to people.
4:29
But they are often Ideologies that separate people
4:33
create more divisions, and also, they are
4:38
stances that have not been able to deliver
4:41
any kind of welfare or improvement to people.
4:46
So we really would be looking at both bad ideas,
4:52
sometimes hateful ideas,
4:54
but would also have to put up with worse
4:57
corruption, more arbitrary rule,
5:00
no equality before the law, and ultimately nothing
5:03
like shared prosperity, which we are still aspiring for.
5:08
The fault lines started around 50 years ago,
5:10
you started to see them.
5:11
Let's talk a little bit about that to understand what's shaping
5:13
this, what's really putting shared prosperity into great risk.
5:17
Can you take us through what changed
5:19
at the end of the industrial era to the post-industrial era?
5:23
What was this change that we saw?
5:24
I think if you want to understand what's going on with liberal democracy,
5:28
you have to start with diagnosis.
5:30
I would say the roots of our crisis go back at least 50 years.
5:35
Liberal democracy's problems are rooted in the transition
5:38
from industrial society to a post-industrial society.
5:43
By post-industrial, I mean an economy in which
5:47
digital technologies play an important role.
5:51
Globalization
5:53
combines supply chains
5:55
and flow of goods much more seamlessly around the world.
5:59
And these processes fundamentally
6:03
change the nature of the economy, in particular
6:06
making it much harder to reach shared prosperity.
6:10
Not impossible,
6:12
and there's nothing inevitable about either globalization
6:14
or digital technologies.
6:16
But the way that things have transpired,
6:19
in the United States and many other industrialized economies,
6:23
was that one of the major channels for creating shared prosperity,
6:28
which went from businesses innovating, expanding, producing more,
6:34
to workers also benefiting,
6:36
which was through businesses' desire to hire workers.
6:40
And in a democratic framework with strong labor unions,
6:44
that meant they had to pay higher wages, as well as create jobs.
6:47
And that was something that created a powerful drive towards shared prosperity
6:52
because it meant that greater profits and greater revenues for businesses
6:57
came hand-in-hand with higher wages.
7:01
Digital technologies changed every aspect of that bargain.
7:05
First of all, they started favoring more college-educated workers.
7:08
Who could take managerial positions,
7:12
work with data, administrative positions.
7:15
But something else more important was also at play.
7:18
That was the use of digital technologies as a platform for automation.
7:24
By automation, I mean
7:27
the use of machinery and algorithms
7:30
to replace tasks that were previously done by workers.
7:34
And automation is a game changer because it severs
7:36
the link between business expansion and shared prosperity.
7:40
I want to be very clear that this is not in the nature
7:44
of digital technologies that they always have to increase inequality.
7:47
It's the way that we developed and used these technologies.
7:50
Equally important,
7:51
post-industrial society brought very different social relations.
7:56
It elevated the social status and political power of college-educated
8:00
workers, especially those integrated into the global economy
8:05
in major urban centers, such as in New York or Boston.
8:08
But at the same time, they started segregating
8:11
into their own neighborhoods, their own social networks,
8:14
their own communities. And diverging
8:17
from the rest in terms of values.
8:21
So, this created a tension in a broad coalition
8:24
that was essential for liberal democracy.
8:27
For example, you can see that in how center-left
8:31
parties, the Labor Party in Britain, social democratic parties in Europe,
8:34
or the Democratic Party in the United States,
8:36
which used to receive the majority of their votes from working class people,
8:40
people without a college degree, working in manual or service jobs.
8:44
Now that demographic group votes for the center-right or the hard-right,
8:49
and center- left parties are overwhelmingly
8:53
the party of choice for people who live in large
8:56
urban areas with college degrees or postgraduate degrees.
8:59
But center-left parties are essential for liberal democracy
9:03
because they have been the pioneers of the welfare state,
9:06
of shared prosperity, supporting the rights of workers.
9:09
And now
9:11
they became less
9:15
concerned with the rising inequality
9:18
or the non-college groups in society falling behind.
9:22
and then ultimately, with the cultural politics of our era,
9:26
even became a part of a big divide
9:30
within society that has created a major backlash.
9:33
There's a lot of very unique examples you have in the book.
9:36
In fact, you can correct me if I'm
9:39
remembering this wrong, but in Ohio, the Anchor Hawking Company,
9:43
where all of the people, whether you worked in the factory,
9:46
you worked at the office, you were a manager,
9:48
everybody's kids went to the same school,
9:49
everybody lived on the same block.
9:51
And then within a few decades, that was very much not the case.
9:54
And certainly all of us sort of understand that there's certain areas
9:56
where elite people live and certain places where they don't.
9:59
How important are those examples to helping drive home
10:02
to people that how these ideas walk and talk in there every day?
10:05
You summarize it so well, thank you for doing that,
10:08
but they really tell the story.
10:11
Not only has there been a big sea change
10:14
that you don't see managers and workers living in the same communities,
10:20
sending their children to the same school, going to the same churches,
10:25
but you have also
10:27
broken one of the major ways in which you learn
10:30
about the plight of other people or learn to empathize with them.
10:35
And that has accelerated in the age of social media
10:38
and online communication because we are interacting
10:42
with each other much less in natural social networks.
10:46
And that's very important for recognizing
10:49
something that really in hindsight is shocking.
10:53
In the 2016 election,
10:57
when Hillary Clinton was defeated by. Donald Trump
11:01
in the United States, or in the 2016 Brexit referendum,
11:05
most people who were very well-educated, living in major cities like London, New
11:10
York, Boston, were genuinely shocked that there was so much reaction
11:17
against establishment. But that showed the extent. Of the discontent,
11:21
but the shocking thing is how such strong feelings
11:26
went unnoticed for almost three decades,
11:29
and that is because we came to live in silos.
11:33
The major newspapers that we read are part of that silo.
11:36
Who we talk to is part of the silo,
11:37
our education system is part of that silos,
11:40
and it's not just a failure of our information ecosystem,
11:43
but it also makes it much harder for us to form bridges with each other, to form
11:47
that's broad consensus that are so important for democracy to work.
11:53
People forget or overlook
11:54
the connections between political institutions and economic prosperity.
11:58
What is the before and after that? How does one feed into the other?
12:01
Institutions are critical.
12:03
Part of the reason why I am drawn to study and want to
12:09
improve
12:10
liberal democracy is because I believe liberal democratic institutions
12:14
are essential for the kinds of lives that we aspire to:
12:18
Peaceful, based on dialog and understanding,
12:22
prosperous, and free.
12:24
If we start sacrificing our institutions,
12:29
we risk losing all of these things very badly.
12:32
That's why the book talks both about liberal ideas,
12:35
how they have evolved, how they worked and why they were important
12:38
for shared prosperity voice and other things.
12:40
You write in the book, you say, liberalism's crisis is intimately linked
12:44
to its failure to adapt to a changing world.
12:47
And you argue that we should remake liberalism.
12:50
Why is a new framework needed?
12:52
First of all, clearly something's not working.
12:56
There is a real crisis here.
12:58
People are willing to put to power, ideas, leaders
13:03
that really are very questionable in terms of their corruption, in terms
13:07
of their incompetence.
13:08
But it shows the strength of the feelings.
13:11
I think liberalism needs a new governing philosophy.
13:15
You cannot remake institutions without something that brings people
13:19
together around their shared moral values and an understanding
13:23
of how you could build a better society.
13:25
That's what I mean by a governing philosophy,
13:28
which is a collection of coherent ideas need to be part of that.
13:33
In the past, for instance, Social democratic ideas played a very important
13:37
role in the development of the European welfare state, in the affirmation
13:42
of the desirability of shared prosperity, of providing much higher quality
13:46
public services to poorer parts of the country.
13:50
So, we do need something like that
13:52
again. Social democracy was for the industrial age,
13:55
so that's why I emphasize something
13:57
related, but different, what I call working class liberalism.
14:01
And that is my rallying cry for this sort of new governing philosophy.
14:06
You say AI, could enable this shift.
14:09
Right.
14:10
You say, AI, if steered correctly, can be complementary to the expansion
14:13
of collective knowledge and the rebuilding of shared prosperity,
14:16
and then you introduce
14:17
this idea that we mentioned earlier in the introduction of pro-worker AI.
14:20
Tell us, what is pro-worker AI?
14:22
Well first let me take a step back
14:23
and again I'm sort of partly responding to your previous question
14:27
because I didn't want to go on and on
14:29
but working class liberalism at its heart has to be about
14:34
shared prosperity, jobs, and better economic conditions for the working class
14:38
people who don't have a college who are going to work in
14:42
manual jobs, service jobs and a variety of trades.
14:46
So that's essential.
14:47
The other very big, important element.
14:50
In working class liberalism is community.
14:54
So, you have to bring communities in, give them room for self-government,
15:01
to make different choices,
15:03
experiment with their different values,
15:06
and do that while being consistent
15:09
with the founding principles of liberal individual freedom,
15:12
individual rights to expression,
15:15
and this drive for people to have voice.
15:19
Now, taken at its face value,
15:22
AI is a threat to both of them.
15:24
If AI goes in the AGI, artificial general intelligence,
15:28
and an artificial superintelligence direction,
15:31
it could displace
15:34
a very large fraction of the workforce from their jobs
15:36
in the United States, in Europe, ultimately in the emerging world.
15:40
And that would make shared prosperity essentially impossible.
15:43
It would make workers dispensable,
15:46
not so important, second-class citizens, so to speak.
15:51
It would usher in an era of inequality
15:54
far surpassing our very, very unequal present.
15:58
But AI would also, could also,
16:01
continue the
16:02
trend of social media and pacify us more,
16:06
take us away from our communities,
16:09
don't encourage us to get into the messy but ultimately essential business
16:14
of making compromises in real world interactions, getting to understand
16:19
different parties, different segments, perspectives, form bridges,
16:25
deliberate with them.
16:26
All of these things are a real
16:29
threat to the future of liberal democracy.
16:33
But here is the catch.
16:34
There's nothing in the nature of AI
16:37
that it should do those things.
16:38
It's how we develop AI and how we use AI
16:41
and the idea of pro-worker AI. It's about how we should conceptualize
16:45
AI and how we could achieve it and why it might actually be complementary
16:49
to building shared prosperity and making a better liberal democracy.
16:52
And how does it walk and talk, pro-worker AI?
16:55
My argument is that AI
16:58
is not just a
16:59
very powerful generative technology that could be used for automation
17:04
or become the basis of something like artificial general or super intelligence,
17:09
but it's also an amazingly versatile and powerful
17:13
information technology, meaning that it can provide
17:17
very relevant, accurate, context
17:21
relevant. Information to individuals in real time.
17:25
What that promises then is that all of us could become much better
17:30
at solving problems in which we specialize, expanding our expertise,
17:34
starting doing new things, new tasks, more innovation, provide better
17:38
or different goods and services in our line of business.
17:42
And those are very different from AGI or artificial super intelligence
17:48
because they still have the worker in a critical part.
17:55
We are making the worker's capabilities expand.
17:58
We're making the workers perform new tasks. As a result,
18:02
rather than being sidelined,
18:04
this could increase demand for workers and re-energize that channel
18:08
that I emphasize.
18:11
Business is doing
18:12
well, but also paying higher wages and creating jobs.
18:16
And that's so important
18:17
for the future of shared prosperity and the future of liberal democracy.
18:22
And what holds us back?
18:23
Oh, that's not the direction in which we're going.
18:26
Right now,
18:28
there is more than a trillion dollars being spent on AI,
18:32
and a very large fraction of that
18:36
goes into building more and more sophisticated foundation models.
18:39
They're very, very impressive,
18:40
by the way.
18:41
I will not sympathize with anybody who says these are just stochastic parrots.
18:45
These are really capable models. And. And we can benefit a lot from them,
18:49
but we also need to develop them in a way that could be complementary to humans.
18:55
That's the pro-worker agenda.
18:56
I give many examples in the book or in my other work, but
19:00
there are so many places where you can see the promise of pro-workers AI.
19:05
You can make nurses far better.
19:07
They can play a much more important role in diagnosis, care,
19:11
prescribing drugs, making follow-up decisions.
19:14
You can made educators much more effective,
19:17
essentially a single teacher can become -- with the help of AI --
19:22
much better at doing new things
19:24
in terms of reorganizing the classroom and coming closer to a cost-effective
19:28
but quite competent version of individualized education programs.
19:33
You can make electricians or plumbers and all the trades much better
19:36
because the problems that they are solving are becoming more complex.
19:39
Now you can give them tools that expand their expertise
19:44
and enable them to do many things than they could do.
19:47
All of us working in knowledge occupations obviously
19:50
can benefit from AI in the same way as well.
19:53
All of us have heard the promise
19:55
of these wonderful things that could happen.
19:56
But still,
19:57
plenty of folks are equally worried, equally, like, gnawing on
20:00
their fingernails. Why is it that this might not happen?
20:03
Oh, they are absolutely 100 percent, 1,000 percent right to be worried.
20:06
This is not the direction we're going.
20:08
And what holds it back?
20:10
The kind of pro-worker AI I'm describing is actually
20:16
not a big technological barrier.
20:21
It is within our capabilities
20:23
right now.
20:24
There are a few companies that are working here,
20:26
but they are marginal companies without the best scientists,
20:29
the best models, the GPUs, and they are making progress.
20:33
So, we need much more of the resources to go in this direction.
20:37
But it's not what's happening.
20:38
It's not happening
20:39
because of the economic model of Silicon Valley. But it is also
20:43
because of ideology of Silicon Valley, where
20:46
all of the leaders are prioritizing
20:50
building more and more human-like or human-improving models.
20:54
You mentioned the algorithmic cage.
20:57
Can you talk a little bit about that and how that impacts how we think
21:00
and maybe also the direction we might go in?
21:03
Communities, almost by their nature create these cages of norms.
21:07
They have their own norms.
21:08
They try not to let people break out of them.
21:11
When communities do that, we still have to protect individual liberty
21:15
and that's the complex relationship between community and freedom
21:18
that is at the heart of the book as well. What I call the “algorithmic cage” is
21:24
that when the same thing is done, say, by social media,
21:26
you don't get many of the benefits, but you get an even more powerful version
21:30
of that cage of norms because now the algorithms
21:34
are policing you
21:35
or the algorithms are finding other people who are going to very strongly police
21:39
you and are going be very, very militant
21:43
if necessary in forcing you into certain modes of behavior.
21:48
And all of that happens without the shared moral values, the shared understanding,
21:53
the development of information exchange, et cetera,
21:57
that community brings.
21:58
There's an interesting, anecdote you had in the book where after the pandemic
22:01
you went to a conference and you were all excited,
22:04
you're like, oh gosh, this is a big deal,
22:05
we're going to talk about wage stagnation, and this is really important thing,
22:08
people need to be digging into this,
22:09
it has all these other implications,
22:11
and then folks didn't talk about that,
22:12
and you're, like, what? What was going on?
22:14
And why am I bringing that up?
22:17
Because do you think that people are focused on the wrong signals?
22:21
To some extent, yes.
22:21
And that's, again, it's because of the algorithmic cage.
22:25
And this is because of
22:28
that division between the college-educated
22:32
groups that have become very dominant in politics
22:35
throughout Europe and the United States and the less educated.
22:40
So, wage inequality
22:42
effects all of us.
22:43
But it's really the burden is on people without a college degree in the
22:47
United States. It's an extreme case, but it's a very telling extreme case.
22:51
While wages of college graduate Americans have increased,
22:55
those of post-graduate Americans have increased very rapidly.
22:58
By the 35 years from 1980 onwards,
23:03
the real inflation-adjusted earnings of non-college
23:08
workers, non-college men declined almost every year.
23:12
So, inequality discussions have to start there.
23:16
That is just phenomenally. Large change.
23:19
But who are those bearing that burden?
23:24
Those
23:25
are the non-college workers.
23:27
And if you start thinking that the noncolleage workers
23:31
have the wrong values, they're not coming part of your coalition,
23:35
you stop empathizing because you're not interacting with them as much,
23:39
then that goes to the second rank of the problems that you focus on.
23:44
And that's exactly what I witnessed.
23:46
You make the point in the book that business models can also influence
23:50
political frameworks.
23:51
Can you tell a little about that connection
23:53
and how maybe that can help unlock better perspectives for AI.
23:57
I think the most important aspect of the business model
24:01
that we need to debate
24:05
is what
24:08
managers, CEOs,
24:11
view labor as.
24:14
If you think that labor is a cost,
24:17
that is going to
24:20
push you more and more towards automation, because if something is a cost,
24:24
you just try to cut it. That's the right thing to do.
24:26
But labor, your workers are also your main asset.
24:30
And rather than just trying to
24:34
limit the costs, what you should do,
24:37
is Ask yourself, how can I make that key asset be more productive?
24:42
That's where pro-worker AI would come.
24:44
That's one aspect.
24:45
Another aspect is the business model of,
24:47
for example, social media.
24:49
The social media mess that we have today,
24:52
and I say very clearly in the book, social media is not responsible
24:57
for the crisis of liberal democracy, but it is exacerbating it.
25:01
But that mess is very related to the business models
25:04
of the social media companies, which is capture people's attention.
25:09
And then make
25:10
money by sending them individualized digital ads.
25:13
That's a
25:13
particular type of business model, but it's not consistent with the promise
25:18
of digital technologies and online technologies that people believed in
25:21
in the end of 1990s or early 2000s when,
25:25
if you remember, everybody was talking about new economic democracy,
25:29
news democracy, more organic forms of journalism, better communication.
25:35
And they were looking at what was happening in the Arab Spring
25:38
and celebrating that as a gift of Facebook and Twitter.
25:42
Well, these people weren't fully naive.
25:44
Of course, their analysis turned out to be incorrect,
25:48
but their aspirations weren't naive
25:50
because digital technologies are fully capable of doing that.
25:53
They can create new public spheres for us.
25:56
They can provide us with better information and more effective
26:00
ways of forming bridges and sharing information with each other.
26:03
But it's just that, social media's business model isn't that.
26:07
If I'm a business leader listening to this, and I read your book,
26:11
and there's that line where you say, you know,
26:13
“we have to get better at identifying pro-worker, sorry, pro-workers technologies,” right?
26:15
We have to keep identifying pro worker technologies.
26:16
What do I do?
26:17
What is my instinct that I'm moving towards.
26:19
and what's the thing I should be doing instead?
26:21
Can you help them sort of right side?
26:23
Business leaders who are actually producing
26:26
things are very sympathetic. But what can they do?
26:30
First of all, they need tools
26:32
from the tech sector to actualize pro-worker AI,
26:36
and they're not getting that.
26:37
Second, we've created expectations in the stock market through their boards
26:43
that unless you're going lockstep
26:46
in automating work using AI, you're falling behind your competitors.
26:50
So, they have to demonstrate to their boards
26:51
since they're shareholders that they're doing that.
26:54
So, they're actually between a rock and a hard place as well.
26:58
That's why the first step is changing the narrative.
27:01
Davos
27:03
has been at the heart of this agenda of automation
27:06
and businesses should make more profits, they should get leaner,
27:09
we should build more complex networks of supply chains.
27:13
I think Davos people should also ask themselves that question.
27:17
Is this the right direction that we're pushing as the aspiration for our future?
27:22
Nobody thinks that they're taking anything into a bad direction, right?
27:25
Everybody is trying to do the right thing.
27:28
I run the Leadership Podcast:
27:29
Everyone's a great listener, everyone's kind.
27:31
To help them have a gut check,
27:33
is there a question that, anyone listening to this or watching
27:36
this can be asking themselves in different moments, different decision
27:39
points to help gut check if they're going in the right direction?
27:42
I can think of two kinds of questions that should help.
27:45
One is, if you actually start spelling out
27:50
the kind of
27:51
future we are going towards with more and more automation,
27:55
I think ultimately, you will realize that it's not a sustainable one.
28:01
You cannot have democracy
28:04
if a growing fraction of the population is poor,
28:08
is dispensable. It feels and is
28:12
viewed in society as not contributing anything to our collective good,
28:17
I think that's a dystopian future.
28:18
Second, when you actually think about your organization,
28:21
what makes a real difference isn't whether you cut cost by 2%, or 3%,
28:24
4%, or 5%. It's whether you do new things.
28:27
The companies we remember from the past, such as Ford
28:30
Motor Company, are not the ones that cut costs.
28:34
They're the ones revolutionize production by doing new things,
28:37
and for that, you need your most creative workers, and you need to give them better tools.
28:41
What are the biggest misconceptions about
28:44
how AI is impacting jobs.
28:48
What’s frustrating you about the conversation?
28:50
I would say mostly two things.
28:53
The first one is the media's
28:56
emphasis on this false dichotomy
29:01
that either AI
29:03
is going to bring boundless prosperity or it's going to create
29:07
killer robots that destroy human civilization.
29:11
The concerns that I'm expressing are somewhere in the middle.
29:14
There will be advances, but we push these advances in the wrong direction.
29:19
The media
29:22
and thought leaders have a role in
29:24
making sure that the relevant kind of information is provided
29:29
to the public on this, and we've failed at that.
29:33
Second is the notion that somehow
29:37
any democratic input
29:40
into the direction of AI is stopping progress.
29:45
So, you hear that from many leaders in the tech sector
29:49
or their supporters that
29:54
if you want a better future, let these geniuses work.
29:59
And that is the worst attitude you can have.
30:02
AI is going to remake every aspect of our lives, and not just our lives.
30:07
The lives of 6 billion people who are outside of the United States and China,
30:12
whose lives are going to be shaped by AI as well.
30:15
They have no voice, and we have a voice. We
30:17
still live
30:17
in a democracy.
30:18
It is our responsibility to have about how our future is being remade,
30:23
even if it's being remaked by two or three companies. It doesn't mean that
30:26
their CEOs have the right to make every decision.
30:30
I think the democratic input is a very important part of that.
30:33
If we get AI right,
30:35
how does our everyday look different?
30:37
In the same way that our everyday started
30:40
looking different after electricity became part of our lives.
30:44
Not everything went smoothly, but we started being much more productive.
30:47
Our factories became much more modern and effective.
30:52
Our lives became much better with household appliances that used
30:55
electricity. In the same way our lives started becoming better
30:59
after the internet, especially before the internet started
31:03
going in the wrong direction, especially with social media,
31:06
where we had access to much better services and information
31:10
with the early internet. We could do many things that we couldn't do before.
31:13
There was more competition from companies to provide goods and services to us.
31:19
And many, many, many new creative companies could be built via the internet.
31:25
So, AI is going to do many of those things if we play our cards right.
31:30
And what's the tell
31:31
it's going in the wrong direction?
31:33
If we start seeing a particular statistic or a particular trend,
31:36
what's something that should make us all set up a little bit straighter?
31:39
One telltale sign of automation
31:43
is that labor's share of value either in industry
31:48
or labor's share in national income is going down.
31:51
We've been experiencing that for the last three decades.
31:54
And if there is an acceleration in that, that's a sign
31:57
that we are putting all of our eggs in the automation basket.
32:00
There is nothing wrong with automation, by the way.
32:02
Automation has been going on since 1750, and it's going to be a
32:06
part of our future as well.
32:07
The problem is if we just do automation and we don't do the other things,
32:11
creating that human complementary AI building on that pro-worker AI.
32:16
And that trend with that statistic, if it goes in the wrong direction,
32:20
you know, why is it so important that we avoid that fate?
32:23
Because it would usher in an era of unimaginable inequality
32:29
and it would destroy our liberal democracy. What else?
32:34
What about job destruction?
32:36
Is there a job destruction sort of statistic, some kind of a turning point
32:40
we would get to where it's like, okay, gosh, now
32:43
there's no turning back or we need to start kind of finding a new way?
32:47
What are your thoughts about that as a tell?
32:49
I don't think there is a job destruction threshold,
32:53
because if you actually look at the US economy, even when it was creating shared
32:58
prosperity, it was destroying a lot of jobs and creating a lot of jobs.
33:02
So, I'm not saying that we should ultimately always have
33:05
3 and a half million customer service representatives in the United States.
33:09
Some of those jobs will go away.
33:11
But here's the problem.
33:13
If we destroy 3.5 million customer
33:16
service representative jobs, and we create no opportunities for these workers, that's
33:20
a very different world. And we actually give them opportunities
33:24
to upskill and do even more meaningful and higher paying jobs.
33:28
And you can see some of that contrast by comparing
33:31
what happened when robots were introduced in Michigan versus
33:36
in Germany. In Michigan. In Detroit, Flint,
33:41
when robots were introduced by Ford, General Motors, Chrysler,
33:44
and their suppliers, they laid off workers,
33:48
operators, welders, painters, et cetera.
33:52
In Germany,
33:52
what they did is that they took those workers
33:54
and they trained them to become technicians and higher-skilled workers.
33:59
The implications
34:00
are very different, both for productivity and for inequality.
34:03
Is there a change that you would like to see happen in the year ahead in 2027
34:08
that gosh, you know
34:09
If we can kind of make this steer that'll be a step in the right direction
34:13
I think the ultimate huge success which is impossible next year
34:17
or even in a few years is that the conversation changes
34:22
such that a notable share of the amazing talent
34:28
that's going into the tech sector right now goes there not to achieve AGI
34:33
and make their science fiction dreams come to reality.
34:37
But they go there because they think that AI can be developed
34:40
in a pro-worker direction to improve our communication,
34:43
to make our agency better rather than worse.
34:47
And what will it take to to get there to make that change?
34:50
I don't know
34:51
I don't know. I wish I knew.
34:52
It's not in here! It's not in there,
34:54
but I'll tell you,
34:56
it's a social change.
34:58
So, it's change of norms, first of all.
35:00
But government policy can help.
35:02
The government cannot steer AI without any constraints.
35:06
The government can not be the entrepreneur that does AI.
35:09
But the government sets the framework.
35:12
Taxes influence what happens. The United States
35:16
and many other advanced economies have, in my opinion,
35:20
a very distorted tax system
35:23
that benefits capital, which benefits many people who are part of Davos
35:27
because they make their money from capital, and punishes labor.
35:32
Capital is taxed at such a lower rate.
35:35
That's not just unfair
35:37
and inequality-inducing. It's actually bad for the future of AI,
35:40
because what it does is that it sends a signal to firms, a powerful signal,
35:45
that they should automate, because the government is subsidizing automation,
35:48
so we have to change those tax policies.
35:51
We also need support for more
35:54
socially productive directions of AI.
35:57
We need new laws so that people's data are protected,
36:00
and you encourage experts to invest in high-quality data. That would be good
36:05
for all kinds of AI models, but especially for pro-worker AI models.
36:09
So, there are many things that governments can do,
36:10
but it's not gonna be a silver bullet.
36:12
I may have a little rapid response to the section here to kind of close us out
36:17
where I'll give you some statement that we've all heard about
36:19
sort of the inevitable direction for work and tech
36:24
and then you'll tell me whether you think it's going to happen or not.
36:26
How about this one:
36:28
One day we'll all work for AI agents.
36:31
I don't think so.
36:32
AGI, I think, is not something we're going to be able to achieve
36:36
anytime soon, even though AI models
36:40
are getting better. And there will always be people who control AI as well.
36:43
Not all of us will work for AI agents.
36:45
We'll have a four-day workweek.
36:48
Possible. I'm not sure.
36:50
But I'm also not sure that's a solution to our problems.
36:53
Okay.
36:54
Because if you want to have a society in which there is wage
36:59
growth and shared prosperity, you need to make labor more
37:03
productive.
37:05
Whereas the four-day drive comes from the idea that there's
37:09
a fixed amount of jobs or a shrinking amount of jobs.
37:11
And we have to divide that and better among the remaining humans.
37:15
And that's why we're going to make each woman work less.
37:17
It's very different from saying,
37:19
you know. We're getting richer and richer.
37:22
We have a lot to contribute to the production process,
37:24
but we also have to do a lot of learning because we have longer lives
37:27
and a more complex economy.
37:29
Let's take that time for leisure or building our skills.
37:32
That's very different.
37:32
If that's the drive for a four-day week, I'm much more sympathetic.
37:36
Or if the drive of a four-day week is that jobs are disappearing,
37:39
there's nothing we can do about it. We're throwing the towel against AI.
37:43
Let's cut it to four days or three days.
37:45
That's a very different proposition and I'm not that much in support of that.
37:48
Generalists, not specialists, will be the kings of the AI era.
37:53
There could be some truth to that,
37:56
if by generalist you mean people who have the flexibility
37:59
to change what they do and holistically combine different perspectives,
38:04
especially in a more innovative and creative way.
38:07
We'll have pro-worker AI.
38:09
I wish, I wish I wish.
38:11
How likely do you think it is?
38:13
I am not giving up hope, but we do need a change of direction.
38:17
Would you put a probability on it?
38:20
I am very bullish
38:22
that liberalism has the energy to recreate itself ultimately.
38:27
And I'm very bullish if we put the effort
38:31
to create pro-worker AI, we would do it with very good social outcomes
38:35
as a consequence. Whether we'll make those investments
38:39
in both liberal democracy and pro-workers AI,
38:42
I would say that's certainly
38:45
not more likely than 30 or 40 percent.
38:49
So, my optimism is about what we can achieve, not whether we'll
38:53
actually get our act together and achieve it.
38:55
And your last question, sir, what should leaders listening and watching this,
38:58
what should they be prioritizing in the year ahead?
39:00
I believe understanding AI is very, very important,
39:05
but caring about liberal democracy's
39:08
future, our ability to come together around
39:12
shared moral values that bring us together rather than divide us apart.
39:16
I think that's very, very important as well.
39:18
And thinking about the working class,
39:21
inequality, jobs, are first steps in that direction.
39:24
Thank you very much,
39:25
Daron. Oh, thank you, Linda. It's an honor for us.
39:27
Oh, it's my pleasure.
39:27
That was great.
39:28
And that is all the time that we have today.
39:31
Thank you so much, Daron.
39:32
Robin Pomeroy will be back as host next week.
39:35
You can check out Radio Davos and my podcast, Meet the Leader, at
39:40
wef.ch/podcasts.
39:41
Thanks again for listening and watching.
39:43
Have a great day.